Period: 1999 – present
Sensor: SPOT-VGT and PROBA-V
Frequency: 10 days
Data can be accessed here.
Algorithm description: This product has been developed to build up on the already existing and widely validated MODIS and CYCLOPES products while removing limitations observed in these products. MODIS and CYCLOPES were chosen as they were more spatially and temporally consistent.
The general principle is: (1) Generation of the training dataset from CYCLOPES and MODIS biophysical variables (2) neural network calibration using the training dataset and the input top of canopy reflectance values (3) application of the trained network to estimate biophysical variables.
In case of PROBA-V, additional steps are involved that apply spectral response corrections to convert PROBA-V reflectance to VGT as the training dataset (developed for version 1 of these products) were generated using VGT sensor only.
The training dataset for biophysical variables (LAI and FAPAR) is made by fusing these products to produce ‘the best estimates’. The algorithm gives more weightage to the MODIS values for the high range LAI values and more to the CYCLOPES at lower values. This was proposed in response to widely published research findings confirming overestimation of LAI by MODIS at lower end and poor performance of CYCLOPES in dense canopies.
Accuracy: Direct validation against BELMANIP2 global sites yielded in RMSE = 0.744 and R2 = 0.807. Results were consistent or even slightly better than MODIS and CYCLOPES based validation.
- Products derived from PROBA-V have shown less accurate results than those from SPOT-VGT, which is mainly due to the fact that the training datasets used were generated for VGT sensor.
- The GEOV1 product is not gap-filled.
- Large gaps in products over northern latitudes in wintertime and equatorial areas. Larger fraction of missing values on PROBA-V products than on SPOT-VGT products due to stricter cloud masking algorithm.
Baret, F., M. Weiss, R. Lacaze, F. Camacho, H. Makhmara, P. Pacholcyzk, and B. Smets. “GEOV1: LAI and FAPAR Essential Climate Variables and FCOVER Global Time Series Capitalizing over Existing Products. Part1: Principles of Development and Production.” Remote Sensing of Environment 137 (October 2013): 299–309.
Verger, A., Baret, F., & Weiss, M. (2011). A multisensor fusion approach to improve LAI time series. Remote Sensing of Enviroment, 115, 2460–2470
Camacho, F. ; Cernicharo, J. ; Lacaze, R. ; Baret, F. ; Weiss, M. GEOV1: LAI, FAPAR Essential Climate Variables and FCover global time series capitalizing over existing products. Part 2: Validation and inter-comparison with reference products. Remote Sensing of Environment 2013, 137, 310-329.