spatio-temporal rainfall patterns in northern ghana€¦ · acknowledgments 2 acknowledgments...

81
Geographische Institute der Rheinischen Friedrich-Wilhelms-Universität Bonn Spatio-temporal Rainfall Patterns in Northern Ghana Diploma Thesis by Jan Friesen Supervised by Prof. Dr. B. Diekkrüger Bonn, June 2002

Upload: others

Post on 05-Jul-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Geographische Institute der Rheinischen Friedrich-Wilhelms-Universität Bonn

Spatio-temporal Rainfall Patterns in Northern Ghana

Diploma Thesis by

Jan Friesen

Supervised by Prof. Dr. B. Diekkrüger

Bonn, June 2002

Page 2: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Acknowledgments

2

Acknowledgments

Special thanks goes to the staff of the Center for Development Research and the GLOWA

Volta Project, above all Dr. Nick van de Giesen, the project coordinator, with whom

uncountable discussions took place.

Many thanks go to the staff and local partners of the GLOWA Volta Project in Ghana,

especially Dr. Marc Andreini, the local coordinator, for his invaluable technical and scientific

support in the field, Mr. Salisu Adams, a project driver, for his extensive local knowledge and

tremendous assistance with the local people, and Mr. C.N. Kasei, the meteorological science

officer of SARI, for numerous talks and insights into the local climate of northern Ghana.

Last but not least I would like to give thanks to my mother for her great support.

Page 3: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Outline

3

Outline Page

Acknowledgments……………………………………………...………………………... 2

Outline…………………………………………………………………………………… 3

Lists of figures & tables…...………………………...……...…………………………… 5

1. Introduction……………………………..…………………………………………… 7

1.1 General introduction…………………………………….………………………… 7

1.2 West African climatology……………………………….………………………… 8

1.2.1 Major air mass systems……………………………….……...……………… 10

1.2.2 Weather zones……………………………….……………….……………… 11

1.2.3 Precipitation regimes and patterns………………………………...………… 14

1.2.3.1 Togo Gap………………………………….………………………… 17

1.2.4 Precipitation types………………………………....………………………… 18

1.2.4.1 Convective systems………………………………..………………… 19

1.2.4.1.1 Rainstorms……………...……….………………………… 19

1.2.4.1.2 Squall lines………………...…….………………………… 19

1.2.4.2 Monsoon rains…...…………………………….………………..…… 22

2. Methodology……………………………….………………………………………… 23

2.1 Test site……………………………….…………………………………………… 23

2.2 Rain gage network……………………………………....………………………… 25

2.3 Rain gages……………………………….………………………………………… 26

2.4 Calibration……………………………….………………………………………… 28

2.5 Measurement uncertainties……………………………….……………………….. 28

2.6 Additional data……………………………….……………………….…………… 29

2.6.1 SARI weather station………………………...……………………….…...… 29

2.6.2 MM5 precipitation data……………………………….…………………...… 29

2.6.3 TRMM precipitation radar……………………………….………………….. 31

2.7 Malfunctions……………………………….…...…………………….…………… 33

Page 4: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Outline

4

3. Data preparation……………………………….…...…………….…….…………… 34

3.1 Data organization……………………………...…………………….…………..… 34

3.2 Pre-processing……………………………….…..………………….………...…… 36

3.2.1 Selection……………………………….…...…………………….…….…… 36

3.2.2 Bias correction……………………………….…...…………………….…… 37

3.2.3 Left/ right error correction……………………………….…...…………...… 39

3.3 External data preparation……………………………….…...…………………..… 41

3.3.1 SARI precipitation……………………………….…...…………………..…. 41

3.3.2 MM5 precipitation……………………………….…...…………………..…. 41

3.3.3 TRMM precipitation radar scenes……………………………….………..… 42

4. Data analysis……………………………….…...…………………….……...….…… 43

4.1 Event analysis……………………………….…...…………………….………..… 43

4.1.1 3D network analysis……………………………………….…………..……. 48

4.2 Intensity analysis………………………………………..…...……………………. 50

4.3 Combination of event and intensity results………………………………..…...…. 53

4.3.1 Precipitation type characteristics……………………………….………...…. 53

4.3.2 Event type classification……………………………….…...…….…………. 54

4.4 External data comparison……………………………….…...……………………. 56

4.4.1 Comparison SARI – Observed……………………………….…...…...……. 56

4.4.2 Comparison MM5 – Observed……………………………….….….………. 57

4.4.3 TRMM analysis………………………….……………….…...…….………. 59

5. Conclusion………………………………………………………..…....…….………. 63

Literature……………………………….………………………………..………………. 64

Appendices…………………………….….…...………………………………...………. 67

Appendix A……………………………….…...………………………………...………. 69

Appendix B……………………………….…...…….……………………………..……. 80

Page 5: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Lists of figures & tables

5

List of figures: PageFigure 1 Volta basin within West Africa……………………………………….. 9

Figure 2 Location of the surface ITD and air mass distribution…………….….. 11

Figure 3 Weather zones……………………………………………………...….. 12

Figure 4 Position of the weather zones in Ghana…………………………...…... 13

Figures 5ab Mean precipitation (mm) and surface ITD position

Figure 5a January…………………………...………………………………...….. 15

Figure 5b August……………………………………………………………...….. 15

Figure 6 Mean annual rainfall distribution in Ghana………………………..….. 16

Figure 7 Precipitation differences Europe tropical Africa……………….….. 18

Figures 8ab Squall line generation

Figure 8a Cross-sectional formation……………..………………………….….... 20

Figure 8b Surface and vertical squall line sketches...……………………….….... 20

Figure 9 Test site location………………………………………………………. 23

Figures 10ab Dry and wet season at the test site

Figure 10a Test site dry season……………………………………………….….... 24

Figure 10b Test site wet season…………………………………...………….….... 24

Figure 11 Rain gage networks A and B………………………………………….. 25

Figure 12 Tipping-bucket gage…………………………………………….…...... 26

Figures 13ab Rain gage design

Figure 13a Rain gage scheme………………………………………………...….... 27

Figure 13b Rain gage in the field……………………………………………..….... 27

Figures 14ab Gage calibration

Figure 14a Field calibration…………………………………………………..….... 28

Figure 14b Indoor calibration………………………………………………...….... 28

Figures 15ab MM5 domains and sample hindcast

Figure 15a MM5 Domains…………………………………………….………....... 30

Figure 15b Sample MM5 hindcast…...……………………………………….….... 30

Figure 16 TRMM orbits 10. Sept. 2001……………….…………………….….... 32

Figures 17ab Data logger and EXCEL gage files

Figure 17a Raw data logger file…...………………………………………….….... 35

Figure 17b EXCEL gage file…...…………………………………………….….... 35

Figures 18ab Master spreadsheet and graph

Figure 18a Master file…...…………………………………………...……….….... 35

Page 6: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Lists of figures & tables

6

Figure 18b Master graph…...……………………………………………...….….... 35

Figure 19 Sample selection…...……………………………………….…….….... 37

Figure 20 Sample Bias…...………………………………………………….….... 38

Figure 21 Error reversal…...……………………..………………………….….... 40

Figure 22 Extracted MM5 data points…...………………………………….….... 41

Figure 23 Rainfall events sorted by size…...………………………………..….... 45

Figure 24 Totals and CV of different network combinations…...…………..….... 47

Figure 25 3D event patterns…...…………………………...………………..….... 48

Figures 26ab Rainstorm intensity graphs

Figure 26a Single cell rainstorm…...………………………….……………..…..... 51

Figure 26b Multiple cell rainstorm…...…………………………...………..……... 51

Figure 27 Monsoon rain intensity graph…...…………………………...…….….. 51

Figure 28 Squall line intensity graph…...…………………………...…………… 52

Figure 29 Comparison - Networks A and B, SARI…...……………………….… 56

Figure 30 MM5 vs. Observed rain days – September 2001…...……………….… 57

Figure 31 Rainfall per class (MM5 – Observed) …...………………………….... 58

Figure 32 Scatterplot MM5 vs. Observed…...…………………………….……… 58

Figure 33 TRMM sample scenes…...…………………………...……….………. 59

Figures 34abc TRMM scenes and corresponding intensity charts

Figure 34a Event 0720, date: 6. Sept. 2001…...…………………………….…….. 61

Figure 34b Event 0940, date 8. Sept. 2001…...…………………………....…….... 61

Figure 34c Event 1487, date 15. Sept. 2001…...……………………………..….... 61

List of tables : PageTable 1 Bias correction…...………………………………………………..…………. 37

Table 2 Calibration results…...……………………………...………………………... 39

Table 3 Event spreadsheet overview…...……………………………...………..……. 44

Table 4 Event distribution…...……………………………...………..…………..…... 46

Table 5 Coefficients of variation…...……………………………...…………..……... 46

Table 6 Precipitation type characteristics..…………………………...………..……... 53

Table 7 Event statistics…...…………………………………...……...………..……... 54

Table 8 Precipitation type statistics…...……………………………...………..……... 57

Table 9 Line Squall velocities…...…………………………………...………..……... 60

Page 7: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 1 – Introduction

7

1. Introduction

1.1 General Introduction

Rainfall in interior West Africa is subject to great variability, as well spatial as temporal,

expressed by dry seasons of 6 months continuation, enormous interannual variability and the

partially extreme small-scale nature of the precipitation. These uncertainties in water

availability and the growing demand through rapid population increase pose a huge hindrance

to agricultural, household and industrial needs.

This study is conducted to analyze small-scale rainfall in the northern region of Ghana. Main

objectives are the descriptions of observed rainfall patterns and their identification into

existent precipitation types within the surveyed region.

It is part of the GLOWA Volta Project that studies the “Sustainable Water Use under

Changing Land Use, Rainfall Reliability, and Water Use in the Volta Basin”. The main

physical research is done at three experimental watersheds located at Navrongo in the

northern part of Ghana, Tamale, depicted in figure 1, and Ejura in the southern part of the

basin.

With an interdisciplinary approach this project focuses on the “development of a scientifically

sound decision support system for the assessment, sustainable use and development of water

resources in the Volta Basin”. This decision support system will be the product of several

socio-economic and physical models that are currently being build and will later be linked

together. Rainfall, as the most important source of water in the region, is one of the main

physical components used as a direct and indirect input into these models, as for example it is

the basis for river flow modeling as well as available household and irrigation water.

Within the project a weather model, MM5, provides the meteorologic parameters. As the

distribution of climatic stations over West Africa is very coarse, the model needs ground data

for optimization and validation which is for example provided by the spatial variability of

rainfall within the observed network.

Page 8: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 1 – Introduction

8

Another focus of the study is the identification of different rainfall types and their

characteristics from a hydrological point of view. Unlike meteorology, hydrology usually

works on much finer scales, and thus has the need for small scale rainfall data. One important

aspect is the spatial variability but especially in terms of soil saturation and resulting surface

runoff rainfall intensity plays a major role. Therefore a second main aspect is the analysis of

intensity curves of the observed rainfall events, as well as the analysis of precipitation types

and their occurrence.

Chapter one concentrates on the introduction into West African climatology, with special

emphasis on the precipitation types and their characteristics.

Chapter two, as the methodological part, includes a deeper insight into the test site and the

installed equipment, as well as descriptions of external data used in the analytical part, such as

additional local rainfall data, simulated precipitation data from the MM5 weather model, and

TRMM precipitation radar data.

The data organization and pre-processing of collected rainfall data, and the external data

preparation are explained in Chapter three.

