Abstract: The non-point phosphorus (P) loss from farmland is the main cause of eutrophication, and identifying the Critical Source Areas (CSAs) and Critical Influence Factors (CIFs) of the P loss from farmland is a pivotal part of controlling agricultural non-point source pollution. The objective of this study was to assess the environmental risk of P loss from farmland based on P index model in the Haihe River Basin. In this study, we took the Haihe River Basin as the research region, and applied the P index model. Thereinto, the soil available P content and fertilizer-P application rate were selected as the source factors. Synchronously, the soil erosion modulus, annual runoff, the Normalized Differential Distance Index (NDDI) between farmland and river network were chosen as the transport factors. Additionally, this research aims at identifying the CSAs combined with GIS technology, distinguishing the CIFs and exploring the relation between P index and influence factors such as soil properties (mainly including soil texture and organic carbon), source and transport factors via Structural Equation Modeling (SEM). The spatial distributions of soil available P content, fertilizer-P application rate, soil erosion modulus, annual runoff and NDDI in the Haihe River Basin indicated that the most areas were at the risk levels of low-medium, medium-high, very low, medium-high and high in the Haihe River Basin, respectively. Namely, these areas accounted for 66.5%, 61.1%, 99.0%, 54.2%, and 64.8% of the total farmland area in the Haihe River Basin, respectively. Moreover, we revealed that the CIF affecting P loss from farmland in the Haihe River Basin is the transport factors in which the key transport factors are the annual runoff and NDDI. Furthermore, SEM prescribed that there was an extremely significant negative correlation between the source factors (P<0.01) and extremely significant positive correlation with transport factors (P<0.01), and soil properties had highly significant negative correlations with source factors and transport factors (P<0.01). Subsequently, the findings illustrated that the CSAs of P loss from farmland in the Haihe River Basin were mainly located in the plain farming areas of Shandong Province and the southeast of Hebei Province on the north bank of the Yellow River, and the mountain areas in the northwestern of the Haihe River Basin. Due to the plain farming areas have high cultivation intensity and fertilizer-P application rate, flat terrain, and close distance to river network, P of soil and fertilizer could easily migrate into the rivers following the surface runoff with the medium-high intensity. Thus, for the plain farming areas, some practices should be taken to decrease the risk level of P loss from farmland, such as appropriate reduction of the fertilizer-P application rate, comprehensive nutrient management, conservation tillage and deployment of buffer zone between farmland and river network and so on. Because mountain areas have severe soil erosion, therefore, the risk level of P loss from farmland is relatively high. For this, proper soil and water conservation measures should be taken to reduce the soil erosion and risk level of P loss from farmland, such as contour tillage and terrace cropping and so on. In conclusion, the results are beneficial to the method innovation of P loss from farmland at the watershed scale, and can provide a scientific reference for controlling agricultural non-point source pollution in the Haihe River Basin.