Field verification of the Nelson dead fuel moisture model and comparisons with National Fire Danger Rating System (NFDRS) predictions

The relationship of dead fuel moisture to fire behavior is especially critical, since it is dead fuels which are typically involved in the start of wildland fires and are the fuels which “carry” the fire. In the National Fire Danger Rating System (NFDRS) of the USA, dead fuels are separated into four “time-lag” classes: 1-hour, 10-hour, 100-hour, and 1000-hour. The algorithms used today in NFDRS to calculate dead fuel moisture for those four classes of dead fuels are essentially the same ones developed in the 1970s. These algorithms were originally developed to use once-a-day weather information (typically around 1400 local time) to model the lowest moisture time of the day (the time of “peak” fire danger). Thus, they do fairly well around those times, but don't do as well at other times or when there is rain or dew.

During the 1990s Ralph Nelson of the USDA Forest Service developed a theoretical model for dead fuel moisture to take advantage of frequent weather observations from automated weather monitoring stations. The model accounts for the physics of moisture and heat transfer, and includes processes of rainfall and dewfall. The model as originally published (Nelson, 2000) was only for 10-hour dead fuels. Since 2000, however, Nelson developed fuel stick parameters to allow the model to be run for the three other size fuel classes using hourly weather data. Inputs to the Nelson model are air temperature, relative humidity, solar radiation, and rainfall.

In conjunction with a funded project from the Missoula Fire Sciences Lab (USDA Forest Service), the Nelson model was tested against dead fuel moisture (DFM) observations from April 1996 to December 1997 at Slapout, Oklahoma. To be able to better handle different weather data intervals, rainfall in the model code was revised to process rainfall rate rather than rainfall amount. In a series of experiments using both 15-minute and hourly weather data, Nelson model parameters were adjusted to improve model performance against observed DFM, resulting in specific parameter sets for use with a given time interval of weather data.

This paper will compare the performance of the “Nelson model” for all four dead fuel classes against measured dead fuel moisture throughout a 21-month period at Slapout, Oklahoma, in the Oklahoma panhandle. Twice-daily weighings of Ponderosa pine fuel sticks representing all four fuel classes were made during this period, from which fuel moisture was later calculated. The site was near an Oklahoma Mesonet weather station, which provided the input data for the Nelson model. During this period monthly average temperatures ranged from 32F (0C) to 80F (27C) and monthly precipitation from 0.01" (0.25 mm) to 6.84" (174 mm), so a wide range of weather conditions were encountered. In addition, comparisons will be made to DFM predictions from the older NFDRS algorithms. Over the 21-month period, the Nelson model outperformed NFDRS in each size fuel class, with improvements in r2 values ranging from 0.09 for 1-hour fuels to 0.24 for 100-hour fuels when compared against the observational data. The Nelson model is targeted to be the dead fuel moisture model used in the next-generation NFDRS.