The new SMOS-IC L-band vegetation index (L-VOD):

Overview and application to monitoring vegetation biomass at global scale

CCI Biomass Workshop, Paris, Spt. 25-26, 2018

J-P Wigneron, L. Fan, et al.With Bordeaux group: A. Al-Yaari, J. Swenson, F. Frappart, X. Li.

VOD (definition)

SMOS-IC L-VOD product (new , simplified, well-suited to applications)

Applications of L-VOD to vegetation monitoring

CCI Biomass Workshop, Paris, Spt. 25-26, 2018

SMOS (Soil Moisture and Ocean Salinity), PI Y Kerr

Spatial resolution: ~ 35-50kmRevisit time: Max. 3 daysSensitivity ~ 2K over land Goal of accuracy in SM: ~ 0.04 m3/m3

Retrieval algorithm: using multiangular and dual polarization TB

Soil moisture & vegetation opacity (VOD)

based on the inversion of L-MEB, (L-band Microwave Emission of the Biosphere)Wigneron et al., 2000-2018 (algorithm in ESA proposal, L-MEB, SMOS-IC product)

Launch : Dec. 2009: ~ a 9-year data set

Extinction

ATMOSPHERE

SOIL

VEGETATION

eS TS TB* S TB*

Tb*(z=H)VegetationEmission(1- )(1-ω)

Tb*(z=0)Soil reflectivityS

Soilemission

Extinction

VEGETATION

Sol

v

RADAR

s / 2

PASSIVE

Extinction

‐soil moisture (SM), determines smooth soil reflectivity smooth‐biomass (VOD), determines vegetation extinction γ = exp(‐VOD/cos())

‐temperature (TB = emissivity x temperature) 

‐canopy type  (ω), soil roughness (rough= C . smooth), and texture

The Brightness temperature (TB) observations are sensitive to:

L-MEB (L-band Microwave Emission of the Biosphere model)

VOD (nadir) = b . VWC Jackson and Schmugge, 1991

with,

VWC = vegetation water content (kg/m2)

b~0.12 (0.1-0.2 for crops)

For vegetation, L-MEB is based on a zero order solution of radiative transfer equations (- model):

- VOD = KE. H, accounts for extinction effectsKExtinction = KAbsorption + KScatteringKAbsorption = KEmission

- accounts for scattering effects (KScattering / KExtinction)

TBveg=(1-e-VOD/cos())(1-)Tveg(1+soile-VOD/cos())radiometer

SOIL

VEG

ATMSKY

H=Height of Crop

Theory

Experimental

In situ:• EMIRAD (TUD) at the INRA Avignon test site (soybean, corn)• Smosrex, 2003-2010, Toulouse, with Lewis (CESBIO, CNRM, INRA, ONERA), soil, fallow• Landes forest, 2004-2007 (INRA), with EMIRAD-1 (TUD), coniferous forest

• Elbara, 2004-2006 (ETH, U. of Bern), grass, deciduous forest• Elbara -2, 2010 (ESA funded) at the Munich, VAS, Sodankyla sites, grass, mattoral, forest

and airborne• Carols, 2007-2010 (Cnes, ESA), Smosmania (France) and Vas sites (Spain), …

L-MEB algorithm development /evaluation

LEWIS at Smosrex site

MELBEX- EMIRADINRA’01 - EMIRAD

EMIRAD, Landes forest

Examples of results : SOYBEAN (INRA-1991) [Wigneron et al., RSE, 1995-2007]

Retrieved soil moisture

Retrieved VOD, VWC and LAI = f(time)

Retrieved VOD

R2=0.96 R2=0.88

VWC LAItime

VOD VOD VOD

VWC

LAI

-multi-angular observations ➡ simultaneous retrievals of SM and VOD (Wigneron et al., 1995, 2000)

-in SMOS-IC SM retrievals, no need of optical indices to estimate VOD (and vice versa)

SMAP: In the SCA algorithm, VOD is estimated from NDVIAMSR-E: iterative approach based on only 2 observations (LPRM algorithm)

-L-band (1.4GHz, ~ 30 cm): higher sensing capabilitiesthrough dense vegetation than C- (6 GHz, ~ 5 cm) and X- bands

-passive observations are much less sensitive to structural effects of vegetation (row, vertical structure), soil (roughness, surface geometry), topography, etc. than radar observations

the b and parameter are relatively constant for varying vegetation conditions

= 0.07 for forests = 0.1 for other vegetation types

(no need for parameter tuning)

