Uncertainty Characterization in Watershed Models of Managed Landscapes

Abstract:

Input data, parameters and model structure are sources of uncertainty in river basin scale water and biogeochemical cycling models. While various studies have addressed these sources of uncertainty, there is still a lack of understanding in simulations, where land management plays a significant role in the water cycle. Diverse approaches have been used in the past to address these problems, however they focus on different mathematical formulations in mostly natural landscapes. Here, we propose to evaluate the three sources of uncertainty (model structure, parameter and input data) in heavily managed watersheds in the midwestern United States. We use two different modeling frameworks that simulate hydrologic fluxes at daily time steps. First, a simple model with five parameters is developed to simulate key water fluxes – surface and subsurface runoff and evapotranspiration (ET). We use this simple model to elucidate the linkages between relative model parameter dependencies. We then go on to develop a complex watershed-scale model that accounts for key land management decisions, soil and topographic variability. The second model partitions precipitation fluxes into multiple runoff and ET fluxes that are influenced by the heterogeneous landscape and land management. In the second model, we develop, test and validate present-day baseline conditions with the North American Land Data Assimilation System (NLDAS) weather data and land management decisions, including watershed conservation practices. We then use the model to test the sensitivity in simulating present-day watershed fluxes by adjusting (i) different weather inputs (e.g. weather station data instead NLDAS), (ii) model structure/formulation (e.g. different runoff and evapotranspiration estimation methods) and (iii) model parameters.

The parameters for the sensitivity analysis are selected from a previous study. We systematically sample the parameter space and perform multiple simulations during the baseline period 2010-2019. Preliminary results show that parameters related to soil moisture redistribution exhibit steep nonlinear responses at certain ranges, but not much elsewhere (Figure 1). In this presentation, we will show the range of simulated runoff and ET responses to the three sources of uncertainties and compare them with observed fluxes. Our results will be valuable for resource planners as they are faced with making timely decisions related to future environmental change.