Abstract: Authors focused on 30 apple producing counties in Shaanxi province, utilizing historical meteorological data and apple yield data (2000−2022) to predict apple climate yields under future climate scenarios (2023−2050, SSP2−RCP4.5 and SSP5−RCP8.5 scenario of the CanESM5 model) using machine learning models. The HP filter method was used to isolate the climate yield of apples. After variable screening through Spearman correlation analysis and grey relational analysis, four machine learning models were trained (i.e., support vector machine model, linear regression model, BP neural network model, and random forest mode). Following model evaluation, the best−performing model was selected for prediction. The results showed that: (1) the annual variation in apple climate yields from 2023 to 2050 was significant, with minor spatial distribution differences. (2) Under the future SSP5−RCP8.5 scenario, apple climate yields were higher, with more significant regional disparities and more drastic trends. Given the impact of future climate change on apple climate yields in different regions of Shaanxi, strategic planning for apple cultivation should be systematically organized, and a scientific and reasonable agricultural policy system should be established to ensure the sustainable, stable, and efficient development of Shaanxi's apple industry in the context of climate change.

Key words: Climate change, CanESM5, Climate yield, Random forest model, Shaanxi