To better manage total phosphorus (TP) pollution from nonpoint sources in the Fuzhou River Basin, this study constructed a Soil and Water Assessment Tool (SWAT) model for the basin. Based on the calibrated SWAT model, we identified the critical source areas (CSAs) of TP pollution and evaluated the reduction effects of individual Best Management Practices (BMPs) and their combinations. Additionally, we optimized the selection of BMPs by balancing their environmental cost-effectiveness. The findings reveal that the SWAT model exhibits robust applicability in the Fuzhou River Basin, with the CSAs accounting for 37.61% of the total basin area and contributing 60.65% of the total TP load. These CSAs are primarily concentrated in agricultural and orchard zones, where high rates of nutrient loss per unit area are prevalent. Combined BMPs outperform individual BMPs in mitigating nonpoint source pollution loads, and there is a positive correlation between the number of combined BMPs and the average reduction rate. However, given a fixed implementation area within the basin, an optimal threshold for the reduction efficiency exists beyond which incremental BMP combinations yield diminishing returns. Among the individual BMPs, FS (filter strips) stands out as the most cost-effective measure for reducing TP loads in the Fuzhou River Basin. In contrast, the combination of FS+GW+FR10 (filter strips, grassed waterways, and 10% phosphate fertilizer reduction) proves to be the most cost-effective among the evaluated BMP combinations.