Development and assessment of Leaf Area Index of Russia's vegetation cover based on multi-angular observations by KMSS (Meteor-M) and neural network inversion of PROSAIL model

The paper describes, for the first time, a methodology for Leaf Area Index (LAI) retrieval using a remote sensing device KMSS mounted onboard the Russian satellite Meteor-M with a 60-m spatial resolution. The method is based on the inversion of a PROSAIL radiative transfer model which ingests boundary conditions of the parameters, KMSS surface reflectance data and scene geometrical properties, including observation and illumination conditions. A parameterized and trained fully connected neural network was used as an inversion algorithm. When creating the training sample set, a complete orthogonal plan was used to account for all interactions between input parameters of the model, as well considering their distributions and co-distributions of linked parameters based on a meta-analysis of literature. In this work, the effectiveness of two different geometrical observation schemes was investigated – the classical nadir and the multi-angular, with angles ±8.67°. A reasonably high model accuracy of LAI retrieval was reached: RMSE=1, MAE=0.705 and R2=0.722. Based on the developed method, KMSS-2-based and 60-meters-resolution LAI product was produced and tested over the territory of Russia using 2022-year data. A pixel-wise comparison of KMSS-2 LAI with NASA MODIS LAI product (MCD15A3H) for the snow-free period of the year 2022 also indicate that the proposed product has sufficiently high-level characteristics: RMSE=1.065, MAE=0.669 and R2=0.668. The method for LAI retrieval based on KMSS data developed within this study will increase the efficiency and operability of applications related to detailed environmental monitoring based on remote sensing data from Russian satellite systems.