Understanding the dynamics of the grain yield gap (YGAP) and its causative factors is essential for optimizing the layout of grain production and addressing the food crisis, especially in countries with a huge population and less cultivated land, such as China. In the study, a spatial analysis- and machine learning-based framework for YGAP analysis was developed, taking Hunan Province, China, as an application. The results showed that the average YGAP in Hunan Province gradually narrowed from 1990 to 2018, and the YGAPs narrowed in 116 counties. Of which, 26 counties narrowed by more than 4 t ha-1, 58 counties narrowed from 2-4 t ha-1, and 32 counties narrowed within 2 t ha-1. Additionally, we found that the GDP per capita (GDPPC), sunshine hours (SH), per capita annual net income of farmers (PCAI), and rural electricity consumption (REC) play a key role in YGAP change, and the importance of human investment to the YGAP decreased, while socioeconomic environment became the dominant factor that influenced grain production. Comprehensively, the relatively great potential for grain yield growth was generated in sixty-four counties, which are mainly located in the northern, central, and southern Hunan. The findings suggest that it is necessary to consider the trends of economic development in rural areas and population migration in agricultural management. This work provides insights into yield gap dynamics and may contribute to sustainable agricultural management in Hunan Province, China, and other similar regions.
Keywords: determinants; food security; machine learning; spatiotemporal variations; yield gap.