The Cryosphere 7 419ndash431 2013wwwthe-cryospherenet74192013doi105194tc-7-419-2013copy Author(s) 2013 CC Attribution 30 License

EGU Journal Logos (RGB)

Advances in Geosciences

Open A

ccess

Natural Hazards and Earth System

Sciences

Open A

ccess

Annales Geophysicae

Open A

ccess

Nonlinear Processes in Geophysics

Open A

ccess

Atmospheric Chemistry

and Physics

Open A

ccess

Atmospheric Chemistry

and Physics

Open A

ccess

Discussions

Atmospheric Measurement

Techniques

Open A

ccess

Atmospheric Measurement

Techniques

Open A

ccess

Discussions

Biogeosciences

Open A

ccess

Open A

ccess

BiogeosciencesDiscussions

Climate of the Past

Open A

ccess

Open A

ccess

Climate of the Past

Discussions

Earth System Dynamics

Open A

ccess

Open A

ccess

Earth System Dynamics

Discussions

GeoscientificInstrumentation

Methods andData Systems

Open A

ccess

GeoscientificInstrumentation

Methods andData Systems

Open A

ccess

Discussions

GeoscientificModel Development

Open A

ccess

Open A

ccess

GeoscientificModel Development

Discussions

Hydrology and Earth System

Sciences

Open A

ccess

Hydrology and Earth System

SciencesO

pen Access

Discussions

Ocean Science

Open A

ccess

Open A

ccess

Ocean ScienceDiscussions

Solid Earth

Open A

ccess

Open A

ccess

Solid EarthDiscussions

The Cryosphere

Open A

ccess

Open A

ccess

The CryosphereDiscussions

Natural Hazards and Earth System

Sciences

Open A

ccess

Discussions

A century of ice retreat on Kilimanjaro the mapping reloaded

N J Cullen1 P Sirguey2 T Molg3 G Kaser4 M Winkler 4 and S J Fitzsimons1

1Department of Geography University of Otago Dunedin New Zealand2School of Surveying University of Otago Dunedin New Zealand3Chair of Climatology Technische Universitat Berlin Germany4Centre of Climate and Cryosphere Institute of Meteorology and Geophysics University of Innsbruck Austria

Correspondence toN J Cullen (njcgeographyotagoacnz)

Received 20 August 2012 ndash Published in The Cryosphere Discuss 1 October 2012Revised 17 December 2012 ndash Accepted 7 February 2013 ndash Published 4 March 2013

Abstract A new and consistent time series of glacier retreaton Kilimanjaro over the last century has been established byre-interpreting two historical maps and processing nine satel-lite images which removes uncertainty about the locationand extent of past and present ice bodies Three-dimensionalvisualization techniques were used in conjunction with aerialand ground-based photography to facilitate the interpreta-tion of ice boundaries over eight epochs between 1912 and2011 The glaciers have retreated from their former extent of1140 km2 in 1912 to 176 km2 in 2011 which represents atotal loss of about 85 of the ice cover over the last 100 yrThe total loss of ice cover is in broad agreement with pre-vious estimates but to further characterize the spatial andtemporal variability of glacier retreat a cluster analysis us-ing topographical information (elevation slope and aspect)was performed to segment the ice cover as observed in 1912which resulted in three glacier zones being identified Lin-ear extrapolation of the retreat in each of the three identi-fied glacier assemblages implies the ice cover on the west-ern slopes of Kilimanjaro will be gone before 2020 whilethe remaining ice bodies on the plateau and southern slopeswill most likely disappear by 2040 It is highly unlikely thatany body of ice will be present on Kilimanjaro after 2060if present-day climatological conditions are maintained Im-portantly the geo-statistical approach developed in this studyprovides us with an additional tool to characterize the physi-cal processes governing glacier retreat on Kilimanjaro It re-mains clear that to use glacier response to unravel past cli-matic conditions on Kilimanjaro the transition from growthto decay of the plateau glaciers must be further resolved inparticular the mechanisms responsible for vertical cliff de-velopment

1 Introduction

The presence of glaciers in tropical Africa has always beenregarded as somewhat of a peculiarity Johann Rebmannwho is acknowledged as the first European to formally rec-ognize the presence of ice and snow on Kilimanjaro in 1848referred to Africarsquos highest mountain as ldquothe mountain whosesnows defy the fierceness of the equatorial sunrdquo (Meyer1890 p 6) Kibo the highest of three peaks on Kilimanjaro(5895 m) still harbours ice remnants today but the recentand well-documented retreat of its glaciers (eg Sampson1971 Downie and Wilkinson 1972 Messerli 1980 Hasten-rath 1984 2006 Hastenrath and Greischar 1997 Thompsonet al 2002 2009 Cullen et al 2006) provides convincingevidence that present-day climatological conditions are notfavourable for glacier growth To deepen our understandingabout past climate conditions that favoured the build-up ofice it is critical that we establish a consistent record of thechanges to the glaciers on Kibo A detailed knowledge oftheir behaviour through time allows insight into past climateconditions to be established if the atmospheric processes con-trolling the observed changes are understood The glaciers onKibo because of their location and exceptional elevation of-fer a unique location to sample atmospheric conditions in thetropical mid-troposphere which is extremely valuable giventhe influence of low latitude weather and climate processeson global circulation (eg Chiang 2009)

Recent field experiments complemented with atmosphericand glaciological modelling have shed new light on the link-ages between atmospheric processes across multiple spaceand time scales and changes to glacier mass on Kiliman-jaro (eg Molg and Hardy 2004 Molg et al 2006 2008a

Published by Copernicus Publications on behalf of the European Geosciences Union

420 N J Cullen et al A century of ice retreat on Kilimanjaro

2009ab 2012 Cullen et al 2007 Winkler et al 2010)It has been carefully demonstrated that the responses ofglaciers on Africarsquos highest mountain are sensitive to bothlocal and large-scale climate dynamics (Molg and Kaser2011) with the former extent of the glaciers providing im-portant insights into past climate conditions (Molg et al2009b) Given the importance of using glacier behaviour toreconstruct climate variability and change it is critical thatwe have a consistent and carefully documented record of thechanges to the extent of all ice bodies on Kilimanjaro throughtime

Since the first detailed observations of Kilimanjarorsquosglaciers by Hans Meyer in the late 1880s the dramatic re-treat has been described in numerous studies but surprisinglyfew have included detailed maps of the glaciers Early assess-ments that did were dependent primarily on ground-basedobservations and terrestrial photogrammetry (eg Meyer1891 1900 Jaeger 1909 Klute 1914 1920 1921 Downieet al 1956 Humphries 1959) More recently remote sens-ing techniques using both aerial photography and satelliteimagery have become the method of choice to characterizethe ongoing retreat (Hastenrath and Greischar 1997 Thomp-son et al 2002 2009 Cullen et al 2006) Beyond the ap-parent ease of using this approach lies the difficulty of inter-preting ice bodies over rugged and steep terrain using remotesensing imagery that is often compromised by snow coverwhich leads to the common problem of establishing bound-aries between snow and ice The problem is further compli-cated by the expectation that ice boundaries determined fromnew imagery will fit within previously published maps If onecomplies with this expectation in an effort to create an up-dated time series of ice retreat that is consistent with previouswork then it is likely that errors associated with disputableinterpretation of ice bodies andor inaccurate positioning ofsource imagery in the past will appear in any new interpreta-tion

To avoid this we have revisited the entire time series ofice retreat starting with an alternative interpretation of the1912 photogrammetric survey of Klute (1920) and later theground and aerial observations used by Downie and Wilkin-son (1972) In addition nine satellite images from 1975 to2011 have been orthorectified in a consistent fashion pro-viding us with a robust platform from which to establish amore reliable time series of ice retreat The use of three-dimensional (3-D) visualization techniques enhanced and fa-cilitated the interpretation of ice boundaries and allowed usto correct errors that have appeared in previous mapping ef-forts Importantly this study goes beyond clarifying some ofthe existing disputes about the location and extent of the re-maining ice bodies on Kibo by presenting the retreat ratesof three different glacier zones which are segmented usingtopographical information This geo-statistical approach fur-ther demonstrates the importance of different glacier regimeson Kibo which have been used successfully as a frameworkto characterize the different physical processes controlling

glacier growth andor decay on Africarsquos highest mountain(eg Nilsson 1931 Geilinger 1936 Kaser et al 2004 2010Cullen et al 2006 Molg et al 2009b)

2 Data and methods

21 Historical maps

211 The 1912 survey

A detailed ground-based photogrammetric survey of Kili-manjaro was conducted in 1912 by Fritz Klute and EduardOehler which in conjunction with information from previ-ous sketches and mapping efforts (eg Meyer 1900 Jaeger1909) led to a 1 50 000 scale map being produced based ona modified Clark 1880 ellipsoid datum (Klute 1920 1921)A scanned version of this product was registered in ESRIArcGIS software using four intersections of the geographi-cal grid shown on the map using a linear transformation toavoid introducing unwanted distortions The registered mapwas then re-projected onto UTM zone 37 South based on theWGS84 ellipsoid (UTM37S) which is the cartographic sys-tem we have chosen to use in this study While the quality ofthe map should be praised given the complexity of the terrainand the technical limitations of the emerging surveying tech-nique used severe planimetric distortions were evident andprevented the immediate mapping of the glaciers

We took advantage of the 50 m contour lines on the Klute(1920) map to develop an empirical correction for the ob-served planimetric distortions This was achieved by creat-ing equivalent 50-m contour lines using a digital elevationmodel (DEM) sourced from Shuttle Radar Topography Mis-sion (SRTM) data processed and distributed by CGIAR-CSI(Jarvis et al 2008) which has a 3 arcsecond resolution (90 mspatial resolution) (Farr et al 2007) The SRTM DEM wasprojected (UTM 37S) to a 30 m pixel size using a bicubicinterpolation The geolocation accuracy of SRTM DEM re-ported to be better than 13 m (Farr et al 2007) provideda strong basis for supporting the geometric correction of thehistorical map A total of 201 tie points were defined to matchcontour lines drawn on the historical map to those derivedfrom the SRTM DEM This process was further supportedthrough the use of a hill-shaded DEM developed to help re-veal geomorphological clues and features that facilitated aconsistent registration of the map with the underlying terrainA detailed map produced by Jaeger (1909 map dated 1906)of the glaciers in the western region of Kibo also proved valu-able in interpreting the ice boundaries established by Klute(1920)

212 The 1953 and 1957 geological surveys

The University of Sheffield conducted geological surveysof the upper slopes of Kilimanjaro at the invitation of theTanganyika Geological Survey in 1953 and 1957 which

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 421

resulted in numerous descriptive observations of the glaciers(eg Humphries 1959 Downie 1964 Downie and Wilkin-son 1972) Humphries (1959) indicated that in 1953 a pho-togrammetry survey was completed that covered the south-western slopes of Kibo Downie and Wilkinson (1972 p 1)also reported that the whole area could not be visited on thefirst expedition and the survey of the glaciers was completedin 1957 Nevertheless after the first expedition Downie etal (1956) produced a map of the Kibo caldera showing thebasic extent of the ice bodies in this area Their efforts cul-minated in the production of the 1963 geological map sheet562 (1 50 000) which depicts the full extent of the ice bod-ies on Kibo as of February 1962 based on their field observa-tions and aerial photography (Downie and Wilkinson 1972)Humphries (1972 p 53) also provided a detailed sketch ofthe glacier remnants on the western slope of Kibo Hastenrathand Greischar (1997) report that they mapped the ice extentin 1953 from a 1 25 000 scale map produced by Humphries(1953) Efforts to obtain a copy of the 1953 map associatedwith the unpublished report by Humphries (1953) proved un-successful which led to our decision to use the insert of the1963 geological map instead The glacier boundaries as de-picted in the insert are representative of the ice extent onKibo at the end of February 1962 Five intersections of thegeographical grid were used to register the map using a lin-ear transformation Though it is not clearly described theinsert of the 1963 geological map appears to be drawn ontoa modified Clark 1880 ellipsoid which is consistent with theArc 1960 geodetic datum used in Tanzania A re-projectionto UTM37SWGS84 allowed the 1963 map to be incorpo-rated into the consolidated GIS created for this study

22 Satellite data

221 Very high resolution sensors

The 06 m spatial resolution pan-sharpened and orthorecti-fied QuickBird image described by Cullen et al (2006) ob-tained on 1 February 2003 provided the basis for the ge-olocation of imagery used in this study as well as being re-tained to map the areal extent of the glaciers A Kompsatimage from 17 June 2011 was obtained at a L1A process-ing level to enable a precise orthorectification Fusion of thefour 4-m resolution multi-spectral bands with the 1-m reso-lution panchromatic band prior to the orthorectification dra-matically improved the potential to interpret spatial detailsin the Kompsat image Ten ground check points (GCPs) and12 independent check points (CPs) were collected from theQuickBird image and elevation data were retrieved from theSRTM DEM to support the orthorectification The latter wasconducted using the sensor model and rational polynomialcoefficients supplied by the image provider in ERDAS Imag-ine 2011 in conjunction with the SRTM DEM and yieldeda root mean square error (RMSE) of 25 m and 28 m for theGCPs and CPs respectively

222 Advanced Spaceborne Thermal Emission andReflection Radiometer (ASTER)

A search through the ASTER archive led to an image from19 August 2004 being identified as suitable to independentlyassess and support the orthorectification of other images A15 m DEM was produced following a stereo triangulationof the ASTER bands 3N and 3B in Leica PhotogrammetrySuite Twenty-one tie points and seven GCPs collected fromthe orthorectified QuickBird image as well as elevation datafrom the SRTM DEM were used and yielded a RMSE of81 44 and 93 m in the easting northing and elevation di-rections respectively A hill-shaded DEM was then createdand compared to that derived from the SRTM DEM This re-vealed no noticeable planimetric bias and demonstrated thatthe rectification of the ASTER image is consistent with thegeolocation accuracy of SRTM data Importantly the corre-sponding orthorectified ASTER image provided an absolutereference (within aboutplusmn10 m) to assess the rectification ofall other images Furthermore the almost snow-free ASTERimage was extremely useful in our efforts to establish theice boundaries in the QuickBird and Kompsat images de-spite its relatively coarse pixel size (ie 15 m) particularlyin the Breach Wall region due to the advantageous illumina-tion and spectral resolution of the sensor The ASTER imagewas not used as an epoch in the new time series presented inthis study because of its temporal proximity to the QuickBirdimage

223 Landsat images

The full archive of Landsat satellite imagery from the Kili-manjaro region available at the Earth Resources Observationand Science (EROS) Center was examined which led to fourepochs being identified as suitable for mapping the remainingice cover (Table 1) A Landsat 2 Multispectral Scanner Sys-tem (MSS) image from 15 August 1975 was preferred to thatof 24 January 1976 used by Hastenrath and Greischar (1997)as it exhibited much less transient snow Two Landsat 5 The-matic Mapper (TM) images acquired only a month apart on24 June and 26 July 1984 were used as they jointly provideda virtually snow free acquisition (the first date was retainedas the epoch of reference) with improved spatial resolution(ie 30 m) The Landsat 4 TM image of November 1989 citedby Hastenrath and Greischar (1997) (no date provided) couldnot be found in the archive and was therefore not used In-stead an image from 31 December 1992 was utilised Fi-nally a Landsat 7 ETM+ image from 21 February 2000 wasobtained to provide an independent assessment of the map-ping epoch characterized by Thompson et al (2002) using anaerial survey An image fusion technique was also completedto take advantage of the enhanced spatial resolution of thepanchromatic band of the ETM+ sensor (ie 15 m)

The selected Landsat images were orthorectified with thesensor model in ERDAS Imagine using GCPs identified in

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

422 N J Cullen et al A century of ice retreat on Kilimanjaro

Table 1Details of the terrestrial photogrammetry and satellite products used to characterize the retreat of the glaciers on Kibo The spatialresolution is given as either a map scale or pixel size The number of ground check points (GCPs) used for image matching and the root meansquare error (RMSE) (m) associated with the orthorectification (m) are also provided Two further Landsat images and one ASTER imagewere also processed but are not listed as they were not formally used for the time series (see Sect 22)

Map year Source Description of the product Pixel size GCP(m) or RMSEmap scale

MayndashOct 1912 Klute (1920 1921) Terrestrial photogrammetry with 74 1 50 000 ndashstand-lines over 2000 measured pointsand additional field sketches

Feb 1962 Downie and Terrestrial photogrammetry of western 1 50 000 ndashWilkinson (1972) glaciers ground observations and aerial

photography15 Aug 1975 Landsat 2 (MSS) Satellite imagery 57 1059724 Jun 1984 Landsat 5 (TM) Satellite imagery 30 1010531 Dec 1992 Landsat 4 (TM) Satellite imagery 30 118121 Feb 2000 Landsat 7 (TM) Satellite imagery 15 11681 Feb 2003 QuickBirda Satellite imagery 06 ndash26b

17 Jun 2011 Kompsata Satellite imagery 1 1025

a Pan-sharpened multi-spectral imageryb Ortho 1 5000 product level

the ASTER image which had already had its absolute ac-curacy assessed MSS and TMETM+ images were resam-pled to 20 and 10 m respectively using a bicubic interpola-tion to facilitate the visual interpretation of the identified iceboundaries The number of GCPs used and the correspondingRMSE for each image are provided in Table 1 Finally theco-registration between the orthorectified images was scru-tinized and revealed extremely good agreement Minor dis-tortions were identified in the QuickBird image in the steepsouth-east region of Kibo which required a local geomet-ric correction using an empirical rational function model andGCPs collected from the orthorectified Kompsat image Wesuspect the source of these distortions may have been fromthe use of incorrect elevation data in voids from the SRTMDEM used by the image providers during the initial orthorec-tification process It is also likely that unfavourable acquisi-tion geometry in conjunction with the steep terrain in this re-gion may have adversely affected the rectification Followingthe correction the ice boundaries for 2003 defined by Cullenet al (2006) were revised which resulted in a minor changein the total areal extent

23 Mapping ice boundaries

231 Previous mapping efforts

The maps created by Hastenrath and Greischar (1997 Fig 4)and Thompson et al (2009 Fig S2) (referred to in the fol-lowing as HG97F4 and T09FS2 respectively) were digitizedto help facilitate a comparison to the ice boundaries iden-tified in this study The single geographical coordinate in-dicated on HG97F4 (ie 305prime S 3720prime E according to a

Clark 1880 ellipsoid) made it difficult to perform a rigor-ous registration and raises the question as to whether glacierboundaries in HG97F4 are suitable for reproduction Insteadlinks were created to match contour lines with those derivedfrom the SRTM data using a linear transformation to avoidunwanted distortions The same process of matching con-tours was used to register T09FS2 which revealed a per-fect match with HG97F4 This enabled us to confirm thatthe glacier outlines prior to 2000 in T09FS2 are a repro-duction and edited version of HG97F4 which is also ac-knowledged by Thompson et al (2002 2009) Once bothmaps were registered in a consistent cartographic system theglacier boundaries at each epoch (ie 1912 1953 1976 and1989 for HG97F4 and T09FS2 2000 and 2007 for T09FS2)were carefully digitized This process revealed numerous in-consistencies between the glacier boundaries of HG97F4 andT09FS2 Hastenrath (2006 p 2) also alluded to a ldquomappingerrorrdquo related to the 2000 ice boundaries created by Thomp-son et al (2002) To ensure the latter fitted within his own1989 epoch Hastenrath (2006) was required to make adjust-ments to the ice boundaries of the northern ice field Cullen etal (2006 Fig 1) were also required to use an ldquoapproximateglacier extent in 1912rdquo as they found it difficult to place someice remnants observed in 2003 within previously establishedoutlines (eg Hastenrath and Greischar 1997 Thompson etal 2002) which was scrutinized by Thompson et al (2009Fig S4) These problems prompted us to make the decisionto revisit the entire sequence of historical maps and satel-lite imagery While it is not our intention to focus on all ofthe mapping inconsistencies we have observed we do de-scribe some key differences in detail (see Sect 31) in order

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 423

to characterize the physical processes controlling glacier re-treat on Kibo

232 Establishing new ice boundaries

The boundaries of ice bodies were visually interpreted anddigitized from the orthorectified satellite images and histor-ical maps Despite the variability in the spatial and spectralresolution of the data products as well as having to accountfor varying amounts of transient snow in the satellite im-ages the use of the high temporal resolution sequence sig-nificantly improved our ability to interpret the location andextent of ice bodies consistently through time To supportand facilitate the interpretation of all data products (Table 1)three-dimensional visualization techniques were extensivelyused in conjunction with aerial and ground-based photogra-phy from others (eg Hastenrath 2008) and our own fieldexpeditions between 2005ndash2012 The vantage points fromwhich historical photographs were taken were reproducedin the 3-D model of Kilimanjaro which proved particularlyuseful when revisiting the glacier outlines mapped by Klute(1920) The use of 3-D visualization also helped us to inter-pret glacier positions in relation to distinctive geomorpho-logical features (eg moraines rock outcrops) which are de-scribed in more detail in the following section Finally tohelp facilitate future efforts all data products from this anal-ysis have been made available on the Global Land Ice Mea-surements from Space (GLIMS) Glacier Database

3 Results and discussion

31 The mapping reloaded

The total amount of ice cover on Kilimanjaro on 17 June2011 was 176 km2 (Table 2) with the location of these icemasses shown in Fig 1 The remaining ice masses are rem-nants of more extensive glaciers which were first mappedand given names by Meyer (1900) with the nomenclaturestill in use despite the character of the ice bodies havingchanged dramatically over the last century The majority ofthe ice cover observed on Kilimanjaro today fails to strictlyadhere to any common definition of a glacier with most enti-ties isolated and stagnant but following the definition sug-gested by Racoviteanu et al (2009) we include any bodyof ice that is observed after transient snow melts It is clearfrom the accounts and photographs of the early explorers thatonly the principal glaciers were named and that a consider-able number of small stagnant ice masses in particular in thecrater area were not formally recognized This is no more ev-ident than in the south-east crater area where the first explor-ers accessed the summit region through a number of weak-nesses in the ice referred to as notches (eg Hans Meyer andJohannes notches) which have been described in detail else-where (eg Geilinger 1936 Hastenrath 1984) Despite thisapparent difficulty it is clear that the remaining ice bodies

today are remnants of the more extensive glaciers mappedby Klute (1920) using terrestrial photogrammetry (Table 1)with the boundaries established from this effort reflecting theice cover in 1912 (Fig 1) Our calculation of the extent ofthe glaciers on Kilimanjaro at the end of October in 1912 is1140 km2 (Table 2) which is 85 more than what is ob-served on the mountain today

The differences between our interpretation of the 1912map produced by Klute (1920) and that produced by Has-tenrath and Greischar (1997) are also shown in Fig 1 Asnoted in Sect 232 our interpretation of the glacier bound-aries was supported by geomorphological evidence observedusing 3-D visualization and historical photographs Impor-tantly the 3-D model allowed us to carefully identify andlocate the most recent moraines Using these geomorpholog-ical features as markers we were able to determine with ahigh degree of accuracy the former location of the glaciersThis approach was particularly useful in establishing the pastlocation and extent of the outlet glaciers feeding from thenorthern ice field (the Credner Drygalski and (Great andLittle) Penck glaciers) the Uhlig glacier and the two distinctglacier streams located in the western breach area (the Lit-tle and Great Barranco glaciers) as well as the four primarysouthern ice field glaciers (the Heim Kersten Decken andRebmann) The map produced by Jaeger (1909 map dated1906) was also valuable in allowing us to identify a num-ber of small ice bodies located above both the Little andGreat Barranco glaciers (Fig 1) Our interpretation of thealtitudinal extent of the southern ice field glaciers at about4600 m (Fig 1) is supported by the photographs of Uhlig(1904 Fig 50 p 641) and Klute (1920 Fig 1) It is impor-tant to note that both these historical photographs reveal acomplexity of the outcropping rock at the lower elevations ofthe southern ice field glaciers that we have made an attemptto account for (Fig 1)

Despite a well-documented and photographic account ofthe historical retreat of Kilimanjarorsquos glaciers which hasbeen carefully reviewed by Hastenrath (1984 2008) therehave been a number of inconsistences in the interpretationof ice bodies in recent mapping efforts (eg Hastenrathand Greischar 1997 Cullen et al 2006 Hastenrath 2006Thompson et al 2002 2009) which require further clarifi-cation as they have an impact on our ability to interpret thephysical processes controlling the demise of the glaciers onKibo To do this one must briefly reflect on the historical ob-servations starting with those of Hans Meyer who provideda sketch (Meyer 1890 p 345) and a map (Meyer 1891) thatindicates that an ice stream which is now referred to as theFurtwangler glacier flowed over the crater rim into the west-ern breach area Meyer (1891 p 319) notes ldquothe ice from thevast caldera falls as a mighty cascade into the great westernfissure where it unites with the accumulations in the fissureitselfrdquo Though Hastenrath (1984 p 78) disputes whetherthis ice stream (the Furtwangler glacier) was connected tothe Little andor Great Barranco glaciers other historical

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

424 N J Cullen et al A century of ice retreat on Kilimanjaro

4800

4600

5000

4400

5200

5400

5600

5700

4200

4000

5800

4800

4400

4600 4200

37deg23E

37deg23E

37deg22E

37deg22E

37deg21E

37deg21E

37deg20E

37deg20E

3deg3

S

3deg3

S

3deg4

S

3deg4

S

3deg5

S

3deg5

S

3deg6

S

3deg6

S

314000

314000

316000

316000

318000

318000

320000

320000

9658000

9658000

9660000

9660000

9662000

9662000

Uhuru Peak

Northern ice field

Heim

Uhlig

Dec

ken

Ratzel

Kersten

Reb

man

n

Credner

Drygalski

Arrow

Diamond

Balletto

Furtwaumlngler

LittlePenck

5895m

IceCastle

Easternice field

Southernice field

WGS84UTM 37S

0 1000 2000

Meters

Revised 1912 boundaries

Credner Glacier name

2011 Ice boundaries

1912 boundaries fromHastenrath and Greischar (1997)

Little Barranco

Fig 1 Glacier extent on Kilimanjaro in 1912 and 2011 Glacier names with an asterisk do not appear in Klute (1920) and are described inthe text The interpretation of the glacier extent for 1912 as defined by Hastenrath and Greischar (1997) is shown as a dashed line (blue) forcomparative purposes The nomenclature for each glacier used by Klute (1920) is given while the names of four glaciers introduced afterthis time are indicated usinglowast

accounts are more supportive (Gillman 1923 Geilinger1936) A more detailed map of Kilimanjarorsquos glaciers pro-duced by Meyer (1900) based on his observations during afield expedition in 1898 shows the connection no longer ex-isted suggesting a break-up occurred sometime in the early1890s Though the timing is uncertain the implications oflosing this connection are not overlooked by Geilinger (1936p 12) who refers to the Little and Great Barranco glaciers asldquoindependent dead glaciersrdquo and made the astute observationthat both ice entities were retreating at their upper and lowermargins

This observation holds some importance as it helps clar-ify the source of the ice remnants still observed today Themaps by Meyer (1900 map dated 1898) Jaeger (1909map dated 1906) and Klute (1920 map dated 1912) clearlyshow distinct ice streams for the Little and Great Barrancoglaciers which as noted may have been fed by ice from theFurtwangler glacier as recently as the late 19th century Ourrevision of the Klute (1920) map from 1912 (Fig 1) honoursthe discrete boundary between the Little and Great Barranco

glaciers which is not as clearly depicted in the Hastenrathand Greischar (1997) interpretation The failure to clearlydistinguish the boundaries and location of these two indepen-dent ice streams has been inherited by Thompson et al (20022009) who used the Hastenrath and Greischar (1997) inter-pretation of the ice extent prior to 2000 (see Sect 231)Thompson et al (2009) argue that ice remnants identified byCullen et al (2006) in both the ice streams of the former andmore extensive Little and Great Barranco glaciers are absentfrom previous observations and therefore most likely tran-sient snow The sketch by Humphries (1972 Fig 46) andthe map produced by Messerli (1980 Fig 6) also support theinterpretation by Cullen et al (2006) while the ice bodiesin question have been identified in each of the satellite im-ages used in the present analysis (Fig 2) Though not labelledby Humphries (1959 1972) or by those preceding his obser-vations the Arrow glacier has been used to refer to one ofthe Little Barranco ice remnants (Fig 1) It appears that thisterm was adopted by climbers on the mountain (eg Samp-son 1971 p 202) but the exact location of this ice body is

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 425

Table 2Areal extents (km2) of the total ice bodies between 1912 and 2011 as well as for zone 1 (western region) zone 2 (plateau region)and zone 3 (southern region) as defined in the text and Fig 4 The annual rates of area change per observation period (10minus2 km2 yrminus1) areshown in parentheses for the total and zoned areas

Epoch Duration Area [km2] (rate of area change [10minus2 km2 yrminus1])

[years] Total Zone 1 Zone 2 Zone 3

191283 11400 4540 4293 2567196216 4936 7320 (minus827) 2520 (minus409) 3161 (minus229) 1639 (minus188)197562 1347 6052 (minus941) 2060 (minus342) 2529 (minus469) 1463 (minus131)198448 887 4815 (minus1395) 1628 (minus487) 1949 (minus654) 1238 (minus254)199300 853 3800 (minus1190) 1142 (minus570) 1638 (minus365) 1020 (minus256)200014 715 2920 (minus1232) 0701 (minus617) 1411 (minus318) 0808 (minus297)200308 295 2500 (minus1425) 0539 (minus550) 1251 (minus543) 0710 (minus332)201146 838 1762 (minus881) 0333 (minus246) 0937 (minus375) 0492 (minus260)

not well documented and we are uncertain that the remnantidentified in Fig 1 and by Cullen et al (2006) is the sameas that referred to by others (eg Sampson 1971 Messerli1980 Hastenrath and Greischar 1997)

There has also been some confusion about the Diamondand Balletto glaciers (Fig 1) which were not formally rec-ognized in the early mapping work (Meyer 1990 Jaeger1909 Klute 1920) or acknowledged by Humphries (19591972) but are both described by Hastenrath (1984) whileSampson (1971) and Messerli (1980) only refer to the for-mer The Breach Wall area where both glaciers are locatedwas mapped by Klute (1920) but as no break up from thesouthern ice field had occurred at this time they were notgiven discrete labels (Fig 1) However what is indisputableis the presence of an ice field (the Balletto glacier) beneaththe headwall of the Diamond glacier which was not mappedby Thompson et al (2002 2009) The presence of this icefield is not only observed in the satellite record but is docu-mented by Reinhold Messner (Messner 1991 Chapter 21)who made the first (and only) direct ascent of the Breach Wallof Kilimanjaro in 1978 Though Cullen et al (2006) mappedthis region of ice (the Balletto glacier) they failed to label itcorrectly placing the label beneath the Heim glacier after fol-lowing the interpretation of Hastenrath and Greischar (1997Fig 2) Further the Diamond glacier is absent after 1976 inthe maps produced by Hastenrath and Greischar (1997) andHastenrath (2006) but appears in the Thompson et al (20022009) maps despite using the former to build their time se-ries Finally the Uhlig glacier which Jaeger (1909 p 130)named after his friend C Uhlig does not appear to have beenconnected to the northern ice field in the past (Fig 1) Whilethe upper boundary of the Uhlig glacier is difficult to inter-pret in the Klute (1920) map it is clearly shown by Jaeger(1909) that it was not joined to the northern ice field

Figure 2 shows the change in ice extent over almost a 100-yr period which is consistent through time and removes mostof the uncertainty about the location and extent of the re-maining ice bodies on Kibo The retreat over the first part of

the 20th century was approximately 41 km2 (Fig 2a and b)while over the last 50 yr 56 km2 of the ice cover has beenlost (Fig 2cndashh) The retreat between 1912 and 1962 (Fig 2aand b) resulted in a distinct break-up of ice in the vicinity ofthe notches (south-east crater region) separating the northernand eastern ice fields from the southern ice field Remnantsof the Ice Castle (Fig 1) still remained in 1962 and were de-scribed by Sampson (1971) as still being present in the early1970s but appeared to have been lost by 1975 (Fig 2c) Othernoticeable features of the retreat in the early 20th century in-clude the development of a distinct separation between theGreat and Little Penck glaciers which formed through theexpansion of the Ravenstein (first mapped by Meyer 1900)the rapid demise of the Uhlig glacier and the split-up of theDiamond and Balletto glaciers from the southern ice field(Fig 2a and b)

Between 1962 and 1975 the persistent retreat of ice led tothe northern and eastern ice fields becoming separated fromeach other as did the Great Penck glacier from the north-ern ice field (Fig 2b and c) The Little Penck proved moreresistant and did not separate from the northern ice field un-til after 1992 while the remnant of the Great Penck com-pletely disappeared during this period (Fig 2e and f) thoughground-based (field) observations in January 2005 revealeddebris-covered ice at the location of the lowest 1984 extent(Fig 2d) As noted by others (eg Hastenrath and Greischar1997 Cullen et al 2006) the northern ice field developed ahole sometime after 1984 (Fig 2d and e) which opened intoa large canyon after 2003 (Fig 2g and h) Field observationsby the first author in October 2012 revealed that the north-ern ice field has separated into two independent ice bodieswhich must have occurred sometime after 17 June 2011 TheFurtwangler glacier has also separated into smaller ice enti-ties which began after 2003 (Fig 2g and h) The southernice field has not escaped the on-going demise with the upperpart of the Kersten glacier separating from its lower reachesbetween 2003 and 2011 (Fig 2g and h) which was some-what anticipated after mass balance modelling showed this

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

426 N J Cullen et al A century of ice retreat on Kilimanjaro

(g) 2003 (h) 2011

(a) 1912 (b) 1962

(c) 1975 (d) 1984

(e) 1992 (f) 2000

Credner

Great Penck

Ratzel

Ker

sten

Northern ice field

Easternice field

Southernice field

GreatBarranco

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Fig 2 Ice retreat over Kilimanjaro based on the revised areal extent for the following years(a) 1912(b) 1962(c) 1975(d) 1984(e)1992(f) 2000(g) 2003 and(h) 2011 An interactive 3-D illustration of the sequence can be sourced in the Supplement

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 427

was likely to occur (Molg et al 2009b) The upper part ofthe Heim has disappeared and the Decken has narrowed con-siderably over the last 8 yr (Fig 2g and h)

32 New interpretation of the ice retreat

A consistent record of the former extent of the glaciers onKibo provides a platform from which the relationship be-tween glaciological and meteorological processes can be fur-ther explored We believe the spatial and temporal variabilityof the ice cover on Kibo can be used to unravel the physi-cal processes responsible for controlling the decay of the icemasses through time as well as providing insights into themechanisms that may have been responsible for growth in thepast Interestingly those who were privileged to make someof the first (formal) observations on Kilimanjaro recognizedthere was an asymmetry in the ice cover (eg Meyer 1891)which pointed to specific climatological drivers controllingglacial retreat (eg Gillman 1923 Nilsson 1931 Geilinger1936) Meyer (1891 p 308ndash314) noted that ldquoan immensemountain mass like Kilimanjaro must considerably modifythe prevailing wind directionrdquo before making the point thatthe primary cause in the variability of the ice limit is ldquotheunequal distribution of moisture on the northern and south-ern sides of the mountainrdquo Though Meyer (1891) recognizedvariability in moisture was the most important control healso speculated that the asymmetry of low reaching glacierson Kibo in the south south-west and western sectors of themountain a pattern that persisted throughout the 20th cen-tury (Fig 2) depended on ldquolocal peculiarities which favourthe preservation of the snow [rather] than on physical condi-tions that favour a large snowfallrdquo (Meyer 1891 p 315) Torevisit this early observation it is useful to reflect on the hyp-sometry of the glaciers in 1912 (Fig 3a) which shows thatthe probability density in relation to the distribution of the icewas greatest above 5700 m This high density reflects the icecover in the crater region with the majority of the ice in 1912being part of the former ice cap that linked the northern andeastern ice fields (Fig 1) Figure 3b and c show that slopeand aspect were also important controls on the density of icecover with the steep south and south-west facing slopes morefavourable which enabled the glaciers to extend to elevationsbelow 4600 m For the glaciers to have extended to the sameelevation on the east-facing slopes they would have had tocover a much greater horizontal distance (Fig 1)

To fully account for the distinct geometry of the ice masseson Kibo and to establish the physical processes controllingtheir location and extent through time Kaser et al (2004)established a framework that included four separate glacierregimes The motivation to do this was partly based on theearly recognition that the tabular shaped glaciers in the craterregion of Kibo behave differently to glaciers on the flanks ofthe mountain (eg Nilsson 1931 Geilinger 1936) Guidedby this distinction Cullen et al (2006) used the 5700 m con-tour which is roughly the height of the outer crater rim to

Density times 10 minus3

2

4

6

8

WEST EAST

SOUTH

NORTH

4600 4800 5000 5200 5400 5600 58000

1

2

3

4

Elevation [m]

Den

sity

times 10

minus3

0 10 20 30 40 50 60 70 800

1

2

3

4

Slope [deg]

Den

sity

times 10

minus2

(a)

(c)

(b)

Fig 3 Distribution of topographic measures from the areal extentof all ice bodies in 1912(a) elevation(b) slope and(c) aspect Themagnitude of the normalised distribution on the y-axis is a proba-bility density function with the integral equal to unity

distinguish between two broad glacial systems (1) plateau(ge 5700 m) and slope (lt 5700 m) glaciers This simple bi-nary classification has proved useful in describing some ofthe key physical processes responsible for the retreat of iceon Kibo (eg Molg et al 2009b Kaser et al 2010) Tofurther characterize the spatial and temporal variability ofglacier retreat on Kibo we have adopted a geo-statisticalapproach in this study that uses elevation slope and aspectof the ice cover in 1912 (the feature space) to identify clus-ters resembling topographical classes This was achieved byperforming a cluster analysis containing maximum likeli-hood estimates of the topographical parameters in a Gaussian

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

428 N J Cullen et al A century of ice retreat on Kilimanjaro

mixture model (McLachlan and Peel 2000) which resultedin three broad glacial zones being identified (1) western re-gion (zone 1) (2) plateau region (zone 2) and (3) southernregion (zone 3) (Fig 4)

Figure 5 shows the rate of retreat of each of the glacialzones through time as well as a least-squares linear regres-sion fit (and 80 confidence intervals) through each of theglacial zone time series As described by Cullen et al (2006)the retreat of the tabular shaped plateau glaciers (zone 2) iscontrolled primarily by solar radiation-induced ablation onthe vertical cliffs that characterize them as demonstratedthrough radiation modelling (Molg et al 2003) Winkleret al (2010) revealed that direct insolation is the reasonfor the predominant zonal alignment of the ice cliffs andtheir bimodal annual recession (sunlit versus shaded peri-ods) Though the shrinkage of the ice cover along the verticalcliffs is fairly constant from year to year the areal retreat ofthe zone 2 region is not entirely linear with the highest ratesoccurring between 1975 and 1984 (Table 2) which corre-spond to the period following the separation of the northernand eastern ice fields (Fig 2) Therefore changes in the ge-ometry of the ice bodies and their exposure to solar radiationappear to have led to variations in retreat rates (Table 2 andFig 5) which has also been described by Kaser et al (2010)

An explanation for the growth and decay of the plateauglaciers including a description of the physical processesthat may have led to the observed steps in the northern icefield has been provided by Kaser et al (2010) To sustaingrowth in the crater rim region (zone 2) it is critical thatsnowfall is sufficient to allow a continuous snow cover todevelop that is able to survive from one year to the next Ifthe annual build-up of snow is sustained through a period oftime it is quite conceivable that the plateau glaciers couldgrow to several tens of metres in thickness quite quicklyraising the question as to whether this accumulation of icehas the potential to become a source of mass for glacierson the western flanks of the mountain (zone 1) Meyer(1891 p 316ndash317) was somewhat sceptical as to whetherthe plateau region acted as a ldquoreservoirrdquo and thought it ldquomuchmore reasonable to suppose that the rocky rim of the crateris the real reservoirrdquo However Meyer (1890 1891) did pro-vide an account and sketches to support a connection of theplateau ice to the Little and Great Barranco glaciers in thepast which as described may have been important in sustain-ing them (Gillman 1923 Geilinger 1936) There is someimportance in investigating the extent of the linkages be-tween these two glacial zones as together they accounted foralmost 80 of the total ice cover in 1912 (Table 2)

The remaining ice bodies in zone 2 (plateau region) aremostly stagnant today with field measurements and prelim-inary modelling results indicating deformation rates of icewithin the crater are small (lt 10 cm yrminus1) (Kaser et al 2004Winkler et al 2010) It seems unlikely that the present or his-torical thickness of the plateau ice has been sufficient to al-low substantial deformation (Kaser et al 2004) suggesting

3140 3150 3160 3170 3180 3190 3200 32109657

9658

9659

9660

9661

9662

9663

9664

4000

4000

4500

4500 4500

45004500

5000

5000

5000

5000

5000

5500

5500

5500

UTM easting in meters times105

UT

M n

orth

ing

in m

eter

s times

106

Fig 4 Classification of regions based on the distribution of topo-graphic measures in 1912 Zone 1 (western region) zone 2 (plateauregion) and zone 3 (southern region) are shown in red orange andyellow respectively

that mass transport from the nearly horizontal plateau downinto the western slope glaciers (zone 1) has most likely beenminor Careful observations by Klute (1920) demonstratedthat the velocity of the centre of the Great Penck glacier was064 m over a 30-day period in 1912 which led him to spec-ulate that the annual downslope movement was about 7 mper year It should be noted that Nilsson (1931) and Osmas-ton (1989) also report on these observations but incorrectlycite Jaeger (1909) as the person responsible for obtaining themeasurements Nonetheless it is improbable that ice fromthe plateau region could have supported the observed flowdynamics of the Great Penck glacier with the upper reachesof the mountain flanks the most likely source of the massThus the growth and decay of the glaciers in zone 1 (west-ern region) do not appear to be directly dependent on thestability of plateau ice (zone 2) To further resolve the phys-ical processes controlling the stability of the plateau ice itwould be of interest to develop a better understanding of themechanisms responsible for vertical cliff development

The glaciers on the southern slopes (zone 3) differ fromthose on the western slopes (zone 1) as they are separatedfrom the large masses of ice within the caldera (zone 2) bybed topography (Humphries 1972) This distinction is im-portant as any assessment of the atmospheric controls onthe behaviour of the southern ice fields is independent ofthe complexity of the vertical cliff retreat in the plateau re-gion (zone 2) That said the cluster analysis that led to thetopographical classes used in this study shows that the up-per sections of the southern ice field exhibit similar phys-ical characteristics to the ice within the caldera (Fig 4)Nevertheless to obtain a full solution for the asymmetry ofthe glaciers on Kibo that the early explorers observed itis critical to establish an understanding of the atmospheric