In the fourth chapter the analytical work is described. It starts with the analysis of the

observed rainfall events, especially their distribution and classification over the observed time

period, then goes on with an analysis of the single events by looking at the intensity patterns

during the events. After a combinatory part of the above conducted steps that classifies the

events into the different rainfall origins a comparison with SARI and MM5 data is

undertaken. The last step focuses on the combination of the observed rainfall patterns with

TRMM precipitation radar scenes.

Chapter five gives a conclusion of the performed analytical steps.

1.2 West African climatology

As this study concentrates on the Volta basin with special emphasis to northern Ghana the

climatological overview is focused on this region.

West Africa as shown in figure 1 below covers an area between 20°W and 20°E longitudal

and 20°N and 0° latitudal spread.

Page 9: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 1 – Introduction

9

Figure 1: Volta Basin within West Africa

West Africa is situated in the tropics covering four tropical regions. Humid areas like the

equatorial forests in the south, semi-humid areas as the Guinea savanna, semi-arid areas like

the Sudan savanna, and the arid Sahel in the north.

The Volta Basin (see figure 1) lies almost at the center of the region and with an extend of

about 15°N to 5°N latitude and 6°W to 3°E longitude it covers parts of the equatorial forest

zone, mainly the Guinea and Sudan savanna, and a small fraction of the Sahel zone. It

comprises an area of roughly 400,000 km² and is considered one of the benchmark watersheds

for Africa. (Rijsberman, 2002)

This chapter concentrates on describing general climatologic processes during the rainy

season. Its main purpose is to give an overview and to point out characteristics of the different

rainfall types found in interior West Africa. As pointed out earlier the Volta Basin lies within

the tropics thus having two climatic seasons, a dry and a wet period. According to

Acheampong (1988) a month considered “wet” has 102 mm of rainfall and the total rainfall

lies above the potential evaporation. For Ghana this means that the wet season varies from 8

months at Kumasi (southern central Ghana) to 2 months in Navrongo (northern Ghana).

However, as will be pointed out in the test site description, precipitation in these regions

underlies a great interannual variability and long-term rainfall changes depending on the

chosen time period, so that the number of wet months does not remain constant.

A more detailed view of the 2001 rainy season will be given in the test site description, where

the emphasis is put on northern Ghana.

Page 10: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 1 – Introduction

10

1.2.1 Major air mass systems

West African climatology is mainly influenced by two major wind or air mass systems. The

south-western maritime air and the north-eastern continental air. Both systems and their

influence for the local climates are controlled by the position of the Inter-tropical

Discontinuity (ITD). Another system, greatly influencing especially the northern parts of

West Africa, is the African Easterly Jet (AEJ), overlying the former two air layers. Special

attention is given to this air flux, as it is one of the main triggers of squall lines, that are the

origins of the most rainfall in the region. Whereas the dry season is mainly influenced by the

north-eastern winds known as Harmattan, the wet season is mainly influenced by the south-

west monsoon, with the changing position of the ITD moving north from January until

August. A good description about the interaction between these Air masses and the AEJ is

given by Kamara.

“One basic feature of the lower atmosphere over West Africa during the rainy season (May -

November) is the thermally-induced low pressure cell over the Sahara with its axis located

between latitudes 18°N and 22°N. At the same time the Subtropical Anticyclone over the

south Atlantic intensifies and extends equator-wards. The resulting gradient between the two

pressure systems induces a south-westerly air stream (south-west monsoon) from the south

Atlantic. This characteristically warm, moist and convectively unstable airmass penetrates

deep into West Africa, reaching latitudes 20 – 22°N in August.

The depth of the south-west monsoon decreases inland from the Guinea Coast. For example at

Abijan it reaches a maximum level of 3000 meters in July while at Bamako it is only 1000

meters deep.

Above the south-west monsoon at a height of 3000 meters the winds assume a predominantly

easterly component. Near the 3km and 12km levels these upper easterly winds concentrate

into narrow bands of jet steams, known respectively as the African Easterly Jet (AEJ) and the

Easterly Tropical Jet (ETJ). The mid-tropospheric AEJ appears to originate from the strong

horizontal temperature gradient between hot Saharan air and the relatively cooler, but still

warm air of equatorial origin which penetrates to latitudes 10°N and 22°N. Its appearance in

the wind-fields coincides with periods of enhanced development of lower lever synoptic

disturbances such as easterly waves, line squalls and vortices. In the lower latitudes the jet

stream appears twice during the year, once during the Early Squall season of April-May and

during the Late Squalls of October-November. In the central and northern parts of West

Page 11: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 1 – Introduction

11

Africa it appears only once but coincides with a single but longer squall season that lasts from

May to September.

During the rainy season the ETJ lies near latitude 10°N, which incidentally is also the axis of

the rainbelt over West Africa at this time of the year.”( Kamara, 1986:49)

Figure 2: Location of the surface ITD and air mass distribution

(elevation in feet – 5000 ft. ≈ 1500 m) (Source: Ojo, 1977:65)

Hence, the migration of the ITD, and its influence on the distribution of the dry continental air

and the moist maritime air is fundamental to the understanding of West African climates. As

can be seen in figure 2 , the dry continental air to the north of the surface ITD overrides the

maritime air of southern origin, while the latter forms a wedge pointing north under the

former. As described by Kamara (1986) it can be seen that the thickness of the monsoonal

layer decreases northwards.

1.2.2 Weather zones

The above described movement of the ITD and the thereby induced change of weather are

associated with different weather zones. These are zones A through D/ E. Literature often

refers to definitions by Walker (1958, referred to in Weischet, 2000:262) or Hayward and

Oguntoyinbo (1987). Their boundaries and characteristics are still under discussion, as can be

seen in the different definitions by Ojo (1977), Leroux (2001), Weischet (2000), Hayward and

Oguntoyinbo (1987). Figure 3 shows an approach by Weischet (2000) illustrating the different

zones and their main characteristics. The following brief zone descriptions are based on the

definition by Hayward and Oguntoyinbo (1987). As the test site mainly lies in zone C during

the measuring period, zones C and D are discussed in more detail.

Page 12: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 1 – Introduction

12

Figure 3: Weather zones (Source: Weischet, 2000:263, modified)

Illustrated in figure 3 are the positions of the different weather zones in August, as defined by

Weischet (2000). As pointed out earlier the position of the ITD is the main axis defining these

zones. Figures 5 a and b, which describe the general rainfall patterns over West Africa and

can be found below, show a plan view of its north – south movement. ITD, as used in this

thesis, is defined as the surface boundary between dry northern air and the humid maritime

air. The humid maritime air forms a wedge under the northern air going from the equatorial

low pressure trench to the surface ITD, located between zones A and B (see figure 3).

As shown in figure 3, zone A lies north of the ITD. Weather is characterized by small clouds,

dry continental air, winds are north-easterly and rainfall is low.

Zone B is just south of the ITD with a spread of about 320 km. It is mostly rainless but due to

the humid maritime air precipitation events are mainly restricted to isolated thunderstorms.

Zone C extends for about 800 km south of zone B and significant changes occur in the rainfall

activity. Precipitation mainly occurs due to local convectional and squall line events.

Zone D stretches approximately 300 km south and with its high humidity and frequent rainfall

events it has a strong monsoonal character. This is shown in the shower times, often lasting

for several hours with less intensity than those of zone C.

Zone E is the southernmost zone, only affecting the coastal areas. Weather conditions are,

apart from notably less rain, comparable to zone D.

Page 13: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 1 – Introduction

13

After these brief overviews of the different zones based on Hayward and Oguntoyinbo (1987)

a more detailed look at zones C and D is provided. These two, as can be seen in figure 4

define the rainy season at the test site and their definitions differ within literature. To

emphasize the different definitions especially in terms of the zone spread the following part

concentrates on zone descriptions by Ojo (1977), Weischet (2000) and Leroux (2001).

Described are the general weather patterns and precipitation types within the zones which will

be referred to in more detail in the following part.

Figure 4: Position of the weather zones in Ghana (Source: Weischet, 2000:263, modified)

“Zone C is bounded on the north by zone B and extends southward by 600-800 km. The

boundaries are not at all well defined. It is periodically traversed by ‘disturbance lines’ which

may range from a well-defined line squall, through a line, possibly broken, of thunderstorms,

to a heavy belt of cloud without rain. These lines move at an average speed of 25 km/h and

move in almost any direction from south through west to northwest, although longer lines lie

roughly north-south and move towards the west. Somewhat shorter lines have a tendency to

move west-southwest. Rain is often associated with disturbance lines. Periods of rain are short

and of high intensity. […] The northern boundary of zone D is extremely ill-defined. The

zone lies south of zone C with an average width of about 322 km. It used to be regarded as

part of zone C, but there now appear to be significant differences between them. Days with

rain are the rule rather than the exception and rainfall tends to be more prolonged and less

intense. The belts of rain are orientated in an east-west direction and the surface winds are

southwesterly.” (Ojo, 1977:65)

Page 14: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 1 – Introduction

14

According to a description from Weischet (2000) zone B evolves in a smooth transition into

zone C, having a longitudinal spread of 500 to 650 km. The height of the lower monsoonal air

layer averages to 1.5-2 km, which makes the convective cloud generation here much more

intensive compared to convective processes in thicker air layers. Rainfall events are generally

connected to two types – convective storms and line squalls which are both discussed later on.

Zone C then develops into zone D with a longitudinal spread of approximately 300-500 km

and the moist maritime air layer adds up to a height of 2.5-3 km. Consequently, the air mass is

much more stable and rainfall is generally more stratiform, meaning that the average duration

is several hours and the intensity rather low compared to convective rainfall.

Another definition is provided by Leroux (2001), who divides zone C into C1 and C2,

basically by combining zones C and D, while C1 is associated with storms and line squall

rainfall whereas C2 links to continuous, monsoonal rainfall. Especially the differentiation

between zones C and D is not always clear, being combined, divided, and combined again

throughout literature. However, the precipitation types associated with them are either

monsoonal or convective whereas their occurrence and spatial spread differs slightly from

source to source.

1.2.3 Precipitation patterns and regimes

There are three climatic regimes in West Africa. Uni- or monomodal, bimodal, and pseudo-

bimodal. The test site and the larger part of the Volta Basin have a monomodal regime. By

definition the latter is characterized by one single rainfall peak. The maximum is usually

reached in August/ September and after a steady rise to the maximum the season has a rapid

decline to a complete cessation. As the monomodal regime has a large longitudinal spread

rainfall ranges from 500 mm to 3000 mm annually.

“The distribution of precipitation in tropical Africa is highly organized, and of great

functional simplicity. Unimodal regimes are essentially linked to the evolution of the

precipitable water potential of a flux which determines the pluviometric mode and the

efficacy of the agents (fixed or moving) utilizing this potential. Rain tends to fall in summer,

and the rainless season is completely dry inland; the lengthening of the rainless period

Page 15: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 1 – Introduction

15

progressively aggravates the vulnerability of marginal regions which experience the greatest

interannual variations.” (Leroux, 2001:360)

Precipitation patterns for West Africa can generally be divided into two large-scale patterns, a

more zonal distribution in the northern part and a distorted southern part. A description can be

found in Ojo (1977).

In July the pattern of rainfall distribution is distorted especially south of about latitude 12° N.

