Agricultural non-point source pollution (AGNPS) risk prevention and control is more conducive to reducing costs than post-treatment. Therefore, how to effectively identify risks is a key problem to be solved. The study area is located in the NANTUO small watershed of the Three Gorges Reservoir Area in China, on the side of the Yangtze River, where agriculture is developed with many kinds of plantings. There is a water level fluctuation zone (WLFZ) with seasonal elevation difference up to 30 m due to the water level regulation of the Three Gorges reservoir, which is a hot area of AGNPS research. On the basis of daily water level observation in NANTUO small watershed, based on the theory of “source” and “sink”, this study explored the land use change in different seasons with the help of high-resolution remote sensing image (According to the change of water level, it can be divided into two periods: non-submergence period (low water level period from the beginning of June to the end of September) and submergence period (high water level period at other time)). With the help of “source” identification (Types of land use that produce pollutants) and resistance surface calculation index system (Minimum cumulative resistance model, MCR model), the risk level and risk transmission path (The easiest path for pollutants to move to water) of AGNPS in WLFZ were analyzed. Combined with the idea of ecological corridor construction and pollution control, the risk change of regional AGNPS under the optimization of land use mode (The cultivated land types within 50 m and 100 m (Landscape optimization zone) around the water area are adjusted to forest land respectively in two scenarios (Q1 and Q2)) was analyzed as well. Based on our results, 1) the difference in the water areas between the non-submergence and submergence periods was 3.79 km2, and the areas of grassland, farmland, and forest land in the non-submergence period increased by 0.62, 0.85, and 0.35 km2, respectively. The “source” land in this region was mainly sloping farmland with slope above 6°. 2) The calculation results of the minimum cumulative resistance base surface for each source land showed the characteristics of “high values in the west and dispersedly distributed, low values in the east and continuously distributed”. 3) The distribution of the resistance surface during the non-submergence and submergence periods was basically the same; the high-risk zones were widely distributed throughout the region, while the low-risk zones were mainly distributed in the eastern mountains. 4) The number of risk transmission paths followed the order of a4 (farmland with slopes between 6° and 15°) > a5 (farmland with slopes between 2° and 6°) > a1 (rural settlements) > a3 (farmland with slopes between 15° and 25°) > a6 (farmland with slopes ≤ 2°) > a2 (farmland with slopes above 25°), and about 90% of the risk transmission paths were distributed in the lower levels (i.e., level 1, level 2, level 3). The proportion of risk transmission paths in the lower levels during the submergence period was higher than that during the non-submergence period. 5) The proportions of high risk, relatively high risk, medium risk, low risk, and no risk in the study area varied respectively following the rules of 72.79%→70.07%→66.56%, 16.85%→18.16%→18.3%, 6.24%→7.21%→9.3%, 2.25%→2.36%→3.32%, and 1.87%→2.2%→2.52% under the intact status of non-submergence period (Q0), Q1 scenario, and Q2 scenario. 6)The maximum values of pollution transmission path resistance of “source” land under the intact status of non-submergence period (Q0), Q1 scenario, and Q2 scenario followed the order of Q0 < Q1 < Q2. The results showed that the seasonal regulation of water level by artificial water conservancy project has a negative impact on AGNPS, and the MCR model was effective in the study of AGNPS risk and transmission path identification in small area. The results indicated that the wider the landscape optimization belt was, the stronger the barrier effects on pollutants would be and the more difficult it was for the pollutants to reach the water. The increase in the width of the landscape optimization belt could significantly reduce the proportion of low-level transmission paths, which indicated that the construction of ecological corridor had a good positive effect on the water environment protection.