Using an ensemble of high-resolution climate model simulations to detect, attribute, and project changes in extreme rainfall over the Northeast U.S.

Extreme precipitation, both the occurrence and intensity, over the Northeast United States has significantly increased since the 1990s, evidenced by observations. The most salient increase has happened in the fall season (September to November). Understanding the attribution and projection of long-term trends in regional extreme precipitation is essential to adaptationion planning such as infrastructure upgrade. However, such work is challenging due to uncertainties caused by internal climate variability and the requirement of medium-to-high model resolution as well as ensemble size. In this work, we leverage the newly-developed GFDL (Geophysical Fluid Dynamics Laboratory) SPEAR (Seamless System for Prediction and EArth System Research) models which generate 25-km high-resolution simulations (ten members; SPEAR-HI) and 50-km large-ensemble simulations (30 members; SPEAR-MED) for both historical simulations from 1921 to 2014 and projections for the Shared Socioeconomic Pathway 5-8.5 (SSP585) from 2014 to 2100. We aim to address two related scientific questions using GFDL-SPEAR: (1) what are the factors that have contributed to the increasing autumn extreme precipitation over the Northeast US since 1990s? How much of the increase could be attributed to anthropogenic forcing? (2) when would the increased extreme precipitation in response to forced climate change emerge from the noise of internal climate variability?

Our preliminary results first suggest that higher atmospheric resolution in climate models is critical to facilitate the simulations of regional extreme precipitation. For example, SPEAR-HI can simulate comparable frequency of extreme precipitation over the Northeast US (rain rate > 50 mm/day), compared to the observation; while SPEAR-MED underestimates the frequency. Second, the recent increasing Northeast US extreme precipitation is unlikely due to the warming North Atlantic sea surface temperature, even though the timing of the abrupt increase in extreme precipitation coincided with the timing when the Atlantic Multidecadal Oscillation shifted from a cold to warm phase in the mid-1990s. Our ongoing work focuses on evaluating the attributions from other factors including internal variability, aerosols, and greenhouse gas. Last, we analyze SPEAR-HI SSP585 projections and extended control simulations starting from the year 1850. We estimate that the anthropogenically forced increase in the Northeast US autumn extreme precipitation would emerge from the noise of internal climate variability around the 2040s. However, ongoing work will employ more systematic methods to estimate the time of emergence.