Retrieving crops Green Area Index from high temporal and spatial resolution remote sensing data

This paper aims at firstly evaluating the correspondence between Normalized Difference Vegetation Index (NDVI) products from Formosat-2 (F2) and SPOT sensors and then to perform a comparative analysis of two methods for retrieving Green Area Index from high spatial and temporal resolution satellite data (F2 and SPOT). For this purpose, an empirical approach using NDVI plus field data and a Neural Network approach using the PROSAIL model are compared over four different crops: wheat, sunflower, maize and soybean. The performance of both methods were evaluated and compared with in-situ direct (destructive) and indirect (hemispherical photos) measurements. Results suggest better performances for the empirical approach (R2, RMSE). Still the physically-based method leads to good results (R2, RMSE). The latter seems to be more promising due to its portability and independence from field measurements.