North of latitude 12°N the usual zonal pattern prevails, but to the south the distribution of

rainfall reflects the influence of local conditions, because these areas are approximately in

Walker’s zone C or D. Distortions in the pattern of rainfall distribution are created by such

local conditions as the highland and mountainous regions which receive more rainfall than for

example the central Ivory Coast and Ghana which receive less than other areas on the same

latitude.

The zonal respectively distorted patterns are illustrated in figures 5a and b. Especially in the

latter it can clearly be seen that the isohyets to the north are much more organized in an east to

west direction whereas the southern isohyets are more clustered and intermitted.

5a: January 5b: August

Figures 5ab: Mean precipitation (mm) and surface ITD position (Source: Hayward and

Oguntoyinbo, 1987:85)

The August pattern of rainfall distribution is similar to that of July. However, by September

the distorted areas have been greatly reduced and only lie to the south of about latitude 10°N.

Thus the pattern is generally more organized than during July and August.

Page 16: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 1 – Introduction

16

In October and November, when the rainy areas are mostly confined to south of

approximately latitudes 10°N and 8°N respectively the pattern becomes more distorted again

resembling the structure at the beginning of the rainy season.

October and November represent the beginning of the dry season and the decrease in rainfall

at this time is mainly because the ITD has moved so far south that only parts of the Guinea

coast are in weather zone B.

It has to be mentioned, that the rain gage density in West Africa is very low and due to this

data scarcity these nationwide maps do not always correspond to national weather patterns.

However, for northern Ghana a zonal distribution, especially for the Volta Basin, seems

acceptable when looking at the annual distribution of rainfall in figure 6 below. As the actual

rainfall rates in this figure are not of particular interest, but more the arrangement of the

isohyets, they were left in inches. Just for reference 1 inch equals 2.54 cm thus the 50 inch

isohyet comes to 1270 mm.

Figure 6: Mean annual rainfall distribution in Ghana (Source: Ojo, 1977:85)

Page 17: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 1 – Introduction

17

1.2.3.1 Togo Gap

As one of the reasons for the above described zonal disruption in the south one climatic

anomaly has to be mentioned, the so-called Togo Gap. It covers the south-eastern corner of

Ghana and southern Togo and is characterized by unusually low rainfall rates along the coast

(see figure 6). The latter being the main anomaly, as the phenomenon is not only limited to

the stretch between Accra, Ghana and Lome, Togo but also to the coast, since about 40 km

inland the situation normalizes again. The following is based on Hayward and Oguntoyinbo,

1987)

One partial explanation might be the low occurrence of squall lines in this area. This, though

reported at the Togo Gap, is not limited to it and other regions, like Axim on the western coast

of Ghana have even less squall line activity but twice as much rainfall.

Another reason could be the position of the coast line, as it is almost parallel to the

predominant south-eastern wind direction. The assumed effect would be reduced convergence

and uplift of the monsoonal air in comparison to areas where the wind direction tends to be

more onshore which has been observed when comparing the Togo Gap to the area between

Abijan, Ivory Coast and Axim, where the rainfall is recognizable higher. However, other

regions with similar coast lines do not experience these significantly lower precipitation rates.

A third irregularity found is the lower sea surface temperature along the coast. The proposed

outcome on this would be a cooling effect on the prevailing winds, leading to lower surface

air temperature and by that a reduced tendency for uplift and rainfall. This is disproved by the

fact, that coastal weather stations did not record significantly lower temperature rates,

although it has to be seen how the temperature in higher air layers behave.

Other approaches look at different land cover patterns, as the region is mainly deforested and

evolved into savanna grassland, and pose the question how much of a feedback is provided by

this.

As a conclusion there are several possible explanations of which none definitely solves the

phenomenon. Whether there a definite reason or whether it is a combination of the above is

not yet answered.

Page 18: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 1 – Introduction

18

1.2.4 Precipitation types

Traditionally rainfall is classified into three main categories of Convectional, Cyclonic/

Frontal, and Relief/ Orographic rainfall, according to the mechanism involved in the vertical

displacement of air and subsequent adiabatic cooling and condensation. In the tropics, with

particular reference to West Africa, this system of classification is inadequate in terms of

current knowledge of the region’s precipitation dynamics. Figure 7 shows a comparison

between precipitation patterns in mid-latitudal Europe and tropical Africa. Depicted on the x-

axes is the timescale, which is the same for both, and on the y-axes the intensities per hour are

shown in different scales. As can be seen, precipitation in Europe (upper graph) is, in general,

more prolonged and less intense than in tropical Africa (lower graph).

Figure 7: Precipitation differences Europe tropical Africa (Source: Lamb, 1998 referred to

in Beven, 2001:242, modified)

Firstly, there is no evidence of cyclonic or frontal activity in West Africa. Secondly, large

amounts of rainfall are produced by atmospheric processes other than convection and

orographic uplift. In some areas over 50 per cent of the total annual rainfall is derived from

these extra sources.

Within the test area there are three rainfall origins to be found. Convective systems divided

into local rainstorms and line squalls as well as convergence rainfall due to monsoonal air

layers. Also characteristic for West Africa are orographic and coastal convergence rainfall.

Both play a minor role at the test site due to the midland locality and the relatively low relief

within the Volta basin.

Page 19: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 1 – Introduction

19

1.2.4.1 Convective systems

As described earlier the monsoonal layer during September has a thickness of only 1.5-2 km

making it a relatively unstable air layer. Convective processes are triggered by surface

heating, which result in an uplift of air. When lifting precipitable air layers such as the moist

monsoonal air subsequently cumulus or cumulonimbus cloud towers are generated that result

either in local rainstorms or squall line formations. The latter being a result of the easterly

wave perturbations that are believed to originate from east central Africa between longitudes

15°E and 30°E.

In the following part these two convective rainfall types will be explained with special

emphasis to their characteristics, such as duration, intensity and total rainfall. Rainstorms are

the most familiar surface disturbances and together with other small-scale disturbances they

account for the characteristic “patchiness” or “showery” nature of tropical rainfall. (Kamara,

1986)

1.2.4.1.1 Rainstorms

This type of rainstorm results from intense heating of open land surfaces such as grasslands,

large urban areas, rocky mountain terrain, and forest clearings. A local rainstorm is a highly

localised and largely stationary weather system affecting a limited area of about 20-50 km²,

depending on the size of the cumulus tower. It is short lived with a lifetime of 1-2 hours.

1.2.4.1.2 Squall lines

The origin of line squalls is not solved yet, although a greater part of the literature sees their

origin in the mixing of the monsoonal air layer with the above lying African easterly jet. One

approach by Leroux (2001) is illustrated in figures 8a and b.

Page 20: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 1 – Introduction

20

8a: Cross-sectional formation 8b: Surface and vertical squall line sketches

Figures 8ab: Squall line generation (Source: Leroux, 2001:253,254, modified)

As depicted in figure 8a the monsoonal layer is disturbed due to i.e. relief changes or surface

heating. Leroux points out several causes that have been suggested and which all still remain

to be proven: “a multiplication of local convective thunderstorms, the orographical factor, a

conflict between two airmasses, intrusion by a cold front, a rise in pressure or anticyclone

core, the arrival of an easterly wave or an easterly pulse, or a centre of acceleration within the

AEJ…”(Leroux, 2001:238). These causes, however, results in a disturbance of the upper

bound of the monsoonal layer allowing the overlying African Easterly Jet (AEJ) to dam up a

cloud tower. The latter is then pushed to the west as a line disturbance or line squall. These

disorders in the air layers are only possible during the beginning and ending of the rainy

season as the thickness of the monsoonal layer at this stage is still reduced, as was earlier

described by the weather zones. This instability allows compression or disturbance of the

monsoonal layer which in general do not happen when the layer is at its peak thickness.

Figure 8b shows a schematic horizontal view of a line squall and the corresponding vertical

view illustrating the cloud cluster layout. The latter illustrates a squall line structure that is

generated by a moving cold core (indicated by the arrow labeled “ak”), that caused the initial

uplift.

Page 21: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 1 – Introduction

21

“A line squall is a 300-500 km long belt of thunderstorms arranged in a north/ south direction

and propagating westwards against the surface wind at speeds exceeding 18 m/s. Periods of

active line squall development coincide with enhanced easterly wave activity in the wind field

over West Africa. Over the central and northern parts of West Africa thunderstorms are the

main rain producers, accounting, in some cases, for over 90 per cent of the annual total. […]

At the mature stage the cumulonimbus tower develops a large oval-shaped trailing anvil

which is 6-10 km thick and about 200 km long. There is also a saturated convective or small-

scale (10-20km wide) downdraught within the ‘active’ part of the cloud system as well as an

unsaturated mesoscale downdraught (100-200 km wide) induced by evaporative cooling of

falling precipitation from the anvil.”(Kamara, 1986:51)

“One of the most common disturbances in West Africa is generated by thermal convection.

[…] When they occur, large areas report showery weather and many stations report rainfall.

Other significant types of disturbances, such as those associated with the southwesterly flow

and line squalls and the disturbance line, are also prime rain bringers.

The disturbance line is characterized by thunderstorms and squall winds, thereby resembling

the phenomena associated with a cold front.[…] However it is not frontal in origin and

structure, but is embedded in a homogenous maritime air mass.” (Ojo, 1977:69)

Squall lines can generally be described as disturbances moving from east to west in the area

covered by the atlantic monsoon, according to Eaker (1945, referred to in Leroux, 2001,

p.234) they are “tropical Africa’s most spectacular meteorological phenomena”. In the

literature a large variety of names for these disturbances exist. Whereas they were formerly

often referred to as Tornados, this term is now taken by: “line of instability, disturbance line,

easterly storm line, easterly disturbance, African disturbance, storm zone, unstable zone,

easterly wave, rapid wave, African wave, low-level whirlwind, low-level cyclone, monsoon

cyclone, African system”(Leroux, 2001:235).

“The term “squall line” is not in itself explicative, but merely descriptive. It offers no insight

into the cause of the phenomenon, but suggests the variations in wind speed, the violent

rainstorms, and the mobile, linear disposition of its dense cloud formations. Here, this time-

honored term will be retained, since “squall” reminds us of its violence and “line” describes

its organization.” (Leroux, 2001:237)

Page 22: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 1 – Introduction

22

As can be extracted from the above cited references squall lines and especially their

derivation are still under discussion. Nevertheless, their characteristics are twofold with a

storm band or area of high convection in the beginning and an almost stratiform part

following it, which can later be found in the event patterns. Although, some authors give

numbers on the spread of the two parts and the velocity of the squall lines in particular in their

descriptions, these are highly variable and will be discussed later in the analytical part. What

can generally be mentioned is that the convective part is much more intense and shorter in

duration and spread and the second much longer and characterized by light, continuous

downpour.

According to Omotosho (1984) there are three areas in West Africa with high squall line

frequency. Southern Chad, the region to the west of the Jos Plateau in Nigeria, and the central

part of Ghana with actually the highest frequency during a 3-5 year average.

1.2.4.2 Monsoon rains

Between July and September most areas of West Africa south of 10°N experience widespread

and prolonged “monsoon” rains. Periods of maximum convergence rainfall coincide with

minimal thunderstorm activity which tends to suggest differences in the mechanisms that

organise the thunderstorm and the massive uplift of monsoon air. Characteristic for this

convergence rainfall is that, apart from affecting wide areas at a time, it occurs in light,

continuous downpours with occasional spells lasting from a few hours to several days.

Page 23: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 2 – Methodology

23

2. Methodology

2.1 Test site

The test site is located in the Northern Region in Ghana near the District Capital Tamale. The

center of the test site is at 9°25’N and 0°58’W. Figure 9 shows that it is at the center of the

Volta Basin which ranges from northern Burkina Faso down to southern Ghana covering an

area of about 400,000 km². Being in the middle of a rather shallow basin, elevation at the test

site ranges from about 160 to 210 meters.

Figure 9: Test site location

As already pointed out in the introductory part, the test site is on the border between two

climatic regions, the Guinea savanna and the Sudan savanna. Ghanaian literature therefore

often refers to a hybrid zone, the Sudan-Guinea Savanna Zone located between 8° and 11°

northern longitude. Rainfall in this area usually varies from 900-1000 mm of annual rainfall

with an interannual variability of 15-20% within the Guinea and 20-30% in the Sudan zone.

(Kasei, 1990)

Page 24: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 2 – Methodology

24

The rainy season stretches from June/ July until September/ October and peaks in September.

Whereas in the beginning rainfall rates raise continuously, the season ends almost abruptly in

the first two weeks of October. Rainfall rates in Tamale usually range between 950 and 1050

mm annually. Squall line activity and associated rainfall ranges from 30% at the Ghanaian

coast and 90% in the northern parts of Ghana (Eldridge, 1957, referred to in Weischet, 2000,

p.268). According to Acheampong (1988) the wet season at Tamale has a length of 4 months

going from July to October. Figures 10a and b show two images from the test site illustrating

the immense difference between the two seasons.

10a: Dry season 10b: Wet season

Figures 10ab: Dry and wet season at the test site

In 2001, the rainy season had an unusual slow start which later normalized and September

rates were with 249.4 mm again slightly over the 1961-90 mean of 231.4 mm. Nevertheless,

the total of the 2001 season was with 859.9 mm much lower compared to the 1033 mm mean

value. (Kranjac-Berisavljevic', 1998)

Landscape within and around the test site is highly mosaicked with small agricultural plots

interspersed by natural vegetation (see Figures 10ab). As can be seen below in figure 11,

several small villages are situated within the observation grid. The farmland is characterized

by small agricultural plots of cassava, rice, yams, maize, gourds, cotton, ground nuts, and

sorghum crops. The natural vegetation is open savanna with a tree density of 10-100 per ha.

Page 25: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 2 – Methodology

25

2.2 Rain gage network

The gage network was organized as a nested grid consisting of three levels – A, B, and C. The

largest grid was A with a total area of 9x9 km and nine 3x3 km cells. Within the center cell

grid B was located with a cell size of 1x1 km. Figure 11 shows a map of the test with the

gages depicted as yellow squares (A) and blue squares (B). Due to environmental obstacles

such as tree clusters or agricultural plots the gages could not always be positioned at the

calculated grid center, as seen in the south by the layout of gages A6, A7, and A8.

Figure 11: Rain gage networks A and B (contour lines are labeled in feet: 1m ≈ 0.3048 ft –

e.g. 200 m equal ~656 ft)

Page 26: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 2 – Methodology

26

2.3 Rain gages The rain gages consist of two types of commercial tipping bucket gages, funnels, rings and

data loggers.

Figure 12: Tipping-bucket gage (Source: Herschy, 1999:87)

Figure 12 shows a standard tipping bucket with the major components. As rainfall enters the

gage it is led through the funnel towards a rocker. This rocker consists of two equal

compartments. Each time one compartment is filled with a certain volume of rainfall it tips to

one side, emptying it. The standard amount of rainfall per tip is between 0.1 and 0.2 mm, but

as will be explained later the exact amounts have to be measured. To the rocker a magnet is

connected which passes a reed relay that closes every time the rocker tips to one side.

This reed relay again is connected to a HOBO Event data logger. The latter is contact driven,

so that at each contact-closure by the relay, date, time, and consecutive number are stored.

Page 27: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 2 – Methodology

27

Provided together with the hardware is a software allowing to read out the ticks via a readout

cable. Then it is possible to export the files to ASCII format and import them into EXCEL.

Further changes like temporal resolution can be changed with the software.

The data loggers have a memory of 8000 values and a temporal resolution of 0.5 seconds. The

latter had to be set to 1 second in order to prevent double-ticks. These double-ticks occur

when the rocker is tipping too slow, thus recording two ticks. By increasing the interval

between the ticks this can be prevented.

13a: Rain gage scheme

13b: Rain gage in the field

Figures 13: Rain gage design

To have a standard catchment area of 270 cm² each gage was equipped with a plastic ring and

a funnel, mounted at a height of 1.5 m on wooden poles with concrete bases. Totalizers were

mounted below the gages to verify the tipping-bucket measurements. Displayed in figures 13a

and b are the rain gage scheme and an installed gage in the field.

Page 28: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 2 – Methodology

28

2.4 Calibration

As tipping-bucket rain gages work on the principle of a rocker they have to be leveled. In the

field this is not always possible, so that the error caused by this inclination to one side of the

rocker has to be measured to ensure later correction. Due to their individual inclinations

rainfall amounts per tip also vary.

14a: Field calibration 14b: Indoor calibration

Figures 14ab : Gage calibration

To calibrate the gages a field and an indoor calibration were conducted. For the indoor

calibration the gages were leveled. For the calibration about 810 ml the equivalent for 30 mm

of rainfall for the given catchment area were poured into the gages during 30 minutes thus

simulating a heavy rainstorm. As shown in figures 14a and b this was done by placing a tripod

with a dripping device over the rain gages.

2.5 Measurement uncertainties

Precipitation measurements are highly variable due to measurement uncertainties and the true

rainfall values of a catch are unknown. Errors may occur because of gage designs, elevation

differences, or wind fields, while the latter causes by far the greatest error. Underestimation of

the gage catch compared to the ground catch may be as high as 100% and more (UNESCO,

1978 referred to in Herschy, 1999:359).

Page 29: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 2 – Methodology

29

An error limited to tipping bucket gages is that the compartments collecting the rainfall might

overflow during heavy rainfall events. This usually happens when water flows into the

compartments while they are tipping. Other uncertainties may be due to insects, such as bugs,

spiders, and bees found inside the gages and disturbing the mechanisms.

As these uncertainties usually do not show regular patterns they cannot be corrected and it just

has to be noticed, that especially in comparison with other rainfall data, where different

equipment and setup standards are applied rainfall values might differ immensely.

2.6 Additional data

During the analysis measured rainfall is compared to additional rainfall data from three

external sources. Weather station data from the SARI weather station, MM5 hindcasting data,

and precipitation radar data from the Tropical Rainfall Measuring Mission (TRMM).

2.6.1 SARI weather station

The SARI rain gage is about 10 km linear distance from the test site and rainfall is measured

by means of a totalizer which is read every morning at 09:00. It is located at Nyankpala, about

12 km west of Tamale.

2.6.2 MM5 precipitation data

As described in the introduction, MM5 provides the climatic data within the GLOWA Volta

Project. It was originally developed in the USA and then adapted for western Europe. Being

configured for the European weather patterns, it had to be adjusted for the West African

climates, as there are major differences especially in the case of precipitation. Whereas in

Europe advective processes plays a major role, West African rainfall comes to 90% from

convective processes. Below the different domains and input data is discussed. Convective

processes can only be resolved at a very small grid size. Such a dense grid is not practical

with current CPU capacity. Instead, a coarser 9x9 km grid is used and convective rainfall

Page 30: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 2 – Methodology

30

within a cell is calculated at the basis of an empirical scheme. Different schemes exist and the

optimal one has been selected (Grell in GLOWA Volta Annual Report, 2002:5). This scheme,

however may still need further adjustment, given the results presented later in this thesis.

MM5 is run using three nested domains (see figures 15a and b), having a horizontal resolution

of 81x81 km, 27x27 km , and 9x9 km respectively and 26 vertical layers extending up to 30

mbar. The largest domain provides the boundary conditions for the second domain, and the

second domain for the third. The third domain covers the Volta Basin. MM5 includes the

Oregon State University Land Surface Model (OSU-LSM) thereby allowing investigation of

feedback mechanisms between land use change and precipitation as well as regional climate

simulations. The OSU-LSM makes use of vegetation and soil type in calculating infiltration,

percolation and evapotranspiration. Inputs are surface-layer exchange coefficients, radiative

forcing. Precipitation rate and surface fluxes are fed back to the atmosphere. Since 1 October

2000, MM5 is used for operational hindcasting. Each night, data are automatically retrieved

from NCEP (global re-analysis) and WMO sources (observation data), and an operationalized

preprocessing chain of MM5 is started. For each of the 8107 grid-points and for each three-

hour interval, the values of 20 variables are stored, such as rainfall, air temperature, latent and

sensible heat flux. (GLOWA Volta Annual Report, 2002)

15a: MM5 domains 15b: Sample MM5 hindcast

Figures 15ab: MM5 domains and sample hindcast (Source: GLOWA Volta Annual Report,

2001:4,5)

Page 31: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 2 – Methodology

31

The main purpose of MM5 is to run climate simulations to investigate how land use change

and global climate change affects precipitation. However, MM5 is designed as weather

prediction model and has to be adjusted to do long-term simulations. Global climate change

enters as boundary conditions for the largest domain, as provided by the ECHAM4 GCM.

Precipitation is modeled empirically for each cell. Rainfall values produced by MM5 are

convective and advective rainfall which were summed and used as total rainfall as advective

rainfall was nearly nonexistent. The extracted data was for September 2001, as August and

October had gaps, and observed wet days did not match. Further data pre-processing is

explained in the data part.

MM5 plays a major role within the GLOWA Volta Project as an input for physical as well as

socio-economic models. For calibration and verification the model needs ground data

provided by the presented measurements, installed weather stations at the three test sites, and

data from the Ghanaian weather station network.

2.6.3 TRMM precipitation radar

Also used for reference and identification of squall line structures is data from the Tropical

Rainfall Measuring Mission (TRMM). Satellite scenes used are from the TRMM Precipitation

Radar. TRMM is a polar orbiting satellite and thus the temporal resolution is very coarse.

Figure 16 provides an overview showing the TRMM scan path for one day. Displayed are

ascending and descending orbits. The pass time for one orbit is about 1:30 hours and because

ascending and descending orbits cross there are two images a day from a region near or

covering the test site.

Since only visual analysis is carried out with this data and as the squall line structures are

usually of a larger extend the test site does not have to be covered exactly to associate a

certain cloud formation with a rainfall event. As can be seen generally two orbits can be of

use which in the displayed example are numbers 22281 in the ascending and 22287 in the

descending overview.

Page 32: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 2 – Methodology

32

Figure 16: TRMM orbits 10. Sept. 2001 (Source: GSFC DAAC, 2002)

Since the imagery will just be used to provide cloud patterns and the distribution of rainfall

rates within these the technical description of the utilized instrument and algorithm will be

brief.

Precipitation Radar

“The Precipitation Radar will be the first spaceborne instrument designed to provide three-

dimensional maps of storm structure. The measurements should yield invaluable information

on the intensity and distribution of the rain, on the rain type, on the storm depth and on the

height at which the snow melts into rain. The estimates of the heat released into the

atmosphere at different heights based on these measurements can be used to improve models

of the global atmospheric circulation. The Precipitation Radar has a horizontal resolution at

the ground of about 2.5 miles (four kilometers) and a swath width of 137 miles (220

kilometers). One of its most important features will be its ability to provide vertical profiles of

the rain and snow from the surface up to a height of about 12 miles (20 kilometers). The

Precipitation Radar will be able to detect fairly light rain rates down to about .027 inches (0.7

millimeters) per hour. At intense rain rates, where the attenuation effects can be strong, new

methods of data processing have been developed that help correct for this effect. The

Page 33: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 2 – Methodology

33

Precipitation Radar is able to separate out rain echoes for vertical sample sizes of about 820

feet (250 meters) when looking straight down. It will carry out all these measurements while

using only 224 watts of electric power—the power of just a few household light bulbs. The

Precipitation Radar was built by the National Space Development Agency (NASDA) of Japan

as part of its contribution to the joint US/Japan Tropical Rainfall Measuring Mission

(TRMM)” (TRMM Homepage, 2002)

2A25 Algorithm

“The objectives of 2A25 is to correct for the rain attenuation in measured radar reflectivity

and to estimate the instantaneous three- dimensional distribution of rain from the TRMM

Precipitation Radar (PR) data. The estimates of attenuation-corrected radar reflectivity factor

and rainfall rate are given at each resolution cell of the PR. The Estimated near-surface

rainfall rate and average rainfall rate between the two predefined altitudes (2 and 4 km) are

also calculated for each beam position.” (TRMM Homepage, 2002)

2.7 Malfunctions

Due to mainly technical problems some malfunctions occurred which will be explained at this

time, to avoid uncertainties about missing data or analysis steps.

Missing gage totals especially for the early events occurred due to later launch of the gages.

This was for instance because renegotiations with the farmers and villages owning the land

had to be done, or due to transportation irregularities during the first launch phase. Gage A5

was completely taken out of the network, as it was destroyed by local farmers and gage B1

was not installed because of the loss of one data logger.

Although the data loggers recorded date and time of each tick the temporal coordination

between the gages did not work because of technical problems. As a result the rainfall events

could not be analyzed according to the starting times of the gages.

Page 34: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 3 – Data preparation

34

3. Data preparation

In the following chapter data organization and different pre-processing steps will be

explained. At first, the data preparation of the raw readout files to the main EXCEL sheets

being used in the different analysis stages is described, and afterwards the necessary

conversions performed on these. The last parts of this chapter describes the preparation and

conversion of external data, being SARI observed rainfall, MM5 modeled rainfall, and

TRMM precipitation radar scenes.

Before starting the actual analysis the data had to be pre-processes. There are different

analysis levels which although based on the same data need different pre-processing. For

event rainfall statistics where only total rainfall per gage is of interest the left/ right error did

not matter as the totals are not triggered by it. Here, the bias correction and the selection of

valid gages needed to be made. For intensity statistics neither bias correction nor selection

played any role but the left/ right error had to be taken care of since the intensity pattern is

being analyzed.

Thus, depending on which analysis is done different pre-processed data is taken, later

resulting in three data files.

3.1 Data organization

As shown in the methodology, the data files are first read out to a datalogger software as

ASCII files (see figure 17a) and can then be converted into EXCEL spreadsheets (see figure

17b). Each gage and readout were saved in separate files. Information in the raw data logger

files are date, tick time, and the consecutive number of each tick.

Page 35: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 3 – Data preparation

35

17a: Raw data logger file 17b: EXCEL gage file

Figures 17ab: Data logger and EXCEL gage files

In order to compare the different gages, the timeline of all gages had to be standardized. To do

this the date and time information provided in the gage files were converted to seconds

starting from the beginning of the measuring phase. While not all gages were launched at the

same time the start point (0 seconds) was set to the first data logger launch.

After the time conversion all gage files from different readouts were merged and then all

gages were saved in one EXCEL spreadsheet. The different gage files were then combined

with the calibration data, as each gage had different tick rainfall.

18a: Master spreadsheet 18b: Master graph

Figures 18ab: Master spreadsheet and graph

From this master spreadsheet (see figure 18a) the different rainfall events were extracted. To

automate the “event cutting” the time span for each event was extracted by using a graph (see

figure 18b) composed of all gages and then these time windows were copied from the master

Page 36: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 3 – Data preparation

36

spreadsheet into files depicting each event. The event files contained all gages for the given

time period.

For the event analysis the rainfall totals from each gage were extracted from these files. These

values were then combined in one event table being the basis for the event analysis. For the

intensity analysis the gage with the highest total rainfall per event was taken. The latter were

then composed as one EXCEL spreadsheet.

Thus, at the end of this conversion and event cutting phase, three files were generated for

further pre-processing and later analysis.

• The “event” spreadsheet including all gages sorted by events

• The “event analysis” spreadsheet including only the totals of all events organized by

gage

• The “intensity” spreadsheet including the gages with the maximum rainfall from each

event.

3.2 Pre-processing

3.2.1 Selection

At first all events below 1 mm were taken out which equaled a tick number of 10 or less ticks.

Second was the selection of malfunctioning gages. Due to technical problems not all gages

were functioning during every event. Including them into the statistical calculations led to

significant changes in mean rainfall/ event and coefficient of variation/ event as can be seen in

figure 19, where especially the coefficient of variation changes significantly. Illustrated in

figure 19 is an exemplary 3D graph of the A network (including mean rainfall B as center

gage) from September 6 on the left and on the right mean rainfall and CV are shown to

emphasis the differences with and without gage A8. Consequently, for those analysis steps

which included the whole event series instead of looking at single gages these had to be taken

out. During the larger events (above 10 mm) that was usually no problem, as the test site was

more or less evenly covered and gages with rainfall below 1 mm could easily be taken out.

For smaller events the intensity graphs and tick numbers of the lowest gages were checked to

see whether the gage was malfunctioning or not.

Page 37: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 3 – Data preparation

37

Event: 6. Sept. 2001

Event statistics

With gage A8 Without gage A8

Mean rainfall 11.37 mm 9.95 mm

CV 0.42 0.11 Figure 19: Sample selection

3.2.2 Bias Correction

Due to environmental differences, such as elevation, wind fields, and vegetation, some rain

gages have a bias, meaning that they constantly measured higher respectively lower rainfall.

The aim at the event analysis level is to compare the results of all gages so that the individual

characteristics of the rain gages should be reduced as much as possible. In order to correct the

bias the rain totals for each storm were normalized with a correction factor.

stormge

enormstorm P

mmP ×=,

Pstorm, norm: Corrected rain totals for each gage and storm

Pstorm: Biased rain totals for each gage and storm

me: mean rainfall over all mge (table 1, “Total mean”)

mge: mean rainfall for one gage mge all events (table 1, column “Mean mm/e”)

me/mge: correction factor

Table 1: Bias correction

Page 38: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 3 – Data preparation

38

This bias correction only had to be done for the spatial analysis of single events where

different gages were to be compared with each other to gain an overview of the total rainfall

distribution over the whole network. Presented in table 1 are the mean rainfall totals per event,

their differences to the total mean rainfall in mm, and the correction factor. As displayed in

table 1 some bias values differ greatly compared to the measured mean value over all stations

and events. As already mentioned during the methodological part, rainfall measurements are

subject to a variety of uncertainties, for which, if possible, they have to be corrected.

In terms of analysis steps, the correction was done for rainfall totals and not for intensities of

single gages.

Figure 20: Sample Bias

Figure 20 shows the bias of selected gages. On the x-axis the events are shown and the y-axis

depicts the associated rainfall totals. Shown are gages B2 (red), B4 (green), and B5 (blue). B2

is the gage with the lowest mean rainfall, and thus the highest correction as can be seen in

table 1. Also displayed are B4 as an example for very low deviation from the mean over all

gages and events, and B5 as the gage with the highest mean rainfall. Depicted by the dotted

blue line are the corrected totals for gage B5.

As can clearly be seen the, biased gages are measuring more respectively less rainfall over

almost all events, which strengthens the need to correct these gages.

Page 39: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 3 – Data preparation

39

3.2.3 Left/ right error correction

In the original data logger files only the tick times together with the consecutive numbers are

reported. In order to get actual mm values rainfall per tick had to be known. These values

were obtained by calibrating each gage as described in the methodological part.

Table 2: Calibration results gage left interval total interval ratio ticks

A1 642.5 1400 0.458928571 294

A2 909 1608 0.565298507 285

A3 472 1520 0.310526316 273

A4 1073.5 1963 0.54686704 293

A6 1079 1854 0.581984898 366

A7 389.5 1324 0.29418429 232

A8 559.5 1735 0.322478386 270

B2 578 1403 0.411974341 111

B3 667.5 1601.5 0.416796753 112

B4 1012.5 1452.5 0.69707401 272

B5 575.5 1463.5 0.393235395 272

B6 1060.5 1511 0.701853077 237

B7 1050 2317 0.453172205 111

B8 863 1705.5 0.506009968 117

Table 2 shows the results of the calibration phase. Column “gage” depicts the gage ID, “left

interval” is the cumulative interval time from the left to the right tick in seconds, “total

interval” is the cumulative interval time between all ticks, “Ratio” is the ratio between the two

interval times, and “ticks” gives the number of ticks that occurred during the calibration

phase. The variation in the number of ticks originates from the use of two different kinds of

tipping-bucket rain gages.

By dividing the cumulated time lag of one side by the total time lag a percentage for that side

was derived, which could then be applied to the rainfall per tick.

To apply these calibrated tick values to the event datasets the time lags had to be observed to

distinguish between longer and shorter time lags and assigning the different tick values.

Page 40: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 3 – Data preparation

40

For the intensity analysis these left/ right errors had to be decimated in order to smoothen the

curve. Therefore, the before calculated ratio (see table 2) between the left and the right side

was applied to the interval times between the ticks.

Figure 21: Error reversal

The left/ right correction posed a problem insofar as the error reversed during events, thus

making it not feasible to correct by the data extracted from table 2.

Figure 21 shows such a reversal. Displayed is an exemplary intensity graph with left/ right

corrected (red line) and averaged (blue line) intensities, that will be explained below. As can

be seen the corrected intensity curve works until about 1700 seconds and suddenly the error is

aggravated instead of minimized. The origin of this aggravation was probably a missed tick to

one side, so that the error changed during the event.

In order to still minimize the error the intensities were instead calculated by a moving average

of 2 from the tick intervals and the mean rainfall per tick. The thereby averaged line is

illustrated by the blue line in figure 21.

Page 41: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 3 – Data preparation

41

3.3 External data preparation

3.3.1 SARI observed rainfall

Data sheets came in hardcopy format. Rainfall values were copied to an EXCEL spreadsheet

and daily totals below 1 mm were left out.

3.3.2 MM5

As described in the methodological part, MM5 works at 3 domains. The daily hindcasting,

consisting of the smallest domain covering the Volta Basin, has a spatial resolution of 9x9

km. Simulation intervals of climatological data are computed eight times a day at 0, 3, 6, 8,

12, 15, 17, and 20 hours. There are eight files a day each covering the whole domain. Out of

these files a spatial window was cut covering 16 MM5 cells around the test site (see figure 22)

and then comprised into one data file consisting of all 16 point values for September 2001.

Figure 22: Extracted MM5 data points (Source: GLOWA Volta Annual Report, 2002:5,

modified)

MM5 models convective and stratiform rainfall, but as stratiform precipitation for September

was only about 10 mm compared to nearly 570 mm of convective both were summed to one

rainfall figure. To compare MM5 to observed rainfall they had to be converted to daily totals.

Page 42: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 3 – Data preparation

42

For the observed data all rain events during one day were added together and for MM5 all

simulation intervals yielding precipitation were considered an event.

As with the observed data, multiple events per day were added and event dates were set to the

first rainfall occurrence respectively gage tick. For days with multiple events the totals were

added for each gage or simulation point and values like mean rainfall, standard deviation, CV,

and Range were recalculated over the day.

3.3.3 TRMM precipitation radar scenes

TRMM scenes could be downloaded as .HDF files containing the complete orbits. The files

were viewed with the TRMM HDF Viewer, cut out to the detected squall line formation and

exported as images. The cut out images all had a size of 10° by 10° with a 5° separation line.

1° at the observed latitudes is roughly 110 km and size measurements were done by this

figure.

Page 43: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 4 – Data analysis

43

4. Data analysis

The analysis chapter consist of five parts, the event analysis, the intensity analysis, the

combination of the two former parts, the comparison with external data from SARI and MM5,

and the comparison with TRMM data.

Event analysis concentrates on the totals per event and gage and summarizes and classifies

them in order to get a general view of the data as well as to point out significant features.

The second part concentrates on the temporal distribution of rainfall during the events and

therefore uses intensity graphs showing the rainfall intensity, the timeline of the event and the

cumulative density function of the rainfall.

In the combination of both parts the characteristics of the different rainfall types are compared

to the statistical and visual analysis results.

MM5 SARI part compares observed mean values for the AB network with observed data

from SARI and modeled precipitation from MM5.

The TRMM part concentrates on selected TRMM scenes from September 2001 and general

squall line features are extracted as well as a comparison between two intensity graphs and the

appending TRMM images.

4.1 Event Analysis

In order to describe the rainfall during the data collection the data was analyzed by totals. To

do this, the totals of all gages ordered by event and network were placed in one spreadsheet.

Starting from this table general statistics like mean rainfall, coefficient of variation, mean

error, bias elimination, and 3D graphs of the events could be done. What can be extracted is a

general description of the later part of the 2001 wet season. As not the whole season but only

September and October were covered no complete seasonal cycle could be presented.

Page 44: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 4 – Data analysis

44

Major parts of this analysis step besides the general statistics are the examination of the

coefficient of variation (CV) and the 3D visualization of the single events. While the CV

gives a statistical value about the spatial variability of the rainfall pattern it does not show any

patterns that can be seen when looking at the actual bar graphs of the different gage totals per

event.

Below an overview of the data file comprising all data used in the first analysis step is given.

Displayed are all events sorted by size in the columns and the single gage totals in the rows.

Table 3: Event spreadsheet overview

The first column shows the gage IDs and the calculation headers, and the top row the event

IDs. On the top part the A network is displayed with the statistics for all A gages and

additionally statistics for all A gages with the mean rainfall of network B embedded at the

center cell. In the lower part the B network with statistics and below the statistics for all gages

A and B.

The events in this overview are sorted by size with the smallest events to the left, the different

shades of blue depict the three classes which will be explained later.

Page 45: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 4 – Data analysis

45

After assembling the gage totals for all extracted events in one table, the different processing

steps, which are explained in more detail within the methodological part were conducted. First

the gage selection, to get rid of non-functioning gages, which were mostly characterized by

values below 1 mm, and then the bias correction to eliminate measuring irregularities. To find

out whether there are irregularities between the two networks, summarized data, as mean

rainfall per event, standard deviation and CV were calculated separately for both, for all, and

for network A with the mean rainfall of network B imbedded as a gage in the center.

A first overview shows that the data is highly variable. One of the most significant values

depicting this are the rainfall totals per gage ranging from slightly above 1 mm till up to 95

mm over all gages.

Figure 23: Rainfall events sorted by size

Figure 23 shows the mean rainfall totals for both networks A and B separately. Also displayed

is the CV while the graph is sorted by mean rainfall over all gages. Although there might be a

slight decline in the CV, the data is not extensive enough relate that to the size. Consequently,

more information on the single events is needed in order to relate them to certain rainfall

types, and a classification will be made.

As can be seen in figure 23 the first eleven events are almost all around 5 mm with no

significant rise or outliers, the second group ranges from event 12 to 18 and although not as

Page 46: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 4 – Data analysis

46

uniform as the first part show considerable more rainfall. The last four events also show a

major rise in rainfall values. Whereas, at first the idea was to classify the events into classes

below 10, between 10 and 20, and above 20, the graph shows that there is a step between the

lower and the middle part that way including three events more associable with those around

10 mm mean rainfall.

Then the data was classified into three classes as described above depicting the distribution of

CV and total rainfall between the three. This was done to see whether the mean values over

the classes show a trend underlying the shown noise above(see figure 23).

Table 4: Event distribution

Class # (%) mm AB (%) total mm AB

I: > 7 50 15 40.04

II: 7-20 32 28 73.28

III: < 20 18 57 151.18

Illustrated in table 4 are from left to right the three classes, the event occurrence per class in

percent, the distribution of the measured rainfall in percent, and the total rainfall per class in

mm. Considerable attention is drawn by class III, depicting only few rainy days but with 57%

accounting for the most total rainfall. As described in the introductory part, these high

intensity rain storms which are, compared, for example, to European mid-latitudal

precipitation rather short are typical for tropical Africa. As will be seen later the events above

20 mm are all due to squall line activity.

Described as well in the introductory part West African rainfall is often characterized as being

“patchy”. To statistically describe this spatial variability the coefficient of variation, using all

gage totals of the single events, was calculated per class. The results are shown below in table

5.

Table 5: Coefficients of variation

Class A B A+B AB

I: > 7 0.44 0.35 0.41 0.41

II: 7-20 0.29 0.24 0.28 0.29

III: < 20 0.28 0.21 0.26 0.25

Page 47: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 4 – Data analysis

47

Values for the different networks and combinations are calculated. Columns A and B show

the coefficients of variation for each network apart from each other. A+B is calculated on the

base of network A with the center cell derived from the mean rainfall total of network B. AB

takes all gages into account and is the value on which later classification as can be seen in the

intensity analysis is done.

As expected, the variability in general is decreasing against the total mean precipitation per

event. Those with high rainfall totals tend to cover the test area more evenly, whereas smaller

events only partially cover the test site or show an increasing pattern described below during

the 3D graph analysis and thus yield high CV values.

Unlike the CV distribution in figure 23, where only a slight trend was recognized the

distribution within the classes (see figure 24) shows a more explicit allocation, with CV

differences between class I and II of about 30%.

Figure 24: Totals and CV of different network combinations

Page 48: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 4 – Data analysis

48

Not only the relation between the coefficient of variation and the event size draws attention

but also the differences between the two gage networks which can be seen in table 5 and

figure 24. Presented in the figure are again the CV values as lines and the total rainfall per

class in mm as bars.

Network A with an area of 9x9 km has a coefficient of variation that is up to 20% higher than

Network B covering an area of 3x3 km. The assumption on this difference between the

networks is that network B, due to its smaller size is more often covered completely by

rainfall fields and thus has a less high spatial variability.

4.1.1 3D network analysis

To get a better understanding of the spatial distribution of the events other than coefficient of

variation values the data was displayed in a 3D bar graph. Presented below are three

examples, whereas all 3D graphs of both networks can be found in appendix A. Displayed is

the total rainfall per gage and event. In order to compare it to the real networks the diagram is

oriented towards north corresponding to the gage positions in the field, as illustrated within

the test site description above (see figure 11). The gages shown in the examples below are

from network A with the mean value of network B in the center. Plotted on top are the CV

values for each event.

When looking at the rainfall totals distribution (see figure 25) three patterns tend to appear

several times.

CV 0.17 CV 0.52 CV 0.50

Pattern I Pattern II Pattern III

Figure 25: 3D event patterns

Page 49: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 4 – Data analysis

49

Pattern I shows a rainfall event covering the whole test site almost evenly. In general, large

events due to line squall or monsoonal precipitation are associated with this pattern. In

numbers this pattern usually has CV values below 0.3.

Pattern II is the de- respectively increasing event pattern. In this case the gage totals are

increasing from one corner or side of the test area. Reasons for this are precipitation fields

passing by the test site. Later, when the TRMM data will be compared to the observed rainfall

one example of such a pattern will be shown. Coefficient of variation values for this pattern

are usually high, so to say above 0.3.

Pattern III has the highest spatial variability and shows no difference in terms of CV values to

pattern II. This is generally associated with local storms consisting of one or more small

precipitation cells and is a good example for the “patchiness” of especially local convection.

As can be seen later in the intensity graphs in more detail some small events (>10 mm)

consist of several small storms. A clear separation of these rainfall cells was not possible,

because the time periods between the ticks are often below 10 min. and the temporal

distribution of those cells differs from gage to gage.

However, as pointed out, it is not possible to distinguish between the different rainfall types

although the different classes show diverse coefficients of variation. Large rainfall events due

to squall line precipitation or heavy rainstorms generally have relatively low spatial variability

with coefficients of variation below 0.3. Nevertheless small precipitation cells yielding low

rainfall may also have small CV values, if the precipitation cell covers the test site, whereas

large rainfall events may have high CV values if the test site is only “touched” by the rainfall

field.

Hence, a more exact differentiation between the rainfall types can only be done by means of

the intensity curve as shown in the next step.

Page 50: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 4 – Data analysis

50

4.2 Intensity Analysis

Although the general pattern of decreasing CV values against increasing rainfall totals can be

seen when classifying the data, a more exact identification of the different rainfall types and

their characteristics needs to be found. In addition to the analysis steps carried out above the

temporal distribution of rainfall is taken into account in the next step.

The intensity analysis is done to define single events and to distinguish between the different

rainfall types. The definition of single events basically works via the cumulative density

function (CDF) of the cumulative rainfall. It shows how the rainfall is temporarily distributed

over the whole event. Thus it can be seen whether there are breaks during the event, or the

events can be divided in high and low intensity parts which then can help connecting them to

different rainfall origins.

The latter is the second emphasis of this analysis step concentrating on the actual origin of the

event. Within the test area different types of rainfall are predominant, such as rainstorms, line

squalls, and monsoonal rainfall which are described in more detail within the introductory

part. Although the events cannot always be definitely coupled with those origins some show

close resemblance to them.

For visualization intensity graphs are used. Displayed on the y-axes are the intensity in mm

per hour and the cumulative density function (CDF) of the rainfall. On the x-axis the time in

seconds showing the event duration is displayed. Together with the 3D graphs the intensity

graphs for all captured events are presented in appendix A. In the upper left corner of each

intensity graph the main statistics for that particular event are listed (also see appendix A).

The visualization of the CDF is taken into account, to better show whether the rainfall is

temporarily even distributed or is divided into high and low intensity parts. On the intensity

axis a maximum scale of 300 mm/hr is chosen to make the comparison between the events

easier whereas the time scale is chosen individually.

Below some examples are shown, which clearly depict the three different precipitation types.

Page 51: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 4 – Data analysis

51

26a: Single cell rainstorm 26b: Multiple cell rainstorm

Figure 26ab: Rainstorm intensity graphs

The two figures 26a and b show typical convective storms being short rainfall events with an

almost abrupt beginning and end. Figure 26a shows a rainstorm with one cell and as can be

seen in the CDF rainfall is evenly distributed. Characteristics that mark this events as a

rainstorm are the short duration and the continuous CDF trend.

Figure 26b shows three short rainfall events, which could be separated for the actual gage, but

are too close together to separate them for the whole network, as referred to during the 3D

analysis.

Figure 27 : Monsoon rain intensity graph

Page 52: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 4 – Data analysis

52

Figure 27 shows a typical large-scale convergence rainfall characterized by a light, continuous

downpour. Main features are the long event duration, starting from 3 hours onwards, and the

low intensity. The CDF curve is evenly distributed and unlike the rainstorm intensity graphs

there are no multiple cells. Altogether three monsoonal patterns were found, all occurring in

the first days of September, and thus strengthening the fact that they generally appear earlier

in the season, as stated in the introductory part.

Figure 28 : Squall line intensity graph

Figure 28 is a convective event probably within a line squall structure. As can be seen the

event can be divided into two parts starting with a relatively heavy downpour which later

evolves into a light drizzle. Unlike the two preceding events the CDF in this figure shows a

different rainfall distribution with 80% falling during the heavy downpour and 20% in the

second part.

Page 53: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 4 – Data analysis

53

4.3 Combining results from event and intensity analysis

After the description of the three main precipitation types in the introductory part and the

preceding analysis steps some characteristics of the different types are reviewed in the

following part, and the observed events will be classified according to these.

4.3.1 Precipitation type features

In summary, the rainfall origins can be identified by visually interpreting the intensity graphs

and under consideration of descriptions found in the literature. When facilitating this on all

events certain features for the three precipitation types, during the observation phase, can

generally be found. For the described types certain significant features can be extracted (see

table 6).

Table 6: Precipitation type characteristics

Type Description Intensity process Duration

Rainstorm Single or multiple

convective cells

short, relatively sudden beginning

and ending

Short – below 1 hour

Monsoon

rain

Stratiform rainfall long, continuous low rainfall Long – 3-4 hours

Squall line Rainfall induced

by squall lines

Strong convective part in the

beginning followed by low rainfall

Highly variable - from

30 min. to 3-4 hours

The distinction between monsoonal and rainstorm/ line squall precipitation is relatively clear,

as monsoonal rainfall has no convective part and a comparably long duration. Thus the

identification using intensity and CDF curves makes it easy to differentiate between those

types. The difference between single or multiple convective cells and line squalls is only the

latter having a determinable stratiform component in addition to the convective part. As the

duration and total rainfall is not as different except for the large line squall events the division

is not as clear as above and some events might be misinterpreted. However, in the following

part all events are classified in order to quantify their occurrence and contribution to total

rainfall.

Page 54: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 4 – Data analysis

54

4.3.2 Event type classification

After looking into the general statistics comprising all events and the implementation of the

intensity graphs the collected information was combined. In the below shown table 7 all

events and selected features are shown.

Displayed are the event ID, the event date, the start time, that means the first tick time, the

mean rainfall over the whole network, the maximum rainfall, the event duration of the gage

with the highest rainfall, the CV over the A network with B mean rainfall embedded as a

center gage, the type as classified, and for the rainstorms the number of cells, as for the events

with more than one cell the duration per cell is actually shorter.

Table 7: Event statistics

The corresponding intensity and 3D charts can be found in appendix A. Despite the above

mentioned difficulties in clearly differentiating between the different events an attempt to

quantify the distribution of the different rainfall types and their features was conducted and

the main results can be seen in table 8. Although, the differences between rainstorm and

squall line events are not as evident as those between convective and stratiform, the statistics

of the three different types are as expected.

Page 55: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 4 – Data analysis

55

Table 8: Precipitation type statistics

Displayed in table 8 is the classified data from table 7 above sorted by type. CV, Mean mm,

and duration are mean values over all events of the corresponding type, whereas the other

fields contain summarized data of the specific type classes.

It can be seen that values like CV, mean mm and duration meet the expectations. Especially

the events duration confirms the described characteristics except for squall lines which were

usually longer, but as can be seen below in the TRMM part the duration, of course, depends

greatly on the position of the squall line. Remarkable is the high contribution of squall line

rainfall to total rainfall.

Page 56: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 4 – Data analysis

56

4.4 External data comparison

4.4.1 Comparison SARI - Observed

Comparisons between observed rainfall within the test site and the rainfall measured at SARI

(see Methodology), which lies about 10 km to the southwest shows that the general

distribution of events and the total rainfall is almost equal. Differences occur due to the

different measuring techniques. Whereas SARI rainfall is read every morning at 9:00 the

event date for networks A and B is set to the time at the first tick of the event.

Figure 29: Comparison - Networks A and B, SARI

Although some differences are due to the different event times it can be seen that some small

events do not occur in both places even though the distance in only about 10 km, which again

stands for the “patchiness” of small events in particular. When looking at the rainfall totals for

September/ October 2001 (see figure 29) the SARI station measured 261 mm whereas the

mean values for networks A and B came up to 265 mm. Though this comparison is based on a

very short observation, it shows that SARI’s rainfall measurements give a very good monthly

total and suggest that their long-term data is representative for the Tamale test site.

Page 57: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 4 – Data analysis

57

4.4.2 Comparison MM5 – Observed

As described in the methodological part, MM5 is the weather model operated in the GLOWA

Volta project. MM5 works on a 9x9 km scale and predicted values are given every 2 to 4

hours, allocated over the day these are at 0, 3, 6, 8, 12, 15, 17, and 20 hours. To compare

MM5 and observed values both had to be brought to the same timescale, where a daily basis

was chosen.

An intraseasonal comparison between MM5 and observed totals is shown below in figure 30.

Shown on the x-axis is the date, on the y-axis the rainfall.

Figure 30: MM5 vs. Observed rain days – September 2001

As well rain days as total rainfall over the selected comparison period differ greatly. In

contrast to observed rain days, MM5 presents an almost daily rainfall occurrence.

Consequently, MM5 is overestimating total rainfall with about 145%. However, when looking

at the class wise distribution, figure 31 illustrates that the overestimation is, though taking

place in all three classes, primarily affecting events over 20 mm.

Page 58: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 4 – Data analysis

58

Figure 31: Rainfall per class (MM5 – Observed)

For the comparison of standard deviation and range between MM5 and observed data only

matching rain days were used. Plotted in figure 32 are simulated MM5 values against the

corresponding observed values. The red bars depict the standard deviation printed thick as

well as negative and positive range printed thin of the simulated MM5 cells. The blue bars

give the standard deviation and range for the observed rain gages. The black line depicts a one

on one line on which all points should lie for a perfect match between model and measure

Figure 32: Scatterplot MM5 vs. Observed

Page 59: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 4 – Data analysis

59

For data points above the black line MM5 overestimates, whereas points below underestimate

the daily rainfall. The green rectangle drawn around two exemplary data points show that the

best rainfall values within the studied MM5 grid lie within the observed range. Although

neither simulated rainfall totals nor the standard deviations are covering the observed rainfall,

6 of 10 data points are within range.

4.4.3 TRMM analysis

Seeing that especially squall lines make up a large amount of the rainfall TRMM data was

sighted in order to further examine their characteristics, and also compared to descriptions

found in the literature on which the actual identification was based. As described in the

methodological part the TRMM satellite is polar orbiting and thus scenes from roughly the

same area are available twice a day. This coarse temporal resolution of the scenes made it

difficult to identify squall line formations that could be associated with measured events.

Three clusters from September 6, 9, and 13 were identified that most likely coincided with a

rainfall event. However, several squall lines moving over West Africa were identified which

helped proving their patterns in the intensity graphs. Displayed in figure 33 below are two

sample squall line structures. The images have a size of 10° by 10° with a dotted line at 5°.

The color scheme represents the rainfall rate decreasing from red to yellow, green, and blue.

All detected squall line structures can be found in appendix B. For measuring their spatial

spread 1 degree was set to 110 km.

Date: 22. Sept. 2001 Date: 27. Sept. 2001

Figure 33: TRMM sample scenes (Source: GSFC DAAC, 2002)

Page 60: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 4 – Data analysis

60

Though this is only a small selection of events and a larger survey would yield more exact

data it can already be seen that the spatial spread varies greatly and numbers found in the

literature match only partly. When splitting up the squall lines into convective and stratiform

elements the latter ranges from about 90 to 370 km (E-W) while the convective part only

varies between 25 and 35 km from east to west. Because TRMM in general only catches

fractions of squall lines these distances can only be given for the east - west direction.

Assuming a velocity of about 60 km/h for squall lines this means approximately 30 min. for

the initial high intensity downpour, but as can be seen in table 9 these velocities may vary.

Despite the facts that these are very few samples and that, as stated above, the squall line

velocity varies greatly (see table 9) the temporal patterns can usually be found in the collected

data as can be seen in the event overview in appendix A. With regard to TRMM scenes, it can

be said that due to the limited orbit width and the intermediate coverage most of the squall

line formations are cut off or are simply too big to be captured completely.

Table 9: Line Squall velocities

Source Average Speed (km/h)

Oyo, 1977 ~ 25

Leroux, 2001 40 – 60

NOAA, 2002 ~ 32 – 64

Weischet, 2000 40 – 50

Kamara, 1986 > ~ 65

Hayward and Oguntoyinbo, 1987 58

When looking at the squall line cluster it can clearly be seen that, of course, the size of the

rainfall event greatly depends on where the squall line passes the network. Here it is shown

that due to the different sizes of squall line structures several of the smaller rainfall events

may also be caused by squall lines and that especially in the marginal areas to the north and

south of the squall lines the distribution between convective and stratiform rainfall is greatly

changed, as is shown below in figures 34a, b, and c.

Page 61: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 4 – Data analysis

61

34a: Event 0720, date: 6. Sept. 2001

34b: Event 0940, date 8. Sept. 2001

34c: Event 1487, date 15. Sept. 2001

Figures 34abc: TRMM scenes and corresponding intensity charts (Source: GSFC DAAC, 2002)

Figures 34a, b, and c show three TRMM scenes together with the intensity graphs from the

associated events. The crosses in the TRMM scenes mark the test site position.

Page 62: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 4 – Data analysis

62

As can be seen in Figure 33 and 34a, b, and c the convective, high intensity part in the

beginning is followed by a stratiform, low intensity part in the end over the whole north-south

spread of the squall line cluster.

Whereas the squall line in figure 34a is too far away to assume a corresponding intensity

profile within the TRMM scene this may be done with the following figures 34b and c.

For figure 34b the gap which is observed in many squall lines (see appendix B) between the

convective and the stratiform segment can also be found in the intensity curve. This gap,

though found in almost any squall line formation, is often only intermittently existent. Hence,

not all intensity graphs show this gap between convective and stratiform segment, which

seems, if existent, to be one of the most significant features of squall line intensity graphs

provided that the gage is more or less in the center of the squall line formation.

Figure 34c shows that the network is right on the southern edge of the squall line, thus not

having a very extensive stratiform part. However, this is one example for an increasing 3D

graph, as described by pattern II (see 3D network analysis).

During data processing some attention was paid towards fractal analysis and possible patterns

within the intensity curves which then were stopped as the effort would have exceeded the

scope of this work. There, it could be seen, that the intensity graphs posses a fractal

dimension, although it could not be differentiated between the rainfall types. A closer look

seems promising, as it could help modeling rainfall intensities at a hydrologically useful scale.

In this sense particularly the fractal behavior of squall line events, of which most rainfall goes

into runoff, and which account for most of the total rainfall, should be examined in more

detail.

Page 63: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Chapter 5 - Conclusion

63

5. Conclusions

The observed rainfall was analyzed in different ways at different spatial and temporal scales.

At the event level, the analysis concentrated on gage and event totals, and at the intensity

level on the rainfall distribution within the event. Selected results from these steps were then

combined and used to classify the different rainfall types predominant for the Volta Basin.

Further steps involved the comparison and analysis of measured rainfall data from the region

with MM5 simulations and TRMM precipitation data.

