Approach To Improve Simulation Performance Of Watershed Model With Scarce Water Quality Data

Watershed hydrological model is the main tool in research of watershedhydrological characteristics. Its result is usually used to formulate plans of watershed’sdevelopment, as well as protection. Researchers are always faced with the problem ofplenty of hydrological monitoring data and scarce water quality data. Therefore, there isa problem to be resolved, that the simulation process of scarce water quality data is infact the fitting of points to line, which is usually difficult and troubling.Through analyzing the mechanism of watershed model and the monitoring data,we found that there is high positive correlation between water quality simulationperformance and the hydrological simulation performance on water quality monitoringday. Traditional post-treatment process is responsible for the overall date, therefore itcan’t improve the above local performance, such as on and around water qualitymonitoring day. Based on traditional method, an improved approach was proposed toimprove hydrological simulation effect on and around water quality monitoring day. Itincludes two main factors: length of period with the center of water quality monitoringday, and Nash-Sutliffe effiencience coefficient. At the same time,16experiment wasdesigned to validate the improved approach and compare the significance of those twofactors.SWAT (Soil and Water Assessment Tool) was selected as the tool. The upper andmiddle Mulanxi Basin in Putian City, Fujian Province, China was selected as theresearch area. Total Phosphorus (TP) was selected as the water quality index.Simulation period covered from Jan.1st2008to Dec.31th, with time step of day. Thebest result of16plans showed improvement of47.3%. Length of period was moreimportant than weights of broad NSE. For the frequency of2months a time of waterquality monitoring, length of no longer than2weeks was efficient. It could beconcluded that the improved method is efficient to improve simulation performance ofwatershed model with scarce water quality data. At the same time, the broad NSE wasfirst proposed and used in this research.At last, it could be found through uncertainty analysis that the TP simulationperformance with improved approach was obviously improved. The NSE withimproved approach was concentrated between-0.6and-0.48, while NSE withtraditional approach was dispersed between-4.5and-1.5. It means that, the improved approach could provide systematical help in improving water quality simulationperformance.