1/13 Development of high level biophysical products from the fusion of medium resolution sensors for regional to global applications: the CYCLOPES project. Reseau Terre Espace F. Baret, M. Leroy, O. Hagolle, P. Bicheron, J. L. Roujean, M. Weiss, C. Bacour, B Geiger, B. Berthelot, R. Lacaze, O. Samain, F. Nino, B. Miras, David Béal & G. Derive
2/ BACKGROUND Large number of medium resolution satellites… but underexploited… Because most users require: High level products (3) and space agencies stop at low level products (1-2) Long series of observations: continuity and consistency of global products Accurate products: Need for cooperation between sensors for improved products validated products associated with confidence intervals
3/13 Objectives of CYCLOPES Demonstrate the capacity of producing operationally consistent global fields of biophysical variables over long and continuous time series Use the products within two applications related to Climate change issues (GMES): –Detection and categorization of land cover change –Introduction within Global carbon cycle models Carbon Cycle and Change in Land Observational Products from an Ensemble of Satellites
4/13 The Products Biophysical variables: Albedo fAPAR fCover LAI Resolution: 1km - 8km Temporal sampling: 10 days Coverage: Global Duration: Sensors: AVHRR VEGETATION POLDER MERIS (MSG)
5/13 Principles Temporal compositing Fusion between sensors Incremental production First version deriving from available algorithmic modules. Up-dates every 6-9 months: inclusion of improvements in each modules and in multi-sensor fusion Validation and accuracy assessment R TOA Cloud Atm-corr R TOC Bio-algo VAR No possibility for Fusion Possibility to fuse For single date Spectral & directional Normalization? Possible ingestion of several configurations Possible fusion (filtering) V1-V2 V3-V4
6/13 Version 1 (May 2004) Using already existing algorithms to meet the deadline (To+12)!! Version 1 will be available by the end of April 2004: global 8km for Africa-Europe for 1km (VGT) VI Roujean/Lacaze Aerosol Climatology Original threshold
7/13 H0V1 H0V0H1V0 H1V1H2V1H3V1H4V1H5V1 H2V0H3V0H4V0H5V0 The First version … + 1 km Europe & Africa 2002 & year global requires about 1 month processing !! 8 km Global 2002 & 2003
8/13 Consistent calibration between sensors (CNES) Consistent cloud screening (CNES-Noveltis) (Evaluation then adjusting threshold) Consistent Atmospheric correction (CNES-Noveltis) (Monthly MODIS min average values) Preprocessing
9/13 Directional Normalization & albedo estimation VEG_NIR VEG_VIS Validation: Broad band Bi-hemispherical albedo Using Roujean & Li-Ross kernel models adjusted over a weighed gaussian window (30days) Outliers removal procedure Linear model for spectral integration Good consistency between VGT-1 and VGT 2 and ground measurements VEG
10/13 Biophysical algorithm first version DVI=k0nir - k0red ; fCove r= c1 DVI+c2 LAI=2.11log(1-fCover) Available simple algorithm to be implemented in version 1: (Roujean-Lacaze 2002) Multisensor fusion first version
11/13 Temporal consistency evaluation O 2003 O 2002 X ECOCLIMAP Very good temporal consistency although gaps (clouds) are observed
12/13 Validation over ground measurements
13/13 CONCLUSION Huge effort in providing consistency between satellite processing Feed-back loop with users and validation activity CYCLOPES will constitute the initial products within GEOLAND: implementation of the GMES vision… Need continuity of algorithmic research efforts and validation… beyond CYCLOPES (after 2006) Need data beyond 2008!!!!! – Improved spatial resolution –Autonomous atmospheric correction …