Real-Time Analysis of Fire Weather Prediction Accuracy

Fire weather forecasters, fire planners, and decision makers do not have easy access to information needed to verify the accuracy of fire weather forecasts and the products that rely on them to predict ignition potential. The Joint Fire Science Program (JFSP) funded development of a system that produces intuitive, easily understandable meteorological model performance assessments and provides end users with real time information about meteorological model bias, model reliability, and overall performance of fire weather forecasts. We have produced a system that ingests data from RAWS and ASOS weather stations; analyzes the difference between observed and forecasted weather data to identify accuracy, bias, and uncertainty; and displays the analysis in an online map system. Users can view current, historical, and forecast weather data from WRF, NAM, NDFD, and GFS weather models for each weather station. In addition, observed and forecasted fuel moisture and fire danger rating indices using the National Fire Danger Rating System (NFDRS) and the Canadian Forest Fire Danger Rating System (CFFDRS) are displayed. Simple statistical analyses allow users to identify model bias and model accuracy. We have found that fire weather accuracy can be improved locally by using a simple mean bias measure. Every day for each station, the system calculates a mean bias measure over the past seven days. We apply this bias measure to the forecasted fire weather variables, such as air temperature and relative humidity, to adjust the fire weather metrics spatially and temporally in light of local model performance.