Agricultural non-point source pollution (NPSP) has become a global concern requiring urgent attention, and high fertilizer application to cash crops often leads to more serious NPSP. We quantified the control effect of best management practices (BMPs) on total nitrogen (TN) and total phosphorus (TP) pollution in the Dongtiaoxi Watershed, with an emphasis on identifying critical source areas (CSAs), using the Soil Water Assessment Tool (SWAT) model. Although CSAs accounted for 11.78% of the whole watershed, they contributed to 23.43 and 21.72% of the TN and TP loads, respectively. In agricultural systems, planting cash crops is a key factor in the emergence of CSAs and NPSP; thus, it is necessary to improve the management of cash crops. Further, combined BMPs outperformed individual measures, achieving a 25.51% reduction in TN at the watershed scale. At the sub-watershed level (e.g., sub-watershed 27), combined BMPs reduced TN and TP by 50.11 and 20.00%, respectively. There are a large number of cash crops such as commercial bamboo forests in the study area that have been fertilized but were previously regarded as forest land. This can also lead to NPSP but may have been ignored in the past. Overall, our study classifies cash crops as agricultural land and improves the reliability of the results. The findings not only help to improve the water quality from the Taihu Basin but also provide technical support and a theoretical basis for government departments to improve the water environment.

Coastal saline–alkaline land has considerable potential for soil carbon sequestration, but how different land-use types affect microbial resource limitation and carbon-use efficiency (CUE) in coastal saline–alkaline soils remains unclear. Four representative land-use types, namely bare land (BL), wetland (WL), grassland (GL), and forest land (FL), were investigated in the coastal saline–alkaline soils of the Yellow River Delta. Soil physicochemical properties, microbial biomass, and extracellular enzyme activities were measured, and ecoenzymatic stoichiometry, microbial resource limitation, and CUE were subsequently calculated. Compared with BL, vegetated land-use types decreased electrical conductivity by 52.1–95.8%, while soil water content, soil nutrient indicators, and microbial biomass indicators increased by 47.1–77.5%, 2.6–136.8%, and 2.2–274.4%, respectively. WL was mainly phosphorus-limited, whereas BL, GL, and FL were primarily nitrogen-limited. Despite relatively high soil organic carbon and nutrient availability, GL showed the strongest N limitation and was the only land-use type showing C limitation. Model-estimated CUE ranged from 0.544 to 0.579 and followed the order FL > BL > WL > GL. Random forest analysis showed that soil physicochemical properties contributed most to CUE variation (42.9%). Structural equation modeling further indicated that soil physicochemical properties were indirectly associated with CUE, mainly through stoichiometric characteristics and microbial resource limitation, showing positive and negative associations, respectively. These findings provide microbial evidence for optimizing land-use patterns, vegetation restoration, and carbon-oriented ecological restoration in coastal saline–alkaline land.