Performance assessment of the Sentinel-2 LAI products and data fusion techniques for developing new LAI datasets over the high-altitude Himalayan forests

The present study evaluates the accuracy of SNAP-Sentinel-2 Prototype Processor (SL2P) derived Leaf Area Index (LAI) and proposes a new simple method to generate new datasets of LAI through data fusion. Rigorous optimization of the data fusion approaches (Kalman filter and Linear weighted) were performed for the generation of new LAI products over the complex hilly terrain of the Himalayan region. The results showed a good correlation (r = 0.79) and low error (RMSE = 1.63) between SNAP-derived (at 20 m) and ground-observed LAI. A lower correlation was obtained between the ground observed LAI data and the corresponding global LAI products for the Moderate Resolution Imaging Spectroradiometer (MODIS) (r = 0.1, RMSE = 1.19), Copernicus Global Land Service (CGLS) (r = 0.1, RMSE = 0.61) and the Visible Infrared Imaging Radiometer Suite (VIIRS) (r = 0.04, RMSE = 1.25). Notably, after implementing the data fusion, both SNAP-derived LAI and Global LAI products exhibited much-improved performance statistics with ground observed data sets.

Automated summary

This study evaluates the accuracy of SNAP-Sentinel-2 Prototype Processor (SL2P) derived Leaf Area Index (LAI) and proposes a new method for generating new LAI datasets through data fusion. The results show a good correlation and low error between SNAP-derived LAI and ground-observed LAI. After implementing data fusion, both SNAP-derived LAI and Global LAI products exhibit improved performance statistics with ground observed data sets.