Spatiotemporal Variations of Gross Primary Productivity for Cropland Using Machine Learning Approach

An enhanced understanding of farmland gross primary productivity (GPP) is critical for coping with rising food demand. The spatiotemporal distribution and patterns of GPP, actual evapotranspiration (ET), net primary productivity (NPP), precipitation (mm), soil moisture, land surface temperature (°C), normalized difference vegetation index, digital elevation model (DEM), leaf area index (LAI), and agricultural grain production for farmlands are measured at the field scale, particularly considering the variances between crop circumstances. In this research, the Moderate Resolution Imaging Spectroradiometer (MODIS) data and the newly developed machine learning analysis method were applied to the analysis and quantity, spatiotemporal variation, and yearly trends in GPP for Samara Oblast croplands from 2000 to 2022. The Random Forest (RF) method has been created for the Russian region. Cropland's annual GPP increased significantly between 2000–2010 but subsequently decreased in 2010–2011. The findings demonstrate that the variations in crop grain production statistics were incompatible with the annual trend. The RF model has been applied to demonstrate cropland GPP prediction in the Samara region for several crop types. A significant value of R2 = 0.82 was measured when compared with the ground truth data. The favorable human-activity influence on GPP is particularly apparent through crops, which dominate in the southwestern farming region and exhibit significant seasonal differences. This research establishes an analytical basis for the sustainable development and ecological preservation of the Samara Oblast in Russia.