Main outcomes were the quantification of the spatial rainfall variability for different classes,

the detection of the three rainfall types, and the classification of the different events to these

types. The comparison with SARI rainfall data yielded a close resemblance especially on a

monthly basis. MM5, at least at the short monthly and daily and rather small spatial basis

needs further improvement.

Particularly with regard to MM5 the high spatial variability, even within a rather small 9x9

km grid, has to be pointed out. Although tropical Africa’s rainfall is known for its

“patchiness”, mean CV values of 0.25 to 0.4 pose an enormous variability, making it

extremely difficult to determine reliable point rainfall.

TRMM images have proven to be a good source for the detection and of cloud clusters if the

coarse temporal resolution allows a match. Further analysis should be conducted towards the

validation of TRMM data over tropical Africa and the quantification of squall line

occurrences as they are major rainfall producers in the midland and northern regions of West

Africa.

Page 64: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Literature

64

Literature

Acheampong, P.K. (1988): Water Balance Analysis for Ghana. In: Geography, 73: 125-131.

Adelekan, I.O. (1998): Spatio-temporal variations in thunderstorm rainfall over Nigeria. In:

International Journal of Climatology 18: 1273-1284.

Adiku, S.G.K., et al. (1997) : An analysis of the within-season rainfall characteristics and

simulation of the daily rainfall in two savanna zones in Ghana. In: Agricultural and

Forest Meteorology, 86: 51-62.

Andreini, M., et al. (2000): Volta Basin Water Balance, ZEF Discussion Paper Nr. 21. ZEF,

Bonn: 2-7.

Beven. K.J. (2001): Rainfall-Runoff Modelling – The Primer. Wiley, Chichester: 240-244.

Caniaux, G., et al. (1994) : A Numerical Study of the Stratiform Region of a Fast-Moving

Squall Line. Part I: General Description and Water and Heat Budgets. In: Journal of

the Atmospheric Sciences, 51 (21). pp.2046-2074.

GSFC DAAC (2002): NASA GSFC Earth Sciences (GES) Distributed Active Archive Center

(DAAC). At: http://lake.nascom.nasa.gov/daac-

bin/whom/mk_page_cgi.pl?PATH=dataset/TRMM/01_Data_Products/01_Orbital/08

_Pr_Prof_2A_25/2001/10/11 (visited: 11.03.2002)

Gunston, H. (1998): Field Hydrology In Tropical Countries. Intermediate Technology

Publications, London.

Hayward, D. and J. Oguntoyinbo (1987): Climatology of West Africa. Hutchinson, London.

Herschey, R.W. (1999): Hydrometric Instruments. In: Herschey, R.W. (ed.) Hydrometry –

Principles and Practices. Wiley, Chichester: 85-142.

Herschey, R.W. (1999): Uncertainties in Hydrometric Measurements. In: Herschey, R.W.

(ed.) Hydrometry – Principles and Practices. Wiley, Chichester: 355-370.

Jones, K.R., et al. (1981) : Arid Zone Hydrology for Agricultural Development. FAO

Irrigation and Drainage Paper 37. FAO, Rome: 12-43.

Kamara, S.I. (1986): The origins and types of rainfall in West Africa. In: Weather, 41: 48-56.

Kasei, C.N. (1988): The Physical Environment of Semiarid Ghana. In: Unger, P.W., et al.

(eds.) Challenges in Dryland Agriculture – A Global Perspective. Texas Agricultural

Experiment Station, Amarillo/ Bushland: 350-354.

Kasei, C.N. (1990): A Synopsis on the Climate of the North of Ghana. Presented at the 2nd

workshop on improving farming systems in the savanna zone of Ghana. April 24-26

1990. Nyankpala Agricultural College. Nyankpala-Tamale, Ghana.

Page 65: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Literature

65

Kranjac-Berisavljevic', G., et al. (1998): Rethinking Natural Resource Degradation

In Semi-Arid Sub-Saharan Africa: The Case of Semi-Arid Ghana. Faculty of

Agriculture, University for Development Studies, Tamale, Ghana, and Overseas

Development Institute (ODI), London, United Kingdom. At:

http://www.odi.org.uk/rpeg/soil_degradation/ghlit.pdf (visited: 10.05.2002): 5-15.

GLOWA Volta Annual Report 2001 (2002): Atmosphere Cluster. ZEF, Bonn: 1-17.

Kunstmann, H. (2000): Numerische Modellierung der Wasserbilanz im Voltabecken. In:

Jahresbericht Fraunhofer Institut für Atmosphärische Umweltforschung.

Kunstmann, H. (2001): Modellierung und Validierung des Niederschlags in West Afrika. In:

Jahresbericht Fraunhofer Institut für Atmosphärische Umweltforschung.

Leduc, S.K., et al. (1980) : Precipitation Patterns in West Africa. In : Monthly Weather

Review, 108: 1567-1578.

Leroux, M. (2001): The Meteorology and Climate of Tropical Africa. Springer – Praxis books

in environmental sciences, London.

NOAA: National Weather Service Forecast Office – Miscellaneous – Weather Glossary. At:

http://www.crh.noaa.gov/dtx/?page=glossary (visited: 01.05.2002)

Ojo, O. (1977): The Climates of West Africa. Heinemann, London: 60-98.

Omotosho, J.B. (1984): Spatial and Seasonal Variation of Line Squalls over West Africa. In:

Archives for Meteorology, Geophysics, and Bioclimatology, Series A, 33: 143-150.

Peters, M., et al. (1989) : Rainfall intensity of West African Squall Lines. In: Annales

Geophysicae, 7 (3): 227-238.

Rijsberman, F. (2002): Developing a full proposal for the CP Water and Food. International

Water and Management Institute (IWMI). At: www.cgiar.org/iwmi/challenge-

program/word/DevelopingFullProposal(21_March_2002).doc. (visited: 25.05.2002)

Sharon, D. (1974): The Spatial Pattern of Convective Rainfall in Sukumaland, Tanzania – A

Statistical Analysis. In: Archives for Meteorology, Geophysics, and Bioclimatology,

Series B, 22: 201-218.

Schmidt-Kallert, E. (1994): Ghana : Fakten, Zahlen, Übersichten. Perthes, Gotha: 30-38.

Sustainable Water Use under Changing Land Use, Rainfall Reliability and Water in the Volta

Basin (1999). GLOWA Volta Project Proposal. ZEF, Bonn.

TRMM Homepage (2002): Tropical Rainfall Measuring Mission Homepage. At:

http://trmm.gsfc.nasa.gov/ (visited: 23.04.2002)

Weischet, W. and W. Endlicher (2000): Regionale Klimatologie - 2. Die Alte Welt : Europa,

Afrika, Asien. Teubner, Stuttgart: 261-283.

Page 66: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Literature

66

Windmeijer, P.N. and W. Andriesse (1993): Inland Valleys in West Africa: An Agro-

Ecological Characterization of Rice-Growing Environments. ILRI publication 52.

International Institute for Land Reclamation and Improvement, Wageningen: 15-29.

WMO (1994): Guide to Hydrological Practices – Data Acquisition and Processing, Analysis,

Forecasting and other Applications. WMO-No. 168.

Page 67: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Appendices

67

Appendix A

Appendix A shows all events sorted by rainfall type and size. Plotted are the intensity curves

for the highest gage of each event together with the 3D graphs. In the upper left corner

summarized information about the particular event is shown. This consists of the event ID, the

date (first tick), the rainfall for the particular gage and the mean rainfall over all gages in mm,

the coefficient of variation (CV) over all gages, the duration of the particular gage in seconds

and hours, and the rainfall type.

Intensity curves:

Shown on the x-axis is the time, and on the left y-axis the intensity in mm/hr, whereas the

right y-axis shows the cumulative density function (CDF) of the cumulative rainfall.

3D graphs:

Displayed are the 3D graphs for both networks. Network A shows all A gages together with

the mean rainfall for network B as the center gage. Network B shows all B gages.

Index – appendix A

ID Date Mean (mm) Max gage (mm) CV Type Page 112 30.08.2001 4.51 6.14 0.23 monsoonal 70 603 04.09.2001 4.73 6.29 0.24 monsoonal 70 500 03.09.2001 8.89 10.74 0.12 monsoonal 71 3776 11.10.2001 2.51 4.61 0.50 rainstorm 71 2219 23.09.2001 2.91 6.91 0.52 rainstorm 72 95 30.08.2001 2.96 4.95 0.51 rainstorm 72

4280 17.10.2001 3.15 3.38 0.12 rainstorm 73 2228 23.09.2001 3.65 7.96 0.54 rainstorm 73 574 04.09.2001 3.91 11.51 0.87 rainstorm 74 101 30.08.2001 4.06 7.57 0.51 rainstorm 74 3531 08.10.2001 4.30 7.37 0.32 rainstorm 75 3380 07.10.2001 6.70 11.98 0.28 rainstorm 75 1223 12.09.2001 12.52 19.33 0.31 rainstorm 76 2073 22.09.2001 13.71 21.42 0.26 rainstorm 76 1818 19.09.2001 3.36 4.83 0.21 squall line 77 2094 22.09.2001 9.57 16.21 0.48 squall line 77 720 06.09.2001 10.40 12.79 0.22 squall line 78 2170 23.09.2001 11.49 18.47 0.34 squall line 78 940 08.09.2001 24.24 33.11 0.28 squall line 79 2577 27.09.2001 27.64 43.11 0.28 squall line 79 1487 15.09.2001 32.94 46.72 0.25 squall line 80 1792 18.09.2001 66.36 89.00 0.17 squall line 80

Page 68: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Appendices

68

Appendix B

Appendix B consists of the TRMM precipitation images extracted for September 2001. The

images are all cut to a size of 10°x10° longitude and latitude with dotted lines at 5°. For

distance calculations a standard length of 110 km for 1° was used. Below each scene the date

is printed.

Index – appendix B

Date Page 06. Sept. 2001 82 09. Sept. 2001 82 12. Sept. 2001 82 13. Sept. 2001 82 14. Sept. 2001 82 15. Sept. 2001 82 19. Sept. 2001 83 22. Sept. 2001 83 27. Sept. 2001 83

Page 69: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Appendix A

69

Page 70: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Appendix A

70

Page 71: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Appendix A

71

Page 72: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Appendix A

72

Page 73: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Appendix A

73

Page 74: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Appendix A

74

Page 75: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Appendix A

75

Page 76: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Appendix A

76

Page 77: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Appendix A

77

Page 78: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Appendix A

78

Page 79: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Appendix A

79

Page 80: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Appendix B – TRMM Precipitation Radar Images

80

06. Sept. 2001 (Source : GSFC DAAC, 2002) 09. Sept. 2001 (Source : GSFC DAAC, 2002)

12. Sept. 2001 (Source : GSFC DAAC, 2002) 13. Sept. 2001 (Source : GSFC DAAC, 2002)

14. Sept. 2001 (Source : GSFC DAAC, 2002) 15. Sept. 2001 (Source : GSFC DAAC, 2002)

Page 81: Spatio-temporal Rainfall Patterns in Northern Ghana€¦ · Acknowledgments 2 Acknowledgments Special thanks goes to the staff of the Center for Development Research and the GLOWA

Appendix B – TRMM Precipitation Radar Images

81

19. Sept. 2001 (Source : GSFC DAAC, 2002) 22. Sept. 2001 (Source : GSFC DAAC, 2002)

27. Sept. 2001 (Source : GSFC DAAC, 2002)