Radiative Transfer Modeling - EO LDAS - assimila.eu · 1 Nadine Gobron Institute for Environment and Sustainability of EC-JRC, 21020 Ispra (VA), Italy Radiative Transfer Modeling

1

Nadine GobronInstitute for Environment and Sustainability of EC-JRC, 21020

Ispra (VA), Italy

Radiative Transfer Modeling - EO LDAS

EO-LDAS Team: Jeff Settle, Thomas Kaminski, Philip Lewis, Sietse Los, Peter North,

Tristan Quaife, Jon Styles.

JRC: B. Pinty & J.-L. Widlowski

2

EO-LDAS

Radiances TOC or TOA Land Properties

Process Model

RT Models

EO-LDAS (Simplest View)

Complexity depends:

1) Spatial resolution

2) number of land state

variables of interest

3

Vertical radiative coupling between geophysical media

Atmosphere

')',()',(~),()],,(~[4

dzIzzIz sOe

Vegetation

Soil

Upper limit (1)

RTE

Lower Limit (1)

Upper limit (2)

RTE

Lower Limit (2)

Upper limit (3)

RTE

Lower Limit (3)

ZA

ZV

ZS

Extinction coefficient Differential scattering coefficient

In each medium, the transfer of radiation may be represented

by the following approximate equation:Lev

el o

f im

ple

men

tati

on

TOA

TOC

Pinty B. and Verstraete M. M. (1998) `Introduction to Radiation Transfer Modeling in Geophysical Media', in From Urban

Air Pollution to Extra-Solar Planets, Vol. 3, Chapter IV, Edited by C. Boutron, 67--87.

4

Atmosphere

Vegetation

Soil

•Non oriented small scatterers

•Infinite number of scatterers

•Low density turbid medium

•Oriented finite-size scatterers

•Finite number of scatterers

•Dense discrete medium

•Oriented small-size scatterers

•Infinite number of clustered scatterers

•Compact semi-infinite medium


(1D view)

RT 1D - no 3D



5

Atmosphere

Vegetation

Soil

Za

ZV

ZS

),(

azI)(),( 0 ota IzI

0),(

zI

The description of the interaction of a radiation field with a

layered geophysical medium implies the solution of radiation

transfer equations and the specification of appropriate

boundary conditions

),(]||

exp[)(),(0

,

bada

tatv zIzIzI

'''

,

'

2

||)(),,(1

),(

dzIzzI babavba

'''

,

'

2

||)(),,(1

),(

dzIzzI bvbvsbv




6

Model Assumption

),,( 'bav z

Parametric

2-Stream

1-D

3-D

Rahman, H., M. M. Verstraete, and B. Pinty (1993) ' Coupled surface-atmosphere reflectance (CSAR) model. 1. Model

description and inversion on synthetic data ', Journal of Geophysical Research, 98, 20,779-20,789.

7

Parametric

2-Stream

1-D

3-D

Model Assumption

),,( 'bav z

8

Parametric

2-Stream

1-D

3-D

Model Assumption

),,( 'bav z

9

Parametric

2-Stream

1-D

3-D

Model Assumption

),,( 'bav z

10

Spatial Resolution ? Process Model ? Coupling RT atmos. ?

Parametric

2-Stream

1-D

3-D

N/A directly

N/A directly (fluxes)

Which type of RT model?

Option?

discussion …

),,( 'bav z

11

3-D versus 1-D

True <LAI> =2.0

3-D heterogeneous system

True <LAI> =2.0True <LAI> =2.0

3-D heterogeneous system

True <LAI> =2.0

Direct transmission at 30 degrees Sun zenith angle,

0.596

)(3

LAITdirect

D


0.596

)(3

LAITdirect

D

1-D system representation

True <LAI> =2.0

1-D system representation

True <LAI> =2.0

0

12

exp)(LAI

LAITdirect

D


= 0.312

Effects induced by internal variability of LAIPinty, B., N. Gobron, J.-L. Widlowski, T. Lavergne and M. M. Verstraete (2004) `Synergy between 1-D and 3-D radiation

transfer models to retrieve vegetation canopy properties from remote sensing data', Journal of Geophysical Research, Vol.109,

D21205 10.1029/2004JD005214.

12

Parametric

2-Stream

1-D

3-D

Model Assumption

Gobron, N., B. Pinty, M. M. Verstraete and Y. Govaerts (1997) ' A semi-discrete model for the scattering of light by vegetation

', Journal of Geophysical Research, 102, 9431-9446.

13

Discrete canopy: 1D representation

H

Parameters in 1D representation:

Height of canopy, Size of a single leaf & Leaves orientation

(leaf angle distribution), LAI, leaf spectral values, soil

albedo (or 3-4 variables if anisotropic).

Df



14

Ex: Semi-discrete model

single-collided-by-soil BRF:

Semi-discrete is a 1-D’ model using analytical & numerical methods.

K

r

r

r

rsoil

K

G

V

Vai

Gai

cos

)(1),(

cos

)(1 0

0

0 K

Ω0Ωr

ai

ai

ai

single-collided-by-leaves BRF:Ω0Ωr

1

0

0

0

0

cos

)(1

cos

)(1

coscos

)(

Ki

i

r

r

r

i

r

r G

V

Vai

Gaiai

Multiple-collided by leaves and soil BRF:

obtained with the discrete ordinates numerical method.

½-discrete splits the BRF into 3 components



16

Parametric

2-Stream

1-D

3-D

Model Assumption

17

3-Dimensional problem

where the plan-parallel concept may be

inappropriate:

•Document the errors due to an oversimplification of the full

3-D situation, i.e. deviations from the 1D case.

•Explore new ways and techniques for representing, at

limited costs, the 3D nature of the medium which basically

require almost an infinity of parameters!

•Address the application issues for geophysical modeling, e.g.

the definition of new “equivalent variables”, and satellite data

interpretation, e.g. the non-uniqueness of the inverse

problem.

18

3-D Radiative Transfer Equations

3-D Models

•Ray tracing models

•Geometrical models

•Hybrid models

I(x,)G(x,)uL (x)I(x,) uL (x)

(x, )I(x, )d

4

19


Feasibility for implementation in EO-LDAS

Parametric

2-Stream

1-D

3-D

Adjoint Code or LUT

Depend on process

model & spatial

resolution

20


Parametric

2-Stream

1-D

3-D

EO-LDAS Prototype

21

RAMI evaluates models

in forward mode

RAdiative transfer Model Intercomparison

http://rami-benchmark.jrc.ec.europa.eu

Purpose:•act as common platform

for intercomparison efforts

•document uncertainties

and errors among models.

•establish protocol for RT

model evaluation.

22

• RAMI-1 (1999):– Turbid medium and discrete

– Solar domain + purist corner

• RAMI-2 (2002):– Topography + true “zoom-in”

• RAMI-3 (2005):– Birch and conifer scene

(GO models)

– Heterogeneous purist corner

– Local transmission and horizontal flux measurements

HOMogeneous HETerogeneous

RAMI-1 RAMI-2 RAMI-3

13

20

42

Fra

cti

on o

f H

ET

[%]

Num

ber

Experi

ments


660715

980

RAdiative transfer Model Intercomparison

Pinty, B., N. Gobron, J.-L. Widlowski , S. A. W. Gerstl, M. M. Verstraete, M. Antunes, C. Bacour, F. Gascon, J.-P.

Gastellu, N. Goel, S. Jacquemoud, P. North, W. Qin, and R. Thompson (2001) 'Radiation Transfer Model Intercomparison

(RAMI) Exercise', Journal of Geophysical Research, 106, 11,937-11,956.

Pinty, B., J-L. Widlowski, M. Taberner, N. Gobron, M. M. Verstraete and the RAMI-2 Participants (2004) ̀ The RAdiation

transfer Model Intercomparison (RAMI) exercise: Results from the second phase', Journal of Geophysical Research, Vol.109,

D06210 10.1029/2003JD004252.y

Widlowski, J.-L., M. Taberner, B. Pinty, and colleagues (2007) `The third RAdiation transfer Model Intercomparison (RAMI)

exercise: Documenting progress in canopy reflectance models', Journal of Geophysical Research, Vol.112, doi:

10.1029/2006JD007821.

23MODEL NAME PARTICIPANT AFFILIATION

ACRM A. Kuusk Tartu Observatory,Estonia

DART J.P. Gastellu, E.Martin CESBIO, France

drat M. Disney, P. Lewis UCL, UK

FLIGHT P. North Univ. Swansea, UK

frat P. Lewis, M. Disney UCL, UK

FRT M. Möttus, A. Kuusk Tartu Observatory,Estonia

Hyemalis R. Ruiloba NOVELTIS, France

MAC R. Fernandes CCRS, Canada

mbrf W. Qin NASA GFSC, USA

RGM D. Xie, W. Qin Beijing N. Univ., China

Rayspread T. Lavergne JRC, Italy

raytran T. Lavergne JRC,Italy

SAIL++ W. Verhoef NLR, Netherlands

½ discret N. Gobron JRC, Italy

Sprint3 R. Thompson Cox, USA

4SAIL2 W. Verhoef NLR, Netherlands

5scale N. Rochdie, R. Fernandes CCRS, Canada

2stream B. Pinty, T. Lavergne JRC, Italy

3-D models

1-D models

new in RAMI-3


8

13

18

Num

ber

of

models


5

10

11/13

3-D

m

odels

24

25

Absorption

Albedo

Transmission

Measurement Types

Measurements include

Flux quantities:

• Albedo

• Transmission

• Absorption

BRF quantities:

•Total BRF (PP+OP)

total BRF

multiple collided

single collided

single un-collided

BRF quantities:

•Total BRF (PP+OP)

•BRF components–multiple collided

–single un-collided(hit soil only once)

–single collided (hit leaves only once)

26

RT Model Intercomparison Caveat

• In general there is no absolute ‘truth’ available! Model results cannot be evaluated against some reference standard

• Laboratory data are difficult to use as reference standard due to incomplete knowledge of the exact illumination, measurement, as well as (structural and spectral) target properties.

but

• Model results can be compared against each other to document their relative differences.

• Model results can be compared over ensembles of test scenarios to establish trends/behaviours in their performance.

• Careful inspection/verification of an ensemble of model results may lead to the establishment of the “most credible solutions” as a surrogate for the “truth”.

27

Participation & Performance

HOM DIS cases

X2 uses σ=0.03<BRF>3D

Most models are indiscernible

(within 3%) from surrogate truth

More models are different (by

≥ 3%) from surrogate truth

Model performance may be

affected by spectral regimes

LAI=1

LAI=5

LAI=2

Relative Intercomparison: Χ2 statistics

EO-LDAS prototype

Widlowski, J.-L., M. Taberner, B. Pinty, and colleagues (2007) `The third RAdiation transfer Model Intercomparison

(RAMI) exercise: Documenting progress in canopy reflectance models', Journal of Geophysical Research, Vol.112, doi:

10.1029/2006JD007821.

28

Homogeneous

X2 uses σ=0.03<BRF>3D

RAMI-2RAMI-3

DiscreteHeterogeneous

RAMI-2RAMI-3

Model performance improved from RAMI-2 to RAMI-3!

Relative Intercomparison: Χ2 statistics

29

FLIGHT: structure and geometry

Diameter

Ez

x

z y

Ez

DBH

Radius

Canopy:

- Leaf Area Index(LAI)

- Crown envelopes

- Leaf angle distribution (LAD)

- Optical properties

North, Peter R. J. (1996) 'Three-Dimensional Forest Light Interaction Model Using a Monte Carlo Method', IEEE

Transactions on Geoscience and Remote Sensing, 34, 946-956

30

P1

Diameter

Ez

x

z y

Ez

DBH

Radius

Light interaction:

- Source(s)

- Sensor

- Photon paths (multiple)

- Scattering

MCRT - random sampling of photon trajectoriesNorth, Peter R. J. (1996) 'Three-Dimensional Forest Light Interaction Model Using a Monte Carlo Method', IEEE


31

FLIGHT - canopy structure



32

European beech

Coniferous forest



FLIGHT – 3D scene

33

Land surface characteristicsAngular and spectral variation

RED (670 nm) NIR (870 nm)

34

RAMI intercomparison

FLIGHT vs ASAS

North 1996; Pinty et al., 2001, 2003; Widlowski et al., 2008

FLIGHT BRF Validation

37

Various RT vegetation model:

Parametric: serve as proxy for BRDF over land but no direct link

with process model except if surface albedo is foreseen.

Main advantage: Can be used for solving coupled RT prb when TOA

data.

2-stream, 1-D (semi-discrete): Always effective

variables. Adjoint code

3-D: Infinity of parameters? LUT

All can be coupled with atmospheric model like 6S .

38

10 m 100 m 1000 m

zero order model (veg. parameters t-1)

dynamic vegetation model (climate

parameters)

Climate constraint model (climate

parameters+veg t-1)

> 10 km

RT 3D

RT 1D’

RT 1D

Main open issues

Where are the boundaries (and needs)?

39

Future analysis … prototype

Implement leaf spectra module into semi-discrete

code.

Coupled with 6S (new version? polarization)

Dickinson‘s model ? (process model)

Test adjoint code (with TK & TQ).

Simulation of like EO data for validation/verification.

Radiative Transfer Modeling - EO LDAS - assimila.eu · 1 Nadine Gobron Institute for Environment and Sustainability of EC-JRC, 21020 Ispra (VA), Italy Radiative Transfer Modeling

Documents