A Parameter Sensitivity Analysis Across Mesoscale Basins Entering the Gulf Mexico Basin

Abstract

Hydrologic models are tools that can quantify the natural flow regime for locations that lack pre-disturbed flow records by matching existing measurements and translating information from areas we have measurements to places that we don’t. With any model application, we try to balance model complexity, the number of model parameters, with our ability to predict a range of hydrologic processes at fine scales. To address over-parameterization issues that arise from complex models, a sensitivity analysis can be employed to determine which parameters are more or less important. The objective of this study is to understand unaltered drainages in the headwater basins of lower Alabama. To understand unaltered drainages we employed the Method of Morris sensitivity analysis for 7 headwater sites within the Gulf of Mexico Basin. At the headwater locations we used the Precipitation Runoff Modeling System (PRMS) model to simulate streamflow and compared to existing measurements. The importance of a model parameter was identified based on the mean (μ) and standard deviation (σ) across multiple elementary effects. By analyzing parameter sensitivity with respect to multiple metrics describing the flow regime, the sensitivity analysis allows us to rank the importance of the 17 model parameters and understand the dominant hydrologic process for unaltered drainages in headwater basins of lower Alabama. In order to account for different flow regimes, performance of watershed models is often evaluated for multiple functions that capture different parts of the hydrograph. The evaluation functions focused on high flow, low flow, and daily flow. Across the 7 mesoscale basins, we were able to identify the dominant parameters for the 6 different evaluation functions. The sensitivity analysis identified 8 PRMS model parameters as highly impactful on streamflow. These model parameters are associated with the soil-zone, subsurface, impervious zone, and the groundwater reservoir of the PRMS model. The main purpose of these parameters is to route water once it hits the land surface either to the stream network or through the soil profile into the groundwater reservoir are the controlling model parameters. Also, we were able to determine the parameters that were considered impactful were dominated by interactions. Due to the interactions, we have difficulty characterizing the model in terms of model parameters because multiple parameter sets are able to produce the same model output. Model interactions complicate the modeling effort and should be considered during calibration. Ultimately, a sensitivity analysis is able aid in model calibration by identifying impactful parameters and reducing the number of parameters to focus on during calibration.