Evaluation of the Nelson dead fuel moisture model in a forecast environment

The “Nelson model” for dead fuel moisture is targeted to be the dead fuel moisture model for inclusion in the next-generation National Fire Danger Rating System (NFDRS). The Nelson model, as NFDRS, will run on weather data from automated weather stations across the country. As it is anticipated that the next-generation NFDRS will include a mid-range forecast component, it is important to investigate how the Nelson model performs in a such an environment.

As part of a funded project between the Missoula Fire Sciences Lab (USDA Forest Service) and Oklahoma State University, the Nelson model was evaluated in a forecast environment covering the state of Oklahoma during the year 2004. The forecast model utilized was the 32-km Eta model of the National Centers for Environmental Prediction (NCEP). Forecast output utilized included both 00Z and 12Z runs and covered an 84-hour prediction period.

The Nelson model was run using 15-minute data from the Oklahoma Mesonet of 116 automated weather stations and hourly forecast output (interpolated from the 3-hourly output) of the NCEP Eta model. The integration of the two environments proceeded as follows. Every new whole hour, the Nelson model was updated using 15-minute Mesonet data over the past hour. The output from this latest hourly update then served as initialization for the Nelson model, which then ran hourly Eta model data through the remainder of the forecast period. When run with 15-minute Mesonet data, the Nelson model utilized the “15-minute” model parameter set; when run with hourly forecast output, it used the “hourly” model parameter set.

The methodology of analysis was to compare the Nelson forecast output every 6 hours throughout the 84-hour forecast period with the Mesonet-based Nelson output at the respective validation times (i.e., when weather observations had “caught up” with the forecast). Eleven Mesonet stations scattered throughout the state were chosen for analysis. This paper presents the results from the combination of all 11 stations for both a dry month (May) and a wet month (April) from 2004. Statistics for 1-, 10-, 100-, and 1000-hour dead fuel moisture (DFM) are presented for each 6-hour interval throughout the 84-hour forecast period. The Nelson model is seen to perform admirably in this forecast environment. Due to the spatial and temporal variability associated with rainfall, comparisons of model DFM with Mesonet-based DFM throughout the 84-hour forecast period are the least favorable for 1-hour fuels, which are most sensitive to rainfall, and best for 1000-hour fuels.