Accurate identification of maize plantation distribution and timely examination of key spatial-temporal drivers is a practice that can support agricultural production estimates and development decisions. Previous studies have rarely used efficient cloud processing methods to extract crop distribution, and meteorological and socioeconomic factors were often considered independently in driving force analysis. In this paper, we extract the spatial distribution of maize using classification and regression tree (CART) and random forest (RF) algorithms based on the Google Earth Engine (GEE) platform. Combining remote sensing, meteorological and statistical data, the spatio-temporal variation characteristics of maize plantation proportion (MPP) at the county scale were analyzed using trend analysis, kernel density estimation, and standard deviation ellipse analysis, and the driving forces of MPP spatio-temporal variation were explored using partial correlation analysis and geodetectors. Our empirical results in Heilongjiang province, China showed that (1) the CART algorithm achieved higher classification accuracy than the RF algorithm; (2) MPP showed an upward trend in more than 75% of counties, especially in high-latitude regions; (3) the main climatic factor affecting the inter-annual fluctuation of MPP was relative humidity; (4) the impact of socioeconomic factors on MPP spatial distribution was significantly larger than meteorological factors, the temperature was the most important meteorological factor, and the number of rural households was the most important socioeconomic factor affecting MPP spatial distribution. The interaction between different factors was greater than a single factor alone; (5) the correlation between meteorological factors and MPP differed across different latitudinal regions and landforms. This research provides a key reference for the optimal adjustment of crop cultivation distribution and agricultural development planning and policy.