Ignition - The beginning of flame production or smouldering combustion; the starting of a fire.
(CIFFC Glossary of Forest Fire Management Terms)
Probability - A measure or estimate of the degree of confidence one may have in the occurrence of an event
Ignition Probability Research
Schroeder et al. (2006) study the effects of fuel stand thinning treatments and how ignition potential is affected. The report describes the methodologies used to model ignitions and the statistical analysis of the data collected.
Ignition tests were performed on 10 days from June to August 2002. They were performed simultaneously in all plots at randomly selected points, usually at two-hour intervals. The completion of the following sequence was recorded as one ignition test:
- A lit wooden match was placed on the forest floor.
- If the fire burned beyond the influence of the match and continued for two minutes with visible flames, it was considered a self-sustaining ignition (Figure 5).
- If the fire went out before two minutes, it was considered a failed ignition.
- If the ignition with a single match failed, the trial was repeated with two, three and four matches, if required, lit simultaneously.
Ignition probability was modelled for each study site using logistic regression. This technique modelled the probability of a binary event occurring (no ignition versus sustained ignition) based on a set of explanatory variables (Freund and Wilson 1998). The variables used for the models were fuel moisture (i.e., moisture content of litter, small twigs, large twigs, bark, and duff), weather data, and fire weather indices from the CFFDRS (Van Wagner 1987).
The fire weather indices were designed to be calculated from open air weather stations (Turner and Lawson 1978) and only data from the MOFR fire weather station were used for this purpose. The fire weather indices used for the analysis were the Fine Fuel Moisture Code (FFMC) and Initial Spread Index (ISI). The FFMC represents litter moisture content and is based on relative humidity, rainfall, temperature and wind speed (minor component). The ISI is based on FFMC and wind speed. The FWI indices were diurnally adjusted based on hourly readings from the weather station.
The logistic regression was performed as follows: 1. The independent variables were tested for multicollinearity. Significantly correlated variables were not used in the same models. 2. The logistic regression models were run for the independent variables. 3. The above models were evaluated to determine the one that best predicts ignition probability. The log-likelihoodchi-square test for model significance (<0.05) and the Wald statistic indicate the significance of an individual variable (<0.05).
Comparisons among models were done with the Akiake Information Criterion (AIC). The lowest AIC value indicates the best model, values within 2 points of the lowest score are considered similarly strong models, values within 2–10 points are good predictors, and values greater than 10 are poor predictors. The AIC is not an absolute predictor, but compares the relative strength among models. Finally, the c-statistic indicates the model’s ability to assign high probabilities to correct cases compared to incorrect cases (a value of 0.5 indicates the model can do no better than random assignment, and 1 indicates the model assigns higher probabilities to correct cases than to incorrect cases). Data from the thinned sites (no slash removed) and control sites were pooled together since the research plots were in close proximity (within 500 m).
This category currently contains no pages or media.