Diffuse pollution abatement has been a challenge for decision-makers because of the intermittent nature and difficulty of identifying impacts of non-point sources. Depending on the degree of complexity of the system processes and constraints related to time, budget and human resources, variety of tools are used in diffuse pollution management. Decision-makers prefer to use rough estimates that require limited time and budget, in the preliminary assessment of diffuse pollution. The unit pollution load method which is based on the pollution generation rate per unit area and time for a given land use can aid decision-makers in the preliminary assessment of diffuse pollution. In this study, a deterministic distributed watershed model, SWAT is used together with nonlinear optimization models to estimate unit nutrient pollution loads during wet periods for different land use classes for the semi-arid Lake Mogan watershed that is dominated by agricultural activities. Extensive data sets including in-stream water quality and flowrate measurements, meteorological data, land use/land cover (LULC) map developed using remote sensing algorithms, information about agricultural activities, and soil data are used to calibrate and verify the hydraulic and water quality components of SWAT model. Results show that the unit total nitrogen (TN) and total phosphorus (TP) loads (0.46 kg TN/ha/yr and 0.07 kg TP/ha/yr) generated from the watershed during wet periods are very close to the minimum values of the loads specified in the literature and highly depend on the variations in rainfall. Estimated unit nutrient loads both at watershed scale and for different land use classes can be used to assess diffuse pollution control measures for similar regions with semi-arid conditions and heavy agricultural activity. (C) 2016 Elsevier B.V. All rights reserved.