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 429

1900 1920 1940 1960 1980 2000 2020 2040 20600

05

1

15

2

25

3

35

4

45

5

Year

Are

al e

xten

t [km

2 ]

zone 1 R2=099

zone 2 R2=097

zone 3 R2=09980 confidence interval

4000

4000

4000

4500

4500

4500

45005000

5000

5000

50005000

5500

5500

5500

Fig 5 The areal change of glaciers and ice bodies based on thethree-member classification Linear regressions (thin solid lines) foreach of the three zones are shown as well as the 80 confidenceintervals for each of these relationships

processes controlling the growth andor decay of the southernice fields This has been achieved by Molg et al (2009b)who used observations and atmospherendashglacier modellingto show how critical precipitation humidity and fractionalcloud cover are on the energy and mass balance of the Ker-sten glacier (zone 3) These results compare favourably toearly accounts of the importance of precipitation (moisturechange) and cloud cover on the asymmetry (Meyer 1891Klute 1920 Gillman 1923 Geilinger 1936) with the diur-nal variability of cloud cover being especially important inproviding protection to the glaciers on the western side ofthe mountain (eg Meyer 1891 Hastenrath and Greischar1997)

Recent regional atmospheric modelling experiments haveprovided a more quantitative assessment to account for theasymmetry in glacier mass on Kibo demonstrating that thewest and south-western areas of the mountain are favourableto moisture convergence as controlled by atmospheric dy-namics (Molg et al 2009a) As noted by Osmaston (1989)the asymmetry is not confined to the most recent glacier ex-tents with the pattern of ice distribution established fromformer moraines indicating that recent climatic controls onglacier behaviour are similar to those in the past This sug-gests that establishing the specific nature of the atmosphericcontrols responsible for the most recent shift from glaciergrowth to decay in particular in zones 2 and 3 (plateau andsouthern slope regions) remains critical if we are to useglacier behaviour to reconstruct climatic conditions of thedeeper past

Finally it is difficult not to speculate on when the ice onKibo will disappear which becomes more achievable if onerecognizes the importance of the different glacial zones onthe mountain If one accepts linear extrapolation based on

changes from the past is suitable to make projections for thefuture it is apparent that the ice cover in zone 1 (western re-gion) will most likely be gone before 2020 and that ice inzone 2 (plateau) and zone 3 (southern slopes) will disappearsometime close to 2040 (Fig 5) It is likely that the plateauwill harbour the final remnants of ice on Kibo which couldremain slightly beyond the middle of the 21st century Inter-estingly our linear extrapolation for the loss of ice is in closeagreement to earlier projections using radiation modelling(Molg et al 2003) Though the loss of ice will have no hy-drological significance to lowland areas (eg Gillman 1923Kaser et al 2004 Molg et al 2008b) there is no doubt thattheir disappearance will be a ldquogrievous aesthetic lossrdquo (Salt1951 p 160) which will undoubtedly impact the experiencetourists have on the mountain (Molg et al 2008b)

4 Conclusions

With reference to the glaciers found on Mt Kenya and Kiboin East Africa C Gillman remarked ldquowe might perhapslearn a good deal as to the general world circulation fromcareful and prolonged observation on these particular sum-mitsrdquo (Mackinder 1923 p 23) This statement still holdstrue today (eg Molg et al 2006) and with the increasingawareness that glaciers on Africarsquos highest mountain can beused to extract climate information from the past it seemedtimely to revisit the mapping of the ice cover on Kibo throughtime Our motivation to do so has been to ensure the histor-ical record of these changes is of the highest possible qual-ity and in a format that should help facilitate on-going andfuture research dedicated to resolving the complexity of cli-mate variability and change in East Africa and beyond Thiswas achieved by using a consistent cartographic frameworkthat allows historical maps and new satellite imagery to beexamined The use of three-dimensional visualization tech-niques greatly enhanced our ability to establish the formerlocation of the glaciers

The revisited time series of glacial retreat on Kiliman-jaro contains eight epochs and spans almost 100 yr (1912ndash2011) The glaciers have retreated from their former extentof 1140 km2 in 1912 to 176 km2 in 2011 which representsa loss of about 85 of the ice cover over almost a 100-yr pe-riod Our estimate of the total retreat is consistent with recentassessments (Cullen et al 2006 Thompson et al 2009) butthe careful re-examination of the location and extent of indi-vidual ice bodies through time has allowed us to identify anumber of important differences from previously publishedmaps (eg Hastenrath and Greischar 1997 Thompson et al2002 2009) Historical evidence has been used to supportour interpretation with the detailed observations and pho-tographs obtained by early explorers on Kilimanjaro found tobe extremely valuable (eg Meyer 1891 1900 Uhlig 1904Jaeger 1909 Klute 1914 1920 1921 Gillman 1923 Nils-son 1931 Geilinger 1936) It has to be acknowledged that

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

430 N J Cullen et al A century of ice retreat on Kilimanjaro

much of what we know about Kilimanjaro today was estab-lished through their efforts

To further explore the importance of different glacierregimes on Kibo (Kaser et al 2004) a geo-statistical ap-proach was applied which resulted in the segmentation ofthe ice cover into three broad glacial zones A careful ex-amination of the retreat of ice in each of these glacier zonesindicates that there is a high probability that most of the re-maining ice cover on Kilimanjaro will disappear by 2040and it is highly unlikely that any ice body will remain af-ter 2060 if present retreat rates are maintained Before thishappens it would still be useful if the remaining ice bodiesin particular the remnants of the northern ice field could befurther examined to help resolve existing uncertainties aboutthe mechanisms responsible for controlling the formation ofthe vertical cliffs that characterize them (eg Winkler et al2010) In addition further efforts are required to scrutinizethe exact timing of the onset of the present glacier retreat firstreported by Hans Meyer in the late 19th century If we canrefine our understanding of the atmospheric controls behindthe most recent shift from glacier growth to decay on Kibowe will be in a stronger position to characterize the climaticcontrols of past and much larger glaciations on Africarsquos high-est mountain

Supplementary material related to this article isavailable online athttpwwwthe-cryospherenet74192013tc-7-419-2013-supplementpdf

AcknowledgementsThis research is funded by the Departmentof Geography and School of Surveying University of OtagoNew Zealand Thomas Molg was supported by the Alexander vonHumboldt Foundation The contributions made by Georg Kaserand Michael Winkler were supported by the Austrian ScienceFund (FWF) Projects P 20089-N10 and P 21288-N10 Thisresearch would not have been possible without the support from thefollowing Tanzanian authorities (COSTECH KINAPA TANAPATAWIRI) We are grateful to the referees that took the time toprovide feedback on this research in particular the careful andthoughtful comments by Henry Brecher

Edited by M Schneebeli

References

Chiang J C H The tropics in paleoclimate Annu Rev EarthPlanet Sci 37 263ndash297 2009

Cullen N J Molg T Kaser G Hussein K Steffen Kand Hardy D R Kilimanjaro Glaciers Recent areal ex-tent from satellite data and new interpretation of observed20th century retreat rates Geophys Res Lett 33 L16502doi1010292006GL027084 2006

Cullen N J Molg T Kaser G Steffen K and Hardy D REnergy-balance model validation on the top of Kilimanjaro Tan-zania using eddy covariance data Ann Glaciol 46 227ndash233doi103189172756407782871224 2007

Downie C Glaciations of Mount Kilimanjaro Northeast Tan-ganyika Geol Soc Am Bull 75 1ndash16doi1011300016-7606(1964)75[1GOMKNT]20CO2 1964

Downie C and Wilkinson P The Geology of KilimanjaroSheffield Geology Department University of Sheffield 1972

Downie C Humphries D W Wilcockson W H and WilkinsonP Geology of Kilimanjaro Nature 178 828ndash830 1956

Farr T G Rosen P A Caro E Crippen R Duren R Hens-ley S Kobrick M Paller M Rodriguez E Roth L Seal DShaffer S Shimada J Umland J Werner M Oskin M Bur-bank D and Alsdorf D The Shuttle Radar Topography Mis-sion Rev Geophys 45 RG2004doi1010292005RG0001832007

Geilinger W The retreat of the Kilimanjaro glaciers TanganyikaNotes and Records 2 7ndash20 1936

Gillman C An ascent of Kilimanjaro Geogr J 61 1ndash27 1923Hastenrath S The Glaciers of Equatorial East Africa Reidel Dor-

drecht 1984Hastenrath S Diagnosing the decaying glaciers of equatorial

East Africa Meteorol Z 15 265ndash271doi1011270941-294820060106 2006

Hastenrath S Recession of equatorial glaciers a photo documen-tation Sundog Publishing Madison Wisconsin 2008

Hastenrath S and Greischar L Glacier recession on KilimanjaroEast Africa 1912ndash89 J Glaciol 43 455ndash459 1997

Humphries D Interim report on the glaciology and meteorology ofKilimanjaro Tanganyika JulyndashSept 1953 Geological Survey ofTanganyika Dodoma 1953

Humphries D Preliminary notes on the glaciology of KilimanjaroJ Glaciol 3 475ndash479 1959

Humphries D Glaciology and glacial history in The Geology ofKilimanjaro edited by Downie C and Wilkinson P SheffieldGeology Department University of Sheffield 31ndash71 1972

Jarvis A Reuter H I Nelson A and Guevara E Hole-filledSRTM for the globe Version 4 available from the CGIAR-CSISRTM 90 m Database (httpsrtmcsicgiarorg) 2008

Jaeger F Forschungen in den Hochregionen des KilimandscharoMitt Deutsch Schutzgebiet 22 113ndash197 1909

Kaser G Hardy D R Molg T Bradley R S and Hyera TM Modern glacier retreat on Kilimanjaro as evidence of climatechange Observations and facts Int J Climatol 24 329ndash3392004

Kaser G Molg T Cullen N J Hardy D R andWinkler M Is the decline of ice on Kilimanjaro un-precedented in the Holocene Holocene 20 1079ndash1091doi1011770959683610369498 2010

Klute F Forschungen am Kilimandscharo im Jahre 1912 GeogrZeitschrift 20 496ndash505 1914

Klute F Ergebnisse der Forschungen am Kilimandscharo 1912ReimerndashVohsen Berlin 1920

Klute F Die stereophotogrammetrische Aufnahme der Hochregio-nen des Kilimandscharo Zeitschrift Ges Erdkunde Berlin 56144ndash151 1921

Mackinder C W Hobley C Gillman C and Johnston H Anascent of Kilimanjaro Discussion Geogr J 61 1ndash27 1923

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

420 N J Cullen et al A century of ice retreat on Kilimanjaro

2009ab 2012 Cullen et al 2007 Winkler et al 2010)It has been carefully demonstrated that the responses ofglaciers on Africarsquos highest mountain are sensitive to bothlocal and large-scale climate dynamics (Molg and Kaser2011) with the former extent of the glaciers providing im-portant insights into past climate conditions (Molg et al2009b) Given the importance of using glacier behaviour toreconstruct climate variability and change it is critical thatwe have a consistent and carefully documented record of thechanges to the extent of all ice bodies on Kilimanjaro throughtime

Since the first detailed observations of Kilimanjarorsquosglaciers by Hans Meyer in the late 1880s the dramatic re-treat has been described in numerous studies but surprisinglyfew have included detailed maps of the glaciers Early assess-ments that did were dependent primarily on ground-basedobservations and terrestrial photogrammetry (eg Meyer1891 1900 Jaeger 1909 Klute 1914 1920 1921 Downieet al 1956 Humphries 1959) More recently remote sens-ing techniques using both aerial photography and satelliteimagery have become the method of choice to characterizethe ongoing retreat (Hastenrath and Greischar 1997 Thomp-son et al 2002 2009 Cullen et al 2006) Beyond the ap-parent ease of using this approach lies the difficulty of inter-preting ice bodies over rugged and steep terrain using remotesensing imagery that is often compromised by snow coverwhich leads to the common problem of establishing bound-aries between snow and ice The problem is further compli-cated by the expectation that ice boundaries determined fromnew imagery will fit within previously published maps If onecomplies with this expectation in an effort to create an up-dated time series of ice retreat that is consistent with previouswork then it is likely that errors associated with disputableinterpretation of ice bodies andor inaccurate positioning ofsource imagery in the past will appear in any new interpreta-tion

To avoid this we have revisited the entire time series ofice retreat starting with an alternative interpretation of the1912 photogrammetric survey of Klute (1920) and later theground and aerial observations used by Downie and Wilkin-son (1972) In addition nine satellite images from 1975 to2011 have been orthorectified in a consistent fashion pro-viding us with a robust platform from which to establish amore reliable time series of ice retreat The use of three-dimensional (3-D) visualization techniques enhanced and fa-cilitated the interpretation of ice boundaries and allowed usto correct errors that have appeared in previous mapping ef-forts Importantly this study goes beyond clarifying some ofthe existing disputes about the location and extent of the re-maining ice bodies on Kibo by presenting the retreat ratesof three different glacier zones which are segmented usingtopographical information This geo-statistical approach fur-ther demonstrates the importance of different glacier regimeson Kibo which have been used successfully as a frameworkto characterize the different physical processes controlling

glacier growth andor decay on Africarsquos highest mountain(eg Nilsson 1931 Geilinger 1936 Kaser et al 2004 2010Cullen et al 2006 Molg et al 2009b)

2 Data and methods

21 Historical maps

211 The 1912 survey

A detailed ground-based photogrammetric survey of Kili-manjaro was conducted in 1912 by Fritz Klute and EduardOehler which in conjunction with information from previ-ous sketches and mapping efforts (eg Meyer 1900 Jaeger1909) led to a 1 50 000 scale map being produced based ona modified Clark 1880 ellipsoid datum (Klute 1920 1921)A scanned version of this product was registered in ESRIArcGIS software using four intersections of the geographi-cal grid shown on the map using a linear transformation toavoid introducing unwanted distortions The registered mapwas then re-projected onto UTM zone 37 South based on theWGS84 ellipsoid (UTM37S) which is the cartographic sys-tem we have chosen to use in this study While the quality ofthe map should be praised given the complexity of the terrainand the technical limitations of the emerging surveying tech-nique used severe planimetric distortions were evident andprevented the immediate mapping of the glaciers

We took advantage of the 50 m contour lines on the Klute(1920) map to develop an empirical correction for the ob-served planimetric distortions This was achieved by creat-ing equivalent 50-m contour lines using a digital elevationmodel (DEM) sourced from Shuttle Radar Topography Mis-sion (SRTM) data processed and distributed by CGIAR-CSI(Jarvis et al 2008) which has a 3 arcsecond resolution (90 mspatial resolution) (Farr et al 2007) The SRTM DEM wasprojected (UTM 37S) to a 30 m pixel size using a bicubicinterpolation The geolocation accuracy of SRTM DEM re-ported to be better than 13 m (Farr et al 2007) provideda strong basis for supporting the geometric correction of thehistorical map A total of 201 tie points were defined to matchcontour lines drawn on the historical map to those derivedfrom the SRTM DEM This process was further supportedthrough the use of a hill-shaded DEM developed to help re-veal geomorphological clues and features that facilitated aconsistent registration of the map with the underlying terrainA detailed map produced by Jaeger (1909 map dated 1906)of the glaciers in the western region of Kibo also proved valu-able in interpreting the ice boundaries established by Klute(1920)

212 The 1953 and 1957 geological surveys

The University of Sheffield conducted geological surveysof the upper slopes of Kilimanjaro at the invitation of theTanganyika Geological Survey in 1953 and 1957 which

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 421

resulted in numerous descriptive observations of the glaciers(eg Humphries 1959 Downie 1964 Downie and Wilkin-son 1972) Humphries (1959) indicated that in 1953 a pho-togrammetry survey was completed that covered the south-western slopes of Kibo Downie and Wilkinson (1972 p 1)also reported that the whole area could not be visited on thefirst expedition and the survey of the glaciers was completedin 1957 Nevertheless after the first expedition Downie etal (1956) produced a map of the Kibo caldera showing thebasic extent of the ice bodies in this area Their efforts cul-minated in the production of the 1963 geological map sheet562 (1 50 000) which depicts the full extent of the ice bod-ies on Kibo as of February 1962 based on their field observa-tions and aerial photography (Downie and Wilkinson 1972)Humphries (1972 p 53) also provided a detailed sketch ofthe glacier remnants on the western slope of Kibo Hastenrathand Greischar (1997) report that they mapped the ice extentin 1953 from a 1 25 000 scale map produced by Humphries(1953) Efforts to obtain a copy of the 1953 map associatedwith the unpublished report by Humphries (1953) proved un-successful which led to our decision to use the insert of the1963 geological map instead The glacier boundaries as de-picted in the insert are representative of the ice extent onKibo at the end of February 1962 Five intersections of thegeographical grid were used to register the map using a lin-ear transformation Though it is not clearly described theinsert of the 1963 geological map appears to be drawn ontoa modified Clark 1880 ellipsoid which is consistent with theArc 1960 geodetic datum used in Tanzania A re-projectionto UTM37SWGS84 allowed the 1963 map to be incorpo-rated into the consolidated GIS created for this study

22 Satellite data

221 Very high resolution sensors

The 06 m spatial resolution pan-sharpened and orthorecti-fied QuickBird image described by Cullen et al (2006) ob-tained on 1 February 2003 provided the basis for the ge-olocation of imagery used in this study as well as being re-tained to map the areal extent of the glaciers A Kompsatimage from 17 June 2011 was obtained at a L1A process-ing level to enable a precise orthorectification Fusion of thefour 4-m resolution multi-spectral bands with the 1-m reso-lution panchromatic band prior to the orthorectification dra-matically improved the potential to interpret spatial detailsin the Kompsat image Ten ground check points (GCPs) and12 independent check points (CPs) were collected from theQuickBird image and elevation data were retrieved from theSRTM DEM to support the orthorectification The latter wasconducted using the sensor model and rational polynomialcoefficients supplied by the image provider in ERDAS Imag-ine 2011 in conjunction with the SRTM DEM and yieldeda root mean square error (RMSE) of 25 m and 28 m for theGCPs and CPs respectively

222 Advanced Spaceborne Thermal Emission andReflection Radiometer (ASTER)

A search through the ASTER archive led to an image from19 August 2004 being identified as suitable to independentlyassess and support the orthorectification of other images A15 m DEM was produced following a stereo triangulationof the ASTER bands 3N and 3B in Leica PhotogrammetrySuite Twenty-one tie points and seven GCPs collected fromthe orthorectified QuickBird image as well as elevation datafrom the SRTM DEM were used and yielded a RMSE of81 44 and 93 m in the easting northing and elevation di-rections respectively A hill-shaded DEM was then createdand compared to that derived from the SRTM DEM This re-vealed no noticeable planimetric bias and demonstrated thatthe rectification of the ASTER image is consistent with thegeolocation accuracy of SRTM data Importantly the corre-sponding orthorectified ASTER image provided an absolutereference (within aboutplusmn10 m) to assess the rectification ofall other images Furthermore the almost snow-free ASTERimage was extremely useful in our efforts to establish theice boundaries in the QuickBird and Kompsat images de-spite its relatively coarse pixel size (ie 15 m) particularlyin the Breach Wall region due to the advantageous illumina-tion and spectral resolution of the sensor The ASTER imagewas not used as an epoch in the new time series presented inthis study because of its temporal proximity to the QuickBirdimage

223 Landsat images

The full archive of Landsat satellite imagery from the Kili-manjaro region available at the Earth Resources Observationand Science (EROS) Center was examined which led to fourepochs being identified as suitable for mapping the remainingice cover (Table 1) A Landsat 2 Multispectral Scanner Sys-tem (MSS) image from 15 August 1975 was preferred to thatof 24 January 1976 used by Hastenrath and Greischar (1997)as it exhibited much less transient snow Two Landsat 5 The-matic Mapper (TM) images acquired only a month apart on24 June and 26 July 1984 were used as they jointly provideda virtually snow free acquisition (the first date was retainedas the epoch of reference) with improved spatial resolution(ie 30 m) The Landsat 4 TM image of November 1989 citedby Hastenrath and Greischar (1997) (no date provided) couldnot be found in the archive and was therefore not used In-stead an image from 31 December 1992 was utilised Fi-nally a Landsat 7 ETM+ image from 21 February 2000 wasobtained to provide an independent assessment of the map-ping epoch characterized by Thompson et al (2002) using anaerial survey An image fusion technique was also completedto take advantage of the enhanced spatial resolution of thepanchromatic band of the ETM+ sensor (ie 15 m)

The selected Landsat images were orthorectified with thesensor model in ERDAS Imagine using GCPs identified in

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

422 N J Cullen et al A century of ice retreat on Kilimanjaro

Table 1Details of the terrestrial photogrammetry and satellite products used to characterize the retreat of the glaciers on Kibo The spatialresolution is given as either a map scale or pixel size The number of ground check points (GCPs) used for image matching and the root meansquare error (RMSE) (m) associated with the orthorectification (m) are also provided Two further Landsat images and one ASTER imagewere also processed but are not listed as they were not formally used for the time series (see Sect 22)

Map year Source Description of the product Pixel size GCP(m) or RMSEmap scale

MayndashOct 1912 Klute (1920 1921) Terrestrial photogrammetry with 74 1 50 000 ndashstand-lines over 2000 measured pointsand additional field sketches

Feb 1962 Downie and Terrestrial photogrammetry of western 1 50 000 ndashWilkinson (1972) glaciers ground observations and aerial

photography15 Aug 1975 Landsat 2 (MSS) Satellite imagery 57 1059724 Jun 1984 Landsat 5 (TM) Satellite imagery 30 1010531 Dec 1992 Landsat 4 (TM) Satellite imagery 30 118121 Feb 2000 Landsat 7 (TM) Satellite imagery 15 11681 Feb 2003 QuickBirda Satellite imagery 06 ndash26b

17 Jun 2011 Kompsata Satellite imagery 1 1025

a Pan-sharpened multi-spectral imageryb Ortho 1 5000 product level

the ASTER image which had already had its absolute ac-curacy assessed MSS and TMETM+ images were resam-pled to 20 and 10 m respectively using a bicubic interpola-tion to facilitate the visual interpretation of the identified iceboundaries The number of GCPs used and the correspondingRMSE for each image are provided in Table 1 Finally theco-registration between the orthorectified images was scru-tinized and revealed extremely good agreement Minor dis-tortions were identified in the QuickBird image in the steepsouth-east region of Kibo which required a local geomet-ric correction using an empirical rational function model andGCPs collected from the orthorectified Kompsat image Wesuspect the source of these distortions may have been fromthe use of incorrect elevation data in voids from the SRTMDEM used by the image providers during the initial orthorec-tification process It is also likely that unfavourable acquisi-tion geometry in conjunction with the steep terrain in this re-gion may have adversely affected the rectification Followingthe correction the ice boundaries for 2003 defined by Cullenet al (2006) were revised which resulted in a minor changein the total areal extent

23 Mapping ice boundaries

231 Previous mapping efforts

The maps created by Hastenrath and Greischar (1997 Fig 4)and Thompson et al (2009 Fig S2) (referred to in the fol-lowing as HG97F4 and T09FS2 respectively) were digitizedto help facilitate a comparison to the ice boundaries iden-tified in this study The single geographical coordinate in-dicated on HG97F4 (ie 305prime S 3720prime E according to a

Clark 1880 ellipsoid) made it difficult to perform a rigor-ous registration and raises the question as to whether glacierboundaries in HG97F4 are suitable for reproduction Insteadlinks were created to match contour lines with those derivedfrom the SRTM data using a linear transformation to avoidunwanted distortions The same process of matching con-tours was used to register T09FS2 which revealed a per-fect match with HG97F4 This enabled us to confirm thatthe glacier outlines prior to 2000 in T09FS2 are a repro-duction and edited version of HG97F4 which is also ac-knowledged by Thompson et al (2002 2009) Once bothmaps were registered in a consistent cartographic system theglacier boundaries at each epoch (ie 1912 1953 1976 and1989 for HG97F4 and T09FS2 2000 and 2007 for T09FS2)were carefully digitized This process revealed numerous in-consistencies between the glacier boundaries of HG97F4 andT09FS2 Hastenrath (2006 p 2) also alluded to a ldquomappingerrorrdquo related to the 2000 ice boundaries created by Thomp-son et al (2002) To ensure the latter fitted within his own1989 epoch Hastenrath (2006) was required to make adjust-ments to the ice boundaries of the northern ice field Cullen etal (2006 Fig 1) were also required to use an ldquoapproximateglacier extent in 1912rdquo as they found it difficult to place someice remnants observed in 2003 within previously establishedoutlines (eg Hastenrath and Greischar 1997 Thompson etal 2002) which was scrutinized by Thompson et al (2009Fig S4) These problems prompted us to make the decisionto revisit the entire sequence of historical maps and satel-lite imagery While it is not our intention to focus on all ofthe mapping inconsistencies we have observed we do de-scribe some key differences in detail (see Sect 31) in order

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 423

to characterize the physical processes controlling glacier re-treat on Kibo

232 Establishing new ice boundaries

The boundaries of ice bodies were visually interpreted anddigitized from the orthorectified satellite images and histor-ical maps Despite the variability in the spatial and spectralresolution of the data products as well as having to accountfor varying amounts of transient snow in the satellite im-ages the use of the high temporal resolution sequence sig-nificantly improved our ability to interpret the location andextent of ice bodies consistently through time To supportand facilitate the interpretation of all data products (Table 1)three-dimensional visualization techniques were extensivelyused in conjunction with aerial and ground-based photogra-phy from others (eg Hastenrath 2008) and our own fieldexpeditions between 2005ndash2012 The vantage points fromwhich historical photographs were taken were reproducedin the 3-D model of Kilimanjaro which proved particularlyuseful when revisiting the glacier outlines mapped by Klute(1920) The use of 3-D visualization also helped us to inter-pret glacier positions in relation to distinctive geomorpho-logical features (eg moraines rock outcrops) which are de-scribed in more detail in the following section Finally tohelp facilitate future efforts all data products from this anal-ysis have been made available on the Global Land Ice Mea-surements from Space (GLIMS) Glacier Database

3 Results and discussion

31 The mapping reloaded

The total amount of ice cover on Kilimanjaro on 17 June2011 was 176 km2 (Table 2) with the location of these icemasses shown in Fig 1 The remaining ice masses are rem-nants of more extensive glaciers which were first mappedand given names by Meyer (1900) with the nomenclaturestill in use despite the character of the ice bodies havingchanged dramatically over the last century The majority ofthe ice cover observed on Kilimanjaro today fails to strictlyadhere to any common definition of a glacier with most enti-ties isolated and stagnant but following the definition sug-gested by Racoviteanu et al (2009) we include any bodyof ice that is observed after transient snow melts It is clearfrom the accounts and photographs of the early explorers thatonly the principal glaciers were named and that a consider-able number of small stagnant ice masses in particular in thecrater area were not formally recognized This is no more ev-ident than in the south-east crater area where the first explor-ers accessed the summit region through a number of weak-nesses in the ice referred to as notches (eg Hans Meyer andJohannes notches) which have been described in detail else-where (eg Geilinger 1936 Hastenrath 1984) Despite thisapparent difficulty it is clear that the remaining ice bodies

today are remnants of the more extensive glaciers mappedby Klute (1920) using terrestrial photogrammetry (Table 1)with the boundaries established from this effort reflecting theice cover in 1912 (Fig 1) Our calculation of the extent ofthe glaciers on Kilimanjaro at the end of October in 1912 is1140 km2 (Table 2) which is 85 more than what is ob-served on the mountain today

The differences between our interpretation of the 1912map produced by Klute (1920) and that produced by Has-tenrath and Greischar (1997) are also shown in Fig 1 Asnoted in Sect 232 our interpretation of the glacier bound-aries was supported by geomorphological evidence observedusing 3-D visualization and historical photographs Impor-tantly the 3-D model allowed us to carefully identify andlocate the most recent moraines Using these geomorpholog-ical features as markers we were able to determine with ahigh degree of accuracy the former location of the glaciersThis approach was particularly useful in establishing the pastlocation and extent of the outlet glaciers feeding from thenorthern ice field (the Credner Drygalski and (Great andLittle) Penck glaciers) the Uhlig glacier and the two distinctglacier streams located in the western breach area (the Lit-tle and Great Barranco glaciers) as well as the four primarysouthern ice field glaciers (the Heim Kersten Decken andRebmann) The map produced by Jaeger (1909 map dated1906) was also valuable in allowing us to identify a num-ber of small ice bodies located above both the Little andGreat Barranco glaciers (Fig 1) Our interpretation of thealtitudinal extent of the southern ice field glaciers at about4600 m (Fig 1) is supported by the photographs of Uhlig(1904 Fig 50 p 641) and Klute (1920 Fig 1) It is impor-tant to note that both these historical photographs reveal acomplexity of the outcropping rock at the lower elevations ofthe southern ice field glaciers that we have made an attemptto account for (Fig 1)

Despite a well-documented and photographic account ofthe historical retreat of Kilimanjarorsquos glaciers which hasbeen carefully reviewed by Hastenrath (1984 2008) therehave been a number of inconsistences in the interpretationof ice bodies in recent mapping efforts (eg Hastenrathand Greischar 1997 Cullen et al 2006 Hastenrath 2006Thompson et al 2002 2009) which require further clarifi-cation as they have an impact on our ability to interpret thephysical processes controlling the demise of the glaciers onKibo To do this one must briefly reflect on the historical ob-servations starting with those of Hans Meyer who provideda sketch (Meyer 1890 p 345) and a map (Meyer 1891) thatindicates that an ice stream which is now referred to as theFurtwangler glacier flowed over the crater rim into the west-ern breach area Meyer (1891 p 319) notes ldquothe ice from thevast caldera falls as a mighty cascade into the great westernfissure where it unites with the accumulations in the fissureitselfrdquo Though Hastenrath (1984 p 78) disputes whetherthis ice stream (the Furtwangler glacier) was connected tothe Little andor Great Barranco glaciers other historical

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

424 N J Cullen et al A century of ice retreat on Kilimanjaro

4800

4600

5000

4400

5200

5400

5600

5700

4200

4000

5800

4800

4400

4600 4200

37deg23E

37deg23E

37deg22E

37deg22E

37deg21E

37deg21E

37deg20E

37deg20E

3deg3

S

3deg3

S

3deg4

S

3deg4

S

3deg5

S

3deg5

S

3deg6

S

3deg6

S

314000

314000

316000

316000

318000

318000

320000

320000

9658000

9658000

9660000

9660000

9662000

9662000

Uhuru Peak

Northern ice field

Heim

Uhlig

Dec

ken

Ratzel

Kersten

Reb

man

n

Credner

Drygalski

Arrow

Diamond

Balletto

Furtwaumlngler

LittlePenck

5895m

IceCastle

Easternice field

Southernice field

WGS84UTM 37S

0 1000 2000

Meters

Revised 1912 boundaries

Credner Glacier name

2011 Ice boundaries

1912 boundaries fromHastenrath and Greischar (1997)

Little Barranco

Fig 1 Glacier extent on Kilimanjaro in 1912 and 2011 Glacier names with an asterisk do not appear in Klute (1920) and are described inthe text The interpretation of the glacier extent for 1912 as defined by Hastenrath and Greischar (1997) is shown as a dashed line (blue) forcomparative purposes The nomenclature for each glacier used by Klute (1920) is given while the names of four glaciers introduced afterthis time are indicated usinglowast

accounts are more supportive (Gillman 1923 Geilinger1936) A more detailed map of Kilimanjarorsquos glaciers pro-duced by Meyer (1900) based on his observations during afield expedition in 1898 shows the connection no longer ex-isted suggesting a break-up occurred sometime in the early1890s Though the timing is uncertain the implications oflosing this connection are not overlooked by Geilinger (1936p 12) who refers to the Little and Great Barranco glaciers asldquoindependent dead glaciersrdquo and made the astute observationthat both ice entities were retreating at their upper and lowermargins

This observation holds some importance as it helps clar-ify the source of the ice remnants still observed today Themaps by Meyer (1900 map dated 1898) Jaeger (1909map dated 1906) and Klute (1920 map dated 1912) clearlyshow distinct ice streams for the Little and Great Barrancoglaciers which as noted may have been fed by ice from theFurtwangler glacier as recently as the late 19th century Ourrevision of the Klute (1920) map from 1912 (Fig 1) honoursthe discrete boundary between the Little and Great Barranco

glaciers which is not as clearly depicted in the Hastenrathand Greischar (1997) interpretation The failure to clearlydistinguish the boundaries and location of these two indepen-dent ice streams has been inherited by Thompson et al (20022009) who used the Hastenrath and Greischar (1997) inter-pretation of the ice extent prior to 2000 (see Sect 231)Thompson et al (2009) argue that ice remnants identified byCullen et al (2006) in both the ice streams of the former andmore extensive Little and Great Barranco glaciers are absentfrom previous observations and therefore most likely tran-sient snow The sketch by Humphries (1972 Fig 46) andthe map produced by Messerli (1980 Fig 6) also support theinterpretation by Cullen et al (2006) while the ice bodiesin question have been identified in each of the satellite im-ages used in the present analysis (Fig 2) Though not labelledby Humphries (1959 1972) or by those preceding his obser-vations the Arrow glacier has been used to refer to one ofthe Little Barranco ice remnants (Fig 1) It appears that thisterm was adopted by climbers on the mountain (eg Samp-son 1971 p 202) but the exact location of this ice body is

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 425

Table 2Areal extents (km2) of the total ice bodies between 1912 and 2011 as well as for zone 1 (western region) zone 2 (plateau region)and zone 3 (southern region) as defined in the text and Fig 4 The annual rates of area change per observation period (10minus2 km2 yrminus1) areshown in parentheses for the total and zoned areas

Epoch Duration Area [km2] (rate of area change [10minus2 km2 yrminus1])

[years] Total Zone 1 Zone 2 Zone 3

191283 11400 4540 4293 2567196216 4936 7320 (minus827) 2520 (minus409) 3161 (minus229) 1639 (minus188)197562 1347 6052 (minus941) 2060 (minus342) 2529 (minus469) 1463 (minus131)198448 887 4815 (minus1395) 1628 (minus487) 1949 (minus654) 1238 (minus254)199300 853 3800 (minus1190) 1142 (minus570) 1638 (minus365) 1020 (minus256)200014 715 2920 (minus1232) 0701 (minus617) 1411 (minus318) 0808 (minus297)200308 295 2500 (minus1425) 0539 (minus550) 1251 (minus543) 0710 (minus332)201146 838 1762 (minus881) 0333 (minus246) 0937 (minus375) 0492 (minus260)

not well documented and we are uncertain that the remnantidentified in Fig 1 and by Cullen et al (2006) is the sameas that referred to by others (eg Sampson 1971 Messerli1980 Hastenrath and Greischar 1997)

There has also been some confusion about the Diamondand Balletto glaciers (Fig 1) which were not formally rec-ognized in the early mapping work (Meyer 1990 Jaeger1909 Klute 1920) or acknowledged by Humphries (19591972) but are both described by Hastenrath (1984) whileSampson (1971) and Messerli (1980) only refer to the for-mer The Breach Wall area where both glaciers are locatedwas mapped by Klute (1920) but as no break up from thesouthern ice field had occurred at this time they were notgiven discrete labels (Fig 1) However what is indisputableis the presence of an ice field (the Balletto glacier) beneaththe headwall of the Diamond glacier which was not mappedby Thompson et al (2002 2009) The presence of this icefield is not only observed in the satellite record but is docu-mented by Reinhold Messner (Messner 1991 Chapter 21)who made the first (and only) direct ascent of the Breach Wallof Kilimanjaro in 1978 Though Cullen et al (2006) mappedthis region of ice (the Balletto glacier) they failed to label itcorrectly placing the label beneath the Heim glacier after fol-lowing the interpretation of Hastenrath and Greischar (1997Fig 2) Further the Diamond glacier is absent after 1976 inthe maps produced by Hastenrath and Greischar (1997) andHastenrath (2006) but appears in the Thompson et al (20022009) maps despite using the former to build their time se-ries Finally the Uhlig glacier which Jaeger (1909 p 130)named after his friend C Uhlig does not appear to have beenconnected to the northern ice field in the past (Fig 1) Whilethe upper boundary of the Uhlig glacier is difficult to inter-pret in the Klute (1920) map it is clearly shown by Jaeger(1909) that it was not joined to the northern ice field

Figure 2 shows the change in ice extent over almost a 100-yr period which is consistent through time and removes mostof the uncertainty about the location and extent of the re-maining ice bodies on Kibo The retreat over the first part of

the 20th century was approximately 41 km2 (Fig 2a and b)while over the last 50 yr 56 km2 of the ice cover has beenlost (Fig 2cndashh) The retreat between 1912 and 1962 (Fig 2aand b) resulted in a distinct break-up of ice in the vicinity ofthe notches (south-east crater region) separating the northernand eastern ice fields from the southern ice field Remnantsof the Ice Castle (Fig 1) still remained in 1962 and were de-scribed by Sampson (1971) as still being present in the early1970s but appeared to have been lost by 1975 (Fig 2c) Othernoticeable features of the retreat in the early 20th century in-clude the development of a distinct separation between theGreat and Little Penck glaciers which formed through theexpansion of the Ravenstein (first mapped by Meyer 1900)the rapid demise of the Uhlig glacier and the split-up of theDiamond and Balletto glaciers from the southern ice field(Fig 2a and b)

Between 1962 and 1975 the persistent retreat of ice led tothe northern and eastern ice fields becoming separated fromeach other as did the Great Penck glacier from the north-ern ice field (Fig 2b and c) The Little Penck proved moreresistant and did not separate from the northern ice field un-til after 1992 while the remnant of the Great Penck com-pletely disappeared during this period (Fig 2e and f) thoughground-based (field) observations in January 2005 revealeddebris-covered ice at the location of the lowest 1984 extent(Fig 2d) As noted by others (eg Hastenrath and Greischar1997 Cullen et al 2006) the northern ice field developed ahole sometime after 1984 (Fig 2d and e) which opened intoa large canyon after 2003 (Fig 2g and h) Field observationsby the first author in October 2012 revealed that the north-ern ice field has separated into two independent ice bodieswhich must have occurred sometime after 17 June 2011 TheFurtwangler glacier has also separated into smaller ice enti-ties which began after 2003 (Fig 2g and h) The southernice field has not escaped the on-going demise with the upperpart of the Kersten glacier separating from its lower reachesbetween 2003 and 2011 (Fig 2g and h) which was some-what anticipated after mass balance modelling showed this

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

426 N J Cullen et al A century of ice retreat on Kilimanjaro

(g) 2003 (h) 2011

(a) 1912 (b) 1962

(c) 1975 (d) 1984

(e) 1992 (f) 2000

Credner

Great Penck

Ratzel

Ker

sten

Northern ice field

Easternice field

Southernice field

GreatBarranco

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Fig 2 Ice retreat over Kilimanjaro based on the revised areal extent for the following years(a) 1912(b) 1962(c) 1975(d) 1984(e)1992(f) 2000(g) 2003 and(h) 2011 An interactive 3-D illustration of the sequence can be sourced in the Supplement

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 427

was likely to occur (Molg et al 2009b) The upper part ofthe Heim has disappeared and the Decken has narrowed con-siderably over the last 8 yr (Fig 2g and h)

32 New interpretation of the ice retreat

A consistent record of the former extent of the glaciers onKibo provides a platform from which the relationship be-tween glaciological and meteorological processes can be fur-ther explored We believe the spatial and temporal variabilityof the ice cover on Kibo can be used to unravel the physi-cal processes responsible for controlling the decay of the icemasses through time as well as providing insights into themechanisms that may have been responsible for growth in thepast Interestingly those who were privileged to make someof the first (formal) observations on Kilimanjaro recognizedthere was an asymmetry in the ice cover (eg Meyer 1891)which pointed to specific climatological drivers controllingglacial retreat (eg Gillman 1923 Nilsson 1931 Geilinger1936) Meyer (1891 p 308ndash314) noted that ldquoan immensemountain mass like Kilimanjaro must considerably modifythe prevailing wind directionrdquo before making the point thatthe primary cause in the variability of the ice limit is ldquotheunequal distribution of moisture on the northern and south-ern sides of the mountainrdquo Though Meyer (1891) recognizedvariability in moisture was the most important control healso speculated that the asymmetry of low reaching glacierson Kibo in the south south-west and western sectors of themountain a pattern that persisted throughout the 20th cen-tury (Fig 2) depended on ldquolocal peculiarities which favourthe preservation of the snow [rather] than on physical condi-tions that favour a large snowfallrdquo (Meyer 1891 p 315) Torevisit this early observation it is useful to reflect on the hyp-sometry of the glaciers in 1912 (Fig 3a) which shows thatthe probability density in relation to the distribution of the icewas greatest above 5700 m This high density reflects the icecover in the crater region with the majority of the ice in 1912being part of the former ice cap that linked the northern andeastern ice fields (Fig 1) Figure 3b and c show that slopeand aspect were also important controls on the density of icecover with the steep south and south-west facing slopes morefavourable which enabled the glaciers to extend to elevationsbelow 4600 m For the glaciers to have extended to the sameelevation on the east-facing slopes they would have had tocover a much greater horizontal distance (Fig 1)

To fully account for the distinct geometry of the ice masseson Kibo and to establish the physical processes controllingtheir location and extent through time Kaser et al (2004)established a framework that included four separate glacierregimes The motivation to do this was partly based on theearly recognition that the tabular shaped glaciers in the craterregion of Kibo behave differently to glaciers on the flanks ofthe mountain (eg Nilsson 1931 Geilinger 1936) Guidedby this distinction Cullen et al (2006) used the 5700 m con-tour which is roughly the height of the outer crater rim to

Density times 10 minus3

2

4

6

8

WEST EAST

SOUTH

NORTH

4600 4800 5000 5200 5400 5600 58000

1

2

3

4

Elevation [m]

Den

sity

times 10

minus3

0 10 20 30 40 50 60 70 800

1

2

3

4

Slope [deg]

Den

sity

times 10

minus2

(a)

(c)

(b)

Fig 3 Distribution of topographic measures from the areal extentof all ice bodies in 1912(a) elevation(b) slope and(c) aspect Themagnitude of the normalised distribution on the y-axis is a proba-bility density function with the integral equal to unity

distinguish between two broad glacial systems (1) plateau(ge 5700 m) and slope (lt 5700 m) glaciers This simple bi-nary classification has proved useful in describing some ofthe key physical processes responsible for the retreat of iceon Kibo (eg Molg et al 2009b Kaser et al 2010) Tofurther characterize the spatial and temporal variability ofglacier retreat on Kibo we have adopted a geo-statisticalapproach in this study that uses elevation slope and aspectof the ice cover in 1912 (the feature space) to identify clus-ters resembling topographical classes This was achieved byperforming a cluster analysis containing maximum likeli-hood estimates of the topographical parameters in a Gaussian

