Abstract 244

Abstract ID: 244

Evaluation of S2S Prediction Project Database Performance in Forecasting U.S. Extreme Precipitation Events

Lead Author: Devin McAfee
University of Oklahoma School of Meteorology, United States of America

Keywords: S2S, prediction, extreme, precipitation, verification

Abstract: Long-term extreme precipitation events are extremely socioeconomically impactful meteorological phenomena, posing significant threats to life and property. With the Prediction of Rainfall Extremes at Subseasonal-to-Seasonal Periods (PRES²iP) research group, we are developing methods to characterize and better predict two-week extreme precipitation events in CONUS, particularly in the subseasonal-to-seasonal (S2S) forecast range. We classify two-week extreme precipitation events as clusters of extreme points—grid points with two-week precipitation exceeding the 99th percentile and above-average daily precipitation for at least half of this two-week window—forming polygons with areas greater than 200,000 square kilometers (Dickinson et al., 2021). This work analyzes the skill of 11 models from the S2S Prediction Project database, bias-corrected using gamma mapping, in forecasting extreme grid points within the polygons of 185 events from 1993 to 2012 at lead times ranging from 11.5 to 53.5 days, while identifying seasonal and spatial variability in skill.

We find, on average, the database is incapable of reproducing past events polygons with measurable skill for lead times beyond two weeks. However, there are rare instances where certain models, particularly ECMWF and Météo-France, generate skillful extreme event forecasts at leads times up to four weeks, most commonly over the Pacific Northwest and during winter, potentially associated with “”windows of opportunity”” for skillful S2S-range CONUS extreme precipitation forecasts in the S2S database. Further skill evaluation of atmospheric variables relevant to the large-scale forcing of these events (e.g. 500mb geopotential height anomalies) is requisite to discern the mechanisms driving the aforementioned geospatial and seasonal trends in skill as well as the large-scale features associated with enhanced and suppressed extreme precipitation event predictability in the S2S database.

Elinor R. Martin (University of Oklahoma School of Meteorology)