Predicting fodder crop yield in semi-arid Mongolia using Landsat data: a multivariate analysis approach

Crop yield prediction enables appropriate crop management in response to climate change and is essential for assessing food security and economic efficiency at all levels, from individual farmers to governments. Mongolian cropland yields face several challenges that remain underexplored through spatial analysis, including insufficient data, planning difficulties, harsh climatic conditions, and limited access to advanced technologies. This study aimed to predict crop yields at the local level using Landsat time-series data (2014–2023), along with climate and soil data. Key variables (NDVI, soil parameters, and hydrothermal coefficient) were selected based on prior research on crop yield prediction. Soil parameters (humus, humus depth, soil reaction pH, nitrogen, phosphorus, and potassium) and climate factors (precipitation, temperature) were analyzed to predict crop yield using regression analysis. Two models were developed: a multiple linear regression (MLR) model incorporating the normalized difference vegetation index (NDVI) and another without NDVI. The results showed that the coefficient of determination (R2) from the multivariate regression analysis was 0.91 for the model including NDVI and 0.74 for the model excluding NDVI. These findings can enhance early yield prediction, optimize resource allocation, and support timely decision-making in agricultural management.