Evaluating model selection impacts on forecast uncertainty in the Upper Colorado River Basin Implications for USGS modeling and monitoring

Objective:

Train a postdoctoral fellow or graduate student

Abstract:

Model uncertainty analyses are critical to understanding equifinality, but opportunities
remain to better understand how uncertainties in major parameter and calibration fields
influence forecast uncertainty. In particular, snow accumulation and melt under the forest
canopy is poorly represented in models, and soil parameters and soil moisture are poorly
constrained due to a lack of fully spatially distributed observations and scaling uncertainties in
soil hydraulic parameters. The proposed work will take new data-driven approaches to estimate
uncertainty in subcanopy snow and soil moisture parameters, and use these new uncertainty
estimates along with others in a Bayesian model uncertainty evaluation framework to
determine the relative importance of a wide suite of model parameters, as well as snow and soil
moisture uncertainties, to streamflow forecasting skill. The final deliverables of this research will
be: (1) quantification of the key measurable sources of uncertainty in streamflow forecasts; and
(2) a reproducible workflow of model parameter uncertainty estimation developed in
consultation with USGS for model refinement and informing decisions concerning USGS
observational networks.