|Chinkuyu, A - UNIV OF CALIFORNIA|
|Mexiner, T - UNIV OF CALIFORNIA|
Submitted to: American Society of Agricultural Engineers Meetings Papers
Publication Type: Proceedings
Publication Acceptance Date: May 12, 2003
Publication Date: July 28, 2003
Citation: Chinkuyu, A.J., Mexiner, T., Gish, T.J., Daughtry, C.S. 2003. Sensitivity analysis of gleams model using multi-objective sensitivity analysis procedure. In: Proceedings of American Society of Agricultural Engineers. 03219. Interpretive Summary: Surface and subsurface water movement as well as chemical leaching processes are complex, being influenced by soil properties, landscape position, crop production practices, and climate. In addition, these processes vary by several orders of magnitude over the agricultural landscape. As a result, no model to date can accurately simulate field-scale water flow or chemical transport. In this study, a sensitivity analysis was performed on the Groundwater Loading Effects of Agricultural Management Systems (GLEAMS) model by comparing simulations to field-scale observations from an intensively instrumented site in Beltsville, Maryland. Results indicate that accurate nitrate and pesticide transport simulations are possible only when the hydrology is accurately characterized and modeled.
Technical Abstract: It is difficult and laborious to find the best parameter set for a model using the traditional manual model calibration. This limitation necessitates the emergence of new, automatic, and robust multi-objective sensitivity analysis and model calibration procedures that will improve models for predicting water quality. The robust Multi Objective Generalized Sensitivity Analysis (MOGSA) technique was used to investigate parameter sensitivity of the Groundwater Loading Effects of Agricultural Management Systems (GLEAMS) model for two agricultural watersheds in Beltsville, MD. Initially all transport parameters were allowed to vary, but when hydrologic parameters were kept constant and chemical parameters were allowed to vary there was a dramatic improvement in model sensitivity. Using comparisons of different model fluxes to observations and calibrating hydrologic parameters separately from chemical parameters revealed more information about the model, sensitive parameters, and the natural system being modeled. We would not have observed this model behavior if we had used only one objective function or varied all parameters simultaneously.