The roles of uncertainty in hydrological models: identifying the sources and impacts in a SWAT model of the Maumee River Watershed
Abstract
The need for effective water quality models which can help inform successful policy options and expand the predictive impact of limited monitoring data has been at the forefront of recent discussions related to watershed management. However, a key limitation for interpreting model results is that some models—called deterministic models—produce one prediction without considering it within the context of uncertainty or confidence bounds, which are traditionally underreported and poorly understood. This is the case with the Soil and Water Assessment Tool (SWAT). Despite successful use of SWAT in several policy-relevant projects in the Maumee River watershed—a major contributing watershed in the Lake Erie region—including projects comparing outputs from five separate SWAT models, stakeholders are asking modelers to put their results in the context of uncertainty or confidence in those results. This project aims to assess the role of two potential drivers of uncertainty: parameter uncertainty and input uncertainty related to farm management. The SWAT model for the Maumee watershed has been updated to include hydrologic response unit (HRU) discretization focused on near field level delineation, allowing for direct incorporation of spatially realistic management practices. We identified 1) a suite of parameter sets representing a range of plausible values for commonly calibrated model parameters influencing hydrology and nutrient transport; and 2) a suite of potential baseline model simulations based on varying levels of location-specific input-driving management practices in the model. Model outputs and results of unique combinations of parameter sets and management simulations were used to develop confidence bounds using likelihood ratios. This range of confidence in model output at several scales allows us to quantify the magnitude and characterize the source of potential uncertainty for models in the region. This provides valuable insight into where our models can be improved and how results might best be interpreted for management and policy development.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFM.U12C..11A
- Keywords:
-
- 0810 Post-secondary education;
- EDUCATION;
- 0815 Informal education;
- EDUCATION