Imaging Spectroscopy in Ecology and the Carbon Cycle - Heading Towards New FrontiersMichael Schaepman, Wageningen University, NL

With contributions from Han van Dobben, Sander Mücher, Wieger Wamelink, Alterra, NLManuel Gloor, Princeton Univ., USAGabriela Schaepman-Strub, WU, NL

Content

1. Introduction2. Imaging spectroscopy based approaches

to ecology3. Challenges4. Outlook5. Discussion

1. Introduction

Introduction

Good ecologists don’t wearcontact lenses and always

a raincoat

State and Temporal Dynamics of Ecosystems Thematic classification

Wildlife habitat, successional stage, forest fragmentation, invasive species, biodiversity

Biophysics LAI, specific leaf area, biomass, canopy moisture content,

canopy cover Temporal dynamics

Greening, senescence, forest change detection, windfall Spatial patterns

Spatial metrics, patches, shapes for disturbed forests Large area mapping

Stratifying forest at national level, GlobCover program

Cohen, W.B., and Goward, S.N. (2004). Landsat’s Role in Ecological Applications of Remote Sensing. BioScience, 54(6): 535-545.

Integration with Ecological Models

Physiological models Integration of multitemporal data in to physiologically

based process models (surface energy balance, evapotranspiration, productivity, Light Use Efficiency (LUE) to derive NPP)

Accounting models Carbon accounting models (estimation of biomass change

over time using RS) Habitat assessments

Habitat characterization and mapping Socioeconomic studies

Landscape ownership patterns, visualization of ecological consequences

Cohen, W.B., and Goward, S.N. (2004). Landsat’s Role in Ecological Applications of Remote Sensing. BioScience, 54(6): 535-545.

2. Imaging Spectroscopy Based Approaches

Imaging spectroscopy based approaches to ecology Classification Examples

Imaging Spectroscopy Approaches

Continuous fields Quantitative, physical methods

• Improvement path clear, long development time

Quantitative, statistical methods• Effective, fast, no cause-effect relationship

Categorical variables Classification based approaches Discrete classes

Base maps Orientation and visualisation

Example I

Continuous Fields – Quantitative Physical Based Method Product: NASA MODIS Product No. 15 (MOD15)

• Leaf Area Index (LAI)• Fraction of Photosynthetically Active Radiation Absorbed by

Vegetation (FPAR)

Radiative transfer based approach derived from reflectances and ancillary data (land cover type, background, etc.)

Continuous Fields – Physical Based

http://earthobservatory.nasa.gov/Newsroom/NasaNews/2001/200112206806.html

Example II

Continuous Fields – Quantitative Statistical Based Method Product: (Effective) Leaf Area Index (LAI) derived

from HyMAp data Correlation based approach, where a

Perpendicular Vegetation Index (PVI) is correlated with LAI (R2=0.77, RMSE=0.54 m2/m2)

Continuous Fields – Statistical Based

Schlerf, M, C. Atzberger and J. Hill (2005), Remote sensing of forest biophysical variables using HyMap imaging spectrometer data. Remote Sensing of Environment, Vol. 95(2), p. 177-194

Example III

Categorical Variables – Classification Based Product: 18 land use/cover classes derived from

EO-1 ALI and Hyperion Two level classification scheme starting with 10

classes in first level and 18 classes in second level using segmented linear discriminant analysis (segLDA).

Categorical Variables – Classification Based

Xu, B., and Gong, P. (2002). Land use/cover classification with multispectral and hyperspectral EO-1 data: a comparison. http://www.cnr.berkely.edu/~bingxu/pub/hyper_101002.pdf

Example IV

Categorical Variables – Discrete Classes Product: Invasive specie map of iceplant

(Carpobrotus edulis) expressed in different density levels.

Continuum removal based approach defining presence/absence of Iceplant in combination with MNF. Discretization of results is based on land managers requirements for detection of invasion fronts.

Categorical Variables – Discrete Classes

Underwood, E., Ustin, S., and DiPietro, D. (2003). Mapping nonnative plants using hyperspectral imagery, Remote Sensing of Environment, Vol. 86(2), p. 150-161.

Example V

Base Map Product: WTC debris analysis by mapping levels

of Chrysotile (asbestiform minerals, potential Asbestos indicators)

AVIRIS data recorded on September 16, 18, 22, 23. 2001 over the WTC area. Base maps used for identification of affected sites and determination of ground sampling analysis. Subsequent SUM approach of AVIRIS data.

Base Map

http://pubs.usgs.gov/of/2001/ofr-01-0429/results.html

3. Challenges

Climate scenarios vs. vegetation scenarios Global Land Use/Cover Monitoring Land-Biosphere Models Validation Biochemistry

There are many more! Integration of multiple sensors Scaling issues etc.

