Satellite Based Gridded Climate Data Driven Wheat Yield Datasets at MODIS Pixel Scale

Adverse climatic events like heat stress, floods, unseasonal rainfall, and droughts frequently hinder crop productivity. Long-term crop yield data plays a crucial role in food security planning. This study presents historical wheat yield data at the satellite pixel level from 2001 to 2019 in Uttar Pradesh, India. We use various satellite indicators to develop wheat yield models, including the normalized difference vegetation index and gridded weather data, such as precipitation, temperature, and evapotranspiration. Additionally, we incorporated district-wise wheat yield data from the Directorate of Economics and Statistics (DES) and combined it with satellite variables to develop wheat yield models using multiple regression analysis. The key findings of the present study confer that the multiple regression-based yield models (developed using DES-based yield statistics and satellite variables) efficiently predict the wheat yield at the satellite pixel level. The developed models exhibited strong performance, with R² values ranging from 0.3 to 0.76 when comparing the satellite-derived and DES-based district-wise mean yield. The mean absolute error ranged from 0.22 to 1.7 t/ha over the study years. This methodology can be applied globally, from local to national scales, to quantitatively map and depict crop yields at the satellite pixel level. Such historical crop yield mapping offers valuable insights into long-term yield patterns in response to climate change, thereby enabling the formulation and implementation of food security measures.Keywords: Crop yield, Satellite data, MODIS, NDVI, Multiple regression