A comparison of precipitation downscaling procedures to guide flooding and water resources studies
In relatively small basins with arid climates, rainfall characteristics are highly variable and stream flow is tightly coupled with the nuances of rainfall events (e.g. hourly precipitation patterns). Reclamation's water supply assessments in these basins typically employ CMIP5 projections downscaled with Bias Corrected Statistical Downscaling and Bias Correction/Constructed Analogs (BCSD-BCCA) methods, but these products do not always meet management needs. While BCSD-BCCA simulations accurately represent historical average daily precipitation, they underestimate daily variability, exaggerate the number of daily events and fail to characterize large storms. Moreover, the BCSD procedure does not explicitly account for localized physical precipitation mechanisms (e.g. monsoon and snowfall). These features are essential to hydrological systems in the Southwest that are dominated by rainfall at a daily or finer scale.
We propose a rigorous comparison of the BCSD-BCCA methods with alternative approaches to guide Reclamation staff conducting hydrologic studies. These include rainfall output from dynamical downscaled Regional Climate Models (RCM), a stochastic rainfall generator forced by either Global Climate Models (GCM) or RCM, and projections using historical records conditioned on either GCM or RCM. We will perform this comparison for two sites. The Lower Colorado Region's River Operations Group identified the Bill Williams River drainage to Alamo Lake, upstream of Lake Havasu, as an area in need of improved flood projections. The Upper Santa Cruz River near the Arizona-Mexico border is the site of an on-going Reclamation water resources appraisal study. Our objectives are to: 1) evaluate the suitability of each method for each study type; 2) develop a test to select the preferred procedure for a given type of study; and 3) quantify the impact of each method on Reclamation's planning activities.
Need and Benefit
Selecting the appropriate data for planning studies and other activities requires staff to understand the advantages and drawbacks of the available projections. While statistically downscaled simulations like the CMIP5 BCSD-BCCA projections are easy to access and encompass a wide range of assumptions, they also have drawbacks. A critical issue is the assumption of stationarity of scale across multiple spatial resolutions. Specifically, 'non-dynamical downscaling assumes simplified statistical relationships between atmospheric conditions over a large area and local surface climate'. This assumption is especially problematic in areas where regional and local scale hydrologic processes are important, such as smaller basins with arid and semi-arid climates. There are many such basins in the Southwest United States that fall within Reclamation's Lower Colorado, Upper Colorado, Mid-Pacific and Great Plains regions. In contrast, dynamically downscaled projections use a physical model to simulate climate processes at higher spatial resolutions. This type of model is more likely to be useful at a local planning scale, but is computationally intensive and may not be available for planners' area of interest. However if necessary, it may be possible to obtain such projections from sources other than Reclamation. Currently, Reclamation's guidance suggests that the study team consider such basic factors as the availability of projections for the time period of interest and whether the relevant climate variables are included in the projections. In the end, the study team must judge which model outputs are sufficient to represent future assumptions in support of key study decisions and meet peer review requirements. At present, little is known about the impact of the choice of precipitation projection method on the outcome of the Reclamations planning processes. The proposed study would provide a quantitative metric on the suitability of several types of projection methods for areas where regional and local scale precipitation phenomena are known to be important. In addition to statistically and dynamically downscaled models, we will evaluate the use of a stochastic rainfall generator forced by either Global Climate Models (GCM) or Regional Climate Models (RCM), and projections using historical records conditioned on either GCM or RCM. As noted in the technical guidance, the choice of projection method will be influenced by the hydrologic processes of interest. To describe a range of study types, we will compare results of hydrologic models for both flooding and water resources analyses. Flooding analyses are concerned with short-term, extreme high precipitation and stream flow values, while water resources studies consider long-term hydrologic patterns. In the case of the effects of drought on water supplies, a particular concern is the continuous duration of low precipitation and stream flow values. We will investigate the impact of precipitation projections on both types of studies and their hydrologic variables of interest. The knowledge generated by this study will provide a foundation for more specific guidance to Reclamation staff on the selection of precipitation projections in hydrologic studies. At present, there is no such resource available to meet this need. In addition, the results can be put to immediate use by Reclamation in operations planning for the Colorado River and a water resources appraisal study in the Upper Santa Cruz River Basin.
Contact the Principal Investigator for information about partners.
Bureau of Reclamation Review
The following documents were reviewed by experts in fields relating to this project's study and findings. The results were determined to be achieved using valid means.
A Comparison of Precipitation Downscaling Procedures to Guidelines of Climate Change Impacts on Flooding and Water Resources (final, PDF, 6.0MB)
By Eve Halper
Research Product completed on September 30, 2019