comparison of green vegetation fraction retrievals from spot-vegetation and msg-seviri sensors

(1) CNRS UMR 8586 PRODIG Pôle de Recherche pour l’Organisation et la Diffusion de l’Information Géographique, Paris, France

(2) Turkish State Meteorological Service, Remote Sensing Division, Ankara, Turkey

Comparison of Green Vegetation Fraction Retrievals from SPOT-VEGETATION and MSG-SEVIRI Sensors Bernard LACAZE (1) and Aydin ERTÜRK (2)

1

SPOT-VEGETATION and MSG-SEVIRI NDVI data

SPOT-VEGETATION : deriving Green Vegetation Fraction (GVF) from scaled NDVI

MSG-SEVIRI : deriving GVF from spectral unmixing or scaled NDVI

Results : Comparison of dekadal NDVI and GVF data (MSG-SEVIRI and SPOT-VEGETATION)

Conclusion

Outline

IGARSS 2011 24-29 July, Vancouver, Canada 2

SPOT-VEGETATION sensor (since 1998)

IGARSS 2011 24-29 July, Vancouver, Canada

sun synchronous orbit

altitude 822 km

return interval 1 day

swath width 2250km

SPOT-4 and SPOT-5

3

SPOT-VEGETATION 10-daily NDVI


S10 :Maximum Value Compositing

4

SPOT-VEGETATION 10-daily NDVI


NDVI data : HDF format, 1 byte/pixel

Real NDVI = 0.004 * Digital Number -0.1

Status map : HDF format, 1 byte /pixel

5

MSG-SEVIRI sensor (since 2004)


MSG-1 and MSG-2

SEVIRI one image every 15mn Spinning Enhanced Visible and Infrared Imager

resolution at nadir 3km (1km for channel 12)6

- Land surface monitoring products are elaborated by EUMETSAT Land Surface Analysis Satellite Application Facility (LSA SAF) Lisbon, Portugal : NDVI is not a standard product of LSA SAF (vegetation parameters = FVC, LAI, FAPAR)

- daily NDVI derived from LSA-SAF spectral albedos is available from AMMASAT data base (area= West Africa, spatial resolution 0.05°, daily since september 2005)

- daily experimental NDVI data available from EUMETSAT, since February 2011 (+ pre-operational data from 2010 : 176 days, elaborated with Turkish State Metorological Service)

MSG-SEVIRI daily NDVI products


covered area: West Africa 25° W to 25° E, 20° N to 5° S

resolution 0.05° 1000 samples x 500 linesdaily data since septembre 2005 NDVI calculated from spectral albedos produced

by LSA-SAF (BRDF model Roujean et al., 1992, k0 parameter), model fitted over 5 days)

NetCDF format, integer, NDVI*10000, NDVI≥0, -1 = ocean

MSG – SEVIRI NDVI daily data since sept. 2005 AMMASAT

database

8

MSG – SEVIRI NDVI monthly synthesis january 2010 AMMASAT

database

9

covered area:

whole MSG disk MSG native projection 3712 samples x 3712 lines176 days (February-August 2010)NDVI calculated from top of atmosphere reflectances,

with BRDF correction (daily mean, maximum and minimum NDVI, number of obervations)

HDF5 format, byte, NDVI*100, NDVI≥0

MSG – SEVIRI NDVI daily dataTurkish State Met. Service pre-

operational data (2010)

10

MSG-SEVIRI NDVI 10-daily synthesis

IGARSS 2011 24-29 July Vancouver, Canada

Maximum NDVI 20-30 March 2010

0

0.7

11

MSG-SEVIRI NDVI 10-daily synthesis

IGARSS 2011 24-29 July Vancouver, Canada

Maximum NDVI 20-30 March 2010

0

0.7

21-31 May 2007

12

covered area: whole MSG disk

MSG native projection 3712 samples x 3712 linesdaily data since February 2011 NDVI calculated from top of atmosphere

reflectances, with BRDF correctionHDF5 format, with 899 bytes header

MSG – SEVIRI NDVI daily dataNDVE :Eumetsat experimental data

(2011)

13

MSG – SEVIRI NDVI daily dataEumetsat experimental data (2011)

surface type = IGBP land cover classification

14

GVF is the fraction of green vegetation covering a unit area of horizontal soil, varying from 0 (bare soil) to 1 (full cover)

GVF is independent of leaf and soil optical properties although it is defined with reference to green elements

It is generally close to FAPAR (varying from 0 to 0.95) with the advantage of being defined independently of illumination conditions, making it an intrinsic canopy attribute

for these reasons, GVF is a very good candidate for substitution of classical vegetation indices like NDVI

From NDVI to Green Vegetation Fraction


SPOT-VEGETATION : GVF is named fCover and available since October 2009 from Geoland 2 BIOPAR products database (archived data to be available end of 2011) ; Geoland2 is carried out in the context of GMES, a initiative of the European Commission, which aims to build up a European capacity for Global Monitoring of Environment and Security; other sites for obtaining or disseminating data : GEOSUCCESS, VGT4Africa, EUMETCAST, DevCoCast

MSG-SEVIRI: daily GVF (named FVC Fraction of Vegetation Cover) is an operational product since March 2007 available from LSA SAF archive http://landsaf.meteo.pt/

Availability of Green Vegetation Fraction data


Deriving Green Vegetation Fraction (Fcover) from SPOT-VEGETATION data


scaled NDVI

BioPar Product User Manual (2010)17

Deriving Green Vegetation Fraction (Fcover) from SPOT-VEGETATION data


BioPar Product User Manual (2009)

NDVIsoil = min (NDVImin, 0.14)

NDVImax = 0.85

18

Green Vegetation Fraction (FCover) from SPOT-VEGETATION data


BioPar Product FCover

ADDITIONAL DATA:FCover-ERR (estimated uncertainty) SMB (Status Map or Quality Flag)NMOD (number of clear observations)LMK (Land cover map = GLC 2000 Global Land Cover Map)

Fcover : 30-days composite, updated every 10 days using a sliding window

19

Green Vegetation Fraction (FCover) from SPOT-VEGETATION data



GVF data available since oct. 2009; archived data available end of 2011

10° x 10° tiles

20

Green Vegetation Fraction (FCover) from SPOT-VEGETATION : West Africa, July 1-10, 2010



Example of SPOT-VEGETATION 10-daily GVF

DN=255 : Invalid

DN 0-250 = FVC*250

21

Deriving Green Vegetation Fraction (FVC) from MSG-SEVIRI data


EUMETSAT

The operational daily product FVC is based on the three short-wave channels (VIS 0.6µm, NIR 0.8µm, SWIR 1.6µm) using as input the k0 parameter of a parametric BRDF model (Roujean et al., 1992)

The algorithm relies on an optimised Spectral Mixture Analysis (SMA) technique. In a first step, an exhaustive training set for the soil and vegetation components is defined. Second, a Gaussian Mixture Model is fit to the training data. Third, a Bayesian model selection is used to compute the relative likelihood of membership in each soil/vegetation single-model. FVC is then estimated using a linear-weighted combination single-model estimate

22

Daily Green Vegetation Fraction (FVC) from MSG-SEVIRI data


EUMETSAT

4 sub-imagesHDF5 format

23

10-day SPOT-VEGETATION S10 NDVI MVC, resampled at 0.025° resolution

Maximum of 10 daily NDVI values, MSG-SEVIRI experimental product, resampled at 0.025° resolution (daily mean of n available data during each day)

Results : comparison of NDVI products


Example : 1-10 July 2010, West Africa area

24

Results : SPOT-VEGETATION NDVI S10


Example : 10-day NDVI (1-10 July 2010), West Africa area

DN

Cloud contaminated pixels

25

Results : SPOT-VEGETATION NDVI


Example : 10-day NDVI (1-10 July 2010), West Africa area cloudy pixels (from status map) in white

DN

26

Results : MSG NDVI


Example : 10-day NDVI (1-10 July 2010), West Africa area maximum of 10 daily mean values

NDVI x NDVI x 100

27

Comparison of SPOT-VEG and MSG NDVI


Example : 10-day NDVI (1-10 July 2010), West Africa area

Histograms of NDVI values (cloud-free pixels only)

SPOT-VEG

MSG-SEVIRI

28

Comparison of SPOT-VEG and MSG NDVI


Example : 10-day NDVI (1-10 July 2010), West Africa areacloud-free pixels only

MSG

SPOT-VEG

29

10-daily SPOT-VEGETATION FVC (Geoland 2 BioPar FVC), resampled at 0.025° resolution

Mean of 10 daily MSG-SEVIRI operational product FVC (LSA-SAF) , resampled at 0.025° resolution

Mean of 10 daily MSG-SEVIRI scaled NDVI, using threshold values NDVImin = 0.098 and NDVI max = 0.71 ( values proposed by NOAA-NESDIS, Bob YU pers. comm.)

Results : comparison of FVC products


Example : 1-10 July 2010, West Africa area

30


Results : SPOT-VEG GVF (scaled NDVI)


Example : 10-day FCOVER (1-10 July 2010), West Africa area


grey= no vegetation cover (GVF = 0) ; white = invalid data

GVF (%)

31

Results : MSG-SEVIRI GVF (LSA SAF FVC)


mean of daily FCOVER (1-10 July 2010), West Africa area

derived from FVC product

Grey= no vegetation cover (GVF=0)

GVF (%)

32

Comparison of SPOT-VEG and MSG GVF


Example : 10-day GVF 1-10 July 2010), West Africa areaCloud-free pixels with GVF>0 only

GVFmsg = 0.85 * GVFspot-veg + 10.49

r2 = 0.86

33

Comparison of SPOT-VEG and MSG GVF


Example : all Africa, years 2008,2009

34

Results: MSG-SEVIRI GVF(scaled NDVI)


mean of daily GVF (1-10 July 2010), West Africa areaGrey= no vegetation cover

GVF (%)

NDVImin = 0.098; NDVImax = 0.71(NOAA-NESDIS, 2011)

GVF (%)

35

Comparison of SPOT-VEG and MSG FVC


Example : 10-day GVF 1-10 July 2010), West Africa areaCloud-free pixels with GVF>0 only

MSG

SPOT-VEG

GVFmsg = 0.71 * GVFspot-veg + 17.88 r2 = 0.67

36

Comparison SPOT-VEG vs MSG GVF


Example : desert pixels (GVF = 0)

yellow : GVF = 0 for both data sourcesorange : only GVF SPOT-VEG=0

SPOT-VEG and MSG (FVC from LSA-SAF)

SPOT-VEG and MSG (GVF from scaled NDVI)

37

It is suggested to make use of Green Vegetation Fraction as a substitute to NDVI

Both SPOT-VEGETATION (10-daily) and MSG-SEVIRI (daily) GVF data can be obtained on a near real-time basis: status of products = operational

MSG-SEVIRI has lower spatial resolution than SPOT-VEG, but provides a higher number of available observations in regions with high cloud occurrence, and can be used for high temporal resolution monitoring (1 to 5 days)

Conclusions 1


Different procedures are used to extract FVC from spectral data: all methods suppose that atmospheric and BRDF corrections are efficient

Discrepancies between GVF results from MSG and SPOT-VEG are observed, mainly overestimation of GVF in arid areas when using MSG data

Further research is needed to validate FVC estimations and to ensure intercomparability of results between MSG, SPOT-VEG, MODIS, METOP,… using preferably the same preprocessing/ processing methods to derive GVF

Conclusions 2


Télédétectionpeer-reviewed electronic open-access journalhttp://www.teledetection.net


Thank you for your attention

[email protected]

40

comparison of green vegetation fraction retrievals from spot-vegetation and msg-seviri sensors

Documents

scaled ndvi msgseviri

minimum ndvi

fapar daily ndvi

gvf data msgseviri

canada maximum ndvi

bytes headermsg seviri

comparison of dekadal

scaled ndvi results