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

428 N J Cullen et al A century of ice retreat on Kilimanjaro

mixture model (McLachlan and Peel 2000) which resultedin three broad glacial zones being identified (1) western re-gion (zone 1) (2) plateau region (zone 2) and (3) southernregion (zone 3) (Fig 4)

Figure 5 shows the rate of retreat of each of the glacialzones through time as well as a least-squares linear regres-sion fit (and 80 confidence intervals) through each of theglacial zone time series As described by Cullen et al (2006)the retreat of the tabular shaped plateau glaciers (zone 2) iscontrolled primarily by solar radiation-induced ablation onthe vertical cliffs that characterize them as demonstratedthrough radiation modelling (Molg et al 2003) Winkleret al (2010) revealed that direct insolation is the reasonfor the predominant zonal alignment of the ice cliffs andtheir bimodal annual recession (sunlit versus shaded peri-ods) Though the shrinkage of the ice cover along the verticalcliffs is fairly constant from year to year the areal retreat ofthe zone 2 region is not entirely linear with the highest ratesoccurring between 1975 and 1984 (Table 2) which corre-spond to the period following the separation of the northernand eastern ice fields (Fig 2) Therefore changes in the ge-ometry of the ice bodies and their exposure to solar radiationappear to have led to variations in retreat rates (Table 2 andFig 5) which has also been described by Kaser et al (2010)

An explanation for the growth and decay of the plateauglaciers including a description of the physical processesthat may have led to the observed steps in the northern icefield has been provided by Kaser et al (2010) To sustaingrowth in the crater rim region (zone 2) it is critical thatsnowfall is sufficient to allow a continuous snow cover todevelop that is able to survive from one year to the next Ifthe annual build-up of snow is sustained through a period oftime it is quite conceivable that the plateau glaciers couldgrow to several tens of metres in thickness quite quicklyraising the question as to whether this accumulation of icehas the potential to become a source of mass for glacierson the western flanks of the mountain (zone 1) Meyer(1891 p 316ndash317) was somewhat sceptical as to whetherthe plateau region acted as a ldquoreservoirrdquo and thought it ldquomuchmore reasonable to suppose that the rocky rim of the crateris the real reservoirrdquo However Meyer (1890 1891) did pro-vide an account and sketches to support a connection of theplateau ice to the Little and Great Barranco glaciers in thepast which as described may have been important in sustain-ing them (Gillman 1923 Geilinger 1936) There is someimportance in investigating the extent of the linkages be-tween these two glacial zones as together they accounted foralmost 80 of the total ice cover in 1912 (Table 2)

The remaining ice bodies in zone 2 (plateau region) aremostly stagnant today with field measurements and prelim-inary modelling results indicating deformation rates of icewithin the crater are small (lt 10 cm yrminus1) (Kaser et al 2004Winkler et al 2010) It seems unlikely that the present or his-torical thickness of the plateau ice has been sufficient to al-low substantial deformation (Kaser et al 2004) suggesting

3140 3150 3160 3170 3180 3190 3200 32109657

9658

9659

9660

9661

9662

9663

9664

4000

4000

4500

4500 4500

45004500

5000

5000

5000

5000

5000

5500

5500

5500

UTM easting in meters times105

UT

M n

orth

ing

in m

eter

s times

106

Fig 4 Classification of regions based on the distribution of topo-graphic measures in 1912 Zone 1 (western region) zone 2 (plateauregion) and zone 3 (southern region) are shown in red orange andyellow respectively

that mass transport from the nearly horizontal plateau downinto the western slope glaciers (zone 1) has most likely beenminor Careful observations by Klute (1920) demonstratedthat the velocity of the centre of the Great Penck glacier was064 m over a 30-day period in 1912 which led him to spec-ulate that the annual downslope movement was about 7 mper year It should be noted that Nilsson (1931) and Osmas-ton (1989) also report on these observations but incorrectlycite Jaeger (1909) as the person responsible for obtaining themeasurements Nonetheless it is improbable that ice fromthe plateau region could have supported the observed flowdynamics of the Great Penck glacier with the upper reachesof the mountain flanks the most likely source of the massThus the growth and decay of the glaciers in zone 1 (west-ern region) do not appear to be directly dependent on thestability of plateau ice (zone 2) To further resolve the phys-ical processes controlling the stability of the plateau ice itwould be of interest to develop a better understanding of themechanisms responsible for vertical cliff development

The glaciers on the southern slopes (zone 3) differ fromthose on the western slopes (zone 1) as they are separatedfrom the large masses of ice within the caldera (zone 2) bybed topography (Humphries 1972) This distinction is im-portant as any assessment of the atmospheric controls onthe behaviour of the southern ice fields is independent ofthe complexity of the vertical cliff retreat in the plateau re-gion (zone 2) That said the cluster analysis that led to thetopographical classes used in this study shows that the up-per sections of the southern ice field exhibit similar phys-ical characteristics to the ice within the caldera (Fig 4)Nevertheless to obtain a full solution for the asymmetry ofthe glaciers on Kibo that the early explorers observed itis critical to establish an understanding of the atmospheric

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 429

1900 1920 1940 1960 1980 2000 2020 2040 20600

05

1

15

2

25

3

35

4

45

5

Year

Are

al e

xten

t [km

2 ]

zone 1 R2=099

zone 2 R2=097

zone 3 R2=09980 confidence interval

4000

4000

4000

4500

4500

4500

45005000

5000

5000

50005000

5500

5500

5500

Fig 5 The areal change of glaciers and ice bodies based on thethree-member classification Linear regressions (thin solid lines) foreach of the three zones are shown as well as the 80 confidenceintervals for each of these relationships

processes controlling the growth andor decay of the southernice fields This has been achieved by Molg et al (2009b)who used observations and atmospherendashglacier modellingto show how critical precipitation humidity and fractionalcloud cover are on the energy and mass balance of the Ker-sten glacier (zone 3) These results compare favourably toearly accounts of the importance of precipitation (moisturechange) and cloud cover on the asymmetry (Meyer 1891Klute 1920 Gillman 1923 Geilinger 1936) with the diur-nal variability of cloud cover being especially important inproviding protection to the glaciers on the western side ofthe mountain (eg Meyer 1891 Hastenrath and Greischar1997)

Recent regional atmospheric modelling experiments haveprovided a more quantitative assessment to account for theasymmetry in glacier mass on Kibo demonstrating that thewest and south-western areas of the mountain are favourableto moisture convergence as controlled by atmospheric dy-namics (Molg et al 2009a) As noted by Osmaston (1989)the asymmetry is not confined to the most recent glacier ex-tents with the pattern of ice distribution established fromformer moraines indicating that recent climatic controls onglacier behaviour are similar to those in the past This sug-gests that establishing the specific nature of the atmosphericcontrols responsible for the most recent shift from glaciergrowth to decay in particular in zones 2 and 3 (plateau andsouthern slope regions) remains critical if we are to useglacier behaviour to reconstruct climatic conditions of thedeeper past

Finally it is difficult not to speculate on when the ice onKibo will disappear which becomes more achievable if onerecognizes the importance of the different glacial zones onthe mountain If one accepts linear extrapolation based on

changes from the past is suitable to make projections for thefuture it is apparent that the ice cover in zone 1 (western re-gion) will most likely be gone before 2020 and that ice inzone 2 (plateau) and zone 3 (southern slopes) will disappearsometime close to 2040 (Fig 5) It is likely that the plateauwill harbour the final remnants of ice on Kibo which couldremain slightly beyond the middle of the 21st century Inter-estingly our linear extrapolation for the loss of ice is in closeagreement to earlier projections using radiation modelling(Molg et al 2003) Though the loss of ice will have no hy-drological significance to lowland areas (eg Gillman 1923Kaser et al 2004 Molg et al 2008b) there is no doubt thattheir disappearance will be a ldquogrievous aesthetic lossrdquo (Salt1951 p 160) which will undoubtedly impact the experiencetourists have on the mountain (Molg et al 2008b)

4 Conclusions

With reference to the glaciers found on Mt Kenya and Kiboin East Africa C Gillman remarked ldquowe might perhapslearn a good deal as to the general world circulation fromcareful and prolonged observation on these particular sum-mitsrdquo (Mackinder 1923 p 23) This statement still holdstrue today (eg Molg et al 2006) and with the increasingawareness that glaciers on Africarsquos highest mountain can beused to extract climate information from the past it seemedtimely to revisit the mapping of the ice cover on Kibo throughtime Our motivation to do so has been to ensure the histor-ical record of these changes is of the highest possible qual-ity and in a format that should help facilitate on-going andfuture research dedicated to resolving the complexity of cli-mate variability and change in East Africa and beyond Thiswas achieved by using a consistent cartographic frameworkthat allows historical maps and new satellite imagery to beexamined The use of three-dimensional visualization tech-niques greatly enhanced our ability to establish the formerlocation of the glaciers

The revisited time series of glacial retreat on Kiliman-jaro contains eight epochs and spans almost 100 yr (1912ndash2011) The glaciers have retreated from their former extentof 1140 km2 in 1912 to 176 km2 in 2011 which representsa loss of about 85 of the ice cover over almost a 100-yr pe-riod Our estimate of the total retreat is consistent with recentassessments (Cullen et al 2006 Thompson et al 2009) butthe careful re-examination of the location and extent of indi-vidual ice bodies through time has allowed us to identify anumber of important differences from previously publishedmaps (eg Hastenrath and Greischar 1997 Thompson et al2002 2009) Historical evidence has been used to supportour interpretation with the detailed observations and pho-tographs obtained by early explorers on Kilimanjaro found tobe extremely valuable (eg Meyer 1891 1900 Uhlig 1904Jaeger 1909 Klute 1914 1920 1921 Gillman 1923 Nils-son 1931 Geilinger 1936) It has to be acknowledged that

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

430 N J Cullen et al A century of ice retreat on Kilimanjaro

much of what we know about Kilimanjaro today was estab-lished through their efforts

To further explore the importance of different glacierregimes on Kibo (Kaser et al 2004) a geo-statistical ap-proach was applied which resulted in the segmentation ofthe ice cover into three broad glacial zones A careful ex-amination of the retreat of ice in each of these glacier zonesindicates that there is a high probability that most of the re-maining ice cover on Kilimanjaro will disappear by 2040and it is highly unlikely that any ice body will remain af-ter 2060 if present retreat rates are maintained Before thishappens it would still be useful if the remaining ice bodiesin particular the remnants of the northern ice field could befurther examined to help resolve existing uncertainties aboutthe mechanisms responsible for controlling the formation ofthe vertical cliffs that characterize them (eg Winkler et al2010) In addition further efforts are required to scrutinizethe exact timing of the onset of the present glacier retreat firstreported by Hans Meyer in the late 19th century If we canrefine our understanding of the atmospheric controls behindthe most recent shift from glacier growth to decay on Kibowe will be in a stronger position to characterize the climaticcontrols of past and much larger glaciations on Africarsquos high-est mountain

Supplementary material related to this article isavailable online athttpwwwthe-cryospherenet74192013tc-7-419-2013-supplementpdf

AcknowledgementsThis research is funded by the Departmentof Geography and School of Surveying University of OtagoNew Zealand Thomas Molg was supported by the Alexander vonHumboldt Foundation The contributions made by Georg Kaserand Michael Winkler were supported by the Austrian ScienceFund (FWF) Projects P 20089-N10 and P 21288-N10 Thisresearch would not have been possible without the support from thefollowing Tanzanian authorities (COSTECH KINAPA TANAPATAWIRI) We are grateful to the referees that took the time toprovide feedback on this research in particular the careful andthoughtful comments by Henry Brecher

Edited by M Schneebeli

References

Chiang J C H The tropics in paleoclimate Annu Rev EarthPlanet Sci 37 263ndash297 2009

Cullen N J Molg T Kaser G Hussein K Steffen Kand Hardy D R Kilimanjaro Glaciers Recent areal ex-tent from satellite data and new interpretation of observed20th century retreat rates Geophys Res Lett 33 L16502doi1010292006GL027084 2006

Cullen N J Molg T Kaser G Steffen K and Hardy D REnergy-balance model validation on the top of Kilimanjaro Tan-zania using eddy covariance data Ann Glaciol 46 227ndash233doi103189172756407782871224 2007

Downie C Glaciations of Mount Kilimanjaro Northeast Tan-ganyika Geol Soc Am Bull 75 1ndash16doi1011300016-7606(1964)75[1GOMKNT]20CO2 1964

Downie C and Wilkinson P The Geology of KilimanjaroSheffield Geology Department University of Sheffield 1972

Downie C Humphries D W Wilcockson W H and WilkinsonP Geology of Kilimanjaro Nature 178 828ndash830 1956

Farr T G Rosen P A Caro E Crippen R Duren R Hens-ley S Kobrick M Paller M Rodriguez E Roth L Seal DShaffer S Shimada J Umland J Werner M Oskin M Bur-bank D and Alsdorf D The Shuttle Radar Topography Mis-sion Rev Geophys 45 RG2004doi1010292005RG0001832007

Geilinger W The retreat of the Kilimanjaro glaciers TanganyikaNotes and Records 2 7ndash20 1936

Gillman C An ascent of Kilimanjaro Geogr J 61 1ndash27 1923Hastenrath S The Glaciers of Equatorial East Africa Reidel Dor-

drecht 1984Hastenrath S Diagnosing the decaying glaciers of equatorial

East Africa Meteorol Z 15 265ndash271doi1011270941-294820060106 2006

Hastenrath S Recession of equatorial glaciers a photo documen-tation Sundog Publishing Madison Wisconsin 2008

Hastenrath S and Greischar L Glacier recession on KilimanjaroEast Africa 1912ndash89 J Glaciol 43 455ndash459 1997

Humphries D Interim report on the glaciology and meteorology ofKilimanjaro Tanganyika JulyndashSept 1953 Geological Survey ofTanganyika Dodoma 1953

Humphries D Preliminary notes on the glaciology of KilimanjaroJ Glaciol 3 475ndash479 1959

Humphries D Glaciology and glacial history in The Geology ofKilimanjaro edited by Downie C and Wilkinson P SheffieldGeology Department University of Sheffield 31ndash71 1972

Jarvis A Reuter H I Nelson A and Guevara E Hole-filledSRTM for the globe Version 4 available from the CGIAR-CSISRTM 90 m Database (httpsrtmcsicgiarorg) 2008

Jaeger F Forschungen in den Hochregionen des KilimandscharoMitt Deutsch Schutzgebiet 22 113ndash197 1909

Kaser G Hardy D R Molg T Bradley R S and Hyera TM Modern glacier retreat on Kilimanjaro as evidence of climatechange Observations and facts Int J Climatol 24 329ndash3392004

Kaser G Molg T Cullen N J Hardy D R andWinkler M Is the decline of ice on Kilimanjaro un-precedented in the Holocene Holocene 20 1079ndash1091doi1011770959683610369498 2010

Klute F Forschungen am Kilimandscharo im Jahre 1912 GeogrZeitschrift 20 496ndash505 1914

Klute F Ergebnisse der Forschungen am Kilimandscharo 1912ReimerndashVohsen Berlin 1920

Klute F Die stereophotogrammetrische Aufnahme der Hochregio-nen des Kilimandscharo Zeitschrift Ges Erdkunde Berlin 56144ndash151 1921

Mackinder C W Hobley C Gillman C and Johnston H Anascent of Kilimanjaro Discussion Geogr J 61 1ndash27 1923

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 421

resulted in numerous descriptive observations of the glaciers(eg Humphries 1959 Downie 1964 Downie and Wilkin-son 1972) Humphries (1959) indicated that in 1953 a pho-togrammetry survey was completed that covered the south-western slopes of Kibo Downie and Wilkinson (1972 p 1)also reported that the whole area could not be visited on thefirst expedition and the survey of the glaciers was completedin 1957 Nevertheless after the first expedition Downie etal (1956) produced a map of the Kibo caldera showing thebasic extent of the ice bodies in this area Their efforts cul-minated in the production of the 1963 geological map sheet562 (1 50 000) which depicts the full extent of the ice bod-ies on Kibo as of February 1962 based on their field observa-tions and aerial photography (Downie and Wilkinson 1972)Humphries (1972 p 53) also provided a detailed sketch ofthe glacier remnants on the western slope of Kibo Hastenrathand Greischar (1997) report that they mapped the ice extentin 1953 from a 1 25 000 scale map produced by Humphries(1953) Efforts to obtain a copy of the 1953 map associatedwith the unpublished report by Humphries (1953) proved un-successful which led to our decision to use the insert of the1963 geological map instead The glacier boundaries as de-picted in the insert are representative of the ice extent onKibo at the end of February 1962 Five intersections of thegeographical grid were used to register the map using a lin-ear transformation Though it is not clearly described theinsert of the 1963 geological map appears to be drawn ontoa modified Clark 1880 ellipsoid which is consistent with theArc 1960 geodetic datum used in Tanzania A re-projectionto UTM37SWGS84 allowed the 1963 map to be incorpo-rated into the consolidated GIS created for this study

22 Satellite data

221 Very high resolution sensors

The 06 m spatial resolution pan-sharpened and orthorecti-fied QuickBird image described by Cullen et al (2006) ob-tained on 1 February 2003 provided the basis for the ge-olocation of imagery used in this study as well as being re-tained to map the areal extent of the glaciers A Kompsatimage from 17 June 2011 was obtained at a L1A process-ing level to enable a precise orthorectification Fusion of thefour 4-m resolution multi-spectral bands with the 1-m reso-lution panchromatic band prior to the orthorectification dra-matically improved the potential to interpret spatial detailsin the Kompsat image Ten ground check points (GCPs) and12 independent check points (CPs) were collected from theQuickBird image and elevation data were retrieved from theSRTM DEM to support the orthorectification The latter wasconducted using the sensor model and rational polynomialcoefficients supplied by the image provider in ERDAS Imag-ine 2011 in conjunction with the SRTM DEM and yieldeda root mean square error (RMSE) of 25 m and 28 m for theGCPs and CPs respectively

222 Advanced Spaceborne Thermal Emission andReflection Radiometer (ASTER)

A search through the ASTER archive led to an image from19 August 2004 being identified as suitable to independentlyassess and support the orthorectification of other images A15 m DEM was produced following a stereo triangulationof the ASTER bands 3N and 3B in Leica PhotogrammetrySuite Twenty-one tie points and seven GCPs collected fromthe orthorectified QuickBird image as well as elevation datafrom the SRTM DEM were used and yielded a RMSE of81 44 and 93 m in the easting northing and elevation di-rections respectively A hill-shaded DEM was then createdand compared to that derived from the SRTM DEM This re-vealed no noticeable planimetric bias and demonstrated thatthe rectification of the ASTER image is consistent with thegeolocation accuracy of SRTM data Importantly the corre-sponding orthorectified ASTER image provided an absolutereference (within aboutplusmn10 m) to assess the rectification ofall other images Furthermore the almost snow-free ASTERimage was extremely useful in our efforts to establish theice boundaries in the QuickBird and Kompsat images de-spite its relatively coarse pixel size (ie 15 m) particularlyin the Breach Wall region due to the advantageous illumina-tion and spectral resolution of the sensor The ASTER imagewas not used as an epoch in the new time series presented inthis study because of its temporal proximity to the QuickBirdimage

223 Landsat images

The full archive of Landsat satellite imagery from the Kili-manjaro region available at the Earth Resources Observationand Science (EROS) Center was examined which led to fourepochs being identified as suitable for mapping the remainingice cover (Table 1) A Landsat 2 Multispectral Scanner Sys-tem (MSS) image from 15 August 1975 was preferred to thatof 24 January 1976 used by Hastenrath and Greischar (1997)as it exhibited much less transient snow Two Landsat 5 The-matic Mapper (TM) images acquired only a month apart on24 June and 26 July 1984 were used as they jointly provideda virtually snow free acquisition (the first date was retainedas the epoch of reference) with improved spatial resolution(ie 30 m) The Landsat 4 TM image of November 1989 citedby Hastenrath and Greischar (1997) (no date provided) couldnot be found in the archive and was therefore not used In-stead an image from 31 December 1992 was utilised Fi-nally a Landsat 7 ETM+ image from 21 February 2000 wasobtained to provide an independent assessment of the map-ping epoch characterized by Thompson et al (2002) using anaerial survey An image fusion technique was also completedto take advantage of the enhanced spatial resolution of thepanchromatic band of the ETM+ sensor (ie 15 m)

The selected Landsat images were orthorectified with thesensor model in ERDAS Imagine using GCPs identified in

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

422 N J Cullen et al A century of ice retreat on Kilimanjaro

Table 1Details of the terrestrial photogrammetry and satellite products used to characterize the retreat of the glaciers on Kibo The spatialresolution is given as either a map scale or pixel size The number of ground check points (GCPs) used for image matching and the root meansquare error (RMSE) (m) associated with the orthorectification (m) are also provided Two further Landsat images and one ASTER imagewere also processed but are not listed as they were not formally used for the time series (see Sect 22)

Map year Source Description of the product Pixel size GCP(m) or RMSEmap scale

MayndashOct 1912 Klute (1920 1921) Terrestrial photogrammetry with 74 1 50 000 ndashstand-lines over 2000 measured pointsand additional field sketches

Feb 1962 Downie and Terrestrial photogrammetry of western 1 50 000 ndashWilkinson (1972) glaciers ground observations and aerial

photography15 Aug 1975 Landsat 2 (MSS) Satellite imagery 57 1059724 Jun 1984 Landsat 5 (TM) Satellite imagery 30 1010531 Dec 1992 Landsat 4 (TM) Satellite imagery 30 118121 Feb 2000 Landsat 7 (TM) Satellite imagery 15 11681 Feb 2003 QuickBirda Satellite imagery 06 ndash26b

17 Jun 2011 Kompsata Satellite imagery 1 1025

a Pan-sharpened multi-spectral imageryb Ortho 1 5000 product level

the ASTER image which had already had its absolute ac-curacy assessed MSS and TMETM+ images were resam-pled to 20 and 10 m respectively using a bicubic interpola-tion to facilitate the visual interpretation of the identified iceboundaries The number of GCPs used and the correspondingRMSE for each image are provided in Table 1 Finally theco-registration between the orthorectified images was scru-tinized and revealed extremely good agreement Minor dis-tortions were identified in the QuickBird image in the steepsouth-east region of Kibo which required a local geomet-ric correction using an empirical rational function model andGCPs collected from the orthorectified Kompsat image Wesuspect the source of these distortions may have been fromthe use of incorrect elevation data in voids from the SRTMDEM used by the image providers during the initial orthorec-tification process It is also likely that unfavourable acquisi-tion geometry in conjunction with the steep terrain in this re-gion may have adversely affected the rectification Followingthe correction the ice boundaries for 2003 defined by Cullenet al (2006) were revised which resulted in a minor changein the total areal extent

23 Mapping ice boundaries

231 Previous mapping efforts

The maps created by Hastenrath and Greischar (1997 Fig 4)and Thompson et al (2009 Fig S2) (referred to in the fol-lowing as HG97F4 and T09FS2 respectively) were digitizedto help facilitate a comparison to the ice boundaries iden-tified in this study The single geographical coordinate in-dicated on HG97F4 (ie 305prime S 3720prime E according to a

Clark 1880 ellipsoid) made it difficult to perform a rigor-ous registration and raises the question as to whether glacierboundaries in HG97F4 are suitable for reproduction Insteadlinks were created to match contour lines with those derivedfrom the SRTM data using a linear transformation to avoidunwanted distortions The same process of matching con-tours was used to register T09FS2 which revealed a per-fect match with HG97F4 This enabled us to confirm thatthe glacier outlines prior to 2000 in T09FS2 are a repro-duction and edited version of HG97F4 which is also ac-knowledged by Thompson et al (2002 2009) Once bothmaps were registered in a consistent cartographic system theglacier boundaries at each epoch (ie 1912 1953 1976 and1989 for HG97F4 and T09FS2 2000 and 2007 for T09FS2)were carefully digitized This process revealed numerous in-consistencies between the glacier boundaries of HG97F4 andT09FS2 Hastenrath (2006 p 2) also alluded to a ldquomappingerrorrdquo related to the 2000 ice boundaries created by Thomp-son et al (2002) To ensure the latter fitted within his own1989 epoch Hastenrath (2006) was required to make adjust-ments to the ice boundaries of the northern ice field Cullen etal (2006 Fig 1) were also required to use an ldquoapproximateglacier extent in 1912rdquo as they found it difficult to place someice remnants observed in 2003 within previously establishedoutlines (eg Hastenrath and Greischar 1997 Thompson etal 2002) which was scrutinized by Thompson et al (2009Fig S4) These problems prompted us to make the decisionto revisit the entire sequence of historical maps and satel-lite imagery While it is not our intention to focus on all ofthe mapping inconsistencies we have observed we do de-scribe some key differences in detail (see Sect 31) in order

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 423

to characterize the physical processes controlling glacier re-treat on Kibo

232 Establishing new ice boundaries

The boundaries of ice bodies were visually interpreted anddigitized from the orthorectified satellite images and histor-ical maps Despite the variability in the spatial and spectralresolution of the data products as well as having to accountfor varying amounts of transient snow in the satellite im-ages the use of the high temporal resolution sequence sig-nificantly improved our ability to interpret the location andextent of ice bodies consistently through time To supportand facilitate the interpretation of all data products (Table 1)three-dimensional visualization techniques were extensivelyused in conjunction with aerial and ground-based photogra-phy from others (eg Hastenrath 2008) and our own fieldexpeditions between 2005ndash2012 The vantage points fromwhich historical photographs were taken were reproducedin the 3-D model of Kilimanjaro which proved particularlyuseful when revisiting the glacier outlines mapped by Klute(1920) The use of 3-D visualization also helped us to inter-pret glacier positions in relation to distinctive geomorpho-logical features (eg moraines rock outcrops) which are de-scribed in more detail in the following section Finally tohelp facilitate future efforts all data products from this anal-ysis have been made available on the Global Land Ice Mea-surements from Space (GLIMS) Glacier Database

3 Results and discussion

31 The mapping reloaded

The total amount of ice cover on Kilimanjaro on 17 June2011 was 176 km2 (Table 2) with the location of these icemasses shown in Fig 1 The remaining ice masses are rem-nants of more extensive glaciers which were first mappedand given names by Meyer (1900) with the nomenclaturestill in use despite the character of the ice bodies havingchanged dramatically over the last century The majority ofthe ice cover observed on Kilimanjaro today fails to strictlyadhere to any common definition of a glacier with most enti-ties isolated and stagnant but following the definition sug-gested by Racoviteanu et al (2009) we include any bodyof ice that is observed after transient snow melts It is clearfrom the accounts and photographs of the early explorers thatonly the principal glaciers were named and that a consider-able number of small stagnant ice masses in particular in thecrater area were not formally recognized This is no more ev-ident than in the south-east crater area where the first explor-ers accessed the summit region through a number of weak-nesses in the ice referred to as notches (eg Hans Meyer andJohannes notches) which have been described in detail else-where (eg Geilinger 1936 Hastenrath 1984) Despite thisapparent difficulty it is clear that the remaining ice bodies

today are remnants of the more extensive glaciers mappedby Klute (1920) using terrestrial photogrammetry (Table 1)with the boundaries established from this effort reflecting theice cover in 1912 (Fig 1) Our calculation of the extent ofthe glaciers on Kilimanjaro at the end of October in 1912 is1140 km2 (Table 2) which is 85 more than what is ob-served on the mountain today

The differences between our interpretation of the 1912map produced by Klute (1920) and that produced by Has-tenrath and Greischar (1997) are also shown in Fig 1 Asnoted in Sect 232 our interpretation of the glacier bound-aries was supported by geomorphological evidence observedusing 3-D visualization and historical photographs Impor-tantly the 3-D model allowed us to carefully identify andlocate the most recent moraines Using these geomorpholog-ical features as markers we were able to determine with ahigh degree of accuracy the former location of the glaciersThis approach was particularly useful in establishing the pastlocation and extent of the outlet glaciers feeding from thenorthern ice field (the Credner Drygalski and (Great andLittle) Penck glaciers) the Uhlig glacier and the two distinctglacier streams located in the western breach area (the Lit-tle and Great Barranco glaciers) as well as the four primarysouthern ice field glaciers (the Heim Kersten Decken andRebmann) The map produced by Jaeger (1909 map dated1906) was also valuable in allowing us to identify a num-ber of small ice bodies located above both the Little andGreat Barranco glaciers (Fig 1) Our interpretation of thealtitudinal extent of the southern ice field glaciers at about4600 m (Fig 1) is supported by the photographs of Uhlig(1904 Fig 50 p 641) and Klute (1920 Fig 1) It is impor-tant to note that both these historical photographs reveal acomplexity of the outcropping rock at the lower elevations ofthe southern ice field glaciers that we have made an attemptto account for (Fig 1)

Despite a well-documented and photographic account ofthe historical retreat of Kilimanjarorsquos glaciers which hasbeen carefully reviewed by Hastenrath (1984 2008) therehave been a number of inconsistences in the interpretationof ice bodies in recent mapping efforts (eg Hastenrathand Greischar 1997 Cullen et al 2006 Hastenrath 2006Thompson et al 2002 2009) which require further clarifi-cation as they have an impact on our ability to interpret thephysical processes controlling the demise of the glaciers onKibo To do this one must briefly reflect on the historical ob-servations starting with those of Hans Meyer who provideda sketch (Meyer 1890 p 345) and a map (Meyer 1891) thatindicates that an ice stream which is now referred to as theFurtwangler glacier flowed over the crater rim into the west-ern breach area Meyer (1891 p 319) notes ldquothe ice from thevast caldera falls as a mighty cascade into the great westernfissure where it unites with the accumulations in the fissureitselfrdquo Though Hastenrath (1984 p 78) disputes whetherthis ice stream (the Furtwangler glacier) was connected tothe Little andor Great Barranco glaciers other historical

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

424 N J Cullen et al A century of ice retreat on Kilimanjaro

4800

4600

5000

4400

5200

5400

5600

5700

4200

4000

5800

4800

4400

4600 4200

37deg23E

37deg23E

37deg22E

37deg22E

37deg21E

37deg21E

37deg20E

37deg20E

3deg3

S

3deg3

S

3deg4

S

3deg4

S

3deg5

S

3deg5

S

3deg6

S

3deg6

S

314000

314000

316000

316000

318000

318000

320000

320000

9658000

9658000

9660000

9660000

9662000

9662000

Uhuru Peak

Northern ice field

Heim

Uhlig

Dec

ken

Ratzel

Kersten

Reb

man

n

Credner

Drygalski

Arrow

Diamond

Balletto

Furtwaumlngler

LittlePenck

5895m

IceCastle

Easternice field

Southernice field

WGS84UTM 37S

0 1000 2000

Meters

Revised 1912 boundaries

Credner Glacier name

2011 Ice boundaries

1912 boundaries fromHastenrath and Greischar (1997)

Little Barranco

Fig 1 Glacier extent on Kilimanjaro in 1912 and 2011 Glacier names with an asterisk do not appear in Klute (1920) and are described inthe text The interpretation of the glacier extent for 1912 as defined by Hastenrath and Greischar (1997) is shown as a dashed line (blue) forcomparative purposes The nomenclature for each glacier used by Klute (1920) is given while the names of four glaciers introduced afterthis time are indicated usinglowast

accounts are more supportive (Gillman 1923 Geilinger1936) A more detailed map of Kilimanjarorsquos glaciers pro-duced by Meyer (1900) based on his observations during afield expedition in 1898 shows the connection no longer ex-isted suggesting a break-up occurred sometime in the early1890s Though the timing is uncertain the implications oflosing this connection are not overlooked by Geilinger (1936p 12) who refers to the Little and Great Barranco glaciers asldquoindependent dead glaciersrdquo and made the astute observationthat both ice entities were retreating at their upper and lowermargins

This observation holds some importance as it helps clar-ify the source of the ice remnants still observed today Themaps by Meyer (1900 map dated 1898) Jaeger (1909map dated 1906) and Klute (1920 map dated 1912) clearlyshow distinct ice streams for the Little and Great Barrancoglaciers which as noted may have been fed by ice from theFurtwangler glacier as recently as the late 19th century Ourrevision of the Klute (1920) map from 1912 (Fig 1) honoursthe discrete boundary between the Little and Great Barranco

glaciers which is not as clearly depicted in the Hastenrathand Greischar (1997) interpretation The failure to clearlydistinguish the boundaries and location of these two indepen-dent ice streams has been inherited by Thompson et al (20022009) who used the Hastenrath and Greischar (1997) inter-pretation of the ice extent prior to 2000 (see Sect 231)Thompson et al (2009) argue that ice remnants identified byCullen et al (2006) in both the ice streams of the former andmore extensive Little and Great Barranco glaciers are absentfrom previous observations and therefore most likely tran-sient snow The sketch by Humphries (1972 Fig 46) andthe map produced by Messerli (1980 Fig 6) also support theinterpretation by Cullen et al (2006) while the ice bodiesin question have been identified in each of the satellite im-ages used in the present analysis (Fig 2) Though not labelledby Humphries (1959 1972) or by those preceding his obser-vations the Arrow glacier has been used to refer to one ofthe Little Barranco ice remnants (Fig 1) It appears that thisterm was adopted by climbers on the mountain (eg Samp-son 1971 p 202) but the exact location of this ice body is

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 425

Table 2Areal extents (km2) of the total ice bodies between 1912 and 2011 as well as for zone 1 (western region) zone 2 (plateau region)and zone 3 (southern region) as defined in the text and Fig 4 The annual rates of area change per observation period (10minus2 km2 yrminus1) areshown in parentheses for the total and zoned areas

Epoch Duration Area [km2] (rate of area change [10minus2 km2 yrminus1])

[years] Total Zone 1 Zone 2 Zone 3

191283 11400 4540 4293 2567196216 4936 7320 (minus827) 2520 (minus409) 3161 (minus229) 1639 (minus188)197562 1347 6052 (minus941) 2060 (minus342) 2529 (minus469) 1463 (minus131)198448 887 4815 (minus1395) 1628 (minus487) 1949 (minus654) 1238 (minus254)199300 853 3800 (minus1190) 1142 (minus570) 1638 (minus365) 1020 (minus256)200014 715 2920 (minus1232) 0701 (minus617) 1411 (minus318) 0808 (minus297)200308 295 2500 (minus1425) 0539 (minus550) 1251 (minus543) 0710 (minus332)201146 838 1762 (minus881) 0333 (minus246) 0937 (minus375) 0492 (minus260)

not well documented and we are uncertain that the remnantidentified in Fig 1 and by Cullen et al (2006) is the sameas that referred to by others (eg Sampson 1971 Messerli1980 Hastenrath and Greischar 1997)

There has also been some confusion about the Diamondand Balletto glaciers (Fig 1) which were not formally rec-ognized in the early mapping work (Meyer 1990 Jaeger1909 Klute 1920) or acknowledged by Humphries (19591972) but are both described by Hastenrath (1984) whileSampson (1971) and Messerli (1980) only refer to the for-mer The Breach Wall area where both glaciers are locatedwas mapped by Klute (1920) but as no break up from thesouthern ice field had occurred at this time they were notgiven discrete labels (Fig 1) However what is indisputableis the presence of an ice field (the Balletto glacier) beneaththe headwall of the Diamond glacier which was not mappedby Thompson et al (2002 2009) The presence of this icefield is not only observed in the satellite record but is docu-mented by Reinhold Messner (Messner 1991 Chapter 21)who made the first (and only) direct ascent of the Breach Wallof Kilimanjaro in 1978 Though Cullen et al (2006) mappedthis region of ice (the Balletto glacier) they failed to label itcorrectly placing the label beneath the Heim glacier after fol-lowing the interpretation of Hastenrath and Greischar (1997Fig 2) Further the Diamond glacier is absent after 1976 inthe maps produced by Hastenrath and Greischar (1997) andHastenrath (2006) but appears in the Thompson et al (20022009) maps despite using the former to build their time se-ries Finally the Uhlig glacier which Jaeger (1909 p 130)named after his friend C Uhlig does not appear to have beenconnected to the northern ice field in the past (Fig 1) Whilethe upper boundary of the Uhlig glacier is difficult to inter-pret in the Klute (1920) map it is clearly shown by Jaeger(1909) that it was not joined to the northern ice field

Figure 2 shows the change in ice extent over almost a 100-yr period which is consistent through time and removes mostof the uncertainty about the location and extent of the re-maining ice bodies on Kibo The retreat over the first part of

the 20th century was approximately 41 km2 (Fig 2a and b)while over the last 50 yr 56 km2 of the ice cover has beenlost (Fig 2cndashh) The retreat between 1912 and 1962 (Fig 2aand b) resulted in a distinct break-up of ice in the vicinity ofthe notches (south-east crater region) separating the northernand eastern ice fields from the southern ice field Remnantsof the Ice Castle (Fig 1) still remained in 1962 and were de-scribed by Sampson (1971) as still being present in the early1970s but appeared to have been lost by 1975 (Fig 2c) Othernoticeable features of the retreat in the early 20th century in-clude the development of a distinct separation between theGreat and Little Penck glaciers which formed through theexpansion of the Ravenstein (first mapped by Meyer 1900)the rapid demise of the Uhlig glacier and the split-up of theDiamond and Balletto glaciers from the southern ice field(Fig 2a and b)

Between 1962 and 1975 the persistent retreat of ice led tothe northern and eastern ice fields becoming separated fromeach other as did the Great Penck glacier from the north-ern ice field (Fig 2b and c) The Little Penck proved moreresistant and did not separate from the northern ice field un-til after 1992 while the remnant of the Great Penck com-pletely disappeared during this period (Fig 2e and f) thoughground-based (field) observations in January 2005 revealeddebris-covered ice at the location of the lowest 1984 extent(Fig 2d) As noted by others (eg Hastenrath and Greischar1997 Cullen et al 2006) the northern ice field developed ahole sometime after 1984 (Fig 2d and e) which opened intoa large canyon after 2003 (Fig 2g and h) Field observationsby the first author in October 2012 revealed that the north-ern ice field has separated into two independent ice bodieswhich must have occurred sometime after 17 June 2011 TheFurtwangler glacier has also separated into smaller ice enti-ties which began after 2003 (Fig 2g and h) The southernice field has not escaped the on-going demise with the upperpart of the Kersten glacier separating from its lower reachesbetween 2003 and 2011 (Fig 2g and h) which was some-what anticipated after mass balance modelling showed this

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

426 N J Cullen et al A century of ice retreat on Kilimanjaro

(g) 2003 (h) 2011

(a) 1912 (b) 1962

(c) 1975 (d) 1984

(e) 1992 (f) 2000

Credner

Great Penck

Ratzel

Ker

sten

Northern ice field

Easternice field

Southernice field

GreatBarranco

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Fig 2 Ice retreat over Kilimanjaro based on the revised areal extent for the following years(a) 1912(b) 1962(c) 1975(d) 1984(e)1992(f) 2000(g) 2003 and(h) 2011 An interactive 3-D illustration of the sequence can be sourced in the Supplement

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 427

was likely to occur (Molg et al 2009b) The upper part ofthe Heim has disappeared and the Decken has narrowed con-siderably over the last 8 yr (Fig 2g and h)

32 New interpretation of the ice retreat

A consistent record of the former extent of the glaciers onKibo provides a platform from which the relationship be-tween glaciological and meteorological processes can be fur-ther explored We believe the spatial and temporal variabilityof the ice cover on Kibo can be used to unravel the physi-cal processes responsible for controlling the decay of the icemasses through time as well as providing insights into themechanisms that may have been responsible for growth in thepast Interestingly those who were privileged to make someof the first (formal) observations on Kilimanjaro recognizedthere was an asymmetry in the ice cover (eg Meyer 1891)which pointed to specific climatological drivers controllingglacial retreat (eg Gillman 1923 Nilsson 1931 Geilinger1936) Meyer (1891 p 308ndash314) noted that ldquoan immensemountain mass like Kilimanjaro must considerably modifythe prevailing wind directionrdquo before making the point thatthe primary cause in the variability of the ice limit is ldquotheunequal distribution of moisture on the northern and south-ern sides of the mountainrdquo Though Meyer (1891) recognizedvariability in moisture was the most important control healso speculated that the asymmetry of low reaching glacierson Kibo in the south south-west and western sectors of themountain a pattern that persisted throughout the 20th cen-tury (Fig 2) depended on ldquolocal peculiarities which favourthe preservation of the snow [rather] than on physical condi-tions that favour a large snowfallrdquo (Meyer 1891 p 315) Torevisit this early observation it is useful to reflect on the hyp-sometry of the glaciers in 1912 (Fig 3a) which shows thatthe probability density in relation to the distribution of the icewas greatest above 5700 m This high density reflects the icecover in the crater region with the majority of the ice in 1912being part of the former ice cap that linked the northern andeastern ice fields (Fig 1) Figure 3b and c show that slopeand aspect were also important controls on the density of icecover with the steep south and south-west facing slopes morefavourable which enabled the glaciers to extend to elevationsbelow 4600 m For the glaciers to have extended to the sameelevation on the east-facing slopes they would have had tocover a much greater horizontal distance (Fig 1)

To fully account for the distinct geometry of the ice masseson Kibo and to establish the physical processes controllingtheir location and extent through time Kaser et al (2004)established a framework that included four separate glacierregimes The motivation to do this was partly based on theearly recognition that the tabular shaped glaciers in the craterregion of Kibo behave differently to glaciers on the flanks ofthe mountain (eg Nilsson 1931 Geilinger 1936) Guidedby this distinction Cullen et al (2006) used the 5700 m con-tour which is roughly the height of the outer crater rim to

Density times 10 minus3

2

4

6

8

WEST EAST

SOUTH

NORTH

4600 4800 5000 5200 5400 5600 58000

1

2

3

4

Elevation [m]

Den

sity

times 10

minus3

0 10 20 30 40 50 60 70 800

1

2

3

4

Slope [deg]

Den

sity

times 10

minus2

(a)

(c)

(b)

Fig 3 Distribution of topographic measures from the areal extentof all ice bodies in 1912(a) elevation(b) slope and(c) aspect Themagnitude of the normalised distribution on the y-axis is a proba-bility density function with the integral equal to unity

distinguish between two broad glacial systems (1) plateau(ge 5700 m) and slope (lt 5700 m) glaciers This simple bi-nary classification has proved useful in describing some ofthe key physical processes responsible for the retreat of iceon Kibo (eg Molg et al 2009b Kaser et al 2010) Tofurther characterize the spatial and temporal variability ofglacier retreat on Kibo we have adopted a geo-statisticalapproach in this study that uses elevation slope and aspectof the ice cover in 1912 (the feature space) to identify clus-ters resembling topographical classes This was achieved byperforming a cluster analysis containing maximum likeli-hood estimates of the topographical parameters in a Gaussian

