COMPARISON BETWEEN THE USE OF SAR AND OPTICAL DATA FOR WHEAT YIELD ESTIMATIONS USING CROP MODEL ASSIMILATION

摘要

The ultimate aim of this work is to develop methods for the assimilation of the biophysical variables estimated by remote sensing in a suitable crop growth model. Two strategies were followed, one based on the use of Leaf ArearnIndex (LAI) estimated by optical data, and the other based on the use of biomass estimated by SAR. The first one estimates LAI from the reflectance measured by the optical sensors on board of HJ1A, HJ1B and Landsat, using arnmethod based on the training of artificial neural networks (ANN) with PROSAIL model simulations. The retrieved LAI is used to improve wheat yield estimation, using assimilation methods based on the Ensemble Kalman Filter, whichrnassimilate the biophysical variables into growth crop model. The second strategy estimates biomass from SAR imagery. Polarimetric decomposition methods were used based on multi-temporal fully polarimetric Radarsat-2 datarnduring the entire growing season. The estimated biomass was assimilating to FAO Aqua crop model for improving the winter wheat yield estimation, with the Particle Swarm Optimization (PSO) method. These procedures were usedrnin a spatial application with data collected in the rural area of Yangling (Shaanxi Province) in 2014 and were validated for a number of wheat fields for which ground yield data had been recorded and according to statistical yield datarnfor the area.