Climate Scenarios vs. Vegetation Scenarios IPCC climate change predictions are only

driven by climate scenarios: Currently DGVMs do not consider (at all/well

enough) the inter-/intra-year variability of vegetation

Vegetation can change at much faster rates due to disturbance, rather than due to climate change

To better explain the uncertainty in the land sink/source domain, vegetation scenarios (analog to climate scenarios) should be developed accounting for disturbance

CO2 and Climate: Model Divergence

Source: M. Rast, Ed., SPECTRA – Surface Processes and Ecosystem Changes Through Response Analysis, ESA SP-1279(2), 2004, pp. 66; Data: IPCC (2001) Climate Change 2001: The Scientific Basis. Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge University Press, Cambridge, U.K., 881pp.

Global Net Carbon Balance

Source: M. Rast, Ed., SPECTRA – Surface Processes and Ecosystem Changes Through Response Analysis, ESA SP-1279(2), 2004, pp. 66; Data: Joos, F., G.-K. Plattner, T.F. Stocker, A. Körtzinger, and D.W.R. Wallace, 2003: Trends in Marine Dissolved Oxygen: Implications for Ocean Circulation Changes and the Carbon Budget, EOS, 84 (21), 197-201.

Bars indicate a decade each

Uncertainties are in black

Biomass Divergence due to Management

c.f. contribution of Schmidt, et al. (2005), and Kooistra et al. (2005) for details (also this EARSeL workshop)

Global Land Use/Cover Monitoring

Millennium Ecosystem Assessment

Conclusions: What are the most important uncertainties hindering decision making concerning ecosystems? There are major gaps in global and national monitoring

systems that result in the absence of well-documented, comparable, time-series information for many ecosystem features and that pose significant barriers in assessing conditions and trends in ecosystem services. Moreover, in a number of cases, including hydrological systems, the condition of the monitoring systems that do exist is declining.

Although for 30 years remote sensing capacity has been available that could enable rigorous global monitoring of land cover change, financial resources have not been available to process this information, and thus accurate measurements of land cover change are only available on a case study basis.

MA Board (2005). Millennium Ecosystem Assessment Synthesis Report, Reid, W.V., (ed.), p. 219, http://www.maweb.org

Existing and Planned Land Cover Maps

Mucher, S., and Schaepman, M. (2005). Inventory of regional, national and global land cover maps produced using satellite data. Unpublished data.

Land Cover Changes in EuropePercentage of transect area changed between 1950-1990

-3,00%

-2,00%

-1,00%

0,00%

1,00%

2,00%

3,00%

Artificial Surfaces Agricultural areas Forest and seminatural area

Weltands Water bodies

ch

ang

ed

are

a p

er d

ec

ade

Increase

Decrease

Netto

Mucher, S. (2005), BIOPRESS – Linking Pan-European Land Cover Change to Pressures on Biodiversity

Land Cover Changes in EuropePercentage of transect area changed between 1990-2000

-3,00%

-2,00%

-1,00%

0,00%

1,00%

2,00%

3,00%

Artificial Surfaces Agricultural areas Forest and seminatural area

Weltands Water bodies

ch

an

ge

d a

rea

pe

r d

eca

de

Increase

Decrease

Netto

Mucher, S. (2005), BIOPRESS – Linking Pan-European Land Cover Change to Pressures on Biodiversity

Land-Biosphere Models

Land-Biosphere ModelsWhat processes are / should be mapped by land biosphere models ? Carbon Engine

carbon fixed per unit time = F(CO2, light, water availability, temperature, nutrients)

Carbon allocation distribution of C fixed to different living tissue (e.g. stems or roots) =

F(geometry, physiology,plant functional type, species) “Remineralisation ”

carbon flow to nonliving forms and decomposition (fast and slow soil pools) = F(plant functional type, physiology, microbiology, molecular structure (e.g. lignin vs. waxes or cellulose))

Hydrology soil water balance, (root depths)

Population dynamics early versus late successional species through competition for

resources (light, nutrients, water) = F(stand height, stand age, physiology)

disturbance: humans (land use), fire, windfall, insects = F(climate, humans)

Gloor, M. (2005), SPECTRA Simulator – Final Report, ESA

CASA BETHY PnET LM3 SMART/SUMO

Carbon Engine

Light use efficiency, PAR,

fPAR

(Farquhar, Ball, Berry) or LUE

Pmax=a+b*N, where N is foliar

Nitrogen

Farquhar, Ball, Berry

C-ass = f(light, N,P,water

availability, temp)

Phenology fPAR ? Predicted fPAR Not relevant (timestep = 1y)

Allocation Globally fixed ratios; leaf, litter,

roots

? Simple allocation rules for tissue

types

Allometries Ratios (root, shoot, leaf) fixed per

vegetation type

Remineralisation

5 litter, 2 organic pools, first order

decay

? No soil carbon component

Fast and slow pools of C and N

Litter + 2 organic pools, fixed ratio +

1st order decay

Hydrology Bucket type Bucket type One soil layer Bucket type Supplied by external

hydrological model (WATBAL, SWAP)

Discretization PFT’s PFT’s Biomass produced only by tissue type

(foliage)

Defined by mortality and

fecundity functions (species build a

continuum)

5 FT's that compete for light,

N, P, water

Demography None None None Core of model None (but tree mortality included)

Reference Potter et al., Global Biogeochem. Cycles, 7(4): 811-841, 1993

Knorr, Global Ecology & Biogeography, 9:225-252, 2000

Aber, Oecologica, 92(4): 463-474, 1992

Carbon Mitigation Initiative, Princeton

Univ., 2005

Wamelink & al. in prep., 2005

Comparison of 5 Different Land-Biosphere Models

Gloor, M. (2005), van Dobben, H. (2005), pers. comm.

RS Derived Variables and ModelsVariable Instrument CASA Model ED Model SMART/

SUMO

fPAR Optical NPP - NPP

LAI Optical - Eval. predicted LAI Eval. predicted LAI

Albedo Optical/Thermal/IS

- Eval. predicted Albedo -

fCover Optical - Eval. predicted fCover -

fLiving/Dead Biomass

Optical - Eval. predicted mortality -

Leaf Chlorophyll

Optical/IS - - -

Leaf Water Optical/IS - ? -

Leaf Dry Matter

Optical/IS - Eval. leaf dry matter -

Foliage Temperature

Thermal - Important for eval. of water loss during photosynthesis

Eval. potential evapotransp.

Soil Temperature

Thermal - Water stress eval. Water stress eval

Leaf Nitrogen IS - - Eval predicted leaf N

Gloor, M. (2005), van Dobben, H. (2005), pers. comm.

NPP Controls

Geographic Variation in Climatic Controls of Terrestrial Net Primary Production (Nemani, 2003)

NPP Change

Nemani (2003): Trends in global net primary production (NPP) anomalies from 1981 through 1999, computed from historical AVHRR-NDVI data

Ecocasting

Ecological forecasting (or 'ecocasting') is the prediction of ecosystem parameters. NASA Ames is developing advanced computing technologies for converting massive streams of satellite remote sensing data into ecocasts that are easy to read and use.

Nemani, R. (2005): http://geo.arc.nasa.gov/sge/ecocast/data/carbon.html

Validation

Geospatial interpolation Validation of products Representativeness of sites

Geospatial Interpolation for Assimilation

BELMANIP

Baret, F. et al. (2005), Validation and inter-comparison of medium resolution LAI and fPAR products derived from MODIS, MERIS and Vegetation, EGU, Geophys. Res. Abstracts, Vol. 7, 01765

Biochemistry

The biochemistry has gotten a bit lost over the past few years – it must come back in the future!

Biochemicals Present in Vegetation Spectra

Source: Schaepman, M. (2005): Spectrodirectional Remote Sensing – From Pixels to Processes, Inaugural Address, Wageningen.

Decay of a Ficus benjamina L. Leaf

Source:Bartholomeus, H., and Schaepman M. (2004)Decay of Ficus benjamina L. in 10 minutes steps over 8 hrs, unpublished

Each time step is 10 mins., total duration 8 hrsMeasurement is reflectance plus reflected transmittance

Undisturbedleaf

Maximal Spectral Resolution

900850800750700650600550500450400Wavelength [nm]

0.20

0.19

0.18

0.17

0.16

0.15

0.14

0.13

0.12

0.11

0.10

0.09

0.08

0.07

0.06

TO

A a

t sen

sor

radi

ance

[W/(

m2 s

r nm

)]

1.4

1.2

1.0

0.8

0.6

Rat

io []

MERIS TOA bands (incl. FWHM) MERIS TOA radiance (± 1 stdev) ASD ground reflectance modelled to TOA radiance using MODTRAN sun ASD ground reflectance modelled to TOA radiance using Thuillier 2002 sun Ratio of MODTRAN sun to Thuillier 2002 sun

MERIS data:MER_FR__1PNIPA20020822_180221_000000872008_00442_02500_0031.N1

MODTRAN standard solar irradiance vs. Thuillier 2002 solar irradiance

Mean deviation (400 nm - 900 nm): 4.66%

Source:Kneubühler, M., Schaepman, M.E., Thome, K.J., & Schläpfer, D.R. (2003) MERIS/ENVISAT vicarious calibration over land. In Sensors, Systems, and Next-Generation Satellites VII, Vol. 5234, pp. 614-623. SPIE, Barcelona.

4. Outlook

Imaging spectroscopy has deepened the physical understanding of the interaction of photons with matter

Certain links between remote sensing and ecology have been significantly advanced because of advances in imaging spectroscopy

Imaging spectroscopy will never solve the challenges alone – integrated solutions will be the future trend

Nevertheless, we need to make sure imaging spectroscopy will soon be a commodity and not remain a fancy technology in the shade of …!

Thank you for your attention!

© Wageningen UR