Exploiting longer wavelength SAR data for the improvement of surface modelling

Abstract

Accurate land surface process modelling might be limited due to lack of reliable model input data. Key surface variables as land cover information or soil moisture conditions have been proven to be observable by remote sensing systems. The integration of remote sensing data into land surface process models might therefore help to improve their simulations skills. Longer wavelength microwave data has a higher sensitivity to soil moisture content than higher frequency systems. Recent (ALOS) and planed (e.g. SMOS, TerraSAR-L) systems are therefore expected to provide valuable information about soil moisture dynamics. However the impact of integrating remote sensing data into process models will depend on the parameters that can be observed and their sensitivity to the model predictions on the one hand side and on the uncertainty of these retrievals on the other hand. The latter depends on the sensor configuration and retrieval methodology. Data assimilation techniques have become a favourable tool to integrate uncertain observations into models and use them to constrain the model predictions. The present study investigates the potential to retrieve land cover information and geophysical parameters from L-band SAR data. New techniques for the inversion of land surface parameters from L-band SAR data are developed and used to retrieve surface soil moisture information based on a multitemporal SAR image database collected throughout the ESA AGRISAR campaign. Different supervised and unsupervised land cover mapping techniques are developed. The potentials and limits to integrate remote sensing products into a state-of-the-art land surface model is beeing assessed and the merit of the remote sensing data for hydrological modelling is quantified.