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

428 N J Cullen et al A century of ice retreat on Kilimanjaro

mixture model (McLachlan and Peel 2000) which resultedin three broad glacial zones being identified (1) western re-gion (zone 1) (2) plateau region (zone 2) and (3) southernregion (zone 3) (Fig 4)

Figure 5 shows the rate of retreat of each of the glacialzones through time as well as a least-squares linear regres-sion fit (and 80 confidence intervals) through each of theglacial zone time series As described by Cullen et al (2006)the retreat of the tabular shaped plateau glaciers (zone 2) iscontrolled primarily by solar radiation-induced ablation onthe vertical cliffs that characterize them as demonstratedthrough radiation modelling (Molg et al 2003) Winkleret al (2010) revealed that direct insolation is the reasonfor the predominant zonal alignment of the ice cliffs andtheir bimodal annual recession (sunlit versus shaded peri-ods) Though the shrinkage of the ice cover along the verticalcliffs is fairly constant from year to year the areal retreat ofthe zone 2 region is not entirely linear with the highest ratesoccurring between 1975 and 1984 (Table 2) which corre-spond to the period following the separation of the northernand eastern ice fields (Fig 2) Therefore changes in the ge-ometry of the ice bodies and their exposure to solar radiationappear to have led to variations in retreat rates (Table 2 andFig 5) which has also been described by Kaser et al (2010)

An explanation for the growth and decay of the plateauglaciers including a description of the physical processesthat may have led to the observed steps in the northern icefield has been provided by Kaser et al (2010) To sustaingrowth in the crater rim region (zone 2) it is critical thatsnowfall is sufficient to allow a continuous snow cover todevelop that is able to survive from one year to the next Ifthe annual build-up of snow is sustained through a period oftime it is quite conceivable that the plateau glaciers couldgrow to several tens of metres in thickness quite quicklyraising the question as to whether this accumulation of icehas the potential to become a source of mass for glacierson the western flanks of the mountain (zone 1) Meyer(1891 p 316ndash317) was somewhat sceptical as to whetherthe plateau region acted as a ldquoreservoirrdquo and thought it ldquomuchmore reasonable to suppose that the rocky rim of the crateris the real reservoirrdquo However Meyer (1890 1891) did pro-vide an account and sketches to support a connection of theplateau ice to the Little and Great Barranco glaciers in thepast which as described may have been important in sustain-ing them (Gillman 1923 Geilinger 1936) There is someimportance in investigating the extent of the linkages be-tween these two glacial zones as together they accounted foralmost 80 of the total ice cover in 1912 (Table 2)

The remaining ice bodies in zone 2 (plateau region) aremostly stagnant today with field measurements and prelim-inary modelling results indicating deformation rates of icewithin the crater are small (lt 10 cm yrminus1) (Kaser et al 2004Winkler et al 2010) It seems unlikely that the present or his-torical thickness of the plateau ice has been sufficient to al-low substantial deformation (Kaser et al 2004) suggesting

3140 3150 3160 3170 3180 3190 3200 32109657

9658

9659

9660

9661

9662

9663

9664

4000

4000

4500

4500 4500

45004500

5000

5000

5000

5000

5000

5500

5500

5500

UTM easting in meters times105

UT

M n

orth

ing

in m

eter

s times

106

Fig 4 Classification of regions based on the distribution of topo-graphic measures in 1912 Zone 1 (western region) zone 2 (plateauregion) and zone 3 (southern region) are shown in red orange andyellow respectively

that mass transport from the nearly horizontal plateau downinto the western slope glaciers (zone 1) has most likely beenminor Careful observations by Klute (1920) demonstratedthat the velocity of the centre of the Great Penck glacier was064 m over a 30-day period in 1912 which led him to spec-ulate that the annual downslope movement was about 7 mper year It should be noted that Nilsson (1931) and Osmas-ton (1989) also report on these observations but incorrectlycite Jaeger (1909) as the person responsible for obtaining themeasurements Nonetheless it is improbable that ice fromthe plateau region could have supported the observed flowdynamics of the Great Penck glacier with the upper reachesof the mountain flanks the most likely source of the massThus the growth and decay of the glaciers in zone 1 (west-ern region) do not appear to be directly dependent on thestability of plateau ice (zone 2) To further resolve the phys-ical processes controlling the stability of the plateau ice itwould be of interest to develop a better understanding of themechanisms responsible for vertical cliff development

The glaciers on the southern slopes (zone 3) differ fromthose on the western slopes (zone 1) as they are separatedfrom the large masses of ice within the caldera (zone 2) bybed topography (Humphries 1972) This distinction is im-portant as any assessment of the atmospheric controls onthe behaviour of the southern ice fields is independent ofthe complexity of the vertical cliff retreat in the plateau re-gion (zone 2) That said the cluster analysis that led to thetopographical classes used in this study shows that the up-per sections of the southern ice field exhibit similar phys-ical characteristics to the ice within the caldera (Fig 4)Nevertheless to obtain a full solution for the asymmetry ofthe glaciers on Kibo that the early explorers observed itis critical to establish an understanding of the atmospheric

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 429

1900 1920 1940 1960 1980 2000 2020 2040 20600

05

1

15

2

25

3

35

4

45

5

Year

Are

al e

xten

t [km

2 ]

zone 1 R2=099

zone 2 R2=097

zone 3 R2=09980 confidence interval

4000

4000

4000

4500

4500

4500

45005000

5000

5000

50005000

5500

5500

5500

Fig 5 The areal change of glaciers and ice bodies based on thethree-member classification Linear regressions (thin solid lines) foreach of the three zones are shown as well as the 80 confidenceintervals for each of these relationships

processes controlling the growth andor decay of the southernice fields This has been achieved by Molg et al (2009b)who used observations and atmospherendashglacier modellingto show how critical precipitation humidity and fractionalcloud cover are on the energy and mass balance of the Ker-sten glacier (zone 3) These results compare favourably toearly accounts of the importance of precipitation (moisturechange) and cloud cover on the asymmetry (Meyer 1891Klute 1920 Gillman 1923 Geilinger 1936) with the diur-nal variability of cloud cover being especially important inproviding protection to the glaciers on the western side ofthe mountain (eg Meyer 1891 Hastenrath and Greischar1997)

Recent regional atmospheric modelling experiments haveprovided a more quantitative assessment to account for theasymmetry in glacier mass on Kibo demonstrating that thewest and south-western areas of the mountain are favourableto moisture convergence as controlled by atmospheric dy-namics (Molg et al 2009a) As noted by Osmaston (1989)the asymmetry is not confined to the most recent glacier ex-tents with the pattern of ice distribution established fromformer moraines indicating that recent climatic controls onglacier behaviour are similar to those in the past This sug-gests that establishing the specific nature of the atmosphericcontrols responsible for the most recent shift from glaciergrowth to decay in particular in zones 2 and 3 (plateau andsouthern slope regions) remains critical if we are to useglacier behaviour to reconstruct climatic conditions of thedeeper past

Finally it is difficult not to speculate on when the ice onKibo will disappear which becomes more achievable if onerecognizes the importance of the different glacial zones onthe mountain If one accepts linear extrapolation based on

changes from the past is suitable to make projections for thefuture it is apparent that the ice cover in zone 1 (western re-gion) will most likely be gone before 2020 and that ice inzone 2 (plateau) and zone 3 (southern slopes) will disappearsometime close to 2040 (Fig 5) It is likely that the plateauwill harbour the final remnants of ice on Kibo which couldremain slightly beyond the middle of the 21st century Inter-estingly our linear extrapolation for the loss of ice is in closeagreement to earlier projections using radiation modelling(Molg et al 2003) Though the loss of ice will have no hy-drological significance to lowland areas (eg Gillman 1923Kaser et al 2004 Molg et al 2008b) there is no doubt thattheir disappearance will be a ldquogrievous aesthetic lossrdquo (Salt1951 p 160) which will undoubtedly impact the experiencetourists have on the mountain (Molg et al 2008b)

4 Conclusions

With reference to the glaciers found on Mt Kenya and Kiboin East Africa C Gillman remarked ldquowe might perhapslearn a good deal as to the general world circulation fromcareful and prolonged observation on these particular sum-mitsrdquo (Mackinder 1923 p 23) This statement still holdstrue today (eg Molg et al 2006) and with the increasingawareness that glaciers on Africarsquos highest mountain can beused to extract climate information from the past it seemedtimely to revisit the mapping of the ice cover on Kibo throughtime Our motivation to do so has been to ensure the histor-ical record of these changes is of the highest possible qual-ity and in a format that should help facilitate on-going andfuture research dedicated to resolving the complexity of cli-mate variability and change in East Africa and beyond Thiswas achieved by using a consistent cartographic frameworkthat allows historical maps and new satellite imagery to beexamined The use of three-dimensional visualization tech-niques greatly enhanced our ability to establish the formerlocation of the glaciers

The revisited time series of glacial retreat on Kiliman-jaro contains eight epochs and spans almost 100 yr (1912ndash2011) The glaciers have retreated from their former extentof 1140 km2 in 1912 to 176 km2 in 2011 which representsa loss of about 85 of the ice cover over almost a 100-yr pe-riod Our estimate of the total retreat is consistent with recentassessments (Cullen et al 2006 Thompson et al 2009) butthe careful re-examination of the location and extent of indi-vidual ice bodies through time has allowed us to identify anumber of important differences from previously publishedmaps (eg Hastenrath and Greischar 1997 Thompson et al2002 2009) Historical evidence has been used to supportour interpretation with the detailed observations and pho-tographs obtained by early explorers on Kilimanjaro found tobe extremely valuable (eg Meyer 1891 1900 Uhlig 1904Jaeger 1909 Klute 1914 1920 1921 Gillman 1923 Nils-son 1931 Geilinger 1936) It has to be acknowledged that

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

430 N J Cullen et al A century of ice retreat on Kilimanjaro

much of what we know about Kilimanjaro today was estab-lished through their efforts

To further explore the importance of different glacierregimes on Kibo (Kaser et al 2004) a geo-statistical ap-proach was applied which resulted in the segmentation ofthe ice cover into three broad glacial zones A careful ex-amination of the retreat of ice in each of these glacier zonesindicates that there is a high probability that most of the re-maining ice cover on Kilimanjaro will disappear by 2040and it is highly unlikely that any ice body will remain af-ter 2060 if present retreat rates are maintained Before thishappens it would still be useful if the remaining ice bodiesin particular the remnants of the northern ice field could befurther examined to help resolve existing uncertainties aboutthe mechanisms responsible for controlling the formation ofthe vertical cliffs that characterize them (eg Winkler et al2010) In addition further efforts are required to scrutinizethe exact timing of the onset of the present glacier retreat firstreported by Hans Meyer in the late 19th century If we canrefine our understanding of the atmospheric controls behindthe most recent shift from glacier growth to decay on Kibowe will be in a stronger position to characterize the climaticcontrols of past and much larger glaciations on Africarsquos high-est mountain

Supplementary material related to this article isavailable online athttpwwwthe-cryospherenet74192013tc-7-419-2013-supplementpdf

AcknowledgementsThis research is funded by the Departmentof Geography and School of Surveying University of OtagoNew Zealand Thomas Molg was supported by the Alexander vonHumboldt Foundation The contributions made by Georg Kaserand Michael Winkler were supported by the Austrian ScienceFund (FWF) Projects P 20089-N10 and P 21288-N10 Thisresearch would not have been possible without the support from thefollowing Tanzanian authorities (COSTECH KINAPA TANAPATAWIRI) We are grateful to the referees that took the time toprovide feedback on this research in particular the careful andthoughtful comments by Henry Brecher

Edited by M Schneebeli

References

Chiang J C H The tropics in paleoclimate Annu Rev EarthPlanet Sci 37 263ndash297 2009

Cullen N J Molg T Kaser G Hussein K Steffen Kand Hardy D R Kilimanjaro Glaciers Recent areal ex-tent from satellite data and new interpretation of observed20th century retreat rates Geophys Res Lett 33 L16502doi1010292006GL027084 2006

Cullen N J Molg T Kaser G Steffen K and Hardy D REnergy-balance model validation on the top of Kilimanjaro Tan-zania using eddy covariance data Ann Glaciol 46 227ndash233doi103189172756407782871224 2007

Downie C Glaciations of Mount Kilimanjaro Northeast Tan-ganyika Geol Soc Am Bull 75 1ndash16doi1011300016-7606(1964)75[1GOMKNT]20CO2 1964

Downie C and Wilkinson P The Geology of KilimanjaroSheffield Geology Department University of Sheffield 1972

Downie C Humphries D W Wilcockson W H and WilkinsonP Geology of Kilimanjaro Nature 178 828ndash830 1956

Farr T G Rosen P A Caro E Crippen R Duren R Hens-ley S Kobrick M Paller M Rodriguez E Roth L Seal DShaffer S Shimada J Umland J Werner M Oskin M Bur-bank D and Alsdorf D The Shuttle Radar Topography Mis-sion Rev Geophys 45 RG2004doi1010292005RG0001832007

Geilinger W The retreat of the Kilimanjaro glaciers TanganyikaNotes and Records 2 7ndash20 1936

Gillman C An ascent of Kilimanjaro Geogr J 61 1ndash27 1923Hastenrath S The Glaciers of Equatorial East Africa Reidel Dor-

drecht 1984Hastenrath S Diagnosing the decaying glaciers of equatorial

East Africa Meteorol Z 15 265ndash271doi1011270941-294820060106 2006

Hastenrath S Recession of equatorial glaciers a photo documen-tation Sundog Publishing Madison Wisconsin 2008

Hastenrath S and Greischar L Glacier recession on KilimanjaroEast Africa 1912ndash89 J Glaciol 43 455ndash459 1997

Humphries D Interim report on the glaciology and meteorology ofKilimanjaro Tanganyika JulyndashSept 1953 Geological Survey ofTanganyika Dodoma 1953

Humphries D Preliminary notes on the glaciology of KilimanjaroJ Glaciol 3 475ndash479 1959

Humphries D Glaciology and glacial history in The Geology ofKilimanjaro edited by Downie C and Wilkinson P SheffieldGeology Department University of Sheffield 31ndash71 1972

Jarvis A Reuter H I Nelson A and Guevara E Hole-filledSRTM for the globe Version 4 available from the CGIAR-CSISRTM 90 m Database (httpsrtmcsicgiarorg) 2008

Jaeger F Forschungen in den Hochregionen des KilimandscharoMitt Deutsch Schutzgebiet 22 113ndash197 1909

Kaser G Hardy D R Molg T Bradley R S and Hyera TM Modern glacier retreat on Kilimanjaro as evidence of climatechange Observations and facts Int J Climatol 24 329ndash3392004

Kaser G Molg T Cullen N J Hardy D R andWinkler M Is the decline of ice on Kilimanjaro un-precedented in the Holocene Holocene 20 1079ndash1091doi1011770959683610369498 2010

Klute F Forschungen am Kilimandscharo im Jahre 1912 GeogrZeitschrift 20 496ndash505 1914

Klute F Ergebnisse der Forschungen am Kilimandscharo 1912ReimerndashVohsen Berlin 1920

Klute F Die stereophotogrammetrische Aufnahme der Hochregio-nen des Kilimandscharo Zeitschrift Ges Erdkunde Berlin 56144ndash151 1921

Mackinder C W Hobley C Gillman C and Johnston H Anascent of Kilimanjaro Discussion Geogr J 61 1ndash27 1923

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

422 N J Cullen et al A century of ice retreat on Kilimanjaro

Table 1Details of the terrestrial photogrammetry and satellite products used to characterize the retreat of the glaciers on Kibo The spatialresolution is given as either a map scale or pixel size The number of ground check points (GCPs) used for image matching and the root meansquare error (RMSE) (m) associated with the orthorectification (m) are also provided Two further Landsat images and one ASTER imagewere also processed but are not listed as they were not formally used for the time series (see Sect 22)

Map year Source Description of the product Pixel size GCP(m) or RMSEmap scale

MayndashOct 1912 Klute (1920 1921) Terrestrial photogrammetry with 74 1 50 000 ndashstand-lines over 2000 measured pointsand additional field sketches

Feb 1962 Downie and Terrestrial photogrammetry of western 1 50 000 ndashWilkinson (1972) glaciers ground observations and aerial

photography15 Aug 1975 Landsat 2 (MSS) Satellite imagery 57 1059724 Jun 1984 Landsat 5 (TM) Satellite imagery 30 1010531 Dec 1992 Landsat 4 (TM) Satellite imagery 30 118121 Feb 2000 Landsat 7 (TM) Satellite imagery 15 11681 Feb 2003 QuickBirda Satellite imagery 06 ndash26b

17 Jun 2011 Kompsata Satellite imagery 1 1025

a Pan-sharpened multi-spectral imageryb Ortho 1 5000 product level

the ASTER image which had already had its absolute ac-curacy assessed MSS and TMETM+ images were resam-pled to 20 and 10 m respectively using a bicubic interpola-tion to facilitate the visual interpretation of the identified iceboundaries The number of GCPs used and the correspondingRMSE for each image are provided in Table 1 Finally theco-registration between the orthorectified images was scru-tinized and revealed extremely good agreement Minor dis-tortions were identified in the QuickBird image in the steepsouth-east region of Kibo which required a local geomet-ric correction using an empirical rational function model andGCPs collected from the orthorectified Kompsat image Wesuspect the source of these distortions may have been fromthe use of incorrect elevation data in voids from the SRTMDEM used by the image providers during the initial orthorec-tification process It is also likely that unfavourable acquisi-tion geometry in conjunction with the steep terrain in this re-gion may have adversely affected the rectification Followingthe correction the ice boundaries for 2003 defined by Cullenet al (2006) were revised which resulted in a minor changein the total areal extent

23 Mapping ice boundaries

231 Previous mapping efforts

The maps created by Hastenrath and Greischar (1997 Fig 4)and Thompson et al (2009 Fig S2) (referred to in the fol-lowing as HG97F4 and T09FS2 respectively) were digitizedto help facilitate a comparison to the ice boundaries iden-tified in this study The single geographical coordinate in-dicated on HG97F4 (ie 305prime S 3720prime E according to a

Clark 1880 ellipsoid) made it difficult to perform a rigor-ous registration and raises the question as to whether glacierboundaries in HG97F4 are suitable for reproduction Insteadlinks were created to match contour lines with those derivedfrom the SRTM data using a linear transformation to avoidunwanted distortions The same process of matching con-tours was used to register T09FS2 which revealed a per-fect match with HG97F4 This enabled us to confirm thatthe glacier outlines prior to 2000 in T09FS2 are a repro-duction and edited version of HG97F4 which is also ac-knowledged by Thompson et al (2002 2009) Once bothmaps were registered in a consistent cartographic system theglacier boundaries at each epoch (ie 1912 1953 1976 and1989 for HG97F4 and T09FS2 2000 and 2007 for T09FS2)were carefully digitized This process revealed numerous in-consistencies between the glacier boundaries of HG97F4 andT09FS2 Hastenrath (2006 p 2) also alluded to a ldquomappingerrorrdquo related to the 2000 ice boundaries created by Thomp-son et al (2002) To ensure the latter fitted within his own1989 epoch Hastenrath (2006) was required to make adjust-ments to the ice boundaries of the northern ice field Cullen etal (2006 Fig 1) were also required to use an ldquoapproximateglacier extent in 1912rdquo as they found it difficult to place someice remnants observed in 2003 within previously establishedoutlines (eg Hastenrath and Greischar 1997 Thompson etal 2002) which was scrutinized by Thompson et al (2009Fig S4) These problems prompted us to make the decisionto revisit the entire sequence of historical maps and satel-lite imagery While it is not our intention to focus on all ofthe mapping inconsistencies we have observed we do de-scribe some key differences in detail (see Sect 31) in order

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 423

to characterize the physical processes controlling glacier re-treat on Kibo

232 Establishing new ice boundaries

The boundaries of ice bodies were visually interpreted anddigitized from the orthorectified satellite images and histor-ical maps Despite the variability in the spatial and spectralresolution of the data products as well as having to accountfor varying amounts of transient snow in the satellite im-ages the use of the high temporal resolution sequence sig-nificantly improved our ability to interpret the location andextent of ice bodies consistently through time To supportand facilitate the interpretation of all data products (Table 1)three-dimensional visualization techniques were extensivelyused in conjunction with aerial and ground-based photogra-phy from others (eg Hastenrath 2008) and our own fieldexpeditions between 2005ndash2012 The vantage points fromwhich historical photographs were taken were reproducedin the 3-D model of Kilimanjaro which proved particularlyuseful when revisiting the glacier outlines mapped by Klute(1920) The use of 3-D visualization also helped us to inter-pret glacier positions in relation to distinctive geomorpho-logical features (eg moraines rock outcrops) which are de-scribed in more detail in the following section Finally tohelp facilitate future efforts all data products from this anal-ysis have been made available on the Global Land Ice Mea-surements from Space (GLIMS) Glacier Database

3 Results and discussion

31 The mapping reloaded

The total amount of ice cover on Kilimanjaro on 17 June2011 was 176 km2 (Table 2) with the location of these icemasses shown in Fig 1 The remaining ice masses are rem-nants of more extensive glaciers which were first mappedand given names by Meyer (1900) with the nomenclaturestill in use despite the character of the ice bodies havingchanged dramatically over the last century The majority ofthe ice cover observed on Kilimanjaro today fails to strictlyadhere to any common definition of a glacier with most enti-ties isolated and stagnant but following the definition sug-gested by Racoviteanu et al (2009) we include any bodyof ice that is observed after transient snow melts It is clearfrom the accounts and photographs of the early explorers thatonly the principal glaciers were named and that a consider-able number of small stagnant ice masses in particular in thecrater area were not formally recognized This is no more ev-ident than in the south-east crater area where the first explor-ers accessed the summit region through a number of weak-nesses in the ice referred to as notches (eg Hans Meyer andJohannes notches) which have been described in detail else-where (eg Geilinger 1936 Hastenrath 1984) Despite thisapparent difficulty it is clear that the remaining ice bodies

today are remnants of the more extensive glaciers mappedby Klute (1920) using terrestrial photogrammetry (Table 1)with the boundaries established from this effort reflecting theice cover in 1912 (Fig 1) Our calculation of the extent ofthe glaciers on Kilimanjaro at the end of October in 1912 is1140 km2 (Table 2) which is 85 more than what is ob-served on the mountain today

The differences between our interpretation of the 1912map produced by Klute (1920) and that produced by Has-tenrath and Greischar (1997) are also shown in Fig 1 Asnoted in Sect 232 our interpretation of the glacier bound-aries was supported by geomorphological evidence observedusing 3-D visualization and historical photographs Impor-tantly the 3-D model allowed us to carefully identify andlocate the most recent moraines Using these geomorpholog-ical features as markers we were able to determine with ahigh degree of accuracy the former location of the glaciersThis approach was particularly useful in establishing the pastlocation and extent of the outlet glaciers feeding from thenorthern ice field (the Credner Drygalski and (Great andLittle) Penck glaciers) the Uhlig glacier and the two distinctglacier streams located in the western breach area (the Lit-tle and Great Barranco glaciers) as well as the four primarysouthern ice field glaciers (the Heim Kersten Decken andRebmann) The map produced by Jaeger (1909 map dated1906) was also valuable in allowing us to identify a num-ber of small ice bodies located above both the Little andGreat Barranco glaciers (Fig 1) Our interpretation of thealtitudinal extent of the southern ice field glaciers at about4600 m (Fig 1) is supported by the photographs of Uhlig(1904 Fig 50 p 641) and Klute (1920 Fig 1) It is impor-tant to note that both these historical photographs reveal acomplexity of the outcropping rock at the lower elevations ofthe southern ice field glaciers that we have made an attemptto account for (Fig 1)

Despite a well-documented and photographic account ofthe historical retreat of Kilimanjarorsquos glaciers which hasbeen carefully reviewed by Hastenrath (1984 2008) therehave been a number of inconsistences in the interpretationof ice bodies in recent mapping efforts (eg Hastenrathand Greischar 1997 Cullen et al 2006 Hastenrath 2006Thompson et al 2002 2009) which require further clarifi-cation as they have an impact on our ability to interpret thephysical processes controlling the demise of the glaciers onKibo To do this one must briefly reflect on the historical ob-servations starting with those of Hans Meyer who provideda sketch (Meyer 1890 p 345) and a map (Meyer 1891) thatindicates that an ice stream which is now referred to as theFurtwangler glacier flowed over the crater rim into the west-ern breach area Meyer (1891 p 319) notes ldquothe ice from thevast caldera falls as a mighty cascade into the great westernfissure where it unites with the accumulations in the fissureitselfrdquo Though Hastenrath (1984 p 78) disputes whetherthis ice stream (the Furtwangler glacier) was connected tothe Little andor Great Barranco glaciers other historical

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

424 N J Cullen et al A century of ice retreat on Kilimanjaro

4800

4600

5000

4400

5200

5400

5600

5700

4200

4000

5800

4800

4400

4600 4200

37deg23E

37deg23E

37deg22E

37deg22E

37deg21E

37deg21E

37deg20E

37deg20E

3deg3

S

3deg3

S

3deg4

S

3deg4

S

3deg5

S

3deg5

S

3deg6

S

3deg6

S

314000

314000

316000

316000

318000

318000

320000

320000

9658000

9658000

9660000

9660000

9662000

9662000

Uhuru Peak

Northern ice field

Heim

Uhlig

Dec

ken

Ratzel

Kersten

Reb

man

n

Credner

Drygalski

Arrow

Diamond

Balletto

Furtwaumlngler

LittlePenck

5895m

IceCastle

Easternice field

Southernice field

WGS84UTM 37S

0 1000 2000

Meters

Revised 1912 boundaries

Credner Glacier name

2011 Ice boundaries

1912 boundaries fromHastenrath and Greischar (1997)

Little Barranco

Fig 1 Glacier extent on Kilimanjaro in 1912 and 2011 Glacier names with an asterisk do not appear in Klute (1920) and are described inthe text The interpretation of the glacier extent for 1912 as defined by Hastenrath and Greischar (1997) is shown as a dashed line (blue) forcomparative purposes The nomenclature for each glacier used by Klute (1920) is given while the names of four glaciers introduced afterthis time are indicated usinglowast

accounts are more supportive (Gillman 1923 Geilinger1936) A more detailed map of Kilimanjarorsquos glaciers pro-duced by Meyer (1900) based on his observations during afield expedition in 1898 shows the connection no longer ex-isted suggesting a break-up occurred sometime in the early1890s Though the timing is uncertain the implications oflosing this connection are not overlooked by Geilinger (1936p 12) who refers to the Little and Great Barranco glaciers asldquoindependent dead glaciersrdquo and made the astute observationthat both ice entities were retreating at their upper and lowermargins

This observation holds some importance as it helps clar-ify the source of the ice remnants still observed today Themaps by Meyer (1900 map dated 1898) Jaeger (1909map dated 1906) and Klute (1920 map dated 1912) clearlyshow distinct ice streams for the Little and Great Barrancoglaciers which as noted may have been fed by ice from theFurtwangler glacier as recently as the late 19th century Ourrevision of the Klute (1920) map from 1912 (Fig 1) honoursthe discrete boundary between the Little and Great Barranco

glaciers which is not as clearly depicted in the Hastenrathand Greischar (1997) interpretation The failure to clearlydistinguish the boundaries and location of these two indepen-dent ice streams has been inherited by Thompson et al (20022009) who used the Hastenrath and Greischar (1997) inter-pretation of the ice extent prior to 2000 (see Sect 231)Thompson et al (2009) argue that ice remnants identified byCullen et al (2006) in both the ice streams of the former andmore extensive Little and Great Barranco glaciers are absentfrom previous observations and therefore most likely tran-sient snow The sketch by Humphries (1972 Fig 46) andthe map produced by Messerli (1980 Fig 6) also support theinterpretation by Cullen et al (2006) while the ice bodiesin question have been identified in each of the satellite im-ages used in the present analysis (Fig 2) Though not labelledby Humphries (1959 1972) or by those preceding his obser-vations the Arrow glacier has been used to refer to one ofthe Little Barranco ice remnants (Fig 1) It appears that thisterm was adopted by climbers on the mountain (eg Samp-son 1971 p 202) but the exact location of this ice body is

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 425

Table 2Areal extents (km2) of the total ice bodies between 1912 and 2011 as well as for zone 1 (western region) zone 2 (plateau region)and zone 3 (southern region) as defined in the text and Fig 4 The annual rates of area change per observation period (10minus2 km2 yrminus1) areshown in parentheses for the total and zoned areas

Epoch Duration Area [km2] (rate of area change [10minus2 km2 yrminus1])

[years] Total Zone 1 Zone 2 Zone 3

191283 11400 4540 4293 2567196216 4936 7320 (minus827) 2520 (minus409) 3161 (minus229) 1639 (minus188)197562 1347 6052 (minus941) 2060 (minus342) 2529 (minus469) 1463 (minus131)198448 887 4815 (minus1395) 1628 (minus487) 1949 (minus654) 1238 (minus254)199300 853 3800 (minus1190) 1142 (minus570) 1638 (minus365) 1020 (minus256)200014 715 2920 (minus1232) 0701 (minus617) 1411 (minus318) 0808 (minus297)200308 295 2500 (minus1425) 0539 (minus550) 1251 (minus543) 0710 (minus332)201146 838 1762 (minus881) 0333 (minus246) 0937 (minus375) 0492 (minus260)

not well documented and we are uncertain that the remnantidentified in Fig 1 and by Cullen et al (2006) is the sameas that referred to by others (eg Sampson 1971 Messerli1980 Hastenrath and Greischar 1997)

There has also been some confusion about the Diamondand Balletto glaciers (Fig 1) which were not formally rec-ognized in the early mapping work (Meyer 1990 Jaeger1909 Klute 1920) or acknowledged by Humphries (19591972) but are both described by Hastenrath (1984) whileSampson (1971) and Messerli (1980) only refer to the for-mer The Breach Wall area where both glaciers are locatedwas mapped by Klute (1920) but as no break up from thesouthern ice field had occurred at this time they were notgiven discrete labels (Fig 1) However what is indisputableis the presence of an ice field (the Balletto glacier) beneaththe headwall of the Diamond glacier which was not mappedby Thompson et al (2002 2009) The presence of this icefield is not only observed in the satellite record but is docu-mented by Reinhold Messner (Messner 1991 Chapter 21)who made the first (and only) direct ascent of the Breach Wallof Kilimanjaro in 1978 Though Cullen et al (2006) mappedthis region of ice (the Balletto glacier) they failed to label itcorrectly placing the label beneath the Heim glacier after fol-lowing the interpretation of Hastenrath and Greischar (1997Fig 2) Further the Diamond glacier is absent after 1976 inthe maps produced by Hastenrath and Greischar (1997) andHastenrath (2006) but appears in the Thompson et al (20022009) maps despite using the former to build their time se-ries Finally the Uhlig glacier which Jaeger (1909 p 130)named after his friend C Uhlig does not appear to have beenconnected to the northern ice field in the past (Fig 1) Whilethe upper boundary of the Uhlig glacier is difficult to inter-pret in the Klute (1920) map it is clearly shown by Jaeger(1909) that it was not joined to the northern ice field

Figure 2 shows the change in ice extent over almost a 100-yr period which is consistent through time and removes mostof the uncertainty about the location and extent of the re-maining ice bodies on Kibo The retreat over the first part of

the 20th century was approximately 41 km2 (Fig 2a and b)while over the last 50 yr 56 km2 of the ice cover has beenlost (Fig 2cndashh) The retreat between 1912 and 1962 (Fig 2aand b) resulted in a distinct break-up of ice in the vicinity ofthe notches (south-east crater region) separating the northernand eastern ice fields from the southern ice field Remnantsof the Ice Castle (Fig 1) still remained in 1962 and were de-scribed by Sampson (1971) as still being present in the early1970s but appeared to have been lost by 1975 (Fig 2c) Othernoticeable features of the retreat in the early 20th century in-clude the development of a distinct separation between theGreat and Little Penck glaciers which formed through theexpansion of the Ravenstein (first mapped by Meyer 1900)the rapid demise of the Uhlig glacier and the split-up of theDiamond and Balletto glaciers from the southern ice field(Fig 2a and b)

Between 1962 and 1975 the persistent retreat of ice led tothe northern and eastern ice fields becoming separated fromeach other as did the Great Penck glacier from the north-ern ice field (Fig 2b and c) The Little Penck proved moreresistant and did not separate from the northern ice field un-til after 1992 while the remnant of the Great Penck com-pletely disappeared during this period (Fig 2e and f) thoughground-based (field) observations in January 2005 revealeddebris-covered ice at the location of the lowest 1984 extent(Fig 2d) As noted by others (eg Hastenrath and Greischar1997 Cullen et al 2006) the northern ice field developed ahole sometime after 1984 (Fig 2d and e) which opened intoa large canyon after 2003 (Fig 2g and h) Field observationsby the first author in October 2012 revealed that the north-ern ice field has separated into two independent ice bodieswhich must have occurred sometime after 17 June 2011 TheFurtwangler glacier has also separated into smaller ice enti-ties which began after 2003 (Fig 2g and h) The southernice field has not escaped the on-going demise with the upperpart of the Kersten glacier separating from its lower reachesbetween 2003 and 2011 (Fig 2g and h) which was some-what anticipated after mass balance modelling showed this

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

426 N J Cullen et al A century of ice retreat on Kilimanjaro

(g) 2003 (h) 2011

(a) 1912 (b) 1962

(c) 1975 (d) 1984

(e) 1992 (f) 2000

Credner

Great Penck

Ratzel

Ker

sten

Northern ice field

Easternice field

Southernice field

GreatBarranco

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Fig 2 Ice retreat over Kilimanjaro based on the revised areal extent for the following years(a) 1912(b) 1962(c) 1975(d) 1984(e)1992(f) 2000(g) 2003 and(h) 2011 An interactive 3-D illustration of the sequence can be sourced in the Supplement

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 427

was likely to occur (Molg et al 2009b) The upper part ofthe Heim has disappeared and the Decken has narrowed con-siderably over the last 8 yr (Fig 2g and h)

32 New interpretation of the ice retreat

A consistent record of the former extent of the glaciers onKibo provides a platform from which the relationship be-tween glaciological and meteorological processes can be fur-ther explored We believe the spatial and temporal variabilityof the ice cover on Kibo can be used to unravel the physi-cal processes responsible for controlling the decay of the icemasses through time as well as providing insights into themechanisms that may have been responsible for growth in thepast Interestingly those who were privileged to make someof the first (formal) observations on Kilimanjaro recognizedthere was an asymmetry in the ice cover (eg Meyer 1891)which pointed to specific climatological drivers controllingglacial retreat (eg Gillman 1923 Nilsson 1931 Geilinger1936) Meyer (1891 p 308ndash314) noted that ldquoan immensemountain mass like Kilimanjaro must considerably modifythe prevailing wind directionrdquo before making the point thatthe primary cause in the variability of the ice limit is ldquotheunequal distribution of moisture on the northern and south-ern sides of the mountainrdquo Though Meyer (1891) recognizedvariability in moisture was the most important control healso speculated that the asymmetry of low reaching glacierson Kibo in the south south-west and western sectors of themountain a pattern that persisted throughout the 20th cen-tury (Fig 2) depended on ldquolocal peculiarities which favourthe preservation of the snow [rather] than on physical condi-tions that favour a large snowfallrdquo (Meyer 1891 p 315) Torevisit this early observation it is useful to reflect on the hyp-sometry of the glaciers in 1912 (Fig 3a) which shows thatthe probability density in relation to the distribution of the icewas greatest above 5700 m This high density reflects the icecover in the crater region with the majority of the ice in 1912being part of the former ice cap that linked the northern andeastern ice fields (Fig 1) Figure 3b and c show that slopeand aspect were also important controls on the density of icecover with the steep south and south-west facing slopes morefavourable which enabled the glaciers to extend to elevationsbelow 4600 m For the glaciers to have extended to the sameelevation on the east-facing slopes they would have had tocover a much greater horizontal distance (Fig 1)

To fully account for the distinct geometry of the ice masseson Kibo and to establish the physical processes controllingtheir location and extent through time Kaser et al (2004)established a framework that included four separate glacierregimes The motivation to do this was partly based on theearly recognition that the tabular shaped glaciers in the craterregion of Kibo behave differently to glaciers on the flanks ofthe mountain (eg Nilsson 1931 Geilinger 1936) Guidedby this distinction Cullen et al (2006) used the 5700 m con-tour which is roughly the height of the outer crater rim to

Density times 10 minus3

2

4

6

8

WEST EAST

SOUTH

NORTH

4600 4800 5000 5200 5400 5600 58000

1

2

3

4

Elevation [m]

Den

sity

times 10

minus3

0 10 20 30 40 50 60 70 800

1

2

3

4

Slope [deg]

Den

sity

times 10

minus2

(a)

(c)

(b)

Fig 3 Distribution of topographic measures from the areal extentof all ice bodies in 1912(a) elevation(b) slope and(c) aspect Themagnitude of the normalised distribution on the y-axis is a proba-bility density function with the integral equal to unity

distinguish between two broad glacial systems (1) plateau(ge 5700 m) and slope (lt 5700 m) glaciers This simple bi-nary classification has proved useful in describing some ofthe key physical processes responsible for the retreat of iceon Kibo (eg Molg et al 2009b Kaser et al 2010) Tofurther characterize the spatial and temporal variability ofglacier retreat on Kibo we have adopted a geo-statisticalapproach in this study that uses elevation slope and aspectof the ice cover in 1912 (the feature space) to identify clus-ters resembling topographical classes This was achieved byperforming a cluster analysis containing maximum likeli-hood estimates of the topographical parameters in a Gaussian

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

428 N J Cullen et al A century of ice retreat on Kilimanjaro

mixture model (McLachlan and Peel 2000) which resultedin three broad glacial zones being identified (1) western re-gion (zone 1) (2) plateau region (zone 2) and (3) southernregion (zone 3) (Fig 4)

Figure 5 shows the rate of retreat of each of the glacialzones through time as well as a least-squares linear regres-sion fit (and 80 confidence intervals) through each of theglacial zone time series As described by Cullen et al (2006)the retreat of the tabular shaped plateau glaciers (zone 2) iscontrolled primarily by solar radiation-induced ablation onthe vertical cliffs that characterize them as demonstratedthrough radiation modelling (Molg et al 2003) Winkleret al (2010) revealed that direct insolation is the reasonfor the predominant zonal alignment of the ice cliffs andtheir bimodal annual recession (sunlit versus shaded peri-ods) Though the shrinkage of the ice cover along the verticalcliffs is fairly constant from year to year the areal retreat ofthe zone 2 region is not entirely linear with the highest ratesoccurring between 1975 and 1984 (Table 2) which corre-spond to the period following the separation of the northernand eastern ice fields (Fig 2) Therefore changes in the ge-ometry of the ice bodies and their exposure to solar radiationappear to have led to variations in retreat rates (Table 2 andFig 5) which has also been described by Kaser et al (2010)

An explanation for the growth and decay of the plateauglaciers including a description of the physical processesthat may have led to the observed steps in the northern icefield has been provided by Kaser et al (2010) To sustaingrowth in the crater rim region (zone 2) it is critical thatsnowfall is sufficient to allow a continuous snow cover todevelop that is able to survive from one year to the next Ifthe annual build-up of snow is sustained through a period oftime it is quite conceivable that the plateau glaciers couldgrow to several tens of metres in thickness quite quicklyraising the question as to whether this accumulation of icehas the potential to become a source of mass for glacierson the western flanks of the mountain (zone 1) Meyer(1891 p 316ndash317) was somewhat sceptical as to whetherthe plateau region acted as a ldquoreservoirrdquo and thought it ldquomuchmore reasonable to suppose that the rocky rim of the crateris the real reservoirrdquo However Meyer (1890 1891) did pro-vide an account and sketches to support a connection of theplateau ice to the Little and Great Barranco glaciers in thepast which as described may have been important in sustain-ing them (Gillman 1923 Geilinger 1936) There is someimportance in investigating the extent of the linkages be-tween these two glacial zones as together they accounted foralmost 80 of the total ice cover in 1912 (Table 2)

The remaining ice bodies in zone 2 (plateau region) aremostly stagnant today with field measurements and prelim-inary modelling results indicating deformation rates of icewithin the crater are small (lt 10 cm yrminus1) (Kaser et al 2004Winkler et al 2010) It seems unlikely that the present or his-torical thickness of the plateau ice has been sufficient to al-low substantial deformation (Kaser et al 2004) suggesting

3140 3150 3160 3170 3180 3190 3200 32109657

9658

9659

9660

9661

9662

9663

9664

4000

4000

4500

4500 4500

45004500

5000

5000

5000

5000

5000

5500

5500

5500

UTM easting in meters times105

UT

M n

orth

ing

in m

eter

s times

106

Fig 4 Classification of regions based on the distribution of topo-graphic measures in 1912 Zone 1 (western region) zone 2 (plateauregion) and zone 3 (southern region) are shown in red orange andyellow respectively

that mass transport from the nearly horizontal plateau downinto the western slope glaciers (zone 1) has most likely beenminor Careful observations by Klute (1920) demonstratedthat the velocity of the centre of the Great Penck glacier was064 m over a 30-day period in 1912 which led him to spec-ulate that the annual downslope movement was about 7 mper year It should be noted that Nilsson (1931) and Osmas-ton (1989) also report on these observations but incorrectlycite Jaeger (1909) as the person responsible for obtaining themeasurements Nonetheless it is improbable that ice fromthe plateau region could have supported the observed flowdynamics of the Great Penck glacier with the upper reachesof the mountain flanks the most likely source of the massThus the growth and decay of the glaciers in zone 1 (west-ern region) do not appear to be directly dependent on thestability of plateau ice (zone 2) To further resolve the phys-ical processes controlling the stability of the plateau ice itwould be of interest to develop a better understanding of themechanisms responsible for vertical cliff development

The glaciers on the southern slopes (zone 3) differ fromthose on the western slopes (zone 1) as they are separatedfrom the large masses of ice within the caldera (zone 2) bybed topography (Humphries 1972) This distinction is im-portant as any assessment of the atmospheric controls onthe behaviour of the southern ice fields is independent ofthe complexity of the vertical cliff retreat in the plateau re-gion (zone 2) That said the cluster analysis that led to thetopographical classes used in this study shows that the up-per sections of the southern ice field exhibit similar phys-ical characteristics to the ice within the caldera (Fig 4)Nevertheless to obtain a full solution for the asymmetry ofthe glaciers on Kibo that the early explorers observed itis critical to establish an understanding of the atmospheric

