Font Size: a A A

Based On Spatial And Non-spatial Model And Influence Factors Analysis Of The Space-time Characteristics Of Fujian Forest Fires

Posted on:2017-05-24Degree:MasterType:Thesis
Country:ChinaCandidate:H L LiangFull Text:PDF
GTID:2283330485967119Subject:Statistics
Abstract/Summary:PDF Full Text Request
Forest fire is a very important influence factor of forest ecosystem. It will not only cause the loss of forest resources, increasing the carbon emissions of air pollution, but also threatening human life and health and property security. Identifying the pattern of forest fires and its major influential factors is the key for effective forest fire prevention. Fujian province is located in the southeast coastal areas in China, the intensive forest resources, is one of the key forest zone in the southern provinces in our country. The forest coverage of Fujian is 65.95%, ranking first in China. Fujian province has intensive fire frequency. Based on SPSS, R, GWR4, ArcGIS and other mathematical statistical software and spatial analysis software, using logistic regression model, the random forest algorithm and geographical weighted logistic regression model for the meteorological factors of forest fire in Fujian province were analyzed, and the applicability of forest fire forecast model in Fujian area and spatial distribution characteristic of forest fire in Fujian province was studied, the results of the study in this area of forest fire prevention work provide the certain scientific reference.In this study, statistical analysis and Ripley’s K-function were carried out to reveal the temporal and spatial distribution of forest fire in Fujian province. The results showed that the annual change of forest fire in Fujian province was obvious, and there was a significant seasonal variation tendency. Forest fire in Fujian province occurs in the spring, autumn and winter. In addition, forest fire of 2000-2005 in Fujian province was clustered distribution.Logistic regression model, random forest algorithm and geographical weighted logistic regression model were used to analyze the main influence meteorological factors on forest fire in Fujian province. The results manifest that "the highest surface temperature", "the minimum surface temperature", "the supreme station pressure", "sunshine", "daily maximum temperature" and "daily minimum relative humidity" are the main meteorological factors affecting the forest fire in Fujian province. LR model parameters fitting results show that, "the highest surface temperature", "the supreme station pressure", "sunshine" and "daily highest temperature" have a positive correlation between forest fire and meteorological factors, "minimum surface temperature" and "minimum relative humidity" are negatively related with forest fires. RF algorithm fitting results show that "the supreme station pressure" effect on forest fire the largest, "sunshine" effect on forest fire times, and "the daily highest temperature" have less effect on the forest fire. "The highest surface temperature" generally positively effects forest fire, after more than 40 degrees Celsius, stabilizing effects on occurrence of fire. Probability of fire occurrence decreases along with the increase of "minimum surface temperature". There is a negative relationship between "day minimum relative humidity" and forest fire. GWLR fitting results show that the main meteorological factors affecting the forest fire in Fujian province are stable in space, and the relationship between meteorological factors and forest fire changes over the geographical position. In addition to "the minimum surface temperature" which negatively impacts fire over the entire study area, the rest of the meteorological factors influence the fire occurrence range positive and negative; however, the estimated coefficients of the model are not significant in the whole research area.Model evaluation results manifest that, compared with the traditional logistic regression model and geographical weighted logistic regression model, the random forest algorithm fits data better and more suitable for predicting forest fire in Fujian province.To sum up, Fujian province is prone to the forest fire, and each season has different spatial distribution characteristics of forest fire. In spring, forest fires mainly concentrated in coastal areas of Fujian province; in summer, the forest fire in Fujian province is less and high fire risk areas are mainly distributed in Fuzhou and Fuqing area; in autumn, forest fire in Fujian province is mainly concentrated in Nanping, Sanming, and Yongan inland, and Zhangzhou is the coastal area which is more serious in the autumn forest fire; in winter, for the most regions of Fujian province have medium and high fire risk.
Keywords/Search Tags:Fujian province, forest fire, meteorological factors, models
PDF Full Text Request
Related items