Skip to main content
ARS Home » Southeast Area » Tifton, Georgia » Southeast Watershed Research » Research » Publications at this Location » Publication #315868
Title: Using expert knowledge of the hydrological system to constrain multi-objective calibration of SWAT models

Author
item PFANNERSTILL, M - University Of Kiel
item BIEGER, K - Texas A&M University
item Bosch, David - Dave
item GUSE, B - University Of Kiel
item FOHRER, N - University Of Kiel

Submitted to: Annual International SWAT Conference
Publication Type: Abstract Only
Publication Acceptance Date: 3/25/2015
Publication Date: 6/24/2015
Citation: Pfannerstill, M., Bieger, K., Bosch, D.D., Guse, B., Fohrer, N. 2015. Using expert knowledge of the hydrological system to constrain multi-objective calibration of SWAT models. [Abstract] Annual International SWAT Conference, 06/24-26/2015, Italy.

Interpretive Summary:

Technical Abstract: The SWAT model is a helpful tool to predict hydrological processes in a study catchment and their impact on the river discharge at the catchment outlet. For reliable discharge predictions, a precise simulation of hydrological processes is required. Therefore, SWAT has to be calibrated accurately to provide reasonable model results not only for discharge at the watershed outlet, but also for the different water balance components. We highlight the relevance of expert knowledge about the water balance components in combination with appropriate performance metrics by applying a new evaluation framework to identify calibration runs with a realistic representation of the whole hydrological system to our study area, the Little River Experimental Watershed (LREW) in Georgia (USA). Previous studies of the LREW and rules of thumb based on general hydrologic knowledge were used to define appropriate ranges for the water balance components as constraints for the SWAT simulations. After using the Nash-Sutcliffe Efficiency (NSE) and the percent bias (PBIAS) to identify the best calibration runs with respect to the simulation of discharge, we used the constraints to select the parameter sets that also result in a reasonable simulation of the different water balance components. Our results show that satisfactory NSE and PBIAS values do not guarantee realistic simulation of water balance components. Several trade-offs between good statistics for discharge simulations at the watershed outlet and reasonable average annual amounts of the water balance components have been found. In general, the optimization of a realistic simulation of water balance components comes at the expense of NSE and PBIAS values. We conclude that the application of NSE and PBIAS is not sufficient to ensure a satisfactory simulation of the whole hydrological system. Therefore, we propose our approach of using expert knowledge during SWAT model calibration to constrain the ranges of water balance components and thereby achieve a meaningful simulation of the entire hydrological system.