A multiscale and multisensor approach of LAI retrieval in a maritime pine ecosystem

Spatial datasets of biophysical parameters at multiple scales can be important for the modeling of landsurface processes. In this study, we compared how decametric resolution and kilometric resolution LAI retrievals vary over a maritime pine dominated ecosystem in Southern France. Firstly, we used atmospherically corrected Landsat ETM+ and SPOT4 HRVIR reflectances along with ground-based LAI measurements to derive empirical relationships between vegetation indices and measured LAI. These algorithms were later inverted to map LAI over the landscape. RSR-based algorithms showed the best performance for both ETM+ (r²= 0.788) and HRVIR (r²= 0.780) sensors and were more stable than SR and NDVI. Further, after upscaling to 1km, comparison with global LAI products revealed that the CYCLOPES product was more robust in capturing the fine scale signatures. These results show that modeling of landsurface processes could be improved by adopting a multiscale and multisensor approach.