A Short-Term Predictor of Satellite-Observed Fire Activity in the North American Boreal Forest: Toward Improving the Prediction of Smoke Emissions

In order to meet the emerging need for better estimates of smoke emissions in air quality and climate models, a statistical model, based on numerical weather prediction (NWP), is developed to predict the following day's satellite observations of fire activity in the North American boreal forest during the fire season (24-hour forecast). In conjunction with the six components of the Canadian Forest Fire Danger Rating System and other NWP outputs, fire data from the MODerate Resolution Imaging Spectroradiometer (MODIS) and the Geostationary Operational Environmental Satellites (GOES) are used to examine the meteorological separability between the largest fire growth and decay events, with a focus on central Alaska during the large fire season of 2004. This combined information is analyzed in three steps including a maximum likelihood classification, multiple regression, and empirical correction, from which the meteorological effects on fire growth and decay are statistically established to construct the fire prediction model. The potential benefits from incorporating scale-based fire detection uncertainty estimates are also explored. Both MODIS and GOES fire observations show that the NWP-based fire prediction model is an improvement over the forecast of persistence commonly used by near-real-time fire emission inventories. Results from an independent test (2005 fire season) show that the RMSE of predicted MODIS fire observations is reduced by 5.2% compared with persistence. Improvements are strongest (RMSE reduction of 11.4%) for cases with observed decay or extinction of fires. Similar results are obtained from additional independent tests using the 2004 and 2005 GOES fire observations. This study uniquely demonstrates the value and importance of combining NWP data and satellite fire observations to predict biomass-burning emissions, which is a critical step toward producing a global short-term fire prediction model and improving operational forecasts of smoke transport at large spatial scales.