Monday, 17 November 2003
Assessing the value of increased model resolution in forecasting fire danger
The fire season of 2000 was used as a case study to assess the value of increasing mesoscale model resolution for fire weather forecasting. With a domain centered on Western Montana and Northern Idaho, MM5 simulations were run at 36, 12, and 4km resolutions for a 30 day period at the height of the fire season. Verification analyses for meteorological parameters that influence fire danger rating were done for observation sites within the model 4km domain. Analysis of Mean Absolute Error, Root Mean Square Error and Bias for temperature, relative humidity and wind speed and direction indicate only minor differences between the model resolutions. Results also highlight known MM5 model biases and suggest a need for adjustments to the land surface model during extended dry periods.
In addition, MM5 model output data was used to calculate National Fire Danger Rating System (NFDRS) indexes. The model-predicted indexes at all three resolutions were compared with those computed at Remote Automatic Weather Stations (RAWS), and with the growth of active wildfires. The model biases in the fire weather variables carried through to the NFDRS indexes, so that in general the MM5 under-predicted fire danger. This represents the first step in a process to generate real time NFDRS forecasts using MM5 model output.
Supplementary URL: http://www.fs.fed.us/pnw/fera/mm5case