Study On The Correlation Between Forest Fire And Meteorological Factors Based On Fuzzy Rough Sets

Forest fire is a kind of devastating natural disasters, and it can destroy the achievement of afforestation and endanger people’s lives and property safety. At the same time, it also affects the healthy and steady development of the economic and ecological environment construction. In recent years, global climate changed significantly and forest fire and climate change had become a hot issue in the field of forest fire research. Fuzzy rough set theory has the ability to deal with the uncertainty and incomplete information, and it also has characteristics that drive by data and without expert knowledge constraints. These abilities and characteristics are adapted to the forest fire’s features that its forming process is complex and occurrence is uncertain. So the fuzzy rough set is applied to the forest fire research which is possible.Taking Hunan Province as an example, this paper randomly selected 1041 forest fire from historical fire data in China forest fire prevention net and collected 4 kinds of data:temperature, relative humidity, wind speed and 24h precipitation. The data was verified in the professional meteorological website. On the basis of these data, the following research had been done in this paper on the combination between meteorological factors and fuzzy rough set and the relationship between meteorological factors and forest fire.(1) The data of each meteorological factor was analyzed by SPSS data processing software, the results showed that:the temperature value when forest fires occurred most frequently is 20℃,70% of the fire occurred at 10-25℃, the humidity value when forest fires occurred most frequently is 50%,90% of forest fires occurred in the relative humidity which is less than 70%, the wind speed value when forest fires occurred most frequently is 2 m/s, The wind speed values are mainly concentrated in the 1-3 m/s, the 24h precipitation value when forest fires occurred most frequently is 0 mm, more than 90% forest fires’ 24h precipitation is 0 mm.(2) The data after checking was normalized and K-means clustered, then the final case number of K-mean clustering was classified by Natural breakpoint method. It obtained the information decision table which regards the final cluster centers as condition attributes and the levels of case number as decision attribute. The table provided the basis for the combination of fuzzy rough set method and the meteorological factors related to forest fires.(3)This paper analyzed the characteristics of meteorological factors, selected the appropriate fuzzy membership function type for each factor based on these attribute characteristics, and constructed a reasonable fuzzy membership functions model based on the data characteristic of each factor. It also used these model condition attributes turned fuzzy and the fuzzy decision table was obtained. The fuzzy decision table provided data for the application of fuzzy rough set method.(4)Under the support of fuzzy decision table date, by calculating the dependence of decision it attributes in the fuzzy rough set method to all condition attributes, and then calculate the influence of various meteorological factors of forest fire after normalization treatment, the results show that:the most important factor is the pre-fire precipitation in 24h, the minimum is temperature, the specific impact of value as follow:the temperature(0.1162), relative humidity(0.265), wind speed (0.2124), precipitation (0.4063), the results showed that in the process of forest fire monitoring, we should be particularly concerned about the factors of precipitation.(5)Compared meteorological factors after clustering with the number of forest fire, some laws between the group of meteorological factors and forest fire were obtained: if the precipitation of 24h before forest fire occurred is more than 1 mm, no matter how mutative other meteorological factors are, forest fire is rarely occurrence. Many forest fires may occur in the conditions of low temperature and low humidity, low temperature and high humidity, but the forest fires’ frequency of low temperature and low humidity was significantly more than that of low temperature and high humidity. In the condition of high temperature, forest fires occur only in low humidity, and forest fire is difficult to occur in the condition of high temperature and high humidity.By using the method of fuzzy rough sets, the effect of meteorological factors to the occurrence of forest fires was calculated, and a new method which analyzed correlation between forest fire and meteorological factors was obtained. In addition to the meteorological factors, forest fire was also influenced by fuel, topography, population, and customs etc. The forest fire research that takes these factors into a system will be a research tendency in the future.