Extreme Precipitation: Resolving the Added Value of High-Resolution Physical and Statistical Modeling

Abstract:

Human-induced climate change is altering the risk of many types of weather extremes, including the frequency and/or severity of heavy precipitation events. The basic science connecting global warming to more frequent heavy precipitation is relative straightforward. It is far more challenging, however, to predict how climate change will affect the magnitude and frequency of these events at the relatively fine spatial scales at which the impacts of extreme rainfall, snow storms, and flooding are typically characterized.

Using a case study based on a set of geographically distributed long-term weather stations located at Dept. of Defense installations across the U.S., we explore the individual and combined contributions of high-resolution regional climate modeling (WRF), station-based statistical downscaling (ARRM), extreme value distributions (GEV), and the use of global mean temperature-based thresholds rather than time slices (an approach that is illustrated Figure 1) to resolve observed trends and narrow the envelope of projected future change. All projections and analyses are based on the CESM1-MOAR simulation driven by the higher RCP 8.5 scenario, a consistency specifically introduced into the experiment in order to better resolve the strengths and limitations of each method in understanding extreme precipitation trends.

Each of these approaches provides clear added value when compared to direct output from the global climate model. We also find that the ability to refine global model output using high-resolution physical modeling, statistics, and observations can all prove useful at different geographic locations and for different parts of the distribution. However, the primary conclusion of this analysis is the utility of combining multiple physical and statistical modeling and analysis approaches when addressing issues such as extreme precipitation that occur at the tails of the distribution.