Generation of week-2 fire-indicator forecasts for the Climate Prediction Center

Rochelle Worsnop

NOAA/PSL

Tuesday, May 25, 2021, 2:00 pm
Webinar

Abstract

PSL has developed algorithms to generate skillful 8–14 day guidance relevant to fire-weather forecasting at the Climate Prediction Center (CPC). This presentation will introduce the framework used to generate these forecasts, which includes a combination of statistical post-processing techniques and the Global ECMWF Fire Forecast System (GEFF). Training and validation of the post-processing models used 20 years of ECMWF reforecasts and ERA5 reanalysis data for 11 meteorological variables related to fire weather, including surface temperature, wind speed, relative humidity, cloud cover, and precipitation. The calibrated meteorological forecasts were then used to predict fire indicators, which characterize wildfire response as it relates to fuels, weather, and topography. Skill scores show that the post-processed fire-indicator forecasts overall have greater positive skill at days 8–14 relative to raw and climatological forecasts. While raw forecasts lose their skill compared to climatological forecasts by day 8–9, the post-processed forecasts maintain positive skill out to 14 days. It will be shown that the post-processed forecasts are more reliable at predicting above- and below-normal probabilities of fire indicators than the raw forecasts. In addition, the greatest skill for days 8–14 is achieved by aggregating forecast days together. Lastly, the transition of this research into operations at the CPC will be discussed.