Research On The