Abstract: Reducing pollution caused by chemical fertilizers while continuously increasing the output value of agriculture at the same time is an inevitable requirement to improve high quality development of agriculture. Actions aimed at achieving zero growth in chemical fertilizer use have been formulated and implemented in provinces across China. However, inconsistencies in progress and effectiveness among provinces may affect each other. The fertilizer loss coefficient method was used to calculate the emission intensity of non-point source pollution in 31 provinces across the Chinese mainland from 1997 to 2018. Spatial autocorrelation and hotspot analysis methods, based on the calculated emission intensity, were used to reveal the emission intensity temporal and spatial characteristics, and to analyze the interaction effect on neighboring provinces. According to these characteristics, the country was grouped into three regions: a hot spot region, a cold spot region, and a non-hot (cold) spot region. Then, the Environmental Kuznets Curve (EKC) trend between the non-point source pollution emission intensity of fertilizers and the per capita agricultural output value in each group was simulated and predicted under spatial heterogeneity and spatial correlation conditions. The results showed that the emission intensity varied considerably among provinces across the country. The emission intensity was spatially positively autocorrelated across the country with a cluster mode. A hot spot analysis showed that the spatiotemporal pattern for non-point source pollution caused by fertilizers was relatively stable. The hot spot region was mainly concentrated in the central and southern parts of China, especially in the middle and lower reaches of the Yangtze River. However, in recent years, the number of hotspots in the Huanghuaihai region had significantly decreased. The cold spot region was mainly concentrated in the western region and Heilongjiang Province. The EKC trend simulation, based on temporal and spatial pattern feature grouping, showed that there was a significant nonlinear EKC relationship between agricultural economic growth and chemical fertilizer non-point source pollution in each group. However, the trend and the inflection point of each curve were clearly different. The hot spot region appeared as an "inverted U-type" curve, but both the cold spot region and the non-hot (cold) spot region had "inverted N-type" curves. Most provinces in each region were in the ascending phase of the curves. Industrial structure adjustment and industrial transfer promoted the spatial spillover effect of fertilizer derived non-point source pollution among the regions. Therefore, coordinated governance among regions needs to be introduced. Based on the above results, we propose that there should be corresponding policy implications, and that suitable fertilization equipment should be developed and promoted to increase fertilizer utilization in the hot spot region. Conservation farming should be applied and the use of organic fertilizer promoted in the cold spot region. The non-hot (cold) spot region should reasonably adjust the structure of its agricultural industry and focus on the crop-livestock and poultry farming cycle. Ecological compensation and emissions trading should be implemented to achieve collaborative governance.