Effects of dynamic land use inputs on improvement of SWAT model performance and uncertainty analysis of outputs

The objective of this study was to evaluate the impacts of static and dynamic land use input conditions on the performance of non-point source (NPS) model and find out whether dynamic land use input can improve the model accuracy. Soil and Water Assessment Tool (SWAT) model was selected as the evaluation model and seven different land use input conditions were set by setting the land use update file in SWAT. The results showed that the land use pattern in the study area changed from 2000 to 2015 due to climate change and human activities, leading to inconsistencies between different land use patterns. The calibrated results indicated that dynamic land use input conditions could apparently improve the simulation accuracy of total nitrogen (TN) and total phosphorus (TP). CE5Y condition had the best calibrated result with R2 and NSE larger than 0.7 and 0.6, respectively. However, for flow simulation, the land use input conditions had no apparent effect on the model calibration and validation results. The deviation analysis of the model outputs indicated that monthly outputs were more affected by the land use input conditions than annual outputs and that deviations in wet seasons were larger than those in normal and dry seasons. The highest MAD occurred in June and August with a value of 82.87 t and 1.56 t for TN and TP, respectively. This study revealed the importance to consider the land use change when simulating the NPS pollution, and could provide support for land use input settings of NPS pollution models.