The RMRS Rocky Mountain Center (RMC) developed a comprehensive georeferenced model for predicting wildfire starts out to 7 days based on forecast weather and data on vegetation cover and recent drought history.
The fire-management community had a long-standing need for quantitative predictions of wildfire-ignition chances up to 10 days in advance based on current weather forecast, local vegetation cover and drought conditions. Existing models for evaluation of wildfire risk such as the National Fire-Danger Rating System (NFDRS) only provide qualitative assessments of the likelihood of fire ignition or growth. Forecasting the actual daily probabilities of wildfire start due to lightning- and/or human-caused ignitions on a unform National grid of high spatial resolution has been lacking, although the development of such a capability was mandated by the Conservation, Management, and Recreation Act (a.k.a. Dingell Act) in 2019.
This R&D project utilized large data sets of gridded daily wildfire occurrences (Short 2021), 3-hour reanalysis weather parameters from NARR, NOAA’s EDDI drought index, and hourly lightning data from VAISALA NLDN spanning more than 25 years to derive probabilistic equations for predicting the numerical chance of cloud-to-ground lightning flashes and wildfire ignitions on a 20-km grid over the Conterminous USA. Logistic equations were derived for each month of the active fire season (May through September). Three separate statistical analyses considering 195 predictors each were carried out to develop predictive equations for the chance of wildfire ignition due to either lightning, humans, or any cause.
Key Findings
- Forecasting of cloud-to-ground lightning flashes is reliable up to 7 days.
- Low-level atmospheric moisture and temperature are the strongest predictors of wildfire ignitions.
- Monthly equations are required to reliably predict the chance of wildfire starts due to lightning- and human ignitions.