RESUMO
Agriculture-related nonpoint source pollution has been a worldwide issue in the past few decades. Estimating pollutant sources at the basin scale remains a challenge due to the complexity of pollutant transport mechanisms affected by land use, variable climatic condition, geological formation, and lack of long-term observation data. This study was based on the long-term plot observational data of surface runoff and water quality and used principal component analysis and other statistical analyses to analyze the variation of water quality affected by different land uses (cropland, tree forest, shrub, grassland, and wildland). The mean concentration method with the local calibrated/verified SWAT (soil and water assessment tool) model was used to quantify the load of nonpoint source (NPS) pollutants on slope areas under different land uses in the Anjiagou Watershed. Our research results determined permanganate index (CODMn), ammonia nitrogen (NH3-N), total nitrogen (TN), fluoride (F-), nitrite nitrogen (NO2-N), total phosphorus (TP), and hexavalent chromium (Cr6+) as the significant pollutants while 5-day biochemical oxygen demand (BOD5) was identified to be below the water quality standards of Grade V (water for agricultural and general landscape use) in the studied watershed. Pollutants were discharged through either hillslope at a total rate of 2.4 kg ha-1, accounting for 67.6% of the total, or through waterway channels (32.4%). The pollutant concentrations were from 23.5 mg L-1 to 37.4 mg L-1, varying with pollutants and land uses. All examined water quality indicators exceeded the minimum safety standards defined by the regulations of the Gansu provincial government by averaging 3.5 times higher than the safety threshold. The pollutants from hillslopes exceeded water quality standards by a factor of 3.4-4.4 times compared with from the waterway channel by 1.9. Implementing soil and water conservation measures can mitigate pollutants to some extent, particularly during the process of highland runoff converging into waterways. At the watershed level, between 33 and 38% of the runoff and pollutants were discharged from croplands, between 59 and 66% from forest land, < 2% from grassland, and 1% from wildland. This study also demonstrates a simple but novel method to estimate NPS pollutants using long-term plot observations in conjunction with SWAT models, which can be used in other watersheds with similar conditions.