| Forest fires are a recurrent environmental and economic emergency worldwide. To date, fire activity and its impact were addressed by most international projects. In China, great economic losses caused by forest fires kept high and its occurrence ratio was higher than developed country, while the status of forest fire research was unsatisfactory. As a big forestry province of China, Heilongjiang province is also a significant part of northern Eurasia boreal forests zonation because of its abundant forests resources. Heilongjiang province, in which average annual forest fire burned area ranks the first in China, is the most serious fire hazard areas. Large amounts of data of Heilongjiang province such as basic geographic data, MODIS, TM and other remote sensing data, historical fire records, et al. were used to research the spatial-temporal variation law of forest fire, to establish a spatial-temporal model of forest fire spread for typical major forest fires, and to create the Heilongjiang provincial forest fire risk model which was used to divide forest fire danger rating on the support of generalized 3S technology using some methods such as mathematical statistical methods, spatial data mining method, ANN-CA model, spatial-temporal model, et al. The results provided new methods and new ideas for forest fire research. Meanwhile, gave prerequisite and basis for the prediction of forest fire, and provide a scientific basis for macro decision-making and guidance of forest fire prevention in order to greatly improve the efficiency of forest fire prevention work, reducing the occurrence of forest fires, the implementation of efficient forest fire fighting, and reducing the loss of forest fires. The research encompasses several major topics:(1) The results on spatiotemporal distribution law of forest fire based on the history of forest fires from 1980 to 2005 statistics showed that:The inter-annual forest fires in Heilongjiang province fluctuated dramatically. The high frequency of fire happened period lied in the early and mid-1980s and low frequency period lied in late 1980s and the early 1990s. The number of forest fire increased from the end of 1990s and after 2000, but less than the early 1980s. In most years, forest fire alarms were more than the number of fires, which reflected timely detection and extinguishing forest fire can do in many cases. The inter-annual fire burned area which was particularly significant influenced by catastrophic forest fire had a weak correlation between the total numbers of fires and the burned areas. Spring fire season sustained 3 to 4 months, during this period, the number of spring fires and its burned areas reached to 72.52% and 84.39% of all of the fires. The number of summer fires and its burned area were less than spring fire season. Fall fire season was only 1 to 2 months, but certain conditions may also lead to catastrophic forest fires. The forest fire season in spring was from the 59th to 180th in Julian date, and in autumn was from 258th to 303rd. There was a total of 170 days of the forest fire prevention period in accordance with the standard of burned area over 100 hectares in five consecutive days. Few forest fires occurred in plain areas, and forest fires decreased with altitude rising in mountain region. The slope distribution law of the forest fire was not obvious as the slope of gentle rolling, but sunny slope was more susceptible to fire and more likely to spread than shady slope.(2) The study on spatial-temporal variation law of forest fires based on MOD14A1/MYD14A1 and MCD45A1 data set showed that "8+9" data sets which extracted from the mask data products MOD14A1/MYD14A1 was most suitable for forestry active fire detection in Heilongjiang province. The coherence between the data sets and forest fire history point reached to 0.83. The spatial distribution and the temporal difference of MCD45A1 burned area was very obvious from 2000 to 2010, in which inter-annual and monthly burned area fluctuated significantly. More than a half of the burned area located in the north temperate deciduous forest, the area ratio was 53.68%, and about 44% of the burned area distributed in the temperate conifer forest area, while burned area of temperate grassland area was only 2.32%. Forest burned area and burning rate in the area which had mid-elevation (200m≤H≤500m), low gradient (less than5°) was larger than other regions.(3) Forest fire spread instance of typical history fire was studied by using MOD14A1 MYD14A1 data and validation data, and dynamic changes in forest fire spread was obtained. The spread instance data and forest fire spread impact factor data were used to establish a cellular automata model based on artificial neural network. The "5.20" catastrophic forest fires occurred at Quanshandiyunzhi, Aihui District of Heihe City and "5.17" largest forest fires occurred at extension of Fulahancc. Daxing'an Mountain in 2003 were randomly selected to simulate and validate the model. The overall accuracy of the model was 84.13% and 83.01%.(4) 2000-2010 MODIS burn scars of remote sensing data sets MCD45A1 of Heilongjiang province was used to build the spatial logistic forest fire risk model between the spatial distribution of forest fire and forest fire impact factors in support of Geographic Information System technology. The 12 forest fire impact factors including climate, terrain type, terrain factors and human activities were obtained by stepwise regression method. Forest fire risk zoning study was processed in a larger temporal scale and provincial spatial scale. The model fit better in the case of the significance level of 0.05, and the mixed test of the model coefficients passed. The relative operating characteristic (ROC) value of ROC curve analysis was 0.753. Factors corresponding Sig. Value of the forest fire was less than 0.05 in the 0.05 significance level, also passed the Wald test. Heilongjiang province forest fire area was divided into none, low, moderate, high, and extremely high fire risk zones. Great Xing'an Mountain cold temperate coniferous forest area was extremely high and high fire risk zone, while Xiaoxing'an Mountain temperate broadleaf mixed forest area was basically a high or moderate fire risk area. Small parts of the eastern mountain temperate coniferous and broadleaved mixed forest region belonged to moderate fire risk area, and other areas were a low or none fire risk zone.In summary, researches including forest fires spatial-temporal variation law, forest fire spread to the spatial-temporal model and the forest fire risk model provide a scientific reference for forest fire prevention and macro decision-making, guidance, and efficient fighting. As we know, the occurrence of forest fires still has randomness and uncertainty; therefore we cannot relax our vigilance. A series of effective measures should be taken to ensure lasting peace and stability of the large forest, including forest fire prevention publicity and education work should carry out to enhance the people's awareness of forest fire prevention, wild fire source should control and manage strictly, fire behavior in production and living should be standardized, inspection and supervision of high fire danger periods and hazardous areas should be strengthened to address the various fire trouble, fire buffer zone should be opened, forest fire prevention infrastructure construction should be strengthened to improve the comprehensive capacity of preventing forest fires.
|
PDF Full Text Request