Extreme precipitation is among the most destructive natural disasters. The Northeast United States has faced the most rapid increasing occurrences of extreme precipitation within the nation in the past few decades (Fourth National Climate Assessment). The remnants of Hurricane Ida poured record-breaking rainfall in the Northeast US on September 1^st 2021, inundating the densely populated region including the New York City and Philadelphia metropolitan area, causing at least 55 casualties and more than 20 billion dollars in losses (NOAA NCEI, 2021). Similar catastrophes have been commonly seen across the globe in recent years, leading to rising concerns about whether existing infrastructure is robust to possible changes in future warmer climate.

While the climate science community has long understood that precipitation intensity increases with global warming, detecting and projecting changes in extreme precipitation on regional scales, however, remain highly challenging and uncertain. One of the primary limitations comes from models’ coarse horizontal resolution. Global climate models, commonly with atmospheric resolution ranging from 100 to 200 km (1–2°), are not sufficient to resolve the most extreme precipitation events.

Now, a high-resolution climate model, SPEAR (Seamless system for Prediction and EArth system Research), newly developed at National Oceanic and Atmospheric Administration (NOAA) Geophysical Fluid Dynamics Laboratory (GFDL), provides a good opportunity for us to detect and project changes in regional extreme precipitation  (Details about GFDL-SPEAR).

To understand extreme precipitation over the Northeast US in the boreal fall season (September to November), we use a collection of 25-km GFDL-SPEAR simulations with historical radiative forcing (observed greenhouse gas, aerosol, and land use) and projected future changes following a moderate greenhouse gas emission scenario (SSP2-4.5) and an extreme greenhouse gas scenario (SSP5-8.5). Using these simulations, we can disentangle the contribution from human-caused climate change and natural climate variability.

Some key findings:

• We show that the newly developed GFDL-SPEAR with 25 km horizontal resolution simulates much more realistic extreme precipitation over the Northeast US than comparable climate models with 50 or 100 km resolution, including frequency, amplitude, and temporal variability. The 25-km model simulated trends are quantitatively consistent with observed trends over recent decades (Figure 1).
• We use the same GFDL-SPEAR model for future projections. Extreme precipitation, defined as the top1% of daily precipitation based upon historical climatology, would double its frequency by the end of 21^st century under the SSP5-8.5 scenario. In the moderate SSP2-4.5 scenario, the frequency would increase 50% by the end of 21^st century (Figure 1a,b). The increasing rate, however, is not uniform for all intensities among the extreme precipitation range: the very extreme events (>150 mm/day) would be five times more likely by 2100 than in the early 21^st century under the SSP5-8.5 scenario (Figure 1c).
• By the mid-21^st century, the GFDL-SPEAR projects unprecedented rainfall events over the region, driven by increasing anthropogenic forcing which is distinguishable from natural variability (Figure 2).

While this study focuses only on the Northeast US, our work can set an example for further assessment of how a high-resolution climate model simulates extreme precipitation in other regions and seasons, providing additional insight into the advantages and limitations of high-resolution climate models.