Effect of Spatial Unit Delineation on Simulating Non-point Source Pollution by SWAT Model

王琳,朱良君,朱阿兴,刘军志,沈琳
DOI: https://doi.org/10.3969/j.issn.1000-1700.2016.04.012
2016-01-01
Abstract:The representation of spatial heterogeneity by distributed models, such as SWAT (Soil and Water Assessment Tool), a comprehensive, semi-distributed, and process-based continuous model, is based on spatial discretization. Therefore, the predicted pollutants of distributed NPS models are directly affected by different spatial unit delineation schemes. To better understand the relationship and achieve a suitable spatial unit delineation scheme of SWAT model, this paper analysed the sensitivity of non-point source pollution simulation to spatial unit delineation. Zhongtianshe watershed (42 km2) located in Jiangsu Province was selected as the study area because of its severe non-point pollution. A total of 25 spatial unit delineation schemes were designed according to different subbasin and HRU delineation methods based on accumulation area threshold and area percentage threshold of landuse, soil and slope gradation, respectively. Each of the 25 schemes were simulated under a same calibrated parameters set derived from the reference scheme, i.e., a scheme with an accumulation area threshold of 1.52 km2 based on real digitalized stream links and HRU area threshold of 0%-0%-0%. The model calibration showed a good result with Nash-Sutcliffe efficiency coefficients of runoff, sediment, total nitrogen and total phosphorus equal to 0.575, 0.671, 0.483 and 0.758 in calibration period, respectively. The predicted pollutants output are more sensitive to HRU delineation method than subbasin, the reason of which may attribute to lumped landuse areas and the parameters changes of relevant watershed processes. Therefore, some recommend issues may be drawn in SWAT modeling of non-point pollution: the proper HRU delineation level would be as detailed as possible under the assurance of the simulation accuracy of runoff and sediment, which may avoid the unreal simulation caused by lumped parameters of rough HRU delineations; as for the proper subbasin delineation level, a moderate accumulation area threshold which could represent the real stream links properly would be suitable enough.
What problem does this paper attempt to address?