Using Downscaling Approaches to Investigate the Future Changes in Winter Storms and Associated Impacts Over the Northeast U.S

The Northeast U.S. is particularly vulnerable to extreme weather from winter storms (heavy precipitation, wind, and storm surge). This presentation will highlight the future changes (up to year 2098) of cool season (November through March) extratropical cyclones over the Northeast U.S. using the Coupled Intercomparison Modeling Project (CMIP5) as well as Weather Research and Forecasting (WRF) members at 20-km grid spacing downscaled using a few CMIP5 models. The 6-h output from CFSR (Climate Forecast System Reanalysis) and CMIP5 models were used to provide initial and boundary conditions for the historical and future WRF runs, respectively. Some of the sensitivity of the 20-km WRF cyclone characteristics to model physics for this downscaling effort will be discussed using several members from an historical ensemble.

There is a shift towards stronger cyclones in some of the CMIP5 models (e.g., CCSM4) and WRF by the middle 21st Century, especially for those storms originating along the Gulf coast or off the U.S. Southeast coast (Miller-A storms). Some processes responsible for the model differences and future changes in cyclone intensity along the U.S. East coast were diagnosed using storm-relative composites of temperature, winds, precipitation, etc… within a 3000 km by 3000 km box around each cyclone center. For example, during 2069-2098 (for RCP8.5), there is a significant increase in cyclone-relative precipitation, especially over the interior Northeast (20-40% by the later 21st century). This coupled with the decrease in cyclone-relative low-level baroclinicity suggests the importance of diabatic heating from precipitation and warmer SSTs . The higher resolution WRF results in stronger intensity storms than most CMIP5 models during the historical period, so this has motivated future WRF runs to look at those changes, which will be presented.

As an application for the future storm projections, we developed a generalized parametric downscaling technique to create a point-based storm surge time series for the cool season using 6-h CMIP5 data (details provided in a separate poster presentation). The statistical model for New York City is trained and evaluated using Oct-March 10-m wind and mean sea level pressure from the NARR renanalysis between 1979-2012. Overall, there is no statistically significant increase in the frequency of >= 1.0 m surge events for New York City by the late 21st Century, but some models have longer tail in the distribution for stronger surge events. If one includes a sea-level rise of at least 0.6 m during the next several decades, most models have an order in magnitude increase in the number of coastal flooding events.