Forecasting Precipitation Extremes in an Age of Changing Climate: Does the Performance of Forecast Methods Vary Across the CONUS?
Abstract
Climate change is expected to increase the intensity and frequency of extreme precipitation events, which will impact both drought and flooding. This poses challenges for developing accurate and operational forecasts, which are critical for water resource management decisions. There is a need to evaluate precipitation forecasting products performance for predicting precipitation extremes, and a need to understand if that performance has varied over time. To address this, we evaluate the performance of three different forecast products - CHIRPS, ERA5 and HRRR - in forecasting 1-3 day ahead precipitation over the CONUS. Given that the number of ensemble members differs across these forecast products, we statistically develop probabilistic precipitation forecasts using three different methods: 1.) K-Nearest Neighbor Bootstrapping 2.) Canonical Correlation Analysis and 3.) Pooled regression with point forecast error. The three probabilistic forecasting methods and the raw forecast ensemble were evaluated for various locations across CONUS based on a) Brier Score b) Reliability plots and c) Rank Probability Skill Score. Understanding how these forecast products and probabilistic forecasting methods perform over different regions will allow us to assess their utility in forecasting rainfall under different lead times. Furthermore, as the climate continues to change, this assessment will provide information on whether precipitation forecasting should stick with the one size fits all approach or if forecasting methods should vary across CONUS to better predict extremes.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2021
- Bibcode:
- 2021AGUFM.H45B1192L