Reclamation Information Sharing Environment (RISE)

Water resources planning requires statistically unbiased predictions of streamflow. Modeled streamflow is commonly used as input to water resources models to evaluate operational scenarios in Reclamation long-term planning studies, for example Basin Studies under the WaterSMART Program. However, systematic errors(e.g., incorrect process descriptions, errors in observed data, and uncertainties around model parameters) are ever-present in our predictions. These biases in modeled streamflow cascade into issues with water resources model simulations, ultimately making them unusable in some cases. To account for these sources of systematic error post-processing techniques, referred to as bias correction, are used to reduce the effects of these errors. Existing streamflow bias correction techniques have several shortcomings which we address in this work. In this project, we developed a bias correction technique that addresses two issues. First, previous bias-correction techniques generally ignore the relationship of streamflow across the river network, in turn destroying the spatial-temporal consistency of streamflow. Second, previous methods are commonly built on simple mappings between reference and simulated flows without accounting for systematic errors that vary in time. This new approach preserves spatial consistency and provides options for process-dependent corrections. Further, this new approach was tested for future climate change scenarios, acknowledging the likelihood of non-stationarity in streamflow biases.