Abstract: Abstract: The study on spatial heterogeneity of the early-warning status of cultivated land system security (CLSS) and its driving factors at county level has an important role in accurately identifying the early-warning status of the CLSS and scientifically formulating the protection plan of cultivated land system. The purpose of this study was to explore and identify spatial heterogeneity characteristics of the early-warning status of the CLSS and its driving factors in Heilongjiang Province, which is located in main grain producing area of Northeast China. In the paper, exploratory spatial data analysis (ESDA) was used to identify spatial heterogeneity characteristics of the early-warning status of CLSS in 72 counties of Heilongjiang Province, and ordinary least squares (OLS) and geographical weighted regression (GWR) method were employed for exploring driving factors of the spatial heterogeneity of the early-warning status. The results indicated that: 1) The early-warning level of CLSS in general was relatively high in the counties of Heilongjiang Province. Of the 72 counties, 69 counties were in the early-warning status, and 28, 32 and 9 counties were low-level warning, medium-level warning and severe-level warning, respectively. 2) The spatial distribution of the early-warning status showed obvious heterogeneity and agglomeration characteristics of the early-warning status in the study area. The severe-level warning areas were mainly found in the southeast and west of the region, while the secure and relatively secure areas were mainly concentrated in the north-south axis belt areas. The counties with positive spatial autocorrelation (high-high type and low-low type) emerged with the "clusters" and had a powerful agglomeration. The counties with negative spatial autocorrelation (high-low type and low-high type) were few and showed sporadic distribution. 3) Twelve variables were preliminarily selected for exploring the driving factors of the early-warning status by OLS model, and 9 variables, including 3 natural-ecological factors (slope, elevation and annual average temperature) and 6 socioeconomic factors (per capita GDP (gross domestic product), urbanization level, investment in agricultural fixed assets, water-soil coordination degree, agricultural wastes index, and road network density) were determined for further analysis by the GWR model. The regression results by GWR model and ArcGIS spatial analysis showed the natural-ecological factors, including terrain and climate, and socioeconomic factors including the land input-output, land-water resources allocation, had a significant effect on the early-warning status of the CLSS. Moreover, the regression coefficients of these driving factors showed the significant spatial heterogeneity characteristics. In conclusion, the study on spatial heterogeneity of the early-warning status of the CLSS provides a reference for accurately identifying the spatial distribution of the early-warning status in Heilongjiang Province. Compared with OLS model, the GWR model visualizes the non-stationary characteristics of driving factors of the spatial heterogeneity of the early-warning status of the CLSS, which can be more intuitive and in-depth to explore the spatial difference of the effect of each driving factor on the early-warning status, and provide a scientific basis for formulating appropriate measures to mitigate future threats of the cultivated land system.