Assimilating remote sensing data into a crop model to improve predictive performance for spatial applications

The use of crop models on large areas for diagnosing crop growing conditions or predicting crop production is hampered by the lack of sufficient spatial information about model inputs. We propose here a way of spatializing the model that assimilates information obtained from remote sensing images made during the growing season. The method was applied to yield estimates from the SUCROS sugar beet model, run for about 50 fields within two sugar factory areas. The assimilation of four to six SPOT and aerial photography data values into the SUCROS model, coupled with the scattering by arbitrarily inclined leaves (SAIL) reflectance model, re-estimated crop establishment and root system settling parameters, to which the model was particularly sensitive. The field-by-field yield estimates were improved (i.e., with remote sensing data assimilation, the relative root mean square error decreased from 20% to about 10%). The key factors of the method are the number and timing of images that determine the number and the type of parameters that can be estimated. The main limitation of this method is the lack of robustness of the crop model in simulating LAI in serious drought conditions.