Study On Forest Fire Danger Forecasting Based On GIS And RS

Forest fire is one of the most important natural disasters. Many forest fires took placeevery year in the world, which affected about millions hectares of forest. Daxing'an Mountainsis the important area of forest distribution and forest fires with 30 per year in our country. Taheforestry bureau locates in the hinterland of Daxing'an Mountains and is one of the importantforestry enterprises in the northeastern China. The work of forest fire prevention is always themost important task in safety management. So how to predict the forest fire danger ratings andoffer decision support has become the key problem to be settled.As complexity of fire happening and regional diversity, it is not reasonable to forecastforest fire danger with simple analysis and models. In this paper, some technologies were used,such as Remote Sensing, Geographic Information System and mathematical statistical methods.Main work included forest fuel distribution and classification using ETM+data, how to obtainforest fire danger factors, model building and the method of forest fire danger forecastingbased on the above work. Summarization of this dissertation is as follows:(l) Detailed analysis on forest fire regime of Tahe forestry bureau showed that 286 forestfires took place from 1975 to 2004, among which 2002 was the high happening year with 54fires. Fire frequency and burned area tended to be increasing in the recent 30 years. Burnedarea caused by common fires and serious fires and fire frequency increased markedlyespecially after 1990s. And the frequency of catastrophic fires was also very high.(2) Based on the review of internationally and domestically published papers, forest typeswere divided by the supervised classification method and then the result accuracy was verified.The whole classification accuracy was 85.61％and the kappa index was 0.81. Theclassification accuracy of coniferous forest and broadleaf forest was high and that of cut areawas not perfect.(3) Through analysis on the relationships between forest fires and forest types, landform,human environment and meteorological factors, some factors was selected to study the forestfire danger ratings. At the same time, humidity index as the index evaluating the extent of dryand humidity of climate was also selected to correct intraday meteorological condition. As aresult, this could improve the forecasting reliability of forest fire danger.(4) The regions with high fire danger rating included coniferous forest, broadleaf forest,elevation (400m-600m), sunshine slope and road buffer area within 100m through analyzingthe relationship between fire frequency and fire danger factors.(5) Fire frequency tended to be increasing with daily maximum temperature and aridityindex increased. When daily maximum temperature was higher than 25.32℃and aridity index was higher than 2.41, the probability of fire happening increased. Fire frequency tended todecrease linearly with the increasing relative humidity. And fire frequency was high as windspeed was within 2-4m per second and that would fall as wind speed was higher than 4m persecond.(6) The forecast model of fire danger was divided two parts, including backgroundcomposite index and meteorological index. Every fire factor was endued with weight by usingthe Analytical Hierarchy Process (AHP) method and analyzing the relationships between allfire danger factors and fire frequency, which could overcome the disadvantages of lackingreliable quantitative analysis and human interfering.(7) The dates of July 17, 1994 and June 16, 2002 were selected to verify the forecastaccuracy because fire frequency in these two days was very high. High and secondary high firedanger area in the whole forestry bureau was 63.50％and 76.30％respectively. All the firesoccurred in the regions with middle and over middle fire danger ratings, and frequency was90％and 78.57％respectively. The result showed that this fire forecast method could be used inpractice. Adding intraday meteorological condion not only can reflect the changing of forestfire danger, but also can improving the forecasting lever.