In this paper

a practical scheme for assimilation of multi-temporal and multi-polarization ENVISAT ASAR data in rice crop model to map rice yield is presented. To achieve this

rice distribution information is obtained first by rice mapping method to retrieve rice fields from ASAR images

and then an assimilation method is applied to use the temporal single-polarized rice backscattering coefficients which are grouped for each rice pixel to re-initialize ORYZA2000. The assimilation method consists of re-initializing the model with optimal input parameters allows a better temporal agreement between the rice backscattering coefficients retrieved from ASAR data and the ones simulated by a coupled model

i.e.

the combination of ORYZA2000 and a semi-empirical rice backscatter model through LAI. The SCE-UA optimization algorithm is employed to determine the optimal set of input parameters. After the re-initialization

rice yield for each rice pixel is calculated

and the yield map over the area of interest is finally produced. The scheme is applied over Xinghua study area located in the middle of Jiangsu Province of China during the 2006 rice season. The result shows that the obtained rice yield map generally overestimates the actual rice production by 13%

with a relative error of 11.2% at validation sites

but the tendency of rice growth status and spatial variation of the rice yield are well predicted and highly consistent with the actual production variation.