Key features of SMOS to retrieve VOD:

Dynamics of carbon stocks in Africa over 2010-2016

X-VOD

SMOS-IC L-VOD

Biomass(Baccini map2007-2008)

-spatial calibration L-VOD / Biomass in 2011

-’space’ for time substitution: L-VOD is used to monitor time changes in carbon stocks in Africa

Biomass(Baccini map2007-2008)

F. Tian,, et al. "Coupling of ecosystem-scale plant water storage and leaf phenology observed by satellite", Nature EE

Pre-rain Miombo forest© C. Ryan, University of Edinburgh

A high temporal decoupling between plant water storage and LAI in dry Tropical forests(especially in Miombo)

L-VOD

LAI

Time variation in L-VOD and LAI (Miombo)© F. Tian, University of Copenhagen

Time lag between L-VOD and LAI© F. Tian, University of Copenhagen

Miombo

THANK YOU!

Green vegetation:gv =f°(LAI) =f°(VWC) (gv =b•VWC, b~0.1)

Retrieving of the # components of optical depth:[Saleh et al., RSE, 2006]

retrieving VOD_GV (standing vegetation):

for dry conditions : dry litter and no interception

Nadir path

Satellite

Spacecraftvelocity

d N

Swath1000 km

30°

= 55°

Local incidenceangle

Earth

m

SMOS : le Système d’Observation:

Champ de vue (FOV)

visées multi-angulaires: un point au sol est vu sous différents angles de visée au fur et à mesure que le satellite avance:

18SMOS Brightness Temperature (L1C product), Browse product at 42.5°

SMOS : un Système d’Observation Multiangulaire

Le principe:

Mesure de la réflectivité du sol = f(constante diélectrique ε)

sCte diélectrique

εHumidité du sol SM (m3/m3)

Texturestructure du sol

s_lisseMesure radarou passifCorrection

rugosité

SM

fort contraste sol sec (ε~5) et sol humide (ε~30)

Soil dielectric constant = f(SM) à 1.4GHZ

Ulaby et al. (1986)

Soil dielectric

constant ε

Mesures dans les Micro‐ondes• Le principe: mesure de la reflectivité du sol =

f(constante dielectrique ε)Il existe un fort contraste entre la constante dielectrique d’un sol sec (ε~5) et celle d’un sol humide (ε~30)

• Mesures dites ‘Actives’ (Radar) ou ‘Passives’• actif →reflectivité• passif →émissivité (= 1-Réflectivité)

• Mesures dans le domaine ‘basse fréquence’ ~ 1.4 Ghz (bande L) et 10 Ghz (bande X):-faible sensibilité aux effets atmosphériques:(mesure tout temps & corrections précises)-faible atténuation du signal sol par la végétation

Soil dielectric constant = f(SM) à 1.4GHZ

Ulaby et al. (1986)

Nadir path

Satellite

Spacecraftvelocity

d N

Swath1000 km

30°

= 55°

Local incidenceangle

Earth

m

SMOS : Multi‐angular observation system

Field of view (FOV)

multi-angular observations:

-a given site on Earth is seen at different incidence angles as the satellite moves ahead.

-larger angle ranges for sites close to the sub-satellite track

soil

vegetation

atmosphere

vegetationemission

Reflected (Гs)vegetationemission

soil emission : (1-Гs) Ts

TB* Гs TB* (1-Гs) Ts

TB

TB*(z=0)

TB*(z=-d)

diffusion absorption

()

diffusion absorption

()

Vegetation attenuation increases as frequency increases

Saturation of Biomass = f(VOD) comes quicker at high frequencies

X-band (~10 GHz, ~3 cm), AMSR-EC-band (~5 GHz, ~6 cm), AMSR-E, ASCAT… ~100 t/haL-band (1.4 GHz, ~30 cm), ESA/SMOS, NASA/SMAP …, ~250 t/haP-band (~0.4 GHz, ~75 cm), ESA/Biomass…, ~ no saturation ?

Quicker saturation for active vs passive systems

Time variations in L-VOD, SM, EVI,

rainfall

(Most) Dense tropical forests in the Amazon basin (Guyana) = an extreme case

-SM can be clearly related to rainfall

Tian et al., 2017