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 429

1900 1920 1940 1960 1980 2000 2020 2040 20600

05

1

15

2

25

3

35

4

45

5

Year

Are

al e

xten

t [km

2 ]

zone 1 R2=099

zone 2 R2=097

zone 3 R2=09980 confidence interval

4000

4000

4000

4500

4500

4500

45005000

5000

5000

50005000

5500

5500

5500

Fig 5 The areal change of glaciers and ice bodies based on thethree-member classification Linear regressions (thin solid lines) foreach of the three zones are shown as well as the 80 confidenceintervals for each of these relationships

processes controlling the growth andor decay of the southernice fields This has been achieved by Molg et al (2009b)who used observations and atmospherendashglacier modellingto show how critical precipitation humidity and fractionalcloud cover are on the energy and mass balance of the Ker-sten glacier (zone 3) These results compare favourably toearly accounts of the importance of precipitation (moisturechange) and cloud cover on the asymmetry (Meyer 1891Klute 1920 Gillman 1923 Geilinger 1936) with the diur-nal variability of cloud cover being especially important inproviding protection to the glaciers on the western side ofthe mountain (eg Meyer 1891 Hastenrath and Greischar1997)

Recent regional atmospheric modelling experiments haveprovided a more quantitative assessment to account for theasymmetry in glacier mass on Kibo demonstrating that thewest and south-western areas of the mountain are favourableto moisture convergence as controlled by atmospheric dy-namics (Molg et al 2009a) As noted by Osmaston (1989)the asymmetry is not confined to the most recent glacier ex-tents with the pattern of ice distribution established fromformer moraines indicating that recent climatic controls onglacier behaviour are similar to those in the past This sug-gests that establishing the specific nature of the atmosphericcontrols responsible for the most recent shift from glaciergrowth to decay in particular in zones 2 and 3 (plateau andsouthern slope regions) remains critical if we are to useglacier behaviour to reconstruct climatic conditions of thedeeper past

Finally it is difficult not to speculate on when the ice onKibo will disappear which becomes more achievable if onerecognizes the importance of the different glacial zones onthe mountain If one accepts linear extrapolation based on

changes from the past is suitable to make projections for thefuture it is apparent that the ice cover in zone 1 (western re-gion) will most likely be gone before 2020 and that ice inzone 2 (plateau) and zone 3 (southern slopes) will disappearsometime close to 2040 (Fig 5) It is likely that the plateauwill harbour the final remnants of ice on Kibo which couldremain slightly beyond the middle of the 21st century Inter-estingly our linear extrapolation for the loss of ice is in closeagreement to earlier projections using radiation modelling(Molg et al 2003) Though the loss of ice will have no hy-drological significance to lowland areas (eg Gillman 1923Kaser et al 2004 Molg et al 2008b) there is no doubt thattheir disappearance will be a ldquogrievous aesthetic lossrdquo (Salt1951 p 160) which will undoubtedly impact the experiencetourists have on the mountain (Molg et al 2008b)

4 Conclusions

With reference to the glaciers found on Mt Kenya and Kiboin East Africa C Gillman remarked ldquowe might perhapslearn a good deal as to the general world circulation fromcareful and prolonged observation on these particular sum-mitsrdquo (Mackinder 1923 p 23) This statement still holdstrue today (eg Molg et al 2006) and with the increasingawareness that glaciers on Africarsquos highest mountain can beused to extract climate information from the past it seemedtimely to revisit the mapping of the ice cover on Kibo throughtime Our motivation to do so has been to ensure the histor-ical record of these changes is of the highest possible qual-ity and in a format that should help facilitate on-going andfuture research dedicated to resolving the complexity of cli-mate variability and change in East Africa and beyond Thiswas achieved by using a consistent cartographic frameworkthat allows historical maps and new satellite imagery to beexamined The use of three-dimensional visualization tech-niques greatly enhanced our ability to establish the formerlocation of the glaciers

The revisited time series of glacial retreat on Kiliman-jaro contains eight epochs and spans almost 100 yr (1912ndash2011) The glaciers have retreated from their former extentof 1140 km2 in 1912 to 176 km2 in 2011 which representsa loss of about 85 of the ice cover over almost a 100-yr pe-riod Our estimate of the total retreat is consistent with recentassessments (Cullen et al 2006 Thompson et al 2009) butthe careful re-examination of the location and extent of indi-vidual ice bodies through time has allowed us to identify anumber of important differences from previously publishedmaps (eg Hastenrath and Greischar 1997 Thompson et al2002 2009) Historical evidence has been used to supportour interpretation with the detailed observations and pho-tographs obtained by early explorers on Kilimanjaro found tobe extremely valuable (eg Meyer 1891 1900 Uhlig 1904Jaeger 1909 Klute 1914 1920 1921 Gillman 1923 Nils-son 1931 Geilinger 1936) It has to be acknowledged that

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

430 N J Cullen et al A century of ice retreat on Kilimanjaro

much of what we know about Kilimanjaro today was estab-lished through their efforts

To further explore the importance of different glacierregimes on Kibo (Kaser et al 2004) a geo-statistical ap-proach was applied which resulted in the segmentation ofthe ice cover into three broad glacial zones A careful ex-amination of the retreat of ice in each of these glacier zonesindicates that there is a high probability that most of the re-maining ice cover on Kilimanjaro will disappear by 2040and it is highly unlikely that any ice body will remain af-ter 2060 if present retreat rates are maintained Before thishappens it would still be useful if the remaining ice bodiesin particular the remnants of the northern ice field could befurther examined to help resolve existing uncertainties aboutthe mechanisms responsible for controlling the formation ofthe vertical cliffs that characterize them (eg Winkler et al2010) In addition further efforts are required to scrutinizethe exact timing of the onset of the present glacier retreat firstreported by Hans Meyer in the late 19th century If we canrefine our understanding of the atmospheric controls behindthe most recent shift from glacier growth to decay on Kibowe will be in a stronger position to characterize the climaticcontrols of past and much larger glaciations on Africarsquos high-est mountain

Supplementary material related to this article isavailable online athttpwwwthe-cryospherenet74192013tc-7-419-2013-supplementpdf

AcknowledgementsThis research is funded by the Departmentof Geography and School of Surveying University of OtagoNew Zealand Thomas Molg was supported by the Alexander vonHumboldt Foundation The contributions made by Georg Kaserand Michael Winkler were supported by the Austrian ScienceFund (FWF) Projects P 20089-N10 and P 21288-N10 Thisresearch would not have been possible without the support from thefollowing Tanzanian authorities (COSTECH KINAPA TANAPATAWIRI) We are grateful to the referees that took the time toprovide feedback on this research in particular the careful andthoughtful comments by Henry Brecher

Edited by M Schneebeli

References

Chiang J C H The tropics in paleoclimate Annu Rev EarthPlanet Sci 37 263ndash297 2009

Cullen N J Molg T Kaser G Hussein K Steffen Kand Hardy D R Kilimanjaro Glaciers Recent areal ex-tent from satellite data and new interpretation of observed20th century retreat rates Geophys Res Lett 33 L16502doi1010292006GL027084 2006

Cullen N J Molg T Kaser G Steffen K and Hardy D REnergy-balance model validation on the top of Kilimanjaro Tan-zania using eddy covariance data Ann Glaciol 46 227ndash233doi103189172756407782871224 2007

Downie C Glaciations of Mount Kilimanjaro Northeast Tan-ganyika Geol Soc Am Bull 75 1ndash16doi1011300016-7606(1964)75[1GOMKNT]20CO2 1964

Downie C and Wilkinson P The Geology of KilimanjaroSheffield Geology Department University of Sheffield 1972

Downie C Humphries D W Wilcockson W H and WilkinsonP Geology of Kilimanjaro Nature 178 828ndash830 1956

Farr T G Rosen P A Caro E Crippen R Duren R Hens-ley S Kobrick M Paller M Rodriguez E Roth L Seal DShaffer S Shimada J Umland J Werner M Oskin M Bur-bank D and Alsdorf D The Shuttle Radar Topography Mis-sion Rev Geophys 45 RG2004doi1010292005RG0001832007

Geilinger W The retreat of the Kilimanjaro glaciers TanganyikaNotes and Records 2 7ndash20 1936

Gillman C An ascent of Kilimanjaro Geogr J 61 1ndash27 1923Hastenrath S The Glaciers of Equatorial East Africa Reidel Dor-

drecht 1984Hastenrath S Diagnosing the decaying glaciers of equatorial

East Africa Meteorol Z 15 265ndash271doi1011270941-294820060106 2006

Hastenrath S Recession of equatorial glaciers a photo documen-tation Sundog Publishing Madison Wisconsin 2008

Hastenrath S and Greischar L Glacier recession on KilimanjaroEast Africa 1912ndash89 J Glaciol 43 455ndash459 1997

Humphries D Interim report on the glaciology and meteorology ofKilimanjaro Tanganyika JulyndashSept 1953 Geological Survey ofTanganyika Dodoma 1953

Humphries D Preliminary notes on the glaciology of KilimanjaroJ Glaciol 3 475ndash479 1959

Humphries D Glaciology and glacial history in The Geology ofKilimanjaro edited by Downie C and Wilkinson P SheffieldGeology Department University of Sheffield 31ndash71 1972

Jarvis A Reuter H I Nelson A and Guevara E Hole-filledSRTM for the globe Version 4 available from the CGIAR-CSISRTM 90 m Database (httpsrtmcsicgiarorg) 2008

Jaeger F Forschungen in den Hochregionen des KilimandscharoMitt Deutsch Schutzgebiet 22 113ndash197 1909

Kaser G Hardy D R Molg T Bradley R S and Hyera TM Modern glacier retreat on Kilimanjaro as evidence of climatechange Observations and facts Int J Climatol 24 329ndash3392004

Kaser G Molg T Cullen N J Hardy D R andWinkler M Is the decline of ice on Kilimanjaro un-precedented in the Holocene Holocene 20 1079ndash1091doi1011770959683610369498 2010

Klute F Forschungen am Kilimandscharo im Jahre 1912 GeogrZeitschrift 20 496ndash505 1914

Klute F Ergebnisse der Forschungen am Kilimandscharo 1912ReimerndashVohsen Berlin 1920

Klute F Die stereophotogrammetrische Aufnahme der Hochregio-nen des Kilimandscharo Zeitschrift Ges Erdkunde Berlin 56144ndash151 1921

Mackinder C W Hobley C Gillman C and Johnston H Anascent of Kilimanjaro Discussion Geogr J 61 1ndash27 1923

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 423

to characterize the physical processes controlling glacier re-treat on Kibo

232 Establishing new ice boundaries

The boundaries of ice bodies were visually interpreted anddigitized from the orthorectified satellite images and histor-ical maps Despite the variability in the spatial and spectralresolution of the data products as well as having to accountfor varying amounts of transient snow in the satellite im-ages the use of the high temporal resolution sequence sig-nificantly improved our ability to interpret the location andextent of ice bodies consistently through time To supportand facilitate the interpretation of all data products (Table 1)three-dimensional visualization techniques were extensivelyused in conjunction with aerial and ground-based photogra-phy from others (eg Hastenrath 2008) and our own fieldexpeditions between 2005ndash2012 The vantage points fromwhich historical photographs were taken were reproducedin the 3-D model of Kilimanjaro which proved particularlyuseful when revisiting the glacier outlines mapped by Klute(1920) The use of 3-D visualization also helped us to inter-pret glacier positions in relation to distinctive geomorpho-logical features (eg moraines rock outcrops) which are de-scribed in more detail in the following section Finally tohelp facilitate future efforts all data products from this anal-ysis have been made available on the Global Land Ice Mea-surements from Space (GLIMS) Glacier Database

3 Results and discussion

31 The mapping reloaded

The total amount of ice cover on Kilimanjaro on 17 June2011 was 176 km2 (Table 2) with the location of these icemasses shown in Fig 1 The remaining ice masses are rem-nants of more extensive glaciers which were first mappedand given names by Meyer (1900) with the nomenclaturestill in use despite the character of the ice bodies havingchanged dramatically over the last century The majority ofthe ice cover observed on Kilimanjaro today fails to strictlyadhere to any common definition of a glacier with most enti-ties isolated and stagnant but following the definition sug-gested by Racoviteanu et al (2009) we include any bodyof ice that is observed after transient snow melts It is clearfrom the accounts and photographs of the early explorers thatonly the principal glaciers were named and that a consider-able number of small stagnant ice masses in particular in thecrater area were not formally recognized This is no more ev-ident than in the south-east crater area where the first explor-ers accessed the summit region through a number of weak-nesses in the ice referred to as notches (eg Hans Meyer andJohannes notches) which have been described in detail else-where (eg Geilinger 1936 Hastenrath 1984) Despite thisapparent difficulty it is clear that the remaining ice bodies

today are remnants of the more extensive glaciers mappedby Klute (1920) using terrestrial photogrammetry (Table 1)with the boundaries established from this effort reflecting theice cover in 1912 (Fig 1) Our calculation of the extent ofthe glaciers on Kilimanjaro at the end of October in 1912 is1140 km2 (Table 2) which is 85 more than what is ob-served on the mountain today

The differences between our interpretation of the 1912map produced by Klute (1920) and that produced by Has-tenrath and Greischar (1997) are also shown in Fig 1 Asnoted in Sect 232 our interpretation of the glacier bound-aries was supported by geomorphological evidence observedusing 3-D visualization and historical photographs Impor-tantly the 3-D model allowed us to carefully identify andlocate the most recent moraines Using these geomorpholog-ical features as markers we were able to determine with ahigh degree of accuracy the former location of the glaciersThis approach was particularly useful in establishing the pastlocation and extent of the outlet glaciers feeding from thenorthern ice field (the Credner Drygalski and (Great andLittle) Penck glaciers) the Uhlig glacier and the two distinctglacier streams located in the western breach area (the Lit-tle and Great Barranco glaciers) as well as the four primarysouthern ice field glaciers (the Heim Kersten Decken andRebmann) The map produced by Jaeger (1909 map dated1906) was also valuable in allowing us to identify a num-ber of small ice bodies located above both the Little andGreat Barranco glaciers (Fig 1) Our interpretation of thealtitudinal extent of the southern ice field glaciers at about4600 m (Fig 1) is supported by the photographs of Uhlig(1904 Fig 50 p 641) and Klute (1920 Fig 1) It is impor-tant to note that both these historical photographs reveal acomplexity of the outcropping rock at the lower elevations ofthe southern ice field glaciers that we have made an attemptto account for (Fig 1)

Despite a well-documented and photographic account ofthe historical retreat of Kilimanjarorsquos glaciers which hasbeen carefully reviewed by Hastenrath (1984 2008) therehave been a number of inconsistences in the interpretationof ice bodies in recent mapping efforts (eg Hastenrathand Greischar 1997 Cullen et al 2006 Hastenrath 2006Thompson et al 2002 2009) which require further clarifi-cation as they have an impact on our ability to interpret thephysical processes controlling the demise of the glaciers onKibo To do this one must briefly reflect on the historical ob-servations starting with those of Hans Meyer who provideda sketch (Meyer 1890 p 345) and a map (Meyer 1891) thatindicates that an ice stream which is now referred to as theFurtwangler glacier flowed over the crater rim into the west-ern breach area Meyer (1891 p 319) notes ldquothe ice from thevast caldera falls as a mighty cascade into the great westernfissure where it unites with the accumulations in the fissureitselfrdquo Though Hastenrath (1984 p 78) disputes whetherthis ice stream (the Furtwangler glacier) was connected tothe Little andor Great Barranco glaciers other historical

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

424 N J Cullen et al A century of ice retreat on Kilimanjaro

4800

4600

5000

4400

5200

5400

5600

5700

4200

4000

5800

4800

4400

4600 4200

37deg23E

37deg23E

37deg22E

37deg22E

37deg21E

37deg21E

37deg20E

37deg20E

3deg3

S

3deg3

S

3deg4

S

3deg4

S

3deg5

S

3deg5

S

3deg6

S

3deg6

S

314000

314000

316000

316000

318000

318000

320000

320000

9658000

9658000

9660000

9660000

9662000

9662000

Uhuru Peak

Northern ice field

Heim

Uhlig

Dec

ken

Ratzel

Kersten

Reb

man

n

Credner

Drygalski

Arrow

Diamond

Balletto

Furtwaumlngler

LittlePenck

5895m

IceCastle

Easternice field

Southernice field

WGS84UTM 37S

0 1000 2000

Meters

Revised 1912 boundaries

Credner Glacier name

2011 Ice boundaries

1912 boundaries fromHastenrath and Greischar (1997)

Little Barranco

Fig 1 Glacier extent on Kilimanjaro in 1912 and 2011 Glacier names with an asterisk do not appear in Klute (1920) and are described inthe text The interpretation of the glacier extent for 1912 as defined by Hastenrath and Greischar (1997) is shown as a dashed line (blue) forcomparative purposes The nomenclature for each glacier used by Klute (1920) is given while the names of four glaciers introduced afterthis time are indicated usinglowast

accounts are more supportive (Gillman 1923 Geilinger1936) A more detailed map of Kilimanjarorsquos glaciers pro-duced by Meyer (1900) based on his observations during afield expedition in 1898 shows the connection no longer ex-isted suggesting a break-up occurred sometime in the early1890s Though the timing is uncertain the implications oflosing this connection are not overlooked by Geilinger (1936p 12) who refers to the Little and Great Barranco glaciers asldquoindependent dead glaciersrdquo and made the astute observationthat both ice entities were retreating at their upper and lowermargins

This observation holds some importance as it helps clar-ify the source of the ice remnants still observed today Themaps by Meyer (1900 map dated 1898) Jaeger (1909map dated 1906) and Klute (1920 map dated 1912) clearlyshow distinct ice streams for the Little and Great Barrancoglaciers which as noted may have been fed by ice from theFurtwangler glacier as recently as the late 19th century Ourrevision of the Klute (1920) map from 1912 (Fig 1) honoursthe discrete boundary between the Little and Great Barranco

glaciers which is not as clearly depicted in the Hastenrathand Greischar (1997) interpretation The failure to clearlydistinguish the boundaries and location of these two indepen-dent ice streams has been inherited by Thompson et al (20022009) who used the Hastenrath and Greischar (1997) inter-pretation of the ice extent prior to 2000 (see Sect 231)Thompson et al (2009) argue that ice remnants identified byCullen et al (2006) in both the ice streams of the former andmore extensive Little and Great Barranco glaciers are absentfrom previous observations and therefore most likely tran-sient snow The sketch by Humphries (1972 Fig 46) andthe map produced by Messerli (1980 Fig 6) also support theinterpretation by Cullen et al (2006) while the ice bodiesin question have been identified in each of the satellite im-ages used in the present analysis (Fig 2) Though not labelledby Humphries (1959 1972) or by those preceding his obser-vations the Arrow glacier has been used to refer to one ofthe Little Barranco ice remnants (Fig 1) It appears that thisterm was adopted by climbers on the mountain (eg Samp-son 1971 p 202) but the exact location of this ice body is

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 425

Table 2Areal extents (km2) of the total ice bodies between 1912 and 2011 as well as for zone 1 (western region) zone 2 (plateau region)and zone 3 (southern region) as defined in the text and Fig 4 The annual rates of area change per observation period (10minus2 km2 yrminus1) areshown in parentheses for the total and zoned areas

Epoch Duration Area [km2] (rate of area change [10minus2 km2 yrminus1])

[years] Total Zone 1 Zone 2 Zone 3

191283 11400 4540 4293 2567196216 4936 7320 (minus827) 2520 (minus409) 3161 (minus229) 1639 (minus188)197562 1347 6052 (minus941) 2060 (minus342) 2529 (minus469) 1463 (minus131)198448 887 4815 (minus1395) 1628 (minus487) 1949 (minus654) 1238 (minus254)199300 853 3800 (minus1190) 1142 (minus570) 1638 (minus365) 1020 (minus256)200014 715 2920 (minus1232) 0701 (minus617) 1411 (minus318) 0808 (minus297)200308 295 2500 (minus1425) 0539 (minus550) 1251 (minus543) 0710 (minus332)201146 838 1762 (minus881) 0333 (minus246) 0937 (minus375) 0492 (minus260)

not well documented and we are uncertain that the remnantidentified in Fig 1 and by Cullen et al (2006) is the sameas that referred to by others (eg Sampson 1971 Messerli1980 Hastenrath and Greischar 1997)

There has also been some confusion about the Diamondand Balletto glaciers (Fig 1) which were not formally rec-ognized in the early mapping work (Meyer 1990 Jaeger1909 Klute 1920) or acknowledged by Humphries (19591972) but are both described by Hastenrath (1984) whileSampson (1971) and Messerli (1980) only refer to the for-mer The Breach Wall area where both glaciers are locatedwas mapped by Klute (1920) but as no break up from thesouthern ice field had occurred at this time they were notgiven discrete labels (Fig 1) However what is indisputableis the presence of an ice field (the Balletto glacier) beneaththe headwall of the Diamond glacier which was not mappedby Thompson et al (2002 2009) The presence of this icefield is not only observed in the satellite record but is docu-mented by Reinhold Messner (Messner 1991 Chapter 21)who made the first (and only) direct ascent of the Breach Wallof Kilimanjaro in 1978 Though Cullen et al (2006) mappedthis region of ice (the Balletto glacier) they failed to label itcorrectly placing the label beneath the Heim glacier after fol-lowing the interpretation of Hastenrath and Greischar (1997Fig 2) Further the Diamond glacier is absent after 1976 inthe maps produced by Hastenrath and Greischar (1997) andHastenrath (2006) but appears in the Thompson et al (20022009) maps despite using the former to build their time se-ries Finally the Uhlig glacier which Jaeger (1909 p 130)named after his friend C Uhlig does not appear to have beenconnected to the northern ice field in the past (Fig 1) Whilethe upper boundary of the Uhlig glacier is difficult to inter-pret in the Klute (1920) map it is clearly shown by Jaeger(1909) that it was not joined to the northern ice field

Figure 2 shows the change in ice extent over almost a 100-yr period which is consistent through time and removes mostof the uncertainty about the location and extent of the re-maining ice bodies on Kibo The retreat over the first part of

the 20th century was approximately 41 km2 (Fig 2a and b)while over the last 50 yr 56 km2 of the ice cover has beenlost (Fig 2cndashh) The retreat between 1912 and 1962 (Fig 2aand b) resulted in a distinct break-up of ice in the vicinity ofthe notches (south-east crater region) separating the northernand eastern ice fields from the southern ice field Remnantsof the Ice Castle (Fig 1) still remained in 1962 and were de-scribed by Sampson (1971) as still being present in the early1970s but appeared to have been lost by 1975 (Fig 2c) Othernoticeable features of the retreat in the early 20th century in-clude the development of a distinct separation between theGreat and Little Penck glaciers which formed through theexpansion of the Ravenstein (first mapped by Meyer 1900)the rapid demise of the Uhlig glacier and the split-up of theDiamond and Balletto glaciers from the southern ice field(Fig 2a and b)

Between 1962 and 1975 the persistent retreat of ice led tothe northern and eastern ice fields becoming separated fromeach other as did the Great Penck glacier from the north-ern ice field (Fig 2b and c) The Little Penck proved moreresistant and did not separate from the northern ice field un-til after 1992 while the remnant of the Great Penck com-pletely disappeared during this period (Fig 2e and f) thoughground-based (field) observations in January 2005 revealeddebris-covered ice at the location of the lowest 1984 extent(Fig 2d) As noted by others (eg Hastenrath and Greischar1997 Cullen et al 2006) the northern ice field developed ahole sometime after 1984 (Fig 2d and e) which opened intoa large canyon after 2003 (Fig 2g and h) Field observationsby the first author in October 2012 revealed that the north-ern ice field has separated into two independent ice bodieswhich must have occurred sometime after 17 June 2011 TheFurtwangler glacier has also separated into smaller ice enti-ties which began after 2003 (Fig 2g and h) The southernice field has not escaped the on-going demise with the upperpart of the Kersten glacier separating from its lower reachesbetween 2003 and 2011 (Fig 2g and h) which was some-what anticipated after mass balance modelling showed this

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

426 N J Cullen et al A century of ice retreat on Kilimanjaro

(g) 2003 (h) 2011

(a) 1912 (b) 1962

(c) 1975 (d) 1984

(e) 1992 (f) 2000

Credner

Great Penck

Ratzel

Ker

sten

Northern ice field

Easternice field

Southernice field

GreatBarranco

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Fig 2 Ice retreat over Kilimanjaro based on the revised areal extent for the following years(a) 1912(b) 1962(c) 1975(d) 1984(e)1992(f) 2000(g) 2003 and(h) 2011 An interactive 3-D illustration of the sequence can be sourced in the Supplement

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 427

was likely to occur (Molg et al 2009b) The upper part ofthe Heim has disappeared and the Decken has narrowed con-siderably over the last 8 yr (Fig 2g and h)

32 New interpretation of the ice retreat

A consistent record of the former extent of the glaciers onKibo provides a platform from which the relationship be-tween glaciological and meteorological processes can be fur-ther explored We believe the spatial and temporal variabilityof the ice cover on Kibo can be used to unravel the physi-cal processes responsible for controlling the decay of the icemasses through time as well as providing insights into themechanisms that may have been responsible for growth in thepast Interestingly those who were privileged to make someof the first (formal) observations on Kilimanjaro recognizedthere was an asymmetry in the ice cover (eg Meyer 1891)which pointed to specific climatological drivers controllingglacial retreat (eg Gillman 1923 Nilsson 1931 Geilinger1936) Meyer (1891 p 308ndash314) noted that ldquoan immensemountain mass like Kilimanjaro must considerably modifythe prevailing wind directionrdquo before making the point thatthe primary cause in the variability of the ice limit is ldquotheunequal distribution of moisture on the northern and south-ern sides of the mountainrdquo Though Meyer (1891) recognizedvariability in moisture was the most important control healso speculated that the asymmetry of low reaching glacierson Kibo in the south south-west and western sectors of themountain a pattern that persisted throughout the 20th cen-tury (Fig 2) depended on ldquolocal peculiarities which favourthe preservation of the snow [rather] than on physical condi-tions that favour a large snowfallrdquo (Meyer 1891 p 315) Torevisit this early observation it is useful to reflect on the hyp-sometry of the glaciers in 1912 (Fig 3a) which shows thatthe probability density in relation to the distribution of the icewas greatest above 5700 m This high density reflects the icecover in the crater region with the majority of the ice in 1912being part of the former ice cap that linked the northern andeastern ice fields (Fig 1) Figure 3b and c show that slopeand aspect were also important controls on the density of icecover with the steep south and south-west facing slopes morefavourable which enabled the glaciers to extend to elevationsbelow 4600 m For the glaciers to have extended to the sameelevation on the east-facing slopes they would have had tocover a much greater horizontal distance (Fig 1)

To fully account for the distinct geometry of the ice masseson Kibo and to establish the physical processes controllingtheir location and extent through time Kaser et al (2004)established a framework that included four separate glacierregimes The motivation to do this was partly based on theearly recognition that the tabular shaped glaciers in the craterregion of Kibo behave differently to glaciers on the flanks ofthe mountain (eg Nilsson 1931 Geilinger 1936) Guidedby this distinction Cullen et al (2006) used the 5700 m con-tour which is roughly the height of the outer crater rim to

Density times 10 minus3

2

4

6

8

WEST EAST

SOUTH

NORTH

4600 4800 5000 5200 5400 5600 58000

1

2

3

4

Elevation [m]

Den

sity

times 10

minus3

0 10 20 30 40 50 60 70 800

1

2

3

4

Slope [deg]

Den

sity

times 10

minus2

(a)

(c)

(b)

Fig 3 Distribution of topographic measures from the areal extentof all ice bodies in 1912(a) elevation(b) slope and(c) aspect Themagnitude of the normalised distribution on the y-axis is a proba-bility density function with the integral equal to unity

distinguish between two broad glacial systems (1) plateau(ge 5700 m) and slope (lt 5700 m) glaciers This simple bi-nary classification has proved useful in describing some ofthe key physical processes responsible for the retreat of iceon Kibo (eg Molg et al 2009b Kaser et al 2010) Tofurther characterize the spatial and temporal variability ofglacier retreat on Kibo we have adopted a geo-statisticalapproach in this study that uses elevation slope and aspectof the ice cover in 1912 (the feature space) to identify clus-ters resembling topographical classes This was achieved byperforming a cluster analysis containing maximum likeli-hood estimates of the topographical parameters in a Gaussian

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

428 N J Cullen et al A century of ice retreat on Kilimanjaro

mixture model (McLachlan and Peel 2000) which resultedin three broad glacial zones being identified (1) western re-gion (zone 1) (2) plateau region (zone 2) and (3) southernregion (zone 3) (Fig 4)

Figure 5 shows the rate of retreat of each of the glacialzones through time as well as a least-squares linear regres-sion fit (and 80 confidence intervals) through each of theglacial zone time series As described by Cullen et al (2006)the retreat of the tabular shaped plateau glaciers (zone 2) iscontrolled primarily by solar radiation-induced ablation onthe vertical cliffs that characterize them as demonstratedthrough radiation modelling (Molg et al 2003) Winkleret al (2010) revealed that direct insolation is the reasonfor the predominant zonal alignment of the ice cliffs andtheir bimodal annual recession (sunlit versus shaded peri-ods) Though the shrinkage of the ice cover along the verticalcliffs is fairly constant from year to year the areal retreat ofthe zone 2 region is not entirely linear with the highest ratesoccurring between 1975 and 1984 (Table 2) which corre-spond to the period following the separation of the northernand eastern ice fields (Fig 2) Therefore changes in the ge-ometry of the ice bodies and their exposure to solar radiationappear to have led to variations in retreat rates (Table 2 andFig 5) which has also been described by Kaser et al (2010)

An explanation for the growth and decay of the plateauglaciers including a description of the physical processesthat may have led to the observed steps in the northern icefield has been provided by Kaser et al (2010) To sustaingrowth in the crater rim region (zone 2) it is critical thatsnowfall is sufficient to allow a continuous snow cover todevelop that is able to survive from one year to the next Ifthe annual build-up of snow is sustained through a period oftime it is quite conceivable that the plateau glaciers couldgrow to several tens of metres in thickness quite quicklyraising the question as to whether this accumulation of icehas the potential to become a source of mass for glacierson the western flanks of the mountain (zone 1) Meyer(1891 p 316ndash317) was somewhat sceptical as to whetherthe plateau region acted as a ldquoreservoirrdquo and thought it ldquomuchmore reasonable to suppose that the rocky rim of the crateris the real reservoirrdquo However Meyer (1890 1891) did pro-vide an account and sketches to support a connection of theplateau ice to the Little and Great Barranco glaciers in thepast which as described may have been important in sustain-ing them (Gillman 1923 Geilinger 1936) There is someimportance in investigating the extent of the linkages be-tween these two glacial zones as together they accounted foralmost 80 of the total ice cover in 1912 (Table 2)

The remaining ice bodies in zone 2 (plateau region) aremostly stagnant today with field measurements and prelim-inary modelling results indicating deformation rates of icewithin the crater are small (lt 10 cm yrminus1) (Kaser et al 2004Winkler et al 2010) It seems unlikely that the present or his-torical thickness of the plateau ice has been sufficient to al-low substantial deformation (Kaser et al 2004) suggesting

3140 3150 3160 3170 3180 3190 3200 32109657

9658

9659

9660

9661

9662

9663

9664

4000

4000

4500

4500 4500

45004500

5000

5000

5000

5000

5000

5500

5500

5500

UTM easting in meters times105

UT

M n

orth

ing

in m

eter

s times

106

Fig 4 Classification of regions based on the distribution of topo-graphic measures in 1912 Zone 1 (western region) zone 2 (plateauregion) and zone 3 (southern region) are shown in red orange andyellow respectively

that mass transport from the nearly horizontal plateau downinto the western slope glaciers (zone 1) has most likely beenminor Careful observations by Klute (1920) demonstratedthat the velocity of the centre of the Great Penck glacier was064 m over a 30-day period in 1912 which led him to spec-ulate that the annual downslope movement was about 7 mper year It should be noted that Nilsson (1931) and Osmas-ton (1989) also report on these observations but incorrectlycite Jaeger (1909) as the person responsible for obtaining themeasurements Nonetheless it is improbable that ice fromthe plateau region could have supported the observed flowdynamics of the Great Penck glacier with the upper reachesof the mountain flanks the most likely source of the massThus the growth and decay of the glaciers in zone 1 (west-ern region) do not appear to be directly dependent on thestability of plateau ice (zone 2) To further resolve the phys-ical processes controlling the stability of the plateau ice itwould be of interest to develop a better understanding of themechanisms responsible for vertical cliff development

The glaciers on the southern slopes (zone 3) differ fromthose on the western slopes (zone 1) as they are separatedfrom the large masses of ice within the caldera (zone 2) bybed topography (Humphries 1972) This distinction is im-portant as any assessment of the atmospheric controls onthe behaviour of the southern ice fields is independent ofthe complexity of the vertical cliff retreat in the plateau re-gion (zone 2) That said the cluster analysis that led to thetopographical classes used in this study shows that the up-per sections of the southern ice field exhibit similar phys-ical characteristics to the ice within the caldera (Fig 4)Nevertheless to obtain a full solution for the asymmetry ofthe glaciers on Kibo that the early explorers observed itis critical to establish an understanding of the atmospheric

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 429

1900 1920 1940 1960 1980 2000 2020 2040 20600

05

1

15

2

25

3

35

4

45

5

Year

Are

al e

xten

t [km

2 ]

zone 1 R2=099

zone 2 R2=097

zone 3 R2=09980 confidence interval

4000

4000

4000

4500

4500

4500

45005000

5000

5000

50005000

5500

5500

5500

Fig 5 The areal change of glaciers and ice bodies based on thethree-member classification Linear regressions (thin solid lines) foreach of the three zones are shown as well as the 80 confidenceintervals for each of these relationships

processes controlling the growth andor decay of the southernice fields This has been achieved by Molg et al (2009b)who used observations and atmospherendashglacier modellingto show how critical precipitation humidity and fractionalcloud cover are on the energy and mass balance of the Ker-sten glacier (zone 3) These results compare favourably toearly accounts of the importance of precipitation (moisturechange) and cloud cover on the asymmetry (Meyer 1891Klute 1920 Gillman 1923 Geilinger 1936) with the diur-nal variability of cloud cover being especially important inproviding protection to the glaciers on the western side ofthe mountain (eg Meyer 1891 Hastenrath and Greischar1997)

Recent regional atmospheric modelling experiments haveprovided a more quantitative assessment to account for theasymmetry in glacier mass on Kibo demonstrating that thewest and south-western areas of the mountain are favourableto moisture convergence as controlled by atmospheric dy-namics (Molg et al 2009a) As noted by Osmaston (1989)the asymmetry is not confined to the most recent glacier ex-tents with the pattern of ice distribution established fromformer moraines indicating that recent climatic controls onglacier behaviour are similar to those in the past This sug-gests that establishing the specific nature of the atmosphericcontrols responsible for the most recent shift from glaciergrowth to decay in particular in zones 2 and 3 (plateau andsouthern slope regions) remains critical if we are to useglacier behaviour to reconstruct climatic conditions of thedeeper past

Finally it is difficult not to speculate on when the ice onKibo will disappear which becomes more achievable if onerecognizes the importance of the different glacial zones onthe mountain If one accepts linear extrapolation based on

changes from the past is suitable to make projections for thefuture it is apparent that the ice cover in zone 1 (western re-gion) will most likely be gone before 2020 and that ice inzone 2 (plateau) and zone 3 (southern slopes) will disappearsometime close to 2040 (Fig 5) It is likely that the plateauwill harbour the final remnants of ice on Kibo which couldremain slightly beyond the middle of the 21st century Inter-estingly our linear extrapolation for the loss of ice is in closeagreement to earlier projections using radiation modelling(Molg et al 2003) Though the loss of ice will have no hy-drological significance to lowland areas (eg Gillman 1923Kaser et al 2004 Molg et al 2008b) there is no doubt thattheir disappearance will be a ldquogrievous aesthetic lossrdquo (Salt1951 p 160) which will undoubtedly impact the experiencetourists have on the mountain (Molg et al 2008b)

4 Conclusions

With reference to the glaciers found on Mt Kenya and Kiboin East Africa C Gillman remarked ldquowe might perhapslearn a good deal as to the general world circulation fromcareful and prolonged observation on these particular sum-mitsrdquo (Mackinder 1923 p 23) This statement still holdstrue today (eg Molg et al 2006) and with the increasingawareness that glaciers on Africarsquos highest mountain can beused to extract climate information from the past it seemedtimely to revisit the mapping of the ice cover on Kibo throughtime Our motivation to do so has been to ensure the histor-ical record of these changes is of the highest possible qual-ity and in a format that should help facilitate on-going andfuture research dedicated to resolving the complexity of cli-mate variability and change in East Africa and beyond Thiswas achieved by using a consistent cartographic frameworkthat allows historical maps and new satellite imagery to beexamined The use of three-dimensional visualization tech-niques greatly enhanced our ability to establish the formerlocation of the glaciers

The revisited time series of glacial retreat on Kiliman-jaro contains eight epochs and spans almost 100 yr (1912ndash2011) The glaciers have retreated from their former extentof 1140 km2 in 1912 to 176 km2 in 2011 which representsa loss of about 85 of the ice cover over almost a 100-yr pe-riod Our estimate of the total retreat is consistent with recentassessments (Cullen et al 2006 Thompson et al 2009) butthe careful re-examination of the location and extent of indi-vidual ice bodies through time has allowed us to identify anumber of important differences from previously publishedmaps (eg Hastenrath and Greischar 1997 Thompson et al2002 2009) Historical evidence has been used to supportour interpretation with the detailed observations and pho-tographs obtained by early explorers on Kilimanjaro found tobe extremely valuable (eg Meyer 1891 1900 Uhlig 1904Jaeger 1909 Klute 1914 1920 1921 Gillman 1923 Nils-son 1931 Geilinger 1936) It has to be acknowledged that

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

430 N J Cullen et al A century of ice retreat on Kilimanjaro

much of what we know about Kilimanjaro today was estab-lished through their efforts

To further explore the importance of different glacierregimes on Kibo (Kaser et al 2004) a geo-statistical ap-proach was applied which resulted in the segmentation ofthe ice cover into three broad glacial zones A careful ex-amination of the retreat of ice in each of these glacier zonesindicates that there is a high probability that most of the re-maining ice cover on Kilimanjaro will disappear by 2040and it is highly unlikely that any ice body will remain af-ter 2060 if present retreat rates are maintained Before thishappens it would still be useful if the remaining ice bodiesin particular the remnants of the northern ice field could befurther examined to help resolve existing uncertainties aboutthe mechanisms responsible for controlling the formation ofthe vertical cliffs that characterize them (eg Winkler et al2010) In addition further efforts are required to scrutinizethe exact timing of the onset of the present glacier retreat firstreported by Hans Meyer in the late 19th century If we canrefine our understanding of the atmospheric controls behindthe most recent shift from glacier growth to decay on Kibowe will be in a stronger position to characterize the climaticcontrols of past and much larger glaciations on Africarsquos high-est mountain

Supplementary material related to this article isavailable online athttpwwwthe-cryospherenet74192013tc-7-419-2013-supplementpdf

AcknowledgementsThis research is funded by the Departmentof Geography and School of Surveying University of OtagoNew Zealand Thomas Molg was supported by the Alexander vonHumboldt Foundation The contributions made by Georg Kaserand Michael Winkler were supported by the Austrian ScienceFund (FWF) Projects P 20089-N10 and P 21288-N10 Thisresearch would not have been possible without the support from thefollowing Tanzanian authorities (COSTECH KINAPA TANAPATAWIRI) We are grateful to the referees that took the time toprovide feedback on this research in particular the careful andthoughtful comments by Henry Brecher

Edited by M Schneebeli

References

Chiang J C H The tropics in paleoclimate Annu Rev EarthPlanet Sci 37 263ndash297 2009

Cullen N J Molg T Kaser G Hussein K Steffen Kand Hardy D R Kilimanjaro Glaciers Recent areal ex-tent from satellite data and new interpretation of observed20th century retreat rates Geophys Res Lett 33 L16502doi1010292006GL027084 2006

Cullen N J Molg T Kaser G Steffen K and Hardy D REnergy-balance model validation on the top of Kilimanjaro Tan-zania using eddy covariance data Ann Glaciol 46 227ndash233doi103189172756407782871224 2007

Downie C Glaciations of Mount Kilimanjaro Northeast Tan-ganyika Geol Soc Am Bull 75 1ndash16doi1011300016-7606(1964)75[1GOMKNT]20CO2 1964

Downie C and Wilkinson P The Geology of KilimanjaroSheffield Geology Department University of Sheffield 1972

Downie C Humphries D W Wilcockson W H and WilkinsonP Geology of Kilimanjaro Nature 178 828ndash830 1956

Farr T G Rosen P A Caro E Crippen R Duren R Hens-ley S Kobrick M Paller M Rodriguez E Roth L Seal DShaffer S Shimada J Umland J Werner M Oskin M Bur-bank D and Alsdorf D The Shuttle Radar Topography Mis-sion Rev Geophys 45 RG2004doi1010292005RG0001832007

Geilinger W The retreat of the Kilimanjaro glaciers TanganyikaNotes and Records 2 7ndash20 1936

Gillman C An ascent of Kilimanjaro Geogr J 61 1ndash27 1923Hastenrath S The Glaciers of Equatorial East Africa Reidel Dor-

drecht 1984Hastenrath S Diagnosing the decaying glaciers of equatorial

East Africa Meteorol Z 15 265ndash271doi1011270941-294820060106 2006

Hastenrath S Recession of equatorial glaciers a photo documen-tation Sundog Publishing Madison Wisconsin 2008

Hastenrath S and Greischar L Glacier recession on KilimanjaroEast Africa 1912ndash89 J Glaciol 43 455ndash459 1997

Humphries D Interim report on the glaciology and meteorology ofKilimanjaro Tanganyika JulyndashSept 1953 Geological Survey ofTanganyika Dodoma 1953

Humphries D Preliminary notes on the glaciology of KilimanjaroJ Glaciol 3 475ndash479 1959

Humphries D Glaciology and glacial history in The Geology ofKilimanjaro edited by Downie C and Wilkinson P SheffieldGeology Department University of Sheffield 31ndash71 1972

Jarvis A Reuter H I Nelson A and Guevara E Hole-filledSRTM for the globe Version 4 available from the CGIAR-CSISRTM 90 m Database (httpsrtmcsicgiarorg) 2008

Jaeger F Forschungen in den Hochregionen des KilimandscharoMitt Deutsch Schutzgebiet 22 113ndash197 1909

Kaser G Hardy D R Molg T Bradley R S and Hyera TM Modern glacier retreat on Kilimanjaro as evidence of climatechange Observations and facts Int J Climatol 24 329ndash3392004

