| Summary: | 博士 === 國立中興大學 === 森林學系所 === 100 === Forest fire behavior model is a logically and statistically significant tool that can be used as a useful guide for fire behavior study and provids reliable predictions for natural resource managers to make decisions on fire suppression activities to minimize damages. The purpose of this study is to investigate the fire behavior of windbreak forests in the Taichung Harbor of Taiwan, focusing on the effects of weather and fuel factors. The research was conducted by fire danger rating, fuel model building, computer simulation, laboratory testing, and field burning testing. A series of forest fire behavior has been analyzed and classified.
In the fire danger rating, 7 indicators of Canadian FWI System were adapted to analyze the fire weather data of Taichung Harbor in the past 20 years. In addition, fuel moisture content of litter and duff layer were also adopted in this model to predict fire danger rating distribution in windbreak forest of Taichung Harbor. Results of analysis show that 11.6% of time in this period can be classified as the extreme fire danger. The fire intensity could be more than 4,000 kW/m. The highest rate of forest fire danger occurs in the duration from September to Januaryof next year, and the highest peaks in October.
Fuel survey was conducted by mixing 10 m×10 m sampling plots and line-intersect sampling, based on geographical features of the windbreak area in Taichung Harbor. The whole area was divided into 4 blocks, and a total of 33 samples of ground fuel were gathered, and then clustered into 4 fuel types, namely, shrub type, herbal-based, non-cover litter types and mixed type. The average amount of fuel loading for the layer of litter and duff is 17.68 t/ha. The collected data were then used to establish fuel loading prediction model. Five regressional models have been built in this study using data including height of shrub, crown width of shrub and ground cover of grass as input parameters. Surface fuel loadings in the windbreak area of Taichung Harbor were estimated to be 21.12-31.66 t/ha according to these models.
The laboratory-scale burning tests were conducted to verify the effects of weather and fuel factors on the level of fire behavior. Results of experiments showed that at wind speed lower than 1 m/s, fire spread was governed mainly by the buoyancy forces induced from burning of fuels (Fc2 = 0.18, Nc = 8.33). However, as wind speed raised to 3 m/s, the inertia force of wind dominated the spread of fire (Fc2 = 1.04, Nc = 0.49). The fire spread speed (ROS) at 3 m/s wind speed was estimated to be 7 times higher than that at windless condition. A regressional fit of experimental data between the ROS and the wind speed showed three different patterns, ranging from linear function to exponential function, and finally becomes independent state. Low moisture content fuel (7.5%) had 1.5-fold of mass loss rate compared with that of high moisture content fuel (17.0%), and the fire temperature rising rate was also more rapidly (1.35-fold). Besides, simulation results obtained with the Rothermel model showed similar tendency.
All fuel types were then used as inputs to the BEHAVEplus fire model to investigate the fire behaviors, and the results of these simulations were then categorized into three levels. It was noted that 79% of windbreaks area in Taichung Harbor was higher than moderate level of fire indicator. As for the fuel type characterization, it was found that at lower wind speed (≦2.25 m/s), the shrub fuel type of windbreak area can be a reasonable alternative of BEHAVEplus TU5 (R2=0.95, p<0.01).
The field burning tests were conducted in both the summer time and the winter time. The scale of fire was limited by cutting off free burning 18 minutes after ignition. All the data, including fuel, weather and fire behavior parameters, were recorded and fed into a regression model to establish a fire behavior prediction model. Measured data were then compared with the predicted values obtained from BEHAVEplus system as well as FBP model. It was noted that the prediction of BEHAVEplus on ROS was acceptable (R2 = 0.85). However, predctions on the burned area perimeter (R2=0.52), the back fire propagation rate (R2=0.50), and flame length (R2=0.55) were not very accurate. The predictions of FBP model showed different results. ROS was overestimated, and fire intensity parameters were underestimated. But predictions in fire perimeter and the area and the ratio of total length to maximum breath of fire area were better than BEHAVEplus.
|
|---|
