Exploiting SMOS signatures over forests: Tests and proposed improvements of the forward model

R. RAHMOUNE; P. FERRAZZOLI; R. MAGAGI; F. GRINGS; V. BARRAZA

Roma

Simposio; Microwave Radiometry and Remote Sensing of the Environment; 2012

Tor Vergata University of Rome

Modeling forest emission at L band is an important issue for the success of spaceborne radiometric missions, such as Soil Moisture and Ocean Salinity (SMOS), which was launched about two years ago, and the Soil Moisture Active Passive (SMAP) mission, which is under development. In order to correctly retrieve soil moisture, a reliable forward model is necessary. For areas covered by forests, specific problems must be solved. The wave penetration within the canopy is limited, which reduces the sensitivity to variations of soil moisture. Moreover, single and multiple scattering processes are appreciable, making the model representation complex.   The retrieval algorithm described in the ATBD (Algorithm Theoretical Basis Document) of SMOS is based on a forward model and a retrieval process. For all vegetated areas, the forward model adopts a simple radiative transfer approach. For the specific case of forests, the albedo and the optical depth are computed by fitting the outputs generated by the theoretical model, and refined by empirical fitting. The procedure assumes to know the forest kind (broadleaf, or coniferous) and the maximum yearly value of Leaf Area Index, made available by ECOCLIMAP.   In this work, the SMOS forward model is tested over the first two years of available signatures. In particular two regions, characterized by different environments, are selected. One region spans the North-East of Argentina, and is covered by the continuous and deciduous Chaco forest, which is xerophytic, that is adapted to leave in an arid environment. The biomass is moderate, typically ranging between 70?110 t/ha. A detailed biomass map and rainfall information are available for this area. The other region is Berms, located in Canada. Here the forest is mainly coniferous, with some areas characterized by re-growing trees, after recent cuts. Over this forest, the CanEX-SM10 experiment took place and high resolution airborne measurements are also available..   In general, the present Level 2 algorithm underestimates the soil moisture, with respect to estimates of the European Centre for Medium-Range Weather Forecasts (ECMWF) and in situ data. For Chaco forest, the brightness temperatures simulated by the forward model show a temporal trend which follows the rainfall events and the variations of ECMWF soil moisture. However, a slight but systematic underestimation is observed. The possible causes of this discrepancy are evaluated and discussed. The ECOCLIMAP map is checked against local biomass maps developed by Unidad de Manejo del Sistema de Evaluación Forestal (UMSEF). Moreover, the influence of herbaceous vegetation and the specific properties of the wood, which here is dry and hard, are considered. For BERMS forest, the optical depth computed using ECOCLIMAP overestimates the real one. The model is then tested using in situ ground truth and airborne measurements.   Finally, suggestions to improve the forward model in order to consider the properties of specific regions are formulated.