Kaser G Molg T Cullen N J Hardy D R andWinkler M Is the decline of ice on Kilimanjaro un-precedented in the Holocene Holocene 20 1079ndash1091doi1011770959683610369498 2010

Klute F Forschungen am Kilimandscharo im Jahre 1912 GeogrZeitschrift 20 496ndash505 1914

Klute F Ergebnisse der Forschungen am Kilimandscharo 1912ReimerndashVohsen Berlin 1920

Klute F Die stereophotogrammetrische Aufnahme der Hochregio-nen des Kilimandscharo Zeitschrift Ges Erdkunde Berlin 56144ndash151 1921

Mackinder C W Hobley C Gillman C and Johnston H Anascent of Kilimanjaro Discussion Geogr J 61 1ndash27 1923

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

424 N J Cullen et al A century of ice retreat on Kilimanjaro

4800

4600

5000

4400

5200

5400

5600

5700

4200

4000

5800

4800

4400

4600 4200

37deg23E

37deg23E

37deg22E

37deg22E

37deg21E

37deg21E

37deg20E

37deg20E

3deg3

S

3deg3

S

3deg4

S

3deg4

S

3deg5

S

3deg5

S

3deg6

S

3deg6

S

314000

314000

316000

316000

318000

318000

320000

320000

9658000

9658000

9660000

9660000

9662000

9662000

Uhuru Peak

Northern ice field

Heim

Uhlig

Dec

ken

Ratzel

Kersten

Reb

man

n

Credner

Drygalski

Arrow

Diamond

Balletto

Furtwaumlngler

LittlePenck

5895m

IceCastle

Easternice field

Southernice field

WGS84UTM 37S

0 1000 2000

Meters

Revised 1912 boundaries

Credner Glacier name

2011 Ice boundaries

1912 boundaries fromHastenrath and Greischar (1997)

Little Barranco

Fig 1 Glacier extent on Kilimanjaro in 1912 and 2011 Glacier names with an asterisk do not appear in Klute (1920) and are described inthe text The interpretation of the glacier extent for 1912 as defined by Hastenrath and Greischar (1997) is shown as a dashed line (blue) forcomparative purposes The nomenclature for each glacier used by Klute (1920) is given while the names of four glaciers introduced afterthis time are indicated usinglowast

accounts are more supportive (Gillman 1923 Geilinger1936) A more detailed map of Kilimanjarorsquos glaciers pro-duced by Meyer (1900) based on his observations during afield expedition in 1898 shows the connection no longer ex-isted suggesting a break-up occurred sometime in the early1890s Though the timing is uncertain the implications oflosing this connection are not overlooked by Geilinger (1936p 12) who refers to the Little and Great Barranco glaciers asldquoindependent dead glaciersrdquo and made the astute observationthat both ice entities were retreating at their upper and lowermargins

This observation holds some importance as it helps clar-ify the source of the ice remnants still observed today Themaps by Meyer (1900 map dated 1898) Jaeger (1909map dated 1906) and Klute (1920 map dated 1912) clearlyshow distinct ice streams for the Little and Great Barrancoglaciers which as noted may have been fed by ice from theFurtwangler glacier as recently as the late 19th century Ourrevision of the Klute (1920) map from 1912 (Fig 1) honoursthe discrete boundary between the Little and Great Barranco

glaciers which is not as clearly depicted in the Hastenrathand Greischar (1997) interpretation The failure to clearlydistinguish the boundaries and location of these two indepen-dent ice streams has been inherited by Thompson et al (20022009) who used the Hastenrath and Greischar (1997) inter-pretation of the ice extent prior to 2000 (see Sect 231)Thompson et al (2009) argue that ice remnants identified byCullen et al (2006) in both the ice streams of the former andmore extensive Little and Great Barranco glaciers are absentfrom previous observations and therefore most likely tran-sient snow The sketch by Humphries (1972 Fig 46) andthe map produced by Messerli (1980 Fig 6) also support theinterpretation by Cullen et al (2006) while the ice bodiesin question have been identified in each of the satellite im-ages used in the present analysis (Fig 2) Though not labelledby Humphries (1959 1972) or by those preceding his obser-vations the Arrow glacier has been used to refer to one ofthe Little Barranco ice remnants (Fig 1) It appears that thisterm was adopted by climbers on the mountain (eg Samp-son 1971 p 202) but the exact location of this ice body is

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 425

Table 2Areal extents (km2) of the total ice bodies between 1912 and 2011 as well as for zone 1 (western region) zone 2 (plateau region)and zone 3 (southern region) as defined in the text and Fig 4 The annual rates of area change per observation period (10minus2 km2 yrminus1) areshown in parentheses for the total and zoned areas

Epoch Duration Area [km2] (rate of area change [10minus2 km2 yrminus1])

[years] Total Zone 1 Zone 2 Zone 3

191283 11400 4540 4293 2567196216 4936 7320 (minus827) 2520 (minus409) 3161 (minus229) 1639 (minus188)197562 1347 6052 (minus941) 2060 (minus342) 2529 (minus469) 1463 (minus131)198448 887 4815 (minus1395) 1628 (minus487) 1949 (minus654) 1238 (minus254)199300 853 3800 (minus1190) 1142 (minus570) 1638 (minus365) 1020 (minus256)200014 715 2920 (minus1232) 0701 (minus617) 1411 (minus318) 0808 (minus297)200308 295 2500 (minus1425) 0539 (minus550) 1251 (minus543) 0710 (minus332)201146 838 1762 (minus881) 0333 (minus246) 0937 (minus375) 0492 (minus260)

not well documented and we are uncertain that the remnantidentified in Fig 1 and by Cullen et al (2006) is the sameas that referred to by others (eg Sampson 1971 Messerli1980 Hastenrath and Greischar 1997)

There has also been some confusion about the Diamondand Balletto glaciers (Fig 1) which were not formally rec-ognized in the early mapping work (Meyer 1990 Jaeger1909 Klute 1920) or acknowledged by Humphries (19591972) but are both described by Hastenrath (1984) whileSampson (1971) and Messerli (1980) only refer to the for-mer The Breach Wall area where both glaciers are locatedwas mapped by Klute (1920) but as no break up from thesouthern ice field had occurred at this time they were notgiven discrete labels (Fig 1) However what is indisputableis the presence of an ice field (the Balletto glacier) beneaththe headwall of the Diamond glacier which was not mappedby Thompson et al (2002 2009) The presence of this icefield is not only observed in the satellite record but is docu-mented by Reinhold Messner (Messner 1991 Chapter 21)who made the first (and only) direct ascent of the Breach Wallof Kilimanjaro in 1978 Though Cullen et al (2006) mappedthis region of ice (the Balletto glacier) they failed to label itcorrectly placing the label beneath the Heim glacier after fol-lowing the interpretation of Hastenrath and Greischar (1997Fig 2) Further the Diamond glacier is absent after 1976 inthe maps produced by Hastenrath and Greischar (1997) andHastenrath (2006) but appears in the Thompson et al (20022009) maps despite using the former to build their time se-ries Finally the Uhlig glacier which Jaeger (1909 p 130)named after his friend C Uhlig does not appear to have beenconnected to the northern ice field in the past (Fig 1) Whilethe upper boundary of the Uhlig glacier is difficult to inter-pret in the Klute (1920) map it is clearly shown by Jaeger(1909) that it was not joined to the northern ice field

Figure 2 shows the change in ice extent over almost a 100-yr period which is consistent through time and removes mostof the uncertainty about the location and extent of the re-maining ice bodies on Kibo The retreat over the first part of

the 20th century was approximately 41 km2 (Fig 2a and b)while over the last 50 yr 56 km2 of the ice cover has beenlost (Fig 2cndashh) The retreat between 1912 and 1962 (Fig 2aand b) resulted in a distinct break-up of ice in the vicinity ofthe notches (south-east crater region) separating the northernand eastern ice fields from the southern ice field Remnantsof the Ice Castle (Fig 1) still remained in 1962 and were de-scribed by Sampson (1971) as still being present in the early1970s but appeared to have been lost by 1975 (Fig 2c) Othernoticeable features of the retreat in the early 20th century in-clude the development of a distinct separation between theGreat and Little Penck glaciers which formed through theexpansion of the Ravenstein (first mapped by Meyer 1900)the rapid demise of the Uhlig glacier and the split-up of theDiamond and Balletto glaciers from the southern ice field(Fig 2a and b)

Between 1962 and 1975 the persistent retreat of ice led tothe northern and eastern ice fields becoming separated fromeach other as did the Great Penck glacier from the north-ern ice field (Fig 2b and c) The Little Penck proved moreresistant and did not separate from the northern ice field un-til after 1992 while the remnant of the Great Penck com-pletely disappeared during this period (Fig 2e and f) thoughground-based (field) observations in January 2005 revealeddebris-covered ice at the location of the lowest 1984 extent(Fig 2d) As noted by others (eg Hastenrath and Greischar1997 Cullen et al 2006) the northern ice field developed ahole sometime after 1984 (Fig 2d and e) which opened intoa large canyon after 2003 (Fig 2g and h) Field observationsby the first author in October 2012 revealed that the north-ern ice field has separated into two independent ice bodieswhich must have occurred sometime after 17 June 2011 TheFurtwangler glacier has also separated into smaller ice enti-ties which began after 2003 (Fig 2g and h) The southernice field has not escaped the on-going demise with the upperpart of the Kersten glacier separating from its lower reachesbetween 2003 and 2011 (Fig 2g and h) which was some-what anticipated after mass balance modelling showed this

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

426 N J Cullen et al A century of ice retreat on Kilimanjaro

(g) 2003 (h) 2011

(a) 1912 (b) 1962

(c) 1975 (d) 1984

(e) 1992 (f) 2000

Credner

Great Penck

Ratzel

Ker

sten

Northern ice field

Easternice field

Southernice field

GreatBarranco

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Fig 2 Ice retreat over Kilimanjaro based on the revised areal extent for the following years(a) 1912(b) 1962(c) 1975(d) 1984(e)1992(f) 2000(g) 2003 and(h) 2011 An interactive 3-D illustration of the sequence can be sourced in the Supplement

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 427

was likely to occur (Molg et al 2009b) The upper part ofthe Heim has disappeared and the Decken has narrowed con-siderably over the last 8 yr (Fig 2g and h)

32 New interpretation of the ice retreat

A consistent record of the former extent of the glaciers onKibo provides a platform from which the relationship be-tween glaciological and meteorological processes can be fur-ther explored We believe the spatial and temporal variabilityof the ice cover on Kibo can be used to unravel the physi-cal processes responsible for controlling the decay of the icemasses through time as well as providing insights into themechanisms that may have been responsible for growth in thepast Interestingly those who were privileged to make someof the first (formal) observations on Kilimanjaro recognizedthere was an asymmetry in the ice cover (eg Meyer 1891)which pointed to specific climatological drivers controllingglacial retreat (eg Gillman 1923 Nilsson 1931 Geilinger1936) Meyer (1891 p 308ndash314) noted that ldquoan immensemountain mass like Kilimanjaro must considerably modifythe prevailing wind directionrdquo before making the point thatthe primary cause in the variability of the ice limit is ldquotheunequal distribution of moisture on the northern and south-ern sides of the mountainrdquo Though Meyer (1891) recognizedvariability in moisture was the most important control healso speculated that the asymmetry of low reaching glacierson Kibo in the south south-west and western sectors of themountain a pattern that persisted throughout the 20th cen-tury (Fig 2) depended on ldquolocal peculiarities which favourthe preservation of the snow [rather] than on physical condi-tions that favour a large snowfallrdquo (Meyer 1891 p 315) Torevisit this early observation it is useful to reflect on the hyp-sometry of the glaciers in 1912 (Fig 3a) which shows thatthe probability density in relation to the distribution of the icewas greatest above 5700 m This high density reflects the icecover in the crater region with the majority of the ice in 1912being part of the former ice cap that linked the northern andeastern ice fields (Fig 1) Figure 3b and c show that slopeand aspect were also important controls on the density of icecover with the steep south and south-west facing slopes morefavourable which enabled the glaciers to extend to elevationsbelow 4600 m For the glaciers to have extended to the sameelevation on the east-facing slopes they would have had tocover a much greater horizontal distance (Fig 1)

To fully account for the distinct geometry of the ice masseson Kibo and to establish the physical processes controllingtheir location and extent through time Kaser et al (2004)established a framework that included four separate glacierregimes The motivation to do this was partly based on theearly recognition that the tabular shaped glaciers in the craterregion of Kibo behave differently to glaciers on the flanks ofthe mountain (eg Nilsson 1931 Geilinger 1936) Guidedby this distinction Cullen et al (2006) used the 5700 m con-tour which is roughly the height of the outer crater rim to

Density times 10 minus3

2

4

6

8

WEST EAST

SOUTH

NORTH

4600 4800 5000 5200 5400 5600 58000

1

2

3

4

Elevation [m]

Den

sity

times 10

minus3

0 10 20 30 40 50 60 70 800

1

2

3

4

Slope [deg]

Den

sity

times 10

minus2

(a)

(c)

(b)

Fig 3 Distribution of topographic measures from the areal extentof all ice bodies in 1912(a) elevation(b) slope and(c) aspect Themagnitude of the normalised distribution on the y-axis is a proba-bility density function with the integral equal to unity

distinguish between two broad glacial systems (1) plateau(ge 5700 m) and slope (lt 5700 m) glaciers This simple bi-nary classification has proved useful in describing some ofthe key physical processes responsible for the retreat of iceon Kibo (eg Molg et al 2009b Kaser et al 2010) Tofurther characterize the spatial and temporal variability ofglacier retreat on Kibo we have adopted a geo-statisticalapproach in this study that uses elevation slope and aspectof the ice cover in 1912 (the feature space) to identify clus-ters resembling topographical classes This was achieved byperforming a cluster analysis containing maximum likeli-hood estimates of the topographical parameters in a Gaussian

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

428 N J Cullen et al A century of ice retreat on Kilimanjaro

mixture model (McLachlan and Peel 2000) which resultedin three broad glacial zones being identified (1) western re-gion (zone 1) (2) plateau region (zone 2) and (3) southernregion (zone 3) (Fig 4)

Figure 5 shows the rate of retreat of each of the glacialzones through time as well as a least-squares linear regres-sion fit (and 80 confidence intervals) through each of theglacial zone time series As described by Cullen et al (2006)the retreat of the tabular shaped plateau glaciers (zone 2) iscontrolled primarily by solar radiation-induced ablation onthe vertical cliffs that characterize them as demonstratedthrough radiation modelling (Molg et al 2003) Winkleret al (2010) revealed that direct insolation is the reasonfor the predominant zonal alignment of the ice cliffs andtheir bimodal annual recession (sunlit versus shaded peri-ods) Though the shrinkage of the ice cover along the verticalcliffs is fairly constant from year to year the areal retreat ofthe zone 2 region is not entirely linear with the highest ratesoccurring between 1975 and 1984 (Table 2) which corre-spond to the period following the separation of the northernand eastern ice fields (Fig 2) Therefore changes in the ge-ometry of the ice bodies and their exposure to solar radiationappear to have led to variations in retreat rates (Table 2 andFig 5) which has also been described by Kaser et al (2010)

An explanation for the growth and decay of the plateauglaciers including a description of the physical processesthat may have led to the observed steps in the northern icefield has been provided by Kaser et al (2010) To sustaingrowth in the crater rim region (zone 2) it is critical thatsnowfall is sufficient to allow a continuous snow cover todevelop that is able to survive from one year to the next Ifthe annual build-up of snow is sustained through a period oftime it is quite conceivable that the plateau glaciers couldgrow to several tens of metres in thickness quite quicklyraising the question as to whether this accumulation of icehas the potential to become a source of mass for glacierson the western flanks of the mountain (zone 1) Meyer(1891 p 316ndash317) was somewhat sceptical as to whetherthe plateau region acted as a ldquoreservoirrdquo and thought it ldquomuchmore reasonable to suppose that the rocky rim of the crateris the real reservoirrdquo However Meyer (1890 1891) did pro-vide an account and sketches to support a connection of theplateau ice to the Little and Great Barranco glaciers in thepast which as described may have been important in sustain-ing them (Gillman 1923 Geilinger 1936) There is someimportance in investigating the extent of the linkages be-tween these two glacial zones as together they accounted foralmost 80 of the total ice cover in 1912 (Table 2)

The remaining ice bodies in zone 2 (plateau region) aremostly stagnant today with field measurements and prelim-inary modelling results indicating deformation rates of icewithin the crater are small (lt 10 cm yrminus1) (Kaser et al 2004Winkler et al 2010) It seems unlikely that the present or his-torical thickness of the plateau ice has been sufficient to al-low substantial deformation (Kaser et al 2004) suggesting

3140 3150 3160 3170 3180 3190 3200 32109657

9658

9659

9660

9661

9662

9663

9664

4000

4000

4500

4500 4500

45004500

5000

5000

5000

5000

5000

5500

5500

5500

UTM easting in meters times105

UT

M n

orth

ing

in m

eter

s times

106

Fig 4 Classification of regions based on the distribution of topo-graphic measures in 1912 Zone 1 (western region) zone 2 (plateauregion) and zone 3 (southern region) are shown in red orange andyellow respectively

that mass transport from the nearly horizontal plateau downinto the western slope glaciers (zone 1) has most likely beenminor Careful observations by Klute (1920) demonstratedthat the velocity of the centre of the Great Penck glacier was064 m over a 30-day period in 1912 which led him to spec-ulate that the annual downslope movement was about 7 mper year It should be noted that Nilsson (1931) and Osmas-ton (1989) also report on these observations but incorrectlycite Jaeger (1909) as the person responsible for obtaining themeasurements Nonetheless it is improbable that ice fromthe plateau region could have supported the observed flowdynamics of the Great Penck glacier with the upper reachesof the mountain flanks the most likely source of the massThus the growth and decay of the glaciers in zone 1 (west-ern region) do not appear to be directly dependent on thestability of plateau ice (zone 2) To further resolve the phys-ical processes controlling the stability of the plateau ice itwould be of interest to develop a better understanding of themechanisms responsible for vertical cliff development

The glaciers on the southern slopes (zone 3) differ fromthose on the western slopes (zone 1) as they are separatedfrom the large masses of ice within the caldera (zone 2) bybed topography (Humphries 1972) This distinction is im-portant as any assessment of the atmospheric controls onthe behaviour of the southern ice fields is independent ofthe complexity of the vertical cliff retreat in the plateau re-gion (zone 2) That said the cluster analysis that led to thetopographical classes used in this study shows that the up-per sections of the southern ice field exhibit similar phys-ical characteristics to the ice within the caldera (Fig 4)Nevertheless to obtain a full solution for the asymmetry ofthe glaciers on Kibo that the early explorers observed itis critical to establish an understanding of the atmospheric

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 429

1900 1920 1940 1960 1980 2000 2020 2040 20600

05

1

15

2

25

3

35

4

45

5

Year

Are

al e

xten

t [km

2 ]

zone 1 R2=099

zone 2 R2=097

zone 3 R2=09980 confidence interval

4000

4000

4000

4500

4500

4500

45005000

5000

5000

50005000

5500

5500

5500

Fig 5 The areal change of glaciers and ice bodies based on thethree-member classification Linear regressions (thin solid lines) foreach of the three zones are shown as well as the 80 confidenceintervals for each of these relationships

processes controlling the growth andor decay of the southernice fields This has been achieved by Molg et al (2009b)who used observations and atmospherendashglacier modellingto show how critical precipitation humidity and fractionalcloud cover are on the energy and mass balance of the Ker-sten glacier (zone 3) These results compare favourably toearly accounts of the importance of precipitation (moisturechange) and cloud cover on the asymmetry (Meyer 1891Klute 1920 Gillman 1923 Geilinger 1936) with the diur-nal variability of cloud cover being especially important inproviding protection to the glaciers on the western side ofthe mountain (eg Meyer 1891 Hastenrath and Greischar1997)

Recent regional atmospheric modelling experiments haveprovided a more quantitative assessment to account for theasymmetry in glacier mass on Kibo demonstrating that thewest and south-western areas of the mountain are favourableto moisture convergence as controlled by atmospheric dy-namics (Molg et al 2009a) As noted by Osmaston (1989)the asymmetry is not confined to the most recent glacier ex-tents with the pattern of ice distribution established fromformer moraines indicating that recent climatic controls onglacier behaviour are similar to those in the past This sug-gests that establishing the specific nature of the atmosphericcontrols responsible for the most recent shift from glaciergrowth to decay in particular in zones 2 and 3 (plateau andsouthern slope regions) remains critical if we are to useglacier behaviour to reconstruct climatic conditions of thedeeper past

Finally it is difficult not to speculate on when the ice onKibo will disappear which becomes more achievable if onerecognizes the importance of the different glacial zones onthe mountain If one accepts linear extrapolation based on

changes from the past is suitable to make projections for thefuture it is apparent that the ice cover in zone 1 (western re-gion) will most likely be gone before 2020 and that ice inzone 2 (plateau) and zone 3 (southern slopes) will disappearsometime close to 2040 (Fig 5) It is likely that the plateauwill harbour the final remnants of ice on Kibo which couldremain slightly beyond the middle of the 21st century Inter-estingly our linear extrapolation for the loss of ice is in closeagreement to earlier projections using radiation modelling(Molg et al 2003) Though the loss of ice will have no hy-drological significance to lowland areas (eg Gillman 1923Kaser et al 2004 Molg et al 2008b) there is no doubt thattheir disappearance will be a ldquogrievous aesthetic lossrdquo (Salt1951 p 160) which will undoubtedly impact the experiencetourists have on the mountain (Molg et al 2008b)

4 Conclusions

With reference to the glaciers found on Mt Kenya and Kiboin East Africa C Gillman remarked ldquowe might perhapslearn a good deal as to the general world circulation fromcareful and prolonged observation on these particular sum-mitsrdquo (Mackinder 1923 p 23) This statement still holdstrue today (eg Molg et al 2006) and with the increasingawareness that glaciers on Africarsquos highest mountain can beused to extract climate information from the past it seemedtimely to revisit the mapping of the ice cover on Kibo throughtime Our motivation to do so has been to ensure the histor-ical record of these changes is of the highest possible qual-ity and in a format that should help facilitate on-going andfuture research dedicated to resolving the complexity of cli-mate variability and change in East Africa and beyond Thiswas achieved by using a consistent cartographic frameworkthat allows historical maps and new satellite imagery to beexamined The use of three-dimensional visualization tech-niques greatly enhanced our ability to establish the formerlocation of the glaciers

The revisited time series of glacial retreat on Kiliman-jaro contains eight epochs and spans almost 100 yr (1912ndash2011) The glaciers have retreated from their former extentof 1140 km2 in 1912 to 176 km2 in 2011 which representsa loss of about 85 of the ice cover over almost a 100-yr pe-riod Our estimate of the total retreat is consistent with recentassessments (Cullen et al 2006 Thompson et al 2009) butthe careful re-examination of the location and extent of indi-vidual ice bodies through time has allowed us to identify anumber of important differences from previously publishedmaps (eg Hastenrath and Greischar 1997 Thompson et al2002 2009) Historical evidence has been used to supportour interpretation with the detailed observations and pho-tographs obtained by early explorers on Kilimanjaro found tobe extremely valuable (eg Meyer 1891 1900 Uhlig 1904Jaeger 1909 Klute 1914 1920 1921 Gillman 1923 Nils-son 1931 Geilinger 1936) It has to be acknowledged that

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

430 N J Cullen et al A century of ice retreat on Kilimanjaro

much of what we know about Kilimanjaro today was estab-lished through their efforts

To further explore the importance of different glacierregimes on Kibo (Kaser et al 2004) a geo-statistical ap-proach was applied which resulted in the segmentation ofthe ice cover into three broad glacial zones A careful ex-amination of the retreat of ice in each of these glacier zonesindicates that there is a high probability that most of the re-maining ice cover on Kilimanjaro will disappear by 2040and it is highly unlikely that any ice body will remain af-ter 2060 if present retreat rates are maintained Before thishappens it would still be useful if the remaining ice bodiesin particular the remnants of the northern ice field could befurther examined to help resolve existing uncertainties aboutthe mechanisms responsible for controlling the formation ofthe vertical cliffs that characterize them (eg Winkler et al2010) In addition further efforts are required to scrutinizethe exact timing of the onset of the present glacier retreat firstreported by Hans Meyer in the late 19th century If we canrefine our understanding of the atmospheric controls behindthe most recent shift from glacier growth to decay on Kibowe will be in a stronger position to characterize the climaticcontrols of past and much larger glaciations on Africarsquos high-est mountain

Supplementary material related to this article isavailable online athttpwwwthe-cryospherenet74192013tc-7-419-2013-supplementpdf

AcknowledgementsThis research is funded by the Departmentof Geography and School of Surveying University of OtagoNew Zealand Thomas Molg was supported by the Alexander vonHumboldt Foundation The contributions made by Georg Kaserand Michael Winkler were supported by the Austrian ScienceFund (FWF) Projects P 20089-N10 and P 21288-N10 Thisresearch would not have been possible without the support from thefollowing Tanzanian authorities (COSTECH KINAPA TANAPATAWIRI) We are grateful to the referees that took the time toprovide feedback on this research in particular the careful andthoughtful comments by Henry Brecher

Edited by M Schneebeli

References

Chiang J C H The tropics in paleoclimate Annu Rev EarthPlanet Sci 37 263ndash297 2009

Cullen N J Molg T Kaser G Hussein K Steffen Kand Hardy D R Kilimanjaro Glaciers Recent areal ex-tent from satellite data and new interpretation of observed20th century retreat rates Geophys Res Lett 33 L16502doi1010292006GL027084 2006

Cullen N J Molg T Kaser G Steffen K and Hardy D REnergy-balance model validation on the top of Kilimanjaro Tan-zania using eddy covariance data Ann Glaciol 46 227ndash233doi103189172756407782871224 2007

Downie C Glaciations of Mount Kilimanjaro Northeast Tan-ganyika Geol Soc Am Bull 75 1ndash16doi1011300016-7606(1964)75[1GOMKNT]20CO2 1964

Downie C and Wilkinson P The Geology of KilimanjaroSheffield Geology Department University of Sheffield 1972

Downie C Humphries D W Wilcockson W H and WilkinsonP Geology of Kilimanjaro Nature 178 828ndash830 1956

Farr T G Rosen P A Caro E Crippen R Duren R Hens-ley S Kobrick M Paller M Rodriguez E Roth L Seal DShaffer S Shimada J Umland J Werner M Oskin M Bur-bank D and Alsdorf D The Shuttle Radar Topography Mis-sion Rev Geophys 45 RG2004doi1010292005RG0001832007

Geilinger W The retreat of the Kilimanjaro glaciers TanganyikaNotes and Records 2 7ndash20 1936

Gillman C An ascent of Kilimanjaro Geogr J 61 1ndash27 1923Hastenrath S The Glaciers of Equatorial East Africa Reidel Dor-

drecht 1984Hastenrath S Diagnosing the decaying glaciers of equatorial

East Africa Meteorol Z 15 265ndash271doi1011270941-294820060106 2006

Hastenrath S Recession of equatorial glaciers a photo documen-tation Sundog Publishing Madison Wisconsin 2008

Hastenrath S and Greischar L Glacier recession on KilimanjaroEast Africa 1912ndash89 J Glaciol 43 455ndash459 1997

Humphries D Interim report on the glaciology and meteorology ofKilimanjaro Tanganyika JulyndashSept 1953 Geological Survey ofTanganyika Dodoma 1953

Humphries D Preliminary notes on the glaciology of KilimanjaroJ Glaciol 3 475ndash479 1959

Humphries D Glaciology and glacial history in The Geology ofKilimanjaro edited by Downie C and Wilkinson P SheffieldGeology Department University of Sheffield 31ndash71 1972

Jarvis A Reuter H I Nelson A and Guevara E Hole-filledSRTM for the globe Version 4 available from the CGIAR-CSISRTM 90 m Database (httpsrtmcsicgiarorg) 2008

Jaeger F Forschungen in den Hochregionen des KilimandscharoMitt Deutsch Schutzgebiet 22 113ndash197 1909

Kaser G Hardy D R Molg T Bradley R S and Hyera TM Modern glacier retreat on Kilimanjaro as evidence of climatechange Observations and facts Int J Climatol 24 329ndash3392004

Kaser G Molg T Cullen N J Hardy D R andWinkler M Is the decline of ice on Kilimanjaro un-precedented in the Holocene Holocene 20 1079ndash1091doi1011770959683610369498 2010

Klute F Forschungen am Kilimandscharo im Jahre 1912 GeogrZeitschrift 20 496ndash505 1914

Klute F Ergebnisse der Forschungen am Kilimandscharo 1912ReimerndashVohsen Berlin 1920

Klute F Die stereophotogrammetrische Aufnahme der Hochregio-nen des Kilimandscharo Zeitschrift Ges Erdkunde Berlin 56144ndash151 1921

Mackinder C W Hobley C Gillman C and Johnston H Anascent of Kilimanjaro Discussion Geogr J 61 1ndash27 1923

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 425

Table 2Areal extents (km2) of the total ice bodies between 1912 and 2011 as well as for zone 1 (western region) zone 2 (plateau region)and zone 3 (southern region) as defined in the text and Fig 4 The annual rates of area change per observation period (10minus2 km2 yrminus1) areshown in parentheses for the total and zoned areas

Epoch Duration Area [km2] (rate of area change [10minus2 km2 yrminus1])

[years] Total Zone 1 Zone 2 Zone 3

191283 11400 4540 4293 2567196216 4936 7320 (minus827) 2520 (minus409) 3161 (minus229) 1639 (minus188)197562 1347 6052 (minus941) 2060 (minus342) 2529 (minus469) 1463 (minus131)198448 887 4815 (minus1395) 1628 (minus487) 1949 (minus654) 1238 (minus254)199300 853 3800 (minus1190) 1142 (minus570) 1638 (minus365) 1020 (minus256)200014 715 2920 (minus1232) 0701 (minus617) 1411 (minus318) 0808 (minus297)200308 295 2500 (minus1425) 0539 (minus550) 1251 (minus543) 0710 (minus332)201146 838 1762 (minus881) 0333 (minus246) 0937 (minus375) 0492 (minus260)

not well documented and we are uncertain that the remnantidentified in Fig 1 and by Cullen et al (2006) is the sameas that referred to by others (eg Sampson 1971 Messerli1980 Hastenrath and Greischar 1997)

There has also been some confusion about the Diamondand Balletto glaciers (Fig 1) which were not formally rec-ognized in the early mapping work (Meyer 1990 Jaeger1909 Klute 1920) or acknowledged by Humphries (19591972) but are both described by Hastenrath (1984) whileSampson (1971) and Messerli (1980) only refer to the for-mer The Breach Wall area where both glaciers are locatedwas mapped by Klute (1920) but as no break up from thesouthern ice field had occurred at this time they were notgiven discrete labels (Fig 1) However what is indisputableis the presence of an ice field (the Balletto glacier) beneaththe headwall of the Diamond glacier which was not mappedby Thompson et al (2002 2009) The presence of this icefield is not only observed in the satellite record but is docu-mented by Reinhold Messner (Messner 1991 Chapter 21)who made the first (and only) direct ascent of the Breach Wallof Kilimanjaro in 1978 Though Cullen et al (2006) mappedthis region of ice (the Balletto glacier) they failed to label itcorrectly placing the label beneath the Heim glacier after fol-lowing the interpretation of Hastenrath and Greischar (1997Fig 2) Further the Diamond glacier is absent after 1976 inthe maps produced by Hastenrath and Greischar (1997) andHastenrath (2006) but appears in the Thompson et al (20022009) maps despite using the former to build their time se-ries Finally the Uhlig glacier which Jaeger (1909 p 130)named after his friend C Uhlig does not appear to have beenconnected to the northern ice field in the past (Fig 1) Whilethe upper boundary of the Uhlig glacier is difficult to inter-pret in the Klute (1920) map it is clearly shown by Jaeger(1909) that it was not joined to the northern ice field

Figure 2 shows the change in ice extent over almost a 100-yr period which is consistent through time and removes mostof the uncertainty about the location and extent of the re-maining ice bodies on Kibo The retreat over the first part of

the 20th century was approximately 41 km2 (Fig 2a and b)while over the last 50 yr 56 km2 of the ice cover has beenlost (Fig 2cndashh) The retreat between 1912 and 1962 (Fig 2aand b) resulted in a distinct break-up of ice in the vicinity ofthe notches (south-east crater region) separating the northernand eastern ice fields from the southern ice field Remnantsof the Ice Castle (Fig 1) still remained in 1962 and were de-scribed by Sampson (1971) as still being present in the early1970s but appeared to have been lost by 1975 (Fig 2c) Othernoticeable features of the retreat in the early 20th century in-clude the development of a distinct separation between theGreat and Little Penck glaciers which formed through theexpansion of the Ravenstein (first mapped by Meyer 1900)the rapid demise of the Uhlig glacier and the split-up of theDiamond and Balletto glaciers from the southern ice field(Fig 2a and b)

Between 1962 and 1975 the persistent retreat of ice led tothe northern and eastern ice fields becoming separated fromeach other as did the Great Penck glacier from the north-ern ice field (Fig 2b and c) The Little Penck proved moreresistant and did not separate from the northern ice field un-til after 1992 while the remnant of the Great Penck com-pletely disappeared during this period (Fig 2e and f) thoughground-based (field) observations in January 2005 revealeddebris-covered ice at the location of the lowest 1984 extent(Fig 2d) As noted by others (eg Hastenrath and Greischar1997 Cullen et al 2006) the northern ice field developed ahole sometime after 1984 (Fig 2d and e) which opened intoa large canyon after 2003 (Fig 2g and h) Field observationsby the first author in October 2012 revealed that the north-ern ice field has separated into two independent ice bodieswhich must have occurred sometime after 17 June 2011 TheFurtwangler glacier has also separated into smaller ice enti-ties which began after 2003 (Fig 2g and h) The southernice field has not escaped the on-going demise with the upperpart of the Kersten glacier separating from its lower reachesbetween 2003 and 2011 (Fig 2g and h) which was some-what anticipated after mass balance modelling showed this

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

426 N J Cullen et al A century of ice retreat on Kilimanjaro

(g) 2003 (h) 2011

(a) 1912 (b) 1962

(c) 1975 (d) 1984

(e) 1992 (f) 2000

Credner

Great Penck

Ratzel

Ker

sten

Northern ice field

Easternice field

Southernice field

GreatBarranco

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Fig 2 Ice retreat over Kilimanjaro based on the revised areal extent for the following years(a) 1912(b) 1962(c) 1975(d) 1984(e)1992(f) 2000(g) 2003 and(h) 2011 An interactive 3-D illustration of the sequence can be sourced in the Supplement

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 427

was likely to occur (Molg et al 2009b) The upper part ofthe Heim has disappeared and the Decken has narrowed con-siderably over the last 8 yr (Fig 2g and h)

32 New interpretation of the ice retreat

A consistent record of the former extent of the glaciers onKibo provides a platform from which the relationship be-tween glaciological and meteorological processes can be fur-ther explored We believe the spatial and temporal variabilityof the ice cover on Kibo can be used to unravel the physi-cal processes responsible for controlling the decay of the icemasses through time as well as providing insights into themechanisms that may have been responsible for growth in thepast Interestingly those who were privileged to make someof the first (formal) observations on Kilimanjaro recognizedthere was an asymmetry in the ice cover (eg Meyer 1891)which pointed to specific climatological drivers controllingglacial retreat (eg Gillman 1923 Nilsson 1931 Geilinger1936) Meyer (1891 p 308ndash314) noted that ldquoan immensemountain mass like Kilimanjaro must considerably modifythe prevailing wind directionrdquo before making the point thatthe primary cause in the variability of the ice limit is ldquotheunequal distribution of moisture on the northern and south-ern sides of the mountainrdquo Though Meyer (1891) recognizedvariability in moisture was the most important control healso speculated that the asymmetry of low reaching glacierson Kibo in the south south-west and western sectors of themountain a pattern that persisted throughout the 20th cen-tury (Fig 2) depended on ldquolocal peculiarities which favourthe preservation of the snow [rather] than on physical condi-tions that favour a large snowfallrdquo (Meyer 1891 p 315) Torevisit this early observation it is useful to reflect on the hyp-sometry of the glaciers in 1912 (Fig 3a) which shows thatthe probability density in relation to the distribution of the icewas greatest above 5700 m This high density reflects the icecover in the crater region with the majority of the ice in 1912being part of the former ice cap that linked the northern andeastern ice fields (Fig 1) Figure 3b and c show that slopeand aspect were also important controls on the density of icecover with the steep south and south-west facing slopes morefavourable which enabled the glaciers to extend to elevationsbelow 4600 m For the glaciers to have extended to the sameelevation on the east-facing slopes they would have had tocover a much greater horizontal distance (Fig 1)

To fully account for the distinct geometry of the ice masseson Kibo and to establish the physical processes controllingtheir location and extent through time Kaser et al (2004)established a framework that included four separate glacierregimes The motivation to do this was partly based on theearly recognition that the tabular shaped glaciers in the craterregion of Kibo behave differently to glaciers on the flanks ofthe mountain (eg Nilsson 1931 Geilinger 1936) Guidedby this distinction Cullen et al (2006) used the 5700 m con-tour which is roughly the height of the outer crater rim to

Density times 10 minus3

2

4

6

8

WEST EAST

SOUTH

NORTH

4600 4800 5000 5200 5400 5600 58000

1

2

3

4

Elevation [m]

Den

sity

times 10

minus3

0 10 20 30 40 50 60 70 800

1

2

3

4

Slope [deg]

Den

sity

times 10

minus2

(a)

(c)

(b)

Fig 3 Distribution of topographic measures from the areal extentof all ice bodies in 1912(a) elevation(b) slope and(c) aspect Themagnitude of the normalised distribution on the y-axis is a proba-bility density function with the integral equal to unity

distinguish between two broad glacial systems (1) plateau(ge 5700 m) and slope (lt 5700 m) glaciers This simple bi-nary classification has proved useful in describing some ofthe key physical processes responsible for the retreat of iceon Kibo (eg Molg et al 2009b Kaser et al 2010) Tofurther characterize the spatial and temporal variability ofglacier retreat on Kibo we have adopted a geo-statisticalapproach in this study that uses elevation slope and aspectof the ice cover in 1912 (the feature space) to identify clus-ters resembling topographical classes This was achieved byperforming a cluster analysis containing maximum likeli-hood estimates of the topographical parameters in a Gaussian

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

428 N J Cullen et al A century of ice retreat on Kilimanjaro

mixture model (McLachlan and Peel 2000) which resultedin three broad glacial zones being identified (1) western re-gion (zone 1) (2) plateau region (zone 2) and (3) southernregion (zone 3) (Fig 4)

Figure 5 shows the rate of retreat of each of the glacialzones through time as well as a least-squares linear regres-sion fit (and 80 confidence intervals) through each of theglacial zone time series As described by Cullen et al (2006)the retreat of the tabular shaped plateau glaciers (zone 2) iscontrolled primarily by solar radiation-induced ablation onthe vertical cliffs that characterize them as demonstratedthrough radiation modelling (Molg et al 2003) Winkleret al (2010) revealed that direct insolation is the reasonfor the predominant zonal alignment of the ice cliffs andtheir bimodal annual recession (sunlit versus shaded peri-ods) Though the shrinkage of the ice cover along the verticalcliffs is fairly constant from year to year the areal retreat ofthe zone 2 region is not entirely linear with the highest ratesoccurring between 1975 and 1984 (Table 2) which corre-spond to the period following the separation of the northernand eastern ice fields (Fig 2) Therefore changes in the ge-ometry of the ice bodies and their exposure to solar radiationappear to have led to variations in retreat rates (Table 2 andFig 5) which has also been described by Kaser et al (2010)

An explanation for the growth and decay of the plateauglaciers including a description of the physical processesthat may have led to the observed steps in the northern icefield has been provided by Kaser et al (2010) To sustaingrowth in the crater rim region (zone 2) it is critical thatsnowfall is sufficient to allow a continuous snow cover todevelop that is able to survive from one year to the next Ifthe annual build-up of snow is sustained through a period oftime it is quite conceivable that the plateau glaciers couldgrow to several tens of metres in thickness quite quicklyraising the question as to whether this accumulation of icehas the potential to become a source of mass for glacierson the western flanks of the mountain (zone 1) Meyer(1891 p 316ndash317) was somewhat sceptical as to whetherthe plateau region acted as a ldquoreservoirrdquo and thought it ldquomuchmore reasonable to suppose that the rocky rim of the crateris the real reservoirrdquo However Meyer (1890 1891) did pro-vide an account and sketches to support a connection of theplateau ice to the Little and Great Barranco glaciers in thepast which as described may have been important in sustain-ing them (Gillman 1923 Geilinger 1936) There is someimportance in investigating the extent of the linkages be-tween these two glacial zones as together they accounted foralmost 80 of the total ice cover in 1912 (Table 2)

The remaining ice bodies in zone 2 (plateau region) aremostly stagnant today with field measurements and prelim-inary modelling results indicating deformation rates of icewithin the crater are small (lt 10 cm yrminus1) (Kaser et al 2004Winkler et al 2010) It seems unlikely that the present or his-torical thickness of the plateau ice has been sufficient to al-low substantial deformation (Kaser et al 2004) suggesting

3140 3150 3160 3170 3180 3190 3200 32109657

9658

9659

9660

9661

9662

9663

9664

4000

4000

4500

4500 4500

45004500

5000

5000

5000

5000

5000

5500

5500

5500

UTM easting in meters times105

UT

M n

orth

ing

in m

eter

s times

106

Fig 4 Classification of regions based on the distribution of topo-graphic measures in 1912 Zone 1 (western region) zone 2 (plateauregion) and zone 3 (southern region) are shown in red orange andyellow respectively

that mass transport from the nearly horizontal plateau downinto the western slope glaciers (zone 1) has most likely beenminor Careful observations by Klute (1920) demonstratedthat the velocity of the centre of the Great Penck glacier was064 m over a 30-day period in 1912 which led him to spec-ulate that the annual downslope movement was about 7 mper year It should be noted that Nilsson (1931) and Osmas-ton (1989) also report on these observations but incorrectlycite Jaeger (1909) as the person responsible for obtaining themeasurements Nonetheless it is improbable that ice fromthe plateau region could have supported the observed flowdynamics of the Great Penck glacier with the upper reachesof the mountain flanks the most likely source of the massThus the growth and decay of the glaciers in zone 1 (west-ern region) do not appear to be directly dependent on thestability of plateau ice (zone 2) To further resolve the phys-ical processes controlling the stability of the plateau ice itwould be of interest to develop a better understanding of themechanisms responsible for vertical cliff development

The glaciers on the southern slopes (zone 3) differ fromthose on the western slopes (zone 1) as they are separatedfrom the large masses of ice within the caldera (zone 2) bybed topography (Humphries 1972) This distinction is im-portant as any assessment of the atmospheric controls onthe behaviour of the southern ice fields is independent ofthe complexity of the vertical cliff retreat in the plateau re-gion (zone 2) That said the cluster analysis that led to thetopographical classes used in this study shows that the up-per sections of the southern ice field exhibit similar phys-ical characteristics to the ice within the caldera (Fig 4)Nevertheless to obtain a full solution for the asymmetry ofthe glaciers on Kibo that the early explorers observed itis critical to establish an understanding of the atmospheric

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 429

1900 1920 1940 1960 1980 2000 2020 2040 20600

05

1

15

2

25

3

35

4

45

5

Year

Are

al e

xten

t [km

2 ]

zone 1 R2=099

zone 2 R2=097

zone 3 R2=09980 confidence interval

4000

4000

4000

4500

4500

4500

45005000

5000

5000

50005000

5500

5500

5500

Fig 5 The areal change of glaciers and ice bodies based on thethree-member classification Linear regressions (thin solid lines) foreach of the three zones are shown as well as the 80 confidenceintervals for each of these relationships

processes controlling the growth andor decay of the southernice fields This has been achieved by Molg et al (2009b)who used observations and atmospherendashglacier modellingto show how critical precipitation humidity and fractionalcloud cover are on the energy and mass balance of the Ker-sten glacier (zone 3) These results compare favourably toearly accounts of the importance of precipitation (moisturechange) and cloud cover on the asymmetry (Meyer 1891Klute 1920 Gillman 1923 Geilinger 1936) with the diur-nal variability of cloud cover being especially important inproviding protection to the glaciers on the western side ofthe mountain (eg Meyer 1891 Hastenrath and Greischar1997)

Recent regional atmospheric modelling experiments haveprovided a more quantitative assessment to account for theasymmetry in glacier mass on Kibo demonstrating that thewest and south-western areas of the mountain are favourableto moisture convergence as controlled by atmospheric dy-namics (Molg et al 2009a) As noted by Osmaston (1989)the asymmetry is not confined to the most recent glacier ex-tents with the pattern of ice distribution established fromformer moraines indicating that recent climatic controls onglacier behaviour are similar to those in the past This sug-gests that establishing the specific nature of the atmosphericcontrols responsible for the most recent shift from glaciergrowth to decay in particular in zones 2 and 3 (plateau andsouthern slope regions) remains critical if we are to useglacier behaviour to reconstruct climatic conditions of thedeeper past

Finally it is difficult not to speculate on when the ice onKibo will disappear which becomes more achievable if onerecognizes the importance of the different glacial zones onthe mountain If one accepts linear extrapolation based on

changes from the past is suitable to make projections for thefuture it is apparent that the ice cover in zone 1 (western re-gion) will most likely be gone before 2020 and that ice inzone 2 (plateau) and zone 3 (southern slopes) will disappearsometime close to 2040 (Fig 5) It is likely that the plateauwill harbour the final remnants of ice on Kibo which couldremain slightly beyond the middle of the 21st century Inter-estingly our linear extrapolation for the loss of ice is in closeagreement to earlier projections using radiation modelling(Molg et al 2003) Though the loss of ice will have no hy-drological significance to lowland areas (eg Gillman 1923Kaser et al 2004 Molg et al 2008b) there is no doubt thattheir disappearance will be a ldquogrievous aesthetic lossrdquo (Salt1951 p 160) which will undoubtedly impact the experiencetourists have on the mountain (Molg et al 2008b)

4 Conclusions

With reference to the glaciers found on Mt Kenya and Kiboin East Africa C Gillman remarked ldquowe might perhapslearn a good deal as to the general world circulation fromcareful and prolonged observation on these particular sum-mitsrdquo (Mackinder 1923 p 23) This statement still holdstrue today (eg Molg et al 2006) and with the increasingawareness that glaciers on Africarsquos highest mountain can beused to extract climate information from the past it seemedtimely to revisit the mapping of the ice cover on Kibo throughtime Our motivation to do so has been to ensure the histor-ical record of these changes is of the highest possible qual-ity and in a format that should help facilitate on-going andfuture research dedicated to resolving the complexity of cli-mate variability and change in East Africa and beyond Thiswas achieved by using a consistent cartographic frameworkthat allows historical maps and new satellite imagery to beexamined The use of three-dimensional visualization tech-niques greatly enhanced our ability to establish the formerlocation of the glaciers

The revisited time series of glacial retreat on Kiliman-jaro contains eight epochs and spans almost 100 yr (1912ndash2011) The glaciers have retreated from their former extentof 1140 km2 in 1912 to 176 km2 in 2011 which representsa loss of about 85 of the ice cover over almost a 100-yr pe-riod Our estimate of the total retreat is consistent with recentassessments (Cullen et al 2006 Thompson et al 2009) butthe careful re-examination of the location and extent of indi-vidual ice bodies through time has allowed us to identify anumber of important differences from previously publishedmaps (eg Hastenrath and Greischar 1997 Thompson et al2002 2009) Historical evidence has been used to supportour interpretation with the detailed observations and pho-tographs obtained by early explorers on Kilimanjaro found tobe extremely valuable (eg Meyer 1891 1900 Uhlig 1904Jaeger 1909 Klute 1914 1920 1921 Gillman 1923 Nils-son 1931 Geilinger 1936) It has to be acknowledged that

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

430 N J Cullen et al A century of ice retreat on Kilimanjaro

much of what we know about Kilimanjaro today was estab-lished through their efforts

To further explore the importance of different glacierregimes on Kibo (Kaser et al 2004) a geo-statistical ap-proach was applied which resulted in the segmentation ofthe ice cover into three broad glacial zones A careful ex-amination of the retreat of ice in each of these glacier zonesindicates that there is a high probability that most of the re-maining ice cover on Kilimanjaro will disappear by 2040and it is highly unlikely that any ice body will remain af-ter 2060 if present retreat rates are maintained Before thishappens it would still be useful if the remaining ice bodiesin particular the remnants of the northern ice field could befurther examined to help resolve existing uncertainties aboutthe mechanisms responsible for controlling the formation ofthe vertical cliffs that characterize them (eg Winkler et al2010) In addition further efforts are required to scrutinizethe exact timing of the onset of the present glacier retreat firstreported by Hans Meyer in the late 19th century If we canrefine our understanding of the atmospheric controls behindthe most recent shift from glacier growth to decay on Kibowe will be in a stronger position to characterize the climaticcontrols of past and much larger glaciations on Africarsquos high-est mountain

Supplementary material related to this article isavailable online athttpwwwthe-cryospherenet74192013tc-7-419-2013-supplementpdf

AcknowledgementsThis research is funded by the Departmentof Geography and School of Surveying University of OtagoNew Zealand Thomas Molg was supported by the Alexander vonHumboldt Foundation The contributions made by Georg Kaserand Michael Winkler were supported by the Austrian ScienceFund (FWF) Projects P 20089-N10 and P 21288-N10 Thisresearch would not have been possible without the support from thefollowing Tanzanian authorities (COSTECH KINAPA TANAPATAWIRI) We are grateful to the referees that took the time toprovide feedback on this research in particular the careful andthoughtful comments by Henry Brecher

Edited by M Schneebeli

References

Chiang J C H The tropics in paleoclimate Annu Rev EarthPlanet Sci 37 263ndash297 2009

Cullen N J Molg T Kaser G Hussein K Steffen Kand Hardy D R Kilimanjaro Glaciers Recent areal ex-tent from satellite data and new interpretation of observed20th century retreat rates Geophys Res Lett 33 L16502doi1010292006GL027084 2006

Cullen N J Molg T Kaser G Steffen K and Hardy D REnergy-balance model validation on the top of Kilimanjaro Tan-zania using eddy covariance data Ann Glaciol 46 227ndash233doi103189172756407782871224 2007

Downie C Glaciations of Mount Kilimanjaro Northeast Tan-ganyika Geol Soc Am Bull 75 1ndash16doi1011300016-7606(1964)75[1GOMKNT]20CO2 1964

Downie C and Wilkinson P The Geology of KilimanjaroSheffield Geology Department University of Sheffield 1972

Downie C Humphries D W Wilcockson W H and WilkinsonP Geology of Kilimanjaro Nature 178 828ndash830 1956

Farr T G Rosen P A Caro E Crippen R Duren R Hens-ley S Kobrick M Paller M Rodriguez E Roth L Seal DShaffer S Shimada J Umland J Werner M Oskin M Bur-bank D and Alsdorf D The Shuttle Radar Topography Mis-sion Rev Geophys 45 RG2004doi1010292005RG0001832007

Geilinger W The retreat of the Kilimanjaro glaciers TanganyikaNotes and Records 2 7ndash20 1936

Gillman C An ascent of Kilimanjaro Geogr J 61 1ndash27 1923Hastenrath S The Glaciers of Equatorial East Africa Reidel Dor-

drecht 1984Hastenrath S Diagnosing the decaying glaciers of equatorial

East Africa Meteorol Z 15 265ndash271doi1011270941-294820060106 2006

Hastenrath S Recession of equatorial glaciers a photo documen-tation Sundog Publishing Madison Wisconsin 2008

Hastenrath S and Greischar L Glacier recession on KilimanjaroEast Africa 1912ndash89 J Glaciol 43 455ndash459 1997

Humphries D Interim report on the glaciology and meteorology ofKilimanjaro Tanganyika JulyndashSept 1953 Geological Survey ofTanganyika Dodoma 1953

Humphries D Preliminary notes on the glaciology of KilimanjaroJ Glaciol 3 475ndash479 1959

Humphries D Glaciology and glacial history in The Geology ofKilimanjaro edited by Downie C and Wilkinson P SheffieldGeology Department University of Sheffield 31ndash71 1972

Jarvis A Reuter H I Nelson A and Guevara E Hole-filledSRTM for the globe Version 4 available from the CGIAR-CSISRTM 90 m Database (httpsrtmcsicgiarorg) 2008

Jaeger F Forschungen in den Hochregionen des KilimandscharoMitt Deutsch Schutzgebiet 22 113ndash197 1909

Kaser G Hardy D R Molg T Bradley R S and Hyera TM Modern glacier retreat on Kilimanjaro as evidence of climatechange Observations and facts Int J Climatol 24 329ndash3392004

Kaser G Molg T Cullen N J Hardy D R andWinkler M Is the decline of ice on Kilimanjaro un-precedented in the Holocene Holocene 20 1079ndash1091doi1011770959683610369498 2010

Klute F Forschungen am Kilimandscharo im Jahre 1912 GeogrZeitschrift 20 496ndash505 1914

Klute F Ergebnisse der Forschungen am Kilimandscharo 1912ReimerndashVohsen Berlin 1920

Klute F Die stereophotogrammetrische Aufnahme der Hochregio-nen des Kilimandscharo Zeitschrift Ges Erdkunde Berlin 56144ndash151 1921

Mackinder C W Hobley C Gillman C and Johnston H Anascent of Kilimanjaro Discussion Geogr J 61 1ndash27 1923

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

426 N J Cullen et al A century of ice retreat on Kilimanjaro

(g) 2003 (h) 2011

(a) 1912 (b) 1962

(c) 1975 (d) 1984

(e) 1992 (f) 2000

Credner

Great Penck

Ratzel

Ker

sten

Northern ice field

Easternice field

Southernice field

GreatBarranco

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Northern ice field

Easternice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Southernice field

Fig 2 Ice retreat over Kilimanjaro based on the revised areal extent for the following years(a) 1912(b) 1962(c) 1975(d) 1984(e)1992(f) 2000(g) 2003 and(h) 2011 An interactive 3-D illustration of the sequence can be sourced in the Supplement

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 427

was likely to occur (Molg et al 2009b) The upper part ofthe Heim has disappeared and the Decken has narrowed con-siderably over the last 8 yr (Fig 2g and h)

32 New interpretation of the ice retreat

A consistent record of the former extent of the glaciers onKibo provides a platform from which the relationship be-tween glaciological and meteorological processes can be fur-ther explored We believe the spatial and temporal variabilityof the ice cover on Kibo can be used to unravel the physi-cal processes responsible for controlling the decay of the icemasses through time as well as providing insights into themechanisms that may have been responsible for growth in thepast Interestingly those who were privileged to make someof the first (formal) observations on Kilimanjaro recognizedthere was an asymmetry in the ice cover (eg Meyer 1891)which pointed to specific climatological drivers controllingglacial retreat (eg Gillman 1923 Nilsson 1931 Geilinger1936) Meyer (1891 p 308ndash314) noted that ldquoan immensemountain mass like Kilimanjaro must considerably modifythe prevailing wind directionrdquo before making the point thatthe primary cause in the variability of the ice limit is ldquotheunequal distribution of moisture on the northern and south-ern sides of the mountainrdquo Though Meyer (1891) recognizedvariability in moisture was the most important control healso speculated that the asymmetry of low reaching glacierson Kibo in the south south-west and western sectors of themountain a pattern that persisted throughout the 20th cen-tury (Fig 2) depended on ldquolocal peculiarities which favourthe preservation of the snow [rather] than on physical condi-tions that favour a large snowfallrdquo (Meyer 1891 p 315) Torevisit this early observation it is useful to reflect on the hyp-sometry of the glaciers in 1912 (Fig 3a) which shows thatthe probability density in relation to the distribution of the icewas greatest above 5700 m This high density reflects the icecover in the crater region with the majority of the ice in 1912being part of the former ice cap that linked the northern andeastern ice fields (Fig 1) Figure 3b and c show that slopeand aspect were also important controls on the density of icecover with the steep south and south-west facing slopes morefavourable which enabled the glaciers to extend to elevationsbelow 4600 m For the glaciers to have extended to the sameelevation on the east-facing slopes they would have had tocover a much greater horizontal distance (Fig 1)

To fully account for the distinct geometry of the ice masseson Kibo and to establish the physical processes controllingtheir location and extent through time Kaser et al (2004)established a framework that included four separate glacierregimes The motivation to do this was partly based on theearly recognition that the tabular shaped glaciers in the craterregion of Kibo behave differently to glaciers on the flanks ofthe mountain (eg Nilsson 1931 Geilinger 1936) Guidedby this distinction Cullen et al (2006) used the 5700 m con-tour which is roughly the height of the outer crater rim to

Density times 10 minus3

2

4

6

8

WEST EAST

SOUTH

NORTH

4600 4800 5000 5200 5400 5600 58000

1

2

3

4

Elevation [m]

Den

sity

times 10

minus3

0 10 20 30 40 50 60 70 800

1

2

3

4

Slope [deg]

Den

sity

times 10

minus2

(a)

(c)

(b)

Fig 3 Distribution of topographic measures from the areal extentof all ice bodies in 1912(a) elevation(b) slope and(c) aspect Themagnitude of the normalised distribution on the y-axis is a proba-bility density function with the integral equal to unity

distinguish between two broad glacial systems (1) plateau(ge 5700 m) and slope (lt 5700 m) glaciers This simple bi-nary classification has proved useful in describing some ofthe key physical processes responsible for the retreat of iceon Kibo (eg Molg et al 2009b Kaser et al 2010) Tofurther characterize the spatial and temporal variability ofglacier retreat on Kibo we have adopted a geo-statisticalapproach in this study that uses elevation slope and aspectof the ice cover in 1912 (the feature space) to identify clus-ters resembling topographical classes This was achieved byperforming a cluster analysis containing maximum likeli-hood estimates of the topographical parameters in a Gaussian

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

428 N J Cullen et al A century of ice retreat on Kilimanjaro

mixture model (McLachlan and Peel 2000) which resultedin three broad glacial zones being identified (1) western re-gion (zone 1) (2) plateau region (zone 2) and (3) southernregion (zone 3) (Fig 4)

Figure 5 shows the rate of retreat of each of the glacialzones through time as well as a least-squares linear regres-sion fit (and 80 confidence intervals) through each of theglacial zone time series As described by Cullen et al (2006)the retreat of the tabular shaped plateau glaciers (zone 2) iscontrolled primarily by solar radiation-induced ablation onthe vertical cliffs that characterize them as demonstratedthrough radiation modelling (Molg et al 2003) Winkleret al (2010) revealed that direct insolation is the reasonfor the predominant zonal alignment of the ice cliffs andtheir bimodal annual recession (sunlit versus shaded peri-ods) Though the shrinkage of the ice cover along the verticalcliffs is fairly constant from year to year the areal retreat ofthe zone 2 region is not entirely linear with the highest ratesoccurring between 1975 and 1984 (Table 2) which corre-spond to the period following the separation of the northernand eastern ice fields (Fig 2) Therefore changes in the ge-ometry of the ice bodies and their exposure to solar radiationappear to have led to variations in retreat rates (Table 2 andFig 5) which has also been described by Kaser et al (2010)

An explanation for the growth and decay of the plateauglaciers including a description of the physical processesthat may have led to the observed steps in the northern icefield has been provided by Kaser et al (2010) To sustaingrowth in the crater rim region (zone 2) it is critical thatsnowfall is sufficient to allow a continuous snow cover todevelop that is able to survive from one year to the next Ifthe annual build-up of snow is sustained through a period oftime it is quite conceivable that the plateau glaciers couldgrow to several tens of metres in thickness quite quicklyraising the question as to whether this accumulation of icehas the potential to become a source of mass for glacierson the western flanks of the mountain (zone 1) Meyer(1891 p 316ndash317) was somewhat sceptical as to whetherthe plateau region acted as a ldquoreservoirrdquo and thought it ldquomuchmore reasonable to suppose that the rocky rim of the crateris the real reservoirrdquo However Meyer (1890 1891) did pro-vide an account and sketches to support a connection of theplateau ice to the Little and Great Barranco glaciers in thepast which as described may have been important in sustain-ing them (Gillman 1923 Geilinger 1936) There is someimportance in investigating the extent of the linkages be-tween these two glacial zones as together they accounted foralmost 80 of the total ice cover in 1912 (Table 2)

The remaining ice bodies in zone 2 (plateau region) aremostly stagnant today with field measurements and prelim-inary modelling results indicating deformation rates of icewithin the crater are small (lt 10 cm yrminus1) (Kaser et al 2004Winkler et al 2010) It seems unlikely that the present or his-torical thickness of the plateau ice has been sufficient to al-low substantial deformation (Kaser et al 2004) suggesting

3140 3150 3160 3170 3180 3190 3200 32109657

9658

9659

9660

9661

9662

9663

9664

4000

4000

4500

4500 4500

45004500

5000

5000

5000

5000

5000

5500

5500

5500

UTM easting in meters times105

UT

M n

orth

ing

in m

eter

s times

106

Fig 4 Classification of regions based on the distribution of topo-graphic measures in 1912 Zone 1 (western region) zone 2 (plateauregion) and zone 3 (southern region) are shown in red orange andyellow respectively

that mass transport from the nearly horizontal plateau downinto the western slope glaciers (zone 1) has most likely beenminor Careful observations by Klute (1920) demonstratedthat the velocity of the centre of the Great Penck glacier was064 m over a 30-day period in 1912 which led him to spec-ulate that the annual downslope movement was about 7 mper year It should be noted that Nilsson (1931) and Osmas-ton (1989) also report on these observations but incorrectlycite Jaeger (1909) as the person responsible for obtaining themeasurements Nonetheless it is improbable that ice fromthe plateau region could have supported the observed flowdynamics of the Great Penck glacier with the upper reachesof the mountain flanks the most likely source of the massThus the growth and decay of the glaciers in zone 1 (west-ern region) do not appear to be directly dependent on thestability of plateau ice (zone 2) To further resolve the phys-ical processes controlling the stability of the plateau ice itwould be of interest to develop a better understanding of themechanisms responsible for vertical cliff development

The glaciers on the southern slopes (zone 3) differ fromthose on the western slopes (zone 1) as they are separatedfrom the large masses of ice within the caldera (zone 2) bybed topography (Humphries 1972) This distinction is im-portant as any assessment of the atmospheric controls onthe behaviour of the southern ice fields is independent ofthe complexity of the vertical cliff retreat in the plateau re-gion (zone 2) That said the cluster analysis that led to thetopographical classes used in this study shows that the up-per sections of the southern ice field exhibit similar phys-ical characteristics to the ice within the caldera (Fig 4)Nevertheless to obtain a full solution for the asymmetry ofthe glaciers on Kibo that the early explorers observed itis critical to establish an understanding of the atmospheric

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 429

1900 1920 1940 1960 1980 2000 2020 2040 20600

05

1

15

2

25

3

35

4

45

5

Year

Are

al e

xten

t [km

2 ]

zone 1 R2=099

zone 2 R2=097

zone 3 R2=09980 confidence interval

4000

4000

4000

4500

4500

4500

45005000

5000

5000

50005000

5500

5500

5500

Fig 5 The areal change of glaciers and ice bodies based on thethree-member classification Linear regressions (thin solid lines) foreach of the three zones are shown as well as the 80 confidenceintervals for each of these relationships

processes controlling the growth andor decay of the southernice fields This has been achieved by Molg et al (2009b)who used observations and atmospherendashglacier modellingto show how critical precipitation humidity and fractionalcloud cover are on the energy and mass balance of the Ker-sten glacier (zone 3) These results compare favourably toearly accounts of the importance of precipitation (moisturechange) and cloud cover on the asymmetry (Meyer 1891Klute 1920 Gillman 1923 Geilinger 1936) with the diur-nal variability of cloud cover being especially important inproviding protection to the glaciers on the western side ofthe mountain (eg Meyer 1891 Hastenrath and Greischar1997)

Recent regional atmospheric modelling experiments haveprovided a more quantitative assessment to account for theasymmetry in glacier mass on Kibo demonstrating that thewest and south-western areas of the mountain are favourableto moisture convergence as controlled by atmospheric dy-namics (Molg et al 2009a) As noted by Osmaston (1989)the asymmetry is not confined to the most recent glacier ex-tents with the pattern of ice distribution established fromformer moraines indicating that recent climatic controls onglacier behaviour are similar to those in the past This sug-gests that establishing the specific nature of the atmosphericcontrols responsible for the most recent shift from glaciergrowth to decay in particular in zones 2 and 3 (plateau andsouthern slope regions) remains critical if we are to useglacier behaviour to reconstruct climatic conditions of thedeeper past

Finally it is difficult not to speculate on when the ice onKibo will disappear which becomes more achievable if onerecognizes the importance of the different glacial zones onthe mountain If one accepts linear extrapolation based on

changes from the past is suitable to make projections for thefuture it is apparent that the ice cover in zone 1 (western re-gion) will most likely be gone before 2020 and that ice inzone 2 (plateau) and zone 3 (southern slopes) will disappearsometime close to 2040 (Fig 5) It is likely that the plateauwill harbour the final remnants of ice on Kibo which couldremain slightly beyond the middle of the 21st century Inter-estingly our linear extrapolation for the loss of ice is in closeagreement to earlier projections using radiation modelling(Molg et al 2003) Though the loss of ice will have no hy-drological significance to lowland areas (eg Gillman 1923Kaser et al 2004 Molg et al 2008b) there is no doubt thattheir disappearance will be a ldquogrievous aesthetic lossrdquo (Salt1951 p 160) which will undoubtedly impact the experiencetourists have on the mountain (Molg et al 2008b)

4 Conclusions

With reference to the glaciers found on Mt Kenya and Kiboin East Africa C Gillman remarked ldquowe might perhapslearn a good deal as to the general world circulation fromcareful and prolonged observation on these particular sum-mitsrdquo (Mackinder 1923 p 23) This statement still holdstrue today (eg Molg et al 2006) and with the increasingawareness that glaciers on Africarsquos highest mountain can beused to extract climate information from the past it seemedtimely to revisit the mapping of the ice cover on Kibo throughtime Our motivation to do so has been to ensure the histor-ical record of these changes is of the highest possible qual-ity and in a format that should help facilitate on-going andfuture research dedicated to resolving the complexity of cli-mate variability and change in East Africa and beyond Thiswas achieved by using a consistent cartographic frameworkthat allows historical maps and new satellite imagery to beexamined The use of three-dimensional visualization tech-niques greatly enhanced our ability to establish the formerlocation of the glaciers

The revisited time series of glacial retreat on Kiliman-jaro contains eight epochs and spans almost 100 yr (1912ndash2011) The glaciers have retreated from their former extentof 1140 km2 in 1912 to 176 km2 in 2011 which representsa loss of about 85 of the ice cover over almost a 100-yr pe-riod Our estimate of the total retreat is consistent with recentassessments (Cullen et al 2006 Thompson et al 2009) butthe careful re-examination of the location and extent of indi-vidual ice bodies through time has allowed us to identify anumber of important differences from previously publishedmaps (eg Hastenrath and Greischar 1997 Thompson et al2002 2009) Historical evidence has been used to supportour interpretation with the detailed observations and pho-tographs obtained by early explorers on Kilimanjaro found tobe extremely valuable (eg Meyer 1891 1900 Uhlig 1904Jaeger 1909 Klute 1914 1920 1921 Gillman 1923 Nils-son 1931 Geilinger 1936) It has to be acknowledged that

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

430 N J Cullen et al A century of ice retreat on Kilimanjaro

much of what we know about Kilimanjaro today was estab-lished through their efforts

To further explore the importance of different glacierregimes on Kibo (Kaser et al 2004) a geo-statistical ap-proach was applied which resulted in the segmentation ofthe ice cover into three broad glacial zones A careful ex-amination of the retreat of ice in each of these glacier zonesindicates that there is a high probability that most of the re-maining ice cover on Kilimanjaro will disappear by 2040and it is highly unlikely that any ice body will remain af-ter 2060 if present retreat rates are maintained Before thishappens it would still be useful if the remaining ice bodiesin particular the remnants of the northern ice field could befurther examined to help resolve existing uncertainties aboutthe mechanisms responsible for controlling the formation ofthe vertical cliffs that characterize them (eg Winkler et al2010) In addition further efforts are required to scrutinizethe exact timing of the onset of the present glacier retreat firstreported by Hans Meyer in the late 19th century If we canrefine our understanding of the atmospheric controls behindthe most recent shift from glacier growth to decay on Kibowe will be in a stronger position to characterize the climaticcontrols of past and much larger glaciations on Africarsquos high-est mountain

Supplementary material related to this article isavailable online athttpwwwthe-cryospherenet74192013tc-7-419-2013-supplementpdf

AcknowledgementsThis research is funded by the Departmentof Geography and School of Surveying University of OtagoNew Zealand Thomas Molg was supported by the Alexander vonHumboldt Foundation The contributions made by Georg Kaserand Michael Winkler were supported by the Austrian ScienceFund (FWF) Projects P 20089-N10 and P 21288-N10 Thisresearch would not have been possible without the support from thefollowing Tanzanian authorities (COSTECH KINAPA TANAPATAWIRI) We are grateful to the referees that took the time toprovide feedback on this research in particular the careful andthoughtful comments by Henry Brecher

Edited by M Schneebeli

References

Chiang J C H The tropics in paleoclimate Annu Rev EarthPlanet Sci 37 263ndash297 2009

Cullen N J Molg T Kaser G Hussein K Steffen Kand Hardy D R Kilimanjaro Glaciers Recent areal ex-tent from satellite data and new interpretation of observed20th century retreat rates Geophys Res Lett 33 L16502doi1010292006GL027084 2006

Cullen N J Molg T Kaser G Steffen K and Hardy D REnergy-balance model validation on the top of Kilimanjaro Tan-zania using eddy covariance data Ann Glaciol 46 227ndash233doi103189172756407782871224 2007

Downie C Glaciations of Mount Kilimanjaro Northeast Tan-ganyika Geol Soc Am Bull 75 1ndash16doi1011300016-7606(1964)75[1GOMKNT]20CO2 1964

Downie C and Wilkinson P The Geology of KilimanjaroSheffield Geology Department University of Sheffield 1972

Downie C Humphries D W Wilcockson W H and WilkinsonP Geology of Kilimanjaro Nature 178 828ndash830 1956

Farr T G Rosen P A Caro E Crippen R Duren R Hens-ley S Kobrick M Paller M Rodriguez E Roth L Seal DShaffer S Shimada J Umland J Werner M Oskin M Bur-bank D and Alsdorf D The Shuttle Radar Topography Mis-sion Rev Geophys 45 RG2004doi1010292005RG0001832007

Geilinger W The retreat of the Kilimanjaro glaciers TanganyikaNotes and Records 2 7ndash20 1936

Gillman C An ascent of Kilimanjaro Geogr J 61 1ndash27 1923Hastenrath S The Glaciers of Equatorial East Africa Reidel Dor-

drecht 1984Hastenrath S Diagnosing the decaying glaciers of equatorial

East Africa Meteorol Z 15 265ndash271doi1011270941-294820060106 2006

Hastenrath S Recession of equatorial glaciers a photo documen-tation Sundog Publishing Madison Wisconsin 2008

Hastenrath S and Greischar L Glacier recession on KilimanjaroEast Africa 1912ndash89 J Glaciol 43 455ndash459 1997

Humphries D Interim report on the glaciology and meteorology ofKilimanjaro Tanganyika JulyndashSept 1953 Geological Survey ofTanganyika Dodoma 1953

Humphries D Preliminary notes on the glaciology of KilimanjaroJ Glaciol 3 475ndash479 1959

Humphries D Glaciology and glacial history in The Geology ofKilimanjaro edited by Downie C and Wilkinson P SheffieldGeology Department University of Sheffield 31ndash71 1972

Jarvis A Reuter H I Nelson A and Guevara E Hole-filledSRTM for the globe Version 4 available from the CGIAR-CSISRTM 90 m Database (httpsrtmcsicgiarorg) 2008

Jaeger F Forschungen in den Hochregionen des KilimandscharoMitt Deutsch Schutzgebiet 22 113ndash197 1909

Kaser G Hardy D R Molg T Bradley R S and Hyera TM Modern glacier retreat on Kilimanjaro as evidence of climatechange Observations and facts Int J Climatol 24 329ndash3392004

Kaser G Molg T Cullen N J Hardy D R andWinkler M Is the decline of ice on Kilimanjaro un-precedented in the Holocene Holocene 20 1079ndash1091doi1011770959683610369498 2010

Klute F Forschungen am Kilimandscharo im Jahre 1912 GeogrZeitschrift 20 496ndash505 1914

Klute F Ergebnisse der Forschungen am Kilimandscharo 1912ReimerndashVohsen Berlin 1920

Klute F Die stereophotogrammetrische Aufnahme der Hochregio-nen des Kilimandscharo Zeitschrift Ges Erdkunde Berlin 56144ndash151 1921

Mackinder C W Hobley C Gillman C and Johnston H Anascent of Kilimanjaro Discussion Geogr J 61 1ndash27 1923

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 427

was likely to occur (Molg et al 2009b) The upper part ofthe Heim has disappeared and the Decken has narrowed con-siderably over the last 8 yr (Fig 2g and h)

32 New interpretation of the ice retreat

A consistent record of the former extent of the glaciers onKibo provides a platform from which the relationship be-tween glaciological and meteorological processes can be fur-ther explored We believe the spatial and temporal variabilityof the ice cover on Kibo can be used to unravel the physi-cal processes responsible for controlling the decay of the icemasses through time as well as providing insights into themechanisms that may have been responsible for growth in thepast Interestingly those who were privileged to make someof the first (formal) observations on Kilimanjaro recognizedthere was an asymmetry in the ice cover (eg Meyer 1891)which pointed to specific climatological drivers controllingglacial retreat (eg Gillman 1923 Nilsson 1931 Geilinger1936) Meyer (1891 p 308ndash314) noted that ldquoan immensemountain mass like Kilimanjaro must considerably modifythe prevailing wind directionrdquo before making the point thatthe primary cause in the variability of the ice limit is ldquotheunequal distribution of moisture on the northern and south-ern sides of the mountainrdquo Though Meyer (1891) recognizedvariability in moisture was the most important control healso speculated that the asymmetry of low reaching glacierson Kibo in the south south-west and western sectors of themountain a pattern that persisted throughout the 20th cen-tury (Fig 2) depended on ldquolocal peculiarities which favourthe preservation of the snow [rather] than on physical condi-tions that favour a large snowfallrdquo (Meyer 1891 p 315) Torevisit this early observation it is useful to reflect on the hyp-sometry of the glaciers in 1912 (Fig 3a) which shows thatthe probability density in relation to the distribution of the icewas greatest above 5700 m This high density reflects the icecover in the crater region with the majority of the ice in 1912being part of the former ice cap that linked the northern andeastern ice fields (Fig 1) Figure 3b and c show that slopeand aspect were also important controls on the density of icecover with the steep south and south-west facing slopes morefavourable which enabled the glaciers to extend to elevationsbelow 4600 m For the glaciers to have extended to the sameelevation on the east-facing slopes they would have had tocover a much greater horizontal distance (Fig 1)

To fully account for the distinct geometry of the ice masseson Kibo and to establish the physical processes controllingtheir location and extent through time Kaser et al (2004)established a framework that included four separate glacierregimes The motivation to do this was partly based on theearly recognition that the tabular shaped glaciers in the craterregion of Kibo behave differently to glaciers on the flanks ofthe mountain (eg Nilsson 1931 Geilinger 1936) Guidedby this distinction Cullen et al (2006) used the 5700 m con-tour which is roughly the height of the outer crater rim to

Density times 10 minus3

2

4

6

8

WEST EAST

SOUTH

NORTH

4600 4800 5000 5200 5400 5600 58000

1

2

3

4

Elevation [m]

Den

sity

times 10

minus3

0 10 20 30 40 50 60 70 800

1

2

3

4

Slope [deg]

Den

sity

times 10

minus2

(a)

(c)

(b)

Fig 3 Distribution of topographic measures from the areal extentof all ice bodies in 1912(a) elevation(b) slope and(c) aspect Themagnitude of the normalised distribution on the y-axis is a proba-bility density function with the integral equal to unity

distinguish between two broad glacial systems (1) plateau(ge 5700 m) and slope (lt 5700 m) glaciers This simple bi-nary classification has proved useful in describing some ofthe key physical processes responsible for the retreat of iceon Kibo (eg Molg et al 2009b Kaser et al 2010) Tofurther characterize the spatial and temporal variability ofglacier retreat on Kibo we have adopted a geo-statisticalapproach in this study that uses elevation slope and aspectof the ice cover in 1912 (the feature space) to identify clus-ters resembling topographical classes This was achieved byperforming a cluster analysis containing maximum likeli-hood estimates of the topographical parameters in a Gaussian

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

428 N J Cullen et al A century of ice retreat on Kilimanjaro

mixture model (McLachlan and Peel 2000) which resultedin three broad glacial zones being identified (1) western re-gion (zone 1) (2) plateau region (zone 2) and (3) southernregion (zone 3) (Fig 4)

Figure 5 shows the rate of retreat of each of the glacialzones through time as well as a least-squares linear regres-sion fit (and 80 confidence intervals) through each of theglacial zone time series As described by Cullen et al (2006)the retreat of the tabular shaped plateau glaciers (zone 2) iscontrolled primarily by solar radiation-induced ablation onthe vertical cliffs that characterize them as demonstratedthrough radiation modelling (Molg et al 2003) Winkleret al (2010) revealed that direct insolation is the reasonfor the predominant zonal alignment of the ice cliffs andtheir bimodal annual recession (sunlit versus shaded peri-ods) Though the shrinkage of the ice cover along the verticalcliffs is fairly constant from year to year the areal retreat ofthe zone 2 region is not entirely linear with the highest ratesoccurring between 1975 and 1984 (Table 2) which corre-spond to the period following the separation of the northernand eastern ice fields (Fig 2) Therefore changes in the ge-ometry of the ice bodies and their exposure to solar radiationappear to have led to variations in retreat rates (Table 2 andFig 5) which has also been described by Kaser et al (2010)

An explanation for the growth and decay of the plateauglaciers including a description of the physical processesthat may have led to the observed steps in the northern icefield has been provided by Kaser et al (2010) To sustaingrowth in the crater rim region (zone 2) it is critical thatsnowfall is sufficient to allow a continuous snow cover todevelop that is able to survive from one year to the next Ifthe annual build-up of snow is sustained through a period oftime it is quite conceivable that the plateau glaciers couldgrow to several tens of metres in thickness quite quicklyraising the question as to whether this accumulation of icehas the potential to become a source of mass for glacierson the western flanks of the mountain (zone 1) Meyer(1891 p 316ndash317) was somewhat sceptical as to whetherthe plateau region acted as a ldquoreservoirrdquo and thought it ldquomuchmore reasonable to suppose that the rocky rim of the crateris the real reservoirrdquo However Meyer (1890 1891) did pro-vide an account and sketches to support a connection of theplateau ice to the Little and Great Barranco glaciers in thepast which as described may have been important in sustain-ing them (Gillman 1923 Geilinger 1936) There is someimportance in investigating the extent of the linkages be-tween these two glacial zones as together they accounted foralmost 80 of the total ice cover in 1912 (Table 2)

The remaining ice bodies in zone 2 (plateau region) aremostly stagnant today with field measurements and prelim-inary modelling results indicating deformation rates of icewithin the crater are small (lt 10 cm yrminus1) (Kaser et al 2004Winkler et al 2010) It seems unlikely that the present or his-torical thickness of the plateau ice has been sufficient to al-low substantial deformation (Kaser et al 2004) suggesting

3140 3150 3160 3170 3180 3190 3200 32109657

9658

9659

9660

9661

9662

9663

9664

4000

4000

4500

4500 4500

45004500

5000

5000

5000

5000

5000

5500

5500

5500

UTM easting in meters times105

UT

M n

orth

ing

in m

eter

s times

106

Fig 4 Classification of regions based on the distribution of topo-graphic measures in 1912 Zone 1 (western region) zone 2 (plateauregion) and zone 3 (southern region) are shown in red orange andyellow respectively

that mass transport from the nearly horizontal plateau downinto the western slope glaciers (zone 1) has most likely beenminor Careful observations by Klute (1920) demonstratedthat the velocity of the centre of the Great Penck glacier was064 m over a 30-day period in 1912 which led him to spec-ulate that the annual downslope movement was about 7 mper year It should be noted that Nilsson (1931) and Osmas-ton (1989) also report on these observations but incorrectlycite Jaeger (1909) as the person responsible for obtaining themeasurements Nonetheless it is improbable that ice fromthe plateau region could have supported the observed flowdynamics of the Great Penck glacier with the upper reachesof the mountain flanks the most likely source of the massThus the growth and decay of the glaciers in zone 1 (west-ern region) do not appear to be directly dependent on thestability of plateau ice (zone 2) To further resolve the phys-ical processes controlling the stability of the plateau ice itwould be of interest to develop a better understanding of themechanisms responsible for vertical cliff development

The glaciers on the southern slopes (zone 3) differ fromthose on the western slopes (zone 1) as they are separatedfrom the large masses of ice within the caldera (zone 2) bybed topography (Humphries 1972) This distinction is im-portant as any assessment of the atmospheric controls onthe behaviour of the southern ice fields is independent ofthe complexity of the vertical cliff retreat in the plateau re-gion (zone 2) That said the cluster analysis that led to thetopographical classes used in this study shows that the up-per sections of the southern ice field exhibit similar phys-ical characteristics to the ice within the caldera (Fig 4)Nevertheless to obtain a full solution for the asymmetry ofthe glaciers on Kibo that the early explorers observed itis critical to establish an understanding of the atmospheric

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 429

1900 1920 1940 1960 1980 2000 2020 2040 20600

05

1

15

2

25

3

35

4

45

5

Year

Are

al e

xten

t [km

2 ]

zone 1 R2=099

zone 2 R2=097

zone 3 R2=09980 confidence interval

4000

4000

4000

4500

4500

4500

45005000

5000

5000

50005000

5500

5500

5500

Fig 5 The areal change of glaciers and ice bodies based on thethree-member classification Linear regressions (thin solid lines) foreach of the three zones are shown as well as the 80 confidenceintervals for each of these relationships

processes controlling the growth andor decay of the southernice fields This has been achieved by Molg et al (2009b)who used observations and atmospherendashglacier modellingto show how critical precipitation humidity and fractionalcloud cover are on the energy and mass balance of the Ker-sten glacier (zone 3) These results compare favourably toearly accounts of the importance of precipitation (moisturechange) and cloud cover on the asymmetry (Meyer 1891Klute 1920 Gillman 1923 Geilinger 1936) with the diur-nal variability of cloud cover being especially important inproviding protection to the glaciers on the western side ofthe mountain (eg Meyer 1891 Hastenrath and Greischar1997)

Recent regional atmospheric modelling experiments haveprovided a more quantitative assessment to account for theasymmetry in glacier mass on Kibo demonstrating that thewest and south-western areas of the mountain are favourableto moisture convergence as controlled by atmospheric dy-namics (Molg et al 2009a) As noted by Osmaston (1989)the asymmetry is not confined to the most recent glacier ex-tents with the pattern of ice distribution established fromformer moraines indicating that recent climatic controls onglacier behaviour are similar to those in the past This sug-gests that establishing the specific nature of the atmosphericcontrols responsible for the most recent shift from glaciergrowth to decay in particular in zones 2 and 3 (plateau andsouthern slope regions) remains critical if we are to useglacier behaviour to reconstruct climatic conditions of thedeeper past

Finally it is difficult not to speculate on when the ice onKibo will disappear which becomes more achievable if onerecognizes the importance of the different glacial zones onthe mountain If one accepts linear extrapolation based on

changes from the past is suitable to make projections for thefuture it is apparent that the ice cover in zone 1 (western re-gion) will most likely be gone before 2020 and that ice inzone 2 (plateau) and zone 3 (southern slopes) will disappearsometime close to 2040 (Fig 5) It is likely that the plateauwill harbour the final remnants of ice on Kibo which couldremain slightly beyond the middle of the 21st century Inter-estingly our linear extrapolation for the loss of ice is in closeagreement to earlier projections using radiation modelling(Molg et al 2003) Though the loss of ice will have no hy-drological significance to lowland areas (eg Gillman 1923Kaser et al 2004 Molg et al 2008b) there is no doubt thattheir disappearance will be a ldquogrievous aesthetic lossrdquo (Salt1951 p 160) which will undoubtedly impact the experiencetourists have on the mountain (Molg et al 2008b)

4 Conclusions

With reference to the glaciers found on Mt Kenya and Kiboin East Africa C Gillman remarked ldquowe might perhapslearn a good deal as to the general world circulation fromcareful and prolonged observation on these particular sum-mitsrdquo (Mackinder 1923 p 23) This statement still holdstrue today (eg Molg et al 2006) and with the increasingawareness that glaciers on Africarsquos highest mountain can beused to extract climate information from the past it seemedtimely to revisit the mapping of the ice cover on Kibo throughtime Our motivation to do so has been to ensure the histor-ical record of these changes is of the highest possible qual-ity and in a format that should help facilitate on-going andfuture research dedicated to resolving the complexity of cli-mate variability and change in East Africa and beyond Thiswas achieved by using a consistent cartographic frameworkthat allows historical maps and new satellite imagery to beexamined The use of three-dimensional visualization tech-niques greatly enhanced our ability to establish the formerlocation of the glaciers

The revisited time series of glacial retreat on Kiliman-jaro contains eight epochs and spans almost 100 yr (1912ndash2011) The glaciers have retreated from their former extentof 1140 km2 in 1912 to 176 km2 in 2011 which representsa loss of about 85 of the ice cover over almost a 100-yr pe-riod Our estimate of the total retreat is consistent with recentassessments (Cullen et al 2006 Thompson et al 2009) butthe careful re-examination of the location and extent of indi-vidual ice bodies through time has allowed us to identify anumber of important differences from previously publishedmaps (eg Hastenrath and Greischar 1997 Thompson et al2002 2009) Historical evidence has been used to supportour interpretation with the detailed observations and pho-tographs obtained by early explorers on Kilimanjaro found tobe extremely valuable (eg Meyer 1891 1900 Uhlig 1904Jaeger 1909 Klute 1914 1920 1921 Gillman 1923 Nils-son 1931 Geilinger 1936) It has to be acknowledged that

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

430 N J Cullen et al A century of ice retreat on Kilimanjaro

much of what we know about Kilimanjaro today was estab-lished through their efforts

To further explore the importance of different glacierregimes on Kibo (Kaser et al 2004) a geo-statistical ap-proach was applied which resulted in the segmentation ofthe ice cover into three broad glacial zones A careful ex-amination of the retreat of ice in each of these glacier zonesindicates that there is a high probability that most of the re-maining ice cover on Kilimanjaro will disappear by 2040and it is highly unlikely that any ice body will remain af-ter 2060 if present retreat rates are maintained Before thishappens it would still be useful if the remaining ice bodiesin particular the remnants of the northern ice field could befurther examined to help resolve existing uncertainties aboutthe mechanisms responsible for controlling the formation ofthe vertical cliffs that characterize them (eg Winkler et al2010) In addition further efforts are required to scrutinizethe exact timing of the onset of the present glacier retreat firstreported by Hans Meyer in the late 19th century If we canrefine our understanding of the atmospheric controls behindthe most recent shift from glacier growth to decay on Kibowe will be in a stronger position to characterize the climaticcontrols of past and much larger glaciations on Africarsquos high-est mountain

Supplementary material related to this article isavailable online athttpwwwthe-cryospherenet74192013tc-7-419-2013-supplementpdf

AcknowledgementsThis research is funded by the Departmentof Geography and School of Surveying University of OtagoNew Zealand Thomas Molg was supported by the Alexander vonHumboldt Foundation The contributions made by Georg Kaserand Michael Winkler were supported by the Austrian ScienceFund (FWF) Projects P 20089-N10 and P 21288-N10 Thisresearch would not have been possible without the support from thefollowing Tanzanian authorities (COSTECH KINAPA TANAPATAWIRI) We are grateful to the referees that took the time toprovide feedback on this research in particular the careful andthoughtful comments by Henry Brecher

Edited by M Schneebeli

References

Chiang J C H The tropics in paleoclimate Annu Rev EarthPlanet Sci 37 263ndash297 2009

Cullen N J Molg T Kaser G Hussein K Steffen Kand Hardy D R Kilimanjaro Glaciers Recent areal ex-tent from satellite data and new interpretation of observed20th century retreat rates Geophys Res Lett 33 L16502doi1010292006GL027084 2006

Cullen N J Molg T Kaser G Steffen K and Hardy D REnergy-balance model validation on the top of Kilimanjaro Tan-zania using eddy covariance data Ann Glaciol 46 227ndash233doi103189172756407782871224 2007

Downie C Glaciations of Mount Kilimanjaro Northeast Tan-ganyika Geol Soc Am Bull 75 1ndash16doi1011300016-7606(1964)75[1GOMKNT]20CO2 1964

Downie C and Wilkinson P The Geology of KilimanjaroSheffield Geology Department University of Sheffield 1972

Downie C Humphries D W Wilcockson W H and WilkinsonP Geology of Kilimanjaro Nature 178 828ndash830 1956

Farr T G Rosen P A Caro E Crippen R Duren R Hens-ley S Kobrick M Paller M Rodriguez E Roth L Seal DShaffer S Shimada J Umland J Werner M Oskin M Bur-bank D and Alsdorf D The Shuttle Radar Topography Mis-sion Rev Geophys 45 RG2004doi1010292005RG0001832007

Geilinger W The retreat of the Kilimanjaro glaciers TanganyikaNotes and Records 2 7ndash20 1936

Gillman C An ascent of Kilimanjaro Geogr J 61 1ndash27 1923Hastenrath S The Glaciers of Equatorial East Africa Reidel Dor-

drecht 1984Hastenrath S Diagnosing the decaying glaciers of equatorial

East Africa Meteorol Z 15 265ndash271doi1011270941-294820060106 2006

Hastenrath S Recession of equatorial glaciers a photo documen-tation Sundog Publishing Madison Wisconsin 2008

Hastenrath S and Greischar L Glacier recession on KilimanjaroEast Africa 1912ndash89 J Glaciol 43 455ndash459 1997

Humphries D Interim report on the glaciology and meteorology ofKilimanjaro Tanganyika JulyndashSept 1953 Geological Survey ofTanganyika Dodoma 1953

Humphries D Preliminary notes on the glaciology of KilimanjaroJ Glaciol 3 475ndash479 1959

Humphries D Glaciology and glacial history in The Geology ofKilimanjaro edited by Downie C and Wilkinson P SheffieldGeology Department University of Sheffield 31ndash71 1972

Jarvis A Reuter H I Nelson A and Guevara E Hole-filledSRTM for the globe Version 4 available from the CGIAR-CSISRTM 90 m Database (httpsrtmcsicgiarorg) 2008

Jaeger F Forschungen in den Hochregionen des KilimandscharoMitt Deutsch Schutzgebiet 22 113ndash197 1909

Kaser G Hardy D R Molg T Bradley R S and Hyera TM Modern glacier retreat on Kilimanjaro as evidence of climatechange Observations and facts Int J Climatol 24 329ndash3392004

Kaser G Molg T Cullen N J Hardy D R andWinkler M Is the decline of ice on Kilimanjaro un-precedented in the Holocene Holocene 20 1079ndash1091doi1011770959683610369498 2010

Klute F Forschungen am Kilimandscharo im Jahre 1912 GeogrZeitschrift 20 496ndash505 1914

Klute F Ergebnisse der Forschungen am Kilimandscharo 1912ReimerndashVohsen Berlin 1920

Klute F Die stereophotogrammetrische Aufnahme der Hochregio-nen des Kilimandscharo Zeitschrift Ges Erdkunde Berlin 56144ndash151 1921

Mackinder C W Hobley C Gillman C and Johnston H Anascent of Kilimanjaro Discussion Geogr J 61 1ndash27 1923

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

428 N J Cullen et al A century of ice retreat on Kilimanjaro

mixture model (McLachlan and Peel 2000) which resultedin three broad glacial zones being identified (1) western re-gion (zone 1) (2) plateau region (zone 2) and (3) southernregion (zone 3) (Fig 4)

Figure 5 shows the rate of retreat of each of the glacialzones through time as well as a least-squares linear regres-sion fit (and 80 confidence intervals) through each of theglacial zone time series As described by Cullen et al (2006)the retreat of the tabular shaped plateau glaciers (zone 2) iscontrolled primarily by solar radiation-induced ablation onthe vertical cliffs that characterize them as demonstratedthrough radiation modelling (Molg et al 2003) Winkleret al (2010) revealed that direct insolation is the reasonfor the predominant zonal alignment of the ice cliffs andtheir bimodal annual recession (sunlit versus shaded peri-ods) Though the shrinkage of the ice cover along the verticalcliffs is fairly constant from year to year the areal retreat ofthe zone 2 region is not entirely linear with the highest ratesoccurring between 1975 and 1984 (Table 2) which corre-spond to the period following the separation of the northernand eastern ice fields (Fig 2) Therefore changes in the ge-ometry of the ice bodies and their exposure to solar radiationappear to have led to variations in retreat rates (Table 2 andFig 5) which has also been described by Kaser et al (2010)

An explanation for the growth and decay of the plateauglaciers including a description of the physical processesthat may have led to the observed steps in the northern icefield has been provided by Kaser et al (2010) To sustaingrowth in the crater rim region (zone 2) it is critical thatsnowfall is sufficient to allow a continuous snow cover todevelop that is able to survive from one year to the next Ifthe annual build-up of snow is sustained through a period oftime it is quite conceivable that the plateau glaciers couldgrow to several tens of metres in thickness quite quicklyraising the question as to whether this accumulation of icehas the potential to become a source of mass for glacierson the western flanks of the mountain (zone 1) Meyer(1891 p 316ndash317) was somewhat sceptical as to whetherthe plateau region acted as a ldquoreservoirrdquo and thought it ldquomuchmore reasonable to suppose that the rocky rim of the crateris the real reservoirrdquo However Meyer (1890 1891) did pro-vide an account and sketches to support a connection of theplateau ice to the Little and Great Barranco glaciers in thepast which as described may have been important in sustain-ing them (Gillman 1923 Geilinger 1936) There is someimportance in investigating the extent of the linkages be-tween these two glacial zones as together they accounted foralmost 80 of the total ice cover in 1912 (Table 2)

The remaining ice bodies in zone 2 (plateau region) aremostly stagnant today with field measurements and prelim-inary modelling results indicating deformation rates of icewithin the crater are small (lt 10 cm yrminus1) (Kaser et al 2004Winkler et al 2010) It seems unlikely that the present or his-torical thickness of the plateau ice has been sufficient to al-low substantial deformation (Kaser et al 2004) suggesting

3140 3150 3160 3170 3180 3190 3200 32109657

9658

9659

9660

9661

9662

9663

9664

4000

4000

4500

4500 4500

45004500

5000

5000

5000

5000

5000

5500

5500

5500

UTM easting in meters times105

UT

M n

orth

ing

in m

eter

s times

106

Fig 4 Classification of regions based on the distribution of topo-graphic measures in 1912 Zone 1 (western region) zone 2 (plateauregion) and zone 3 (southern region) are shown in red orange andyellow respectively

that mass transport from the nearly horizontal plateau downinto the western slope glaciers (zone 1) has most likely beenminor Careful observations by Klute (1920) demonstratedthat the velocity of the centre of the Great Penck glacier was064 m over a 30-day period in 1912 which led him to spec-ulate that the annual downslope movement was about 7 mper year It should be noted that Nilsson (1931) and Osmas-ton (1989) also report on these observations but incorrectlycite Jaeger (1909) as the person responsible for obtaining themeasurements Nonetheless it is improbable that ice fromthe plateau region could have supported the observed flowdynamics of the Great Penck glacier with the upper reachesof the mountain flanks the most likely source of the massThus the growth and decay of the glaciers in zone 1 (west-ern region) do not appear to be directly dependent on thestability of plateau ice (zone 2) To further resolve the phys-ical processes controlling the stability of the plateau ice itwould be of interest to develop a better understanding of themechanisms responsible for vertical cliff development

The glaciers on the southern slopes (zone 3) differ fromthose on the western slopes (zone 1) as they are separatedfrom the large masses of ice within the caldera (zone 2) bybed topography (Humphries 1972) This distinction is im-portant as any assessment of the atmospheric controls onthe behaviour of the southern ice fields is independent ofthe complexity of the vertical cliff retreat in the plateau re-gion (zone 2) That said the cluster analysis that led to thetopographical classes used in this study shows that the up-per sections of the southern ice field exhibit similar phys-ical characteristics to the ice within the caldera (Fig 4)Nevertheless to obtain a full solution for the asymmetry ofthe glaciers on Kibo that the early explorers observed itis critical to establish an understanding of the atmospheric

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 429

1900 1920 1940 1960 1980 2000 2020 2040 20600

05

1

15

2

25

3

35

4

45

5

Year

Are

al e

xten

t [km

2 ]

zone 1 R2=099

zone 2 R2=097

zone 3 R2=09980 confidence interval

4000

4000

4000

4500

4500

4500

45005000

5000

5000

50005000

5500

5500

5500

Fig 5 The areal change of glaciers and ice bodies based on thethree-member classification Linear regressions (thin solid lines) foreach of the three zones are shown as well as the 80 confidenceintervals for each of these relationships

processes controlling the growth andor decay of the southernice fields This has been achieved by Molg et al (2009b)who used observations and atmospherendashglacier modellingto show how critical precipitation humidity and fractionalcloud cover are on the energy and mass balance of the Ker-sten glacier (zone 3) These results compare favourably toearly accounts of the importance of precipitation (moisturechange) and cloud cover on the asymmetry (Meyer 1891Klute 1920 Gillman 1923 Geilinger 1936) with the diur-nal variability of cloud cover being especially important inproviding protection to the glaciers on the western side ofthe mountain (eg Meyer 1891 Hastenrath and Greischar1997)

Recent regional atmospheric modelling experiments haveprovided a more quantitative assessment to account for theasymmetry in glacier mass on Kibo demonstrating that thewest and south-western areas of the mountain are favourableto moisture convergence as controlled by atmospheric dy-namics (Molg et al 2009a) As noted by Osmaston (1989)the asymmetry is not confined to the most recent glacier ex-tents with the pattern of ice distribution established fromformer moraines indicating that recent climatic controls onglacier behaviour are similar to those in the past This sug-gests that establishing the specific nature of the atmosphericcontrols responsible for the most recent shift from glaciergrowth to decay in particular in zones 2 and 3 (plateau andsouthern slope regions) remains critical if we are to useglacier behaviour to reconstruct climatic conditions of thedeeper past

Finally it is difficult not to speculate on when the ice onKibo will disappear which becomes more achievable if onerecognizes the importance of the different glacial zones onthe mountain If one accepts linear extrapolation based on

changes from the past is suitable to make projections for thefuture it is apparent that the ice cover in zone 1 (western re-gion) will most likely be gone before 2020 and that ice inzone 2 (plateau) and zone 3 (southern slopes) will disappearsometime close to 2040 (Fig 5) It is likely that the plateauwill harbour the final remnants of ice on Kibo which couldremain slightly beyond the middle of the 21st century Inter-estingly our linear extrapolation for the loss of ice is in closeagreement to earlier projections using radiation modelling(Molg et al 2003) Though the loss of ice will have no hy-drological significance to lowland areas (eg Gillman 1923Kaser et al 2004 Molg et al 2008b) there is no doubt thattheir disappearance will be a ldquogrievous aesthetic lossrdquo (Salt1951 p 160) which will undoubtedly impact the experiencetourists have on the mountain (Molg et al 2008b)

4 Conclusions

With reference to the glaciers found on Mt Kenya and Kiboin East Africa C Gillman remarked ldquowe might perhapslearn a good deal as to the general world circulation fromcareful and prolonged observation on these particular sum-mitsrdquo (Mackinder 1923 p 23) This statement still holdstrue today (eg Molg et al 2006) and with the increasingawareness that glaciers on Africarsquos highest mountain can beused to extract climate information from the past it seemedtimely to revisit the mapping of the ice cover on Kibo throughtime Our motivation to do so has been to ensure the histor-ical record of these changes is of the highest possible qual-ity and in a format that should help facilitate on-going andfuture research dedicated to resolving the complexity of cli-mate variability and change in East Africa and beyond Thiswas achieved by using a consistent cartographic frameworkthat allows historical maps and new satellite imagery to beexamined The use of three-dimensional visualization tech-niques greatly enhanced our ability to establish the formerlocation of the glaciers

The revisited time series of glacial retreat on Kiliman-jaro contains eight epochs and spans almost 100 yr (1912ndash2011) The glaciers have retreated from their former extentof 1140 km2 in 1912 to 176 km2 in 2011 which representsa loss of about 85 of the ice cover over almost a 100-yr pe-riod Our estimate of the total retreat is consistent with recentassessments (Cullen et al 2006 Thompson et al 2009) butthe careful re-examination of the location and extent of indi-vidual ice bodies through time has allowed us to identify anumber of important differences from previously publishedmaps (eg Hastenrath and Greischar 1997 Thompson et al2002 2009) Historical evidence has been used to supportour interpretation with the detailed observations and pho-tographs obtained by early explorers on Kilimanjaro found tobe extremely valuable (eg Meyer 1891 1900 Uhlig 1904Jaeger 1909 Klute 1914 1920 1921 Gillman 1923 Nils-son 1931 Geilinger 1936) It has to be acknowledged that

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

430 N J Cullen et al A century of ice retreat on Kilimanjaro

much of what we know about Kilimanjaro today was estab-lished through their efforts

To further explore the importance of different glacierregimes on Kibo (Kaser et al 2004) a geo-statistical ap-proach was applied which resulted in the segmentation ofthe ice cover into three broad glacial zones A careful ex-amination of the retreat of ice in each of these glacier zonesindicates that there is a high probability that most of the re-maining ice cover on Kilimanjaro will disappear by 2040and it is highly unlikely that any ice body will remain af-ter 2060 if present retreat rates are maintained Before thishappens it would still be useful if the remaining ice bodiesin particular the remnants of the northern ice field could befurther examined to help resolve existing uncertainties aboutthe mechanisms responsible for controlling the formation ofthe vertical cliffs that characterize them (eg Winkler et al2010) In addition further efforts are required to scrutinizethe exact timing of the onset of the present glacier retreat firstreported by Hans Meyer in the late 19th century If we canrefine our understanding of the atmospheric controls behindthe most recent shift from glacier growth to decay on Kibowe will be in a stronger position to characterize the climaticcontrols of past and much larger glaciations on Africarsquos high-est mountain

Supplementary material related to this article isavailable online athttpwwwthe-cryospherenet74192013tc-7-419-2013-supplementpdf

AcknowledgementsThis research is funded by the Departmentof Geography and School of Surveying University of OtagoNew Zealand Thomas Molg was supported by the Alexander vonHumboldt Foundation The contributions made by Georg Kaserand Michael Winkler were supported by the Austrian ScienceFund (FWF) Projects P 20089-N10 and P 21288-N10 Thisresearch would not have been possible without the support from thefollowing Tanzanian authorities (COSTECH KINAPA TANAPATAWIRI) We are grateful to the referees that took the time toprovide feedback on this research in particular the careful andthoughtful comments by Henry Brecher

Edited by M Schneebeli

References

Chiang J C H The tropics in paleoclimate Annu Rev EarthPlanet Sci 37 263ndash297 2009

Cullen N J Molg T Kaser G Hussein K Steffen Kand Hardy D R Kilimanjaro Glaciers Recent areal ex-tent from satellite data and new interpretation of observed20th century retreat rates Geophys Res Lett 33 L16502doi1010292006GL027084 2006

Cullen N J Molg T Kaser G Steffen K and Hardy D REnergy-balance model validation on the top of Kilimanjaro Tan-zania using eddy covariance data Ann Glaciol 46 227ndash233doi103189172756407782871224 2007

Downie C Glaciations of Mount Kilimanjaro Northeast Tan-ganyika Geol Soc Am Bull 75 1ndash16doi1011300016-7606(1964)75[1GOMKNT]20CO2 1964

Downie C and Wilkinson P The Geology of KilimanjaroSheffield Geology Department University of Sheffield 1972

Downie C Humphries D W Wilcockson W H and WilkinsonP Geology of Kilimanjaro Nature 178 828ndash830 1956

Farr T G Rosen P A Caro E Crippen R Duren R Hens-ley S Kobrick M Paller M Rodriguez E Roth L Seal DShaffer S Shimada J Umland J Werner M Oskin M Bur-bank D and Alsdorf D The Shuttle Radar Topography Mis-sion Rev Geophys 45 RG2004doi1010292005RG0001832007

Geilinger W The retreat of the Kilimanjaro glaciers TanganyikaNotes and Records 2 7ndash20 1936

Gillman C An ascent of Kilimanjaro Geogr J 61 1ndash27 1923Hastenrath S The Glaciers of Equatorial East Africa Reidel Dor-

drecht 1984Hastenrath S Diagnosing the decaying glaciers of equatorial

East Africa Meteorol Z 15 265ndash271doi1011270941-294820060106 2006

Hastenrath S Recession of equatorial glaciers a photo documen-tation Sundog Publishing Madison Wisconsin 2008

Hastenrath S and Greischar L Glacier recession on KilimanjaroEast Africa 1912ndash89 J Glaciol 43 455ndash459 1997

Humphries D Interim report on the glaciology and meteorology ofKilimanjaro Tanganyika JulyndashSept 1953 Geological Survey ofTanganyika Dodoma 1953

Humphries D Preliminary notes on the glaciology of KilimanjaroJ Glaciol 3 475ndash479 1959

Humphries D Glaciology and glacial history in The Geology ofKilimanjaro edited by Downie C and Wilkinson P SheffieldGeology Department University of Sheffield 31ndash71 1972

Jarvis A Reuter H I Nelson A and Guevara E Hole-filledSRTM for the globe Version 4 available from the CGIAR-CSISRTM 90 m Database (httpsrtmcsicgiarorg) 2008

Jaeger F Forschungen in den Hochregionen des KilimandscharoMitt Deutsch Schutzgebiet 22 113ndash197 1909

Kaser G Hardy D R Molg T Bradley R S and Hyera TM Modern glacier retreat on Kilimanjaro as evidence of climatechange Observations and facts Int J Climatol 24 329ndash3392004

Kaser G Molg T Cullen N J Hardy D R andWinkler M Is the decline of ice on Kilimanjaro un-precedented in the Holocene Holocene 20 1079ndash1091doi1011770959683610369498 2010

Klute F Forschungen am Kilimandscharo im Jahre 1912 GeogrZeitschrift 20 496ndash505 1914

Klute F Ergebnisse der Forschungen am Kilimandscharo 1912ReimerndashVohsen Berlin 1920

Klute F Die stereophotogrammetrische Aufnahme der Hochregio-nen des Kilimandscharo Zeitschrift Ges Erdkunde Berlin 56144ndash151 1921

Mackinder C W Hobley C Gillman C and Johnston H Anascent of Kilimanjaro Discussion Geogr J 61 1ndash27 1923

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 429

1900 1920 1940 1960 1980 2000 2020 2040 20600

05

1

15

2

25

3

35

4

45

5

Year

Are

al e

xten

t [km

2 ]

zone 1 R2=099

zone 2 R2=097

zone 3 R2=09980 confidence interval

4000

4000

4000

4500

4500

4500

45005000

5000

5000

50005000

5500

5500

5500

Fig 5 The areal change of glaciers and ice bodies based on thethree-member classification Linear regressions (thin solid lines) foreach of the three zones are shown as well as the 80 confidenceintervals for each of these relationships

processes controlling the growth andor decay of the southernice fields This has been achieved by Molg et al (2009b)who used observations and atmospherendashglacier modellingto show how critical precipitation humidity and fractionalcloud cover are on the energy and mass balance of the Ker-sten glacier (zone 3) These results compare favourably toearly accounts of the importance of precipitation (moisturechange) and cloud cover on the asymmetry (Meyer 1891Klute 1920 Gillman 1923 Geilinger 1936) with the diur-nal variability of cloud cover being especially important inproviding protection to the glaciers on the western side ofthe mountain (eg Meyer 1891 Hastenrath and Greischar1997)

Recent regional atmospheric modelling experiments haveprovided a more quantitative assessment to account for theasymmetry in glacier mass on Kibo demonstrating that thewest and south-western areas of the mountain are favourableto moisture convergence as controlled by atmospheric dy-namics (Molg et al 2009a) As noted by Osmaston (1989)the asymmetry is not confined to the most recent glacier ex-tents with the pattern of ice distribution established fromformer moraines indicating that recent climatic controls onglacier behaviour are similar to those in the past This sug-gests that establishing the specific nature of the atmosphericcontrols responsible for the most recent shift from glaciergrowth to decay in particular in zones 2 and 3 (plateau andsouthern slope regions) remains critical if we are to useglacier behaviour to reconstruct climatic conditions of thedeeper past

Finally it is difficult not to speculate on when the ice onKibo will disappear which becomes more achievable if onerecognizes the importance of the different glacial zones onthe mountain If one accepts linear extrapolation based on

changes from the past is suitable to make projections for thefuture it is apparent that the ice cover in zone 1 (western re-gion) will most likely be gone before 2020 and that ice inzone 2 (plateau) and zone 3 (southern slopes) will disappearsometime close to 2040 (Fig 5) It is likely that the plateauwill harbour the final remnants of ice on Kibo which couldremain slightly beyond the middle of the 21st century Inter-estingly our linear extrapolation for the loss of ice is in closeagreement to earlier projections using radiation modelling(Molg et al 2003) Though the loss of ice will have no hy-drological significance to lowland areas (eg Gillman 1923Kaser et al 2004 Molg et al 2008b) there is no doubt thattheir disappearance will be a ldquogrievous aesthetic lossrdquo (Salt1951 p 160) which will undoubtedly impact the experiencetourists have on the mountain (Molg et al 2008b)

4 Conclusions

With reference to the glaciers found on Mt Kenya and Kiboin East Africa C Gillman remarked ldquowe might perhapslearn a good deal as to the general world circulation fromcareful and prolonged observation on these particular sum-mitsrdquo (Mackinder 1923 p 23) This statement still holdstrue today (eg Molg et al 2006) and with the increasingawareness that glaciers on Africarsquos highest mountain can beused to extract climate information from the past it seemedtimely to revisit the mapping of the ice cover on Kibo throughtime Our motivation to do so has been to ensure the histor-ical record of these changes is of the highest possible qual-ity and in a format that should help facilitate on-going andfuture research dedicated to resolving the complexity of cli-mate variability and change in East Africa and beyond Thiswas achieved by using a consistent cartographic frameworkthat allows historical maps and new satellite imagery to beexamined The use of three-dimensional visualization tech-niques greatly enhanced our ability to establish the formerlocation of the glaciers

The revisited time series of glacial retreat on Kiliman-jaro contains eight epochs and spans almost 100 yr (1912ndash2011) The glaciers have retreated from their former extentof 1140 km2 in 1912 to 176 km2 in 2011 which representsa loss of about 85 of the ice cover over almost a 100-yr pe-riod Our estimate of the total retreat is consistent with recentassessments (Cullen et al 2006 Thompson et al 2009) butthe careful re-examination of the location and extent of indi-vidual ice bodies through time has allowed us to identify anumber of important differences from previously publishedmaps (eg Hastenrath and Greischar 1997 Thompson et al2002 2009) Historical evidence has been used to supportour interpretation with the detailed observations and pho-tographs obtained by early explorers on Kilimanjaro found tobe extremely valuable (eg Meyer 1891 1900 Uhlig 1904Jaeger 1909 Klute 1914 1920 1921 Gillman 1923 Nils-son 1931 Geilinger 1936) It has to be acknowledged that

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013

430 N J Cullen et al A century of ice retreat on Kilimanjaro

much of what we know about Kilimanjaro today was estab-lished through their efforts

To further explore the importance of different glacierregimes on Kibo (Kaser et al 2004) a geo-statistical ap-proach was applied which resulted in the segmentation ofthe ice cover into three broad glacial zones A careful ex-amination of the retreat of ice in each of these glacier zonesindicates that there is a high probability that most of the re-maining ice cover on Kilimanjaro will disappear by 2040and it is highly unlikely that any ice body will remain af-ter 2060 if present retreat rates are maintained Before thishappens it would still be useful if the remaining ice bodiesin particular the remnants of the northern ice field could befurther examined to help resolve existing uncertainties aboutthe mechanisms responsible for controlling the formation ofthe vertical cliffs that characterize them (eg Winkler et al2010) In addition further efforts are required to scrutinizethe exact timing of the onset of the present glacier retreat firstreported by Hans Meyer in the late 19th century If we canrefine our understanding of the atmospheric controls behindthe most recent shift from glacier growth to decay on Kibowe will be in a stronger position to characterize the climaticcontrols of past and much larger glaciations on Africarsquos high-est mountain

Supplementary material related to this article isavailable online athttpwwwthe-cryospherenet74192013tc-7-419-2013-supplementpdf

AcknowledgementsThis research is funded by the Departmentof Geography and School of Surveying University of OtagoNew Zealand Thomas Molg was supported by the Alexander vonHumboldt Foundation The contributions made by Georg Kaserand Michael Winkler were supported by the Austrian ScienceFund (FWF) Projects P 20089-N10 and P 21288-N10 Thisresearch would not have been possible without the support from thefollowing Tanzanian authorities (COSTECH KINAPA TANAPATAWIRI) We are grateful to the referees that took the time toprovide feedback on this research in particular the careful andthoughtful comments by Henry Brecher

Edited by M Schneebeli

References

Chiang J C H The tropics in paleoclimate Annu Rev EarthPlanet Sci 37 263ndash297 2009

Cullen N J Molg T Kaser G Hussein K Steffen Kand Hardy D R Kilimanjaro Glaciers Recent areal ex-tent from satellite data and new interpretation of observed20th century retreat rates Geophys Res Lett 33 L16502doi1010292006GL027084 2006

Cullen N J Molg T Kaser G Steffen K and Hardy D REnergy-balance model validation on the top of Kilimanjaro Tan-zania using eddy covariance data Ann Glaciol 46 227ndash233doi103189172756407782871224 2007

Downie C Glaciations of Mount Kilimanjaro Northeast Tan-ganyika Geol Soc Am Bull 75 1ndash16doi1011300016-7606(1964)75[1GOMKNT]20CO2 1964

Downie C and Wilkinson P The Geology of KilimanjaroSheffield Geology Department University of Sheffield 1972

Downie C Humphries D W Wilcockson W H and WilkinsonP Geology of Kilimanjaro Nature 178 828ndash830 1956

Farr T G Rosen P A Caro E Crippen R Duren R Hens-ley S Kobrick M Paller M Rodriguez E Roth L Seal DShaffer S Shimada J Umland J Werner M Oskin M Bur-bank D and Alsdorf D The Shuttle Radar Topography Mis-sion Rev Geophys 45 RG2004doi1010292005RG0001832007

Geilinger W The retreat of the Kilimanjaro glaciers TanganyikaNotes and Records 2 7ndash20 1936

Gillman C An ascent of Kilimanjaro Geogr J 61 1ndash27 1923Hastenrath S The Glaciers of Equatorial East Africa Reidel Dor-

drecht 1984Hastenrath S Diagnosing the decaying glaciers of equatorial

East Africa Meteorol Z 15 265ndash271doi1011270941-294820060106 2006

Hastenrath S Recession of equatorial glaciers a photo documen-tation Sundog Publishing Madison Wisconsin 2008

Hastenrath S and Greischar L Glacier recession on KilimanjaroEast Africa 1912ndash89 J Glaciol 43 455ndash459 1997

Humphries D Interim report on the glaciology and meteorology ofKilimanjaro Tanganyika JulyndashSept 1953 Geological Survey ofTanganyika Dodoma 1953

Humphries D Preliminary notes on the glaciology of KilimanjaroJ Glaciol 3 475ndash479 1959

Humphries D Glaciology and glacial history in The Geology ofKilimanjaro edited by Downie C and Wilkinson P SheffieldGeology Department University of Sheffield 31ndash71 1972

Jarvis A Reuter H I Nelson A and Guevara E Hole-filledSRTM for the globe Version 4 available from the CGIAR-CSISRTM 90 m Database (httpsrtmcsicgiarorg) 2008

Jaeger F Forschungen in den Hochregionen des KilimandscharoMitt Deutsch Schutzgebiet 22 113ndash197 1909

Kaser G Hardy D R Molg T Bradley R S and Hyera TM Modern glacier retreat on Kilimanjaro as evidence of climatechange Observations and facts Int J Climatol 24 329ndash3392004

Kaser G Molg T Cullen N J Hardy D R andWinkler M Is the decline of ice on Kilimanjaro un-precedented in the Holocene Holocene 20 1079ndash1091doi1011770959683610369498 2010

Klute F Forschungen am Kilimandscharo im Jahre 1912 GeogrZeitschrift 20 496ndash505 1914

Klute F Ergebnisse der Forschungen am Kilimandscharo 1912ReimerndashVohsen Berlin 1920

Klute F Die stereophotogrammetrische Aufnahme der Hochregio-nen des Kilimandscharo Zeitschrift Ges Erdkunde Berlin 56144ndash151 1921

Mackinder C W Hobley C Gillman C and Johnston H Anascent of Kilimanjaro Discussion Geogr J 61 1ndash27 1923

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

430 N J Cullen et al A century of ice retreat on Kilimanjaro

much of what we know about Kilimanjaro today was estab-lished through their efforts

To further explore the importance of different glacierregimes on Kibo (Kaser et al 2004) a geo-statistical ap-proach was applied which resulted in the segmentation ofthe ice cover into three broad glacial zones A careful ex-amination of the retreat of ice in each of these glacier zonesindicates that there is a high probability that most of the re-maining ice cover on Kilimanjaro will disappear by 2040and it is highly unlikely that any ice body will remain af-ter 2060 if present retreat rates are maintained Before thishappens it would still be useful if the remaining ice bodiesin particular the remnants of the northern ice field could befurther examined to help resolve existing uncertainties aboutthe mechanisms responsible for controlling the formation ofthe vertical cliffs that characterize them (eg Winkler et al2010) In addition further efforts are required to scrutinizethe exact timing of the onset of the present glacier retreat firstreported by Hans Meyer in the late 19th century If we canrefine our understanding of the atmospheric controls behindthe most recent shift from glacier growth to decay on Kibowe will be in a stronger position to characterize the climaticcontrols of past and much larger glaciations on Africarsquos high-est mountain

Supplementary material related to this article isavailable online athttpwwwthe-cryospherenet74192013tc-7-419-2013-supplementpdf

AcknowledgementsThis research is funded by the Departmentof Geography and School of Surveying University of OtagoNew Zealand Thomas Molg was supported by the Alexander vonHumboldt Foundation The contributions made by Georg Kaserand Michael Winkler were supported by the Austrian ScienceFund (FWF) Projects P 20089-N10 and P 21288-N10 Thisresearch would not have been possible without the support from thefollowing Tanzanian authorities (COSTECH KINAPA TANAPATAWIRI) We are grateful to the referees that took the time toprovide feedback on this research in particular the careful andthoughtful comments by Henry Brecher

Edited by M Schneebeli

References

Chiang J C H The tropics in paleoclimate Annu Rev EarthPlanet Sci 37 263ndash297 2009

Cullen N J Molg T Kaser G Hussein K Steffen Kand Hardy D R Kilimanjaro Glaciers Recent areal ex-tent from satellite data and new interpretation of observed20th century retreat rates Geophys Res Lett 33 L16502doi1010292006GL027084 2006

Cullen N J Molg T Kaser G Steffen K and Hardy D REnergy-balance model validation on the top of Kilimanjaro Tan-zania using eddy covariance data Ann Glaciol 46 227ndash233doi103189172756407782871224 2007

Downie C Glaciations of Mount Kilimanjaro Northeast Tan-ganyika Geol Soc Am Bull 75 1ndash16doi1011300016-7606(1964)75[1GOMKNT]20CO2 1964

Downie C and Wilkinson P The Geology of KilimanjaroSheffield Geology Department University of Sheffield 1972

Downie C Humphries D W Wilcockson W H and WilkinsonP Geology of Kilimanjaro Nature 178 828ndash830 1956

Farr T G Rosen P A Caro E Crippen R Duren R Hens-ley S Kobrick M Paller M Rodriguez E Roth L Seal DShaffer S Shimada J Umland J Werner M Oskin M Bur-bank D and Alsdorf D The Shuttle Radar Topography Mis-sion Rev Geophys 45 RG2004doi1010292005RG0001832007

Geilinger W The retreat of the Kilimanjaro glaciers TanganyikaNotes and Records 2 7ndash20 1936

Gillman C An ascent of Kilimanjaro Geogr J 61 1ndash27 1923Hastenrath S The Glaciers of Equatorial East Africa Reidel Dor-

drecht 1984Hastenrath S Diagnosing the decaying glaciers of equatorial

East Africa Meteorol Z 15 265ndash271doi1011270941-294820060106 2006

Hastenrath S Recession of equatorial glaciers a photo documen-tation Sundog Publishing Madison Wisconsin 2008

Hastenrath S and Greischar L Glacier recession on KilimanjaroEast Africa 1912ndash89 J Glaciol 43 455ndash459 1997

Humphries D Interim report on the glaciology and meteorology ofKilimanjaro Tanganyika JulyndashSept 1953 Geological Survey ofTanganyika Dodoma 1953

Humphries D Preliminary notes on the glaciology of KilimanjaroJ Glaciol 3 475ndash479 1959

Humphries D Glaciology and glacial history in The Geology ofKilimanjaro edited by Downie C and Wilkinson P SheffieldGeology Department University of Sheffield 31ndash71 1972

Jarvis A Reuter H I Nelson A and Guevara E Hole-filledSRTM for the globe Version 4 available from the CGIAR-CSISRTM 90 m Database (httpsrtmcsicgiarorg) 2008

Jaeger F Forschungen in den Hochregionen des KilimandscharoMitt Deutsch Schutzgebiet 22 113ndash197 1909

Kaser G Hardy D R Molg T Bradley R S and Hyera TM Modern glacier retreat on Kilimanjaro as evidence of climatechange Observations and facts Int J Climatol 24 329ndash3392004

Kaser G Molg T Cullen N J Hardy D R andWinkler M Is the decline of ice on Kilimanjaro un-precedented in the Holocene Holocene 20 1079ndash1091doi1011770959683610369498 2010

Klute F Forschungen am Kilimandscharo im Jahre 1912 GeogrZeitschrift 20 496ndash505 1914

Klute F Ergebnisse der Forschungen am Kilimandscharo 1912ReimerndashVohsen Berlin 1920

Klute F Die stereophotogrammetrische Aufnahme der Hochregio-nen des Kilimandscharo Zeitschrift Ges Erdkunde Berlin 56144ndash151 1921

Mackinder C W Hobley C Gillman C and Johnston H Anascent of Kilimanjaro Discussion Geogr J 61 1ndash27 1923

The Cryosphere 7 419ndash431 2013 wwwthe-cryospherenet74192013

N J Cullen et al A century of ice retreat on Kilimanjaro 431

McLachlan G and Peel D Finite Mixture Models John Wileyand Sons New York 2000

Messerli B Mountain glaciers in the Mediterranean area and inAfrica in Proceedings of the Riederalp Workshop September1978 IAHS-AISH Publ no 126 197ndash211 1980

Messner R Free spirit a climberrsquos life Hodder and StoughtonGreat Britain 1991

Meyer H Ascent to the Summit of Kilima-Njaro P Roy GeogrSoc 12 331ndash345 1890

Meyer H Across East Africa Glaciers G Philip and Son London1891

Meyer H Der Kilimandscharo Reimer-Vohsen Berlin 1900Molg T and Hardy D R Ablation and associated energy balance

of a horizontal glacier surface on Kilimanjaro J Geophys Res109 D16104doi1010292003JD004338 2004

Molg T and Kaser G A new approach to resolving climate-cryosphere relations Downscaling climate dynamics to glacier-scale mass and energy balance without statistical scale linking JGeophys Res 116 D16101doi1010292011JD015669 2011

Molg T Hardy D R and Kaser G Solar-radiation main-tained glacier recession on Kilimanjaro drawn from combinedice-radiation geometry modeling J Geophys Res 108 4731doi1010292003JD003546 2003

Molg T Renold M Vuille M Cullen N J Stocker TF and Kaser G Indian ocean zonal mode activity ina multicentury-integration of a coupled AOGCM consistentwith climate proxy data Geophys Res Lett 33 L18710doi1010292006GL026384 2006

Molg T Cullen N J Hardy D R Kaser G and Klok LMass balance of a slope glacier on Kilimanjaro and its sensitivityto climate Int J Climatol 28 881ndash892doi101002joc15892008a

Molg T Hardy D R Cullen N J and Kaser G Tropicalglaciers climate change and society Focus on Kilimanjaro (EastAfrica) in The Darkening Peaks Glacial Retreat in Scientificand Social Context edited by Orlove B Wiegandt E andLuckman B University of California Press Berkeley London168ndash182 2008b

Molg T Chiang J H C Gohm A and Cullen N J Tem-poral precipitation variability versus altitude on a tropical highmountain Observations and mesoscale atmospheric modelingQ J Roy Meteorol Soc 135 1439ndash1455doi101002qj4612009a

Molg T Cullen N J Hardy D R Winkler M and Kaser GQuantifying climate change in the tropical mid-troposphere overEast Africa from glacier shrinkage on Kilimanjaro J Clim 224162ndash4181doi1011752009JCLI29541 2009b

Molg T Grosshauser M Hemp A Hofer M and MarzeionB Limited forcing of glacier loss through land coverchange on Kilimanjaro Nat Clim Change 2 254ndash258doi101038nclimate1390 2012

Nilsson E Quaternary glaciations and pluvial lakes in British EastAfrica Geogr Ann 13 249ndash349 1931

Osmaston H Glaciers glaciations and equilibrium line altitudeson Kilimanjaro in Quarternary and environmental research onEast African mountains edited by Mahaney W C BalkemaRotterdam 7ndash30 1989

Racoviteanu A E Paul F Raup B Khalsa S J S and Arm-strong R Challenges and recommendations in mapping ofglacier parameters from space results of the 2008 Global LandIce Measurements from Space (GLIMS) workshop BoulderColorado USA Ann Glaciol 50 53ndash69 2009

Salt G The Shira Plateau of Kilimanjaro Geogr J 117 150ndash1641951

Sampson D N The geology volcanology and glaciology of Kil-imanjaro in Guide Book to Mount Kenya and Kilimanjaroedited by Mitchell J The Mountain Club of Kenya NairobiKenya 155ndash171 1971

Thompson L G Mosley-Thompson E Davis M E HendersonK A Brecher H H Zagorodnov V S Mashiotta T A LinP-N Mikhalenko V N Hardy D R and Beer J Kilimanjaroice core records evidence of Holocene climate change in tropicalAfrica Science 298 589ndash593 2002

Thompson L G Brecher H H Mosley-Thompson E HardyD R and Mark B G Glacier loss on Kilimanjaro continuesunabated P Natl Acad Sci 106 19770ndash19775 2009

Uhlig C Vom Kilimandscharo zum Meru Zeitschrift Ges Erd-kunde Berlin 627ndash650 692ndash718 1904

Winkler M Kaser G Cullen N J Molg T Hardy D R andPfeffer W T Land-based marginal ice cliffs Focus on Kiliman-jaro Erdkunde 64 179ndash193doi103112erdkunde201002052010

wwwthe-cryospherenet74192013 The Cryosphere 7 419ndash431 2013