212 Evaluation of March- and May-initialized Seasonal Forecasts of Buildup Index for the Alaska Fire Season

Monday, 29 January 2024
Hall E (The Baltimore Convention Center)
Cecilia Borries-Strigle, Univ. of Alaska Fairbanks, Fairbanks, AK; and U. S. Bhatt, P. Bieniek, E. Stevens, H. Strader, A. York, and R. Ziel

As temperatures increase and the Arctic climate continues to change, the Alaska wildfire season is expected to have increased fire activity and to become longer than its current period of April 1-September 30. Following the record-breaking 2015 fire season in Alaska, UAF researchers began to work with the Alaska fire management community in predicting the upcoming wildfire season. Fire managers begin to make decisions about the upcoming fire season in March, and additional decisions regarding fire management are made in July, shortly after the peak of the season in late-June. Due to these time windows, a user need for summer fire outlooks created from seasonal forecasts initialized in both March and May was identified. For the past four years, we have created summer fire weather outlooks of Buildup Index (BUI), a fire weather index from the Canadian Forest Fire Danger Rating System, to share with the fire managers at their spring operations meeting in March. May 2023 was the first year we created a May outlook of BUI. Evaluations of the outlooks have also been shared at the end-of-season meeting in October.

Three seasonal forecast models are used to create summer fire outlooks: NOAA CFSv2, ECMWF SEAS5, and Météo-France System8. From these forecasts, temperature, precipitation, and dew point/relative humidity are used to calculate BUI. Temperature and precipitation from global climate models tend to be cooler and wetter at high latitudes compared to observations; therefore, temperature and precipitation anomalies from each model ensemble are added to observational climatology which results in BUI values much closer to observational values. The resulting BUI forecasts are evaluated based on Alaska wildfire subseason (wind-driven, duff-driven, cumulative drought, and diurnal effect), BUI tercile (upper, middle, and lower), and predictive service area subregion in Alaska (Eastern Interior, Western Interior, and Southcentral) with the relative operating characteristic (ROC), Heidke, and mean squared error (MSE) skill scores. The three models are also combined into an equal-weight multimodel ensemble.

According to the three skill scores, the greatest amount of skill is typically found in the wind (April 1 - June 10) and drought (July 21 - August 9) subseasons, in the lower (below average) BUI tercile, and in the Western Interior subregion of Alaska. Combining the models into a multimodel ensemble increases forecast skill by an average of 10% (13%) for the March (May) forecast ROC score and an average of 90% (121%) for the March (May) forecast Heidke skill score. The May forecasts typically have equal or greater skill than the March forecasts, with an average improvement of 2% for the ROC skill score and 20% for the Heidke skill score. The MSE skill score improved by 13% for both years with large fire seasons (>1 million acres burned) and years with strong El Nino events.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner