Scaling Behavior Analysis Of Weather Driven Fire System

The scaling behavior of fire system is one of the difficulties in fire science, because there are many complex influencing parameters, and most of them can't be indentified quantitatively. Now most researchers focus on the forest fire modeling and statistical analysis of real data. It is revealed that forest fires have the spatial-temporal behavior, so we can make the reasonable long-term predict by the Self-Organized Critical (SOC) model, because the occurrence frequency of small and medium fires can be used to quantify the risk of large fires. Therefore, whether the forest fire model is accordant with the real data? What are the temporal scaling characteristics of fire system and its influencing fators? Using modeling and time series analysis in this thesis, the spatial-temporal behavior and scaling characteristics of fire system are studied, as well as the weather parameters, because the weather fluctuation influences forest fire strongly.The forest fire model introduced by Drossel & Schwabl (DS model) has been used to explain the power law frequency-area distribution of real data. But the fire interval distribution of the DS model is an exponential law, different from the power law with periodical fluctuations of actual data. Therefore, a weather driven forest fire model (WD model) is built considering actual hourly weather records, with which the fire igniting probability is calculated. The simulation results indicate that the frequency-interval distribution of the WD model agrees with that of actual forest fire data and, at the same time, the frequency-size distribution of the WD and the DS model are in accordance with each other. In the further analysis of the temporal property of weather data, it is found that the change of related humidity also exhibits a power-law relation with periodic fluctuations, implying that the external driving from weather parameters, not the internal dynamics, is the essential reason for the power-law distribution of fire intervals.Power-law scaling behaviors of the real forest fires and weather parameters are analyzed by means of the detrended fluctuation analysis (DFA) method, because the weather data play key roles in the forest fires. It is found that the forest fire area, temperature, relative humidity and rainfall record all have the similar scaling exponents in relevant timescales and similar crossovers, and exhibit persistent long-range power-law correlations in timescales about larger than 5 days. The results imply that forest fire and weather maybe coupled each other. We also investigate the effects of trends on the DFA results of the four series. It is found that the seasonal period affects the study results and causes the change of scaling exponents. The scaling behavior in relevant timescals may be influenced by different trends.Besides the burned area, the temporal patterns of fire events may have great connections with the correlation structure. It is found that the inter-event time series of both forest and urban fires have persistent long-range power-law correlations, and they all have two scaling exponents despite the different regions and countries. The interval sequences of urban fires closely resemble that of white noise in small timescale, and the correlations are weaker than that of forest fires. Human behavior and human density may affect the long-range correlation in some way. And along with the increase of threshold, the correlation of urban fires becomes weaker in large timescale.The mechanism of fire interval distribution is explained in this thesis, and it is found that the fire area and occurrence time both have long-range correlations, which is similar with the weather parameters. The results would be helpful to indentify the correlation between fire and weather parameters, and to understand the temporal pattern of fire system.