Stochastic Streamflow Generation: A Complementary Approach for Hydroclimate Projections in Hydrologically Complex Basins

Project ID: 23035
Principal Investigator: Subhrendu Gangopadhyay
Research Topic: Water Supply Forecasting
Funded Fiscal Years: 2023 and 2024
Keywords: None

Research Question

Primary Research Question(s)
Can we advance a generalized approach to develop representative streamflow estimates under a non-stationary climate to complement and improve Reclamation's current practices for generating hydroclimate projections?

Secondary Research Question(s)
Can the proposed approach remove steps such as second-order bias correction following streamflow generation and prior to use in water resources system models? Or, will this second-order bias correction step for generated streamflow which is the current practice still be necessary?

Need and Benefit

Streamflow generation under a nonstationary climate continues to be a challenge to the water resources planning community in the semiarid US southwest and broadly across Reclamation regions. Why? The present workflow of using forcings from Global Climate Models (GCMs) that are subsequently downscaled and used as inputs to hydrology models to generate streamflows under climate change conditions are often highly biased, and not directly usable in planning studies using water resources system models. Though second-order bias correction of these generated streamflow is a common practice, there are no satisfactory bias-correction methods currently available to correct for biases in the simulated streamflow under climate change conditions in river basins.

Specifically, in the semi-arid Southwest (e.g., Upper Rio Grande) the primary reason for this challenge to the streamflow bias correction problem is the large number of minimal to practically zero flow conditions that are present in these basins at monthly or seasonal time scales. For our second study area - Upper Deschutes, the Deschutes River is highly connected to the underlying aquifer system, and traditional hydrology models that are used to develop future climate adjusted hydrology do not accurately simulate this behavior, resulting in future climate adjusted hydrology that does not represent the hydrologic characteristics of the basin. Furthermore, developing well calibrated hydrology models that are applicable over a broad range of hydrologic conditions that are projected under a changing climate is also proving to be a challenge for these basins. In addition, downscaled GCM projections of precipitation and temperature continue to raise questions about the efficacy of these global climate models in simulating circulations patterns, specifically the monsoons that are key to the southwest region's water supply needs or large-scale climate processes such as the Pacific Decadal Oscillation (PDO) that impacts Pacific-Northwest's hydroclimate.

Climate impact assessments are complex -- the process starts with climate projection information (forcing data such as precipitation and temperature) which are then translated into hydrologic projections (e.g., streamflow) and processing hydrologic projections such as streamflows through a water resources system model to estimate risk for a range of metrics, e.g., end of water year reservoir content. This flow of information through a chain of interlinked models is the current practice. Lessons learnt from the portfolio of basins across Reclamation's domain shows that this approach generally works well for snow-melt dominated basins, but has presented challenges when applied to the semi-arid southwest basin locations such as the Upper Rio Grande Basin in New Mexico, and the Deschutes River Basin in the Pacific Northwest where subsurface flows are the dominant hydrologic processes. These challenges are due to, (1) the large number of minimal to practically zero flow conditions that are present in these basins at monthly or seasonal time scales, and (2) developing well calibrated hydrology models that are applicable over a broad range of projected hydrologic conditions and the limited success Reclamation has experienced in these hydroclimate settings.

In the proposed approach, we eliminate the need of a hydrology model -- instead, we will use historical climate and streamflow data and climate projections along with the use of a weather generator conditioned on climate projection data (or statistics derived from the climate projections; e.g., Bearup and Gangopadhyay, 2021) to derive probability distributions of streamflow projections. Given the probability distribution of the streamflow projections we can sample a wide range of streamflow values and use the streamflow values as input in the water resources system models.

Contributing Partners

Contact the Principal Investigator for information about partners.

Research Products

Contact the Principal Investigator for information about these documents.

Return to Research Projects

Last Updated: 6/22/20