PUBLICATION HIGHLIGHT: Dynamical downscaling improves upon gridded precipitation products in the Sierra Nevada, California

Mount Whitney in the Sierra Nevada, Credit: Geographer (Wikimedia CC 1.0)
Mount Whitney in the Sierra Nevada, Credit: Geographer (Wikimedia CC 1.0)

Estimating how precipitation is distributed in regions of complex terrain, such as mountains and coasts, is essential for water management and hydrologic predictions, yet is challenging because of limitations in observing systems. In this new study, a researcher from CIRES and the ESRL Physical Sciences Division and collaborators investigated precipitation across California’s Sierra Nevada mountain range using ten datasets—six produced by dynamical downscaling and four produced by mapping rain gauge observations onto a grid. Dynamical downscaling uses a weather model to translate low resolution (regional or larger scale) weather information to a finer (or local) scale. Comparing these datasets to daily estimates of precipitation amounts from a snow pillow (a device that measures the water equivalent of the snowpack) and precipitation estimates inferred from streamflow, the researchers found that estimates from certain configurations of the dynamical downscaling improve upon the gridded estimates.

Many model evaluations use statistically gridded precipitation datasets as truth; had the researchers taken this approach their conclusion would be that all the dynamical downscaling datasets have a wet bias, which the comparisons to observations show is not the case. Investigation of both water year total and single storm precipitation biases revealed that the water year total biases were in some cases quite dependent on biases from one major single storm; therefore, case studies of model configuration performed for individual storm events could lead to incorrect conclusions about the model’s overall tendencies, since precipitation biases are reflected in the higher-order statistics rather than being systematic. The study’s focus of both Sierra-wide and at smaller scales (e.g., watershed scale) reveals that very different biases can exist at highly localized scales.

The results of this study suggest a path forward for improving precipitation estimates in complex terrain, using either dynamical estimates produced from dynamical downscalings, or perhaps hybrid techniques that merge the benefits of dynamical techniques with reduced computational cost.

Authors of Dynamical downscaling improves upon gridded precipitation products in the Sierra Nevada, California are Mimi Hughes of the ESRL Physical Sciences Division, Jessica D. Lundquist of Civil and Environmental Engineering, University of Washington, and Brian Henn of Center for Western Weather and Water Extremes, Scripps Institution of OceanographyUniversity of California, San Diego