Period: 2014 – present
Frequency: 10 days
Data can be accessed here.
Algorithm description: The algorithm is similar to GEOV2 product (neural network trained on CYCLOPES data and temporal compositing using historical time series data). The algorithm is implemented on PROBA-V data at 300m resolution. As the neural nets have been trained on CYCLOPES FCOVER and SPOT-VGT reflectance data, PROBA-V reflectance is converted into SPOT-VGT reflectance.
Accuracy: Global and detailed validation of this product hasn’t been done. However Camacho et al (2016) report some ground based validation results wherein the product was compared against canopy cover measurements for crops. FCOVER product shows systematic overestimation of about 0.12 units.
- Accuracy is slightly less than previous versions when tested over Europe, however global validation has not been done.
- More restrictive cloud screening algorithm introduce more gaps even after temporal compositing steps.
Baret, M. Weiss, A. Verger, and B. Smets, “ATBD for LAI, FAPAR and FCOVE from PROBA-V products at 300m resolution (GEOV3),” ImagineS, Issue 1.73, May 2016.
F. Camacho et al., “Validating GEOV3 LAI, FAPAR and vegetation cover estimates derived from PROBA-V observations at 333m over Europe,” 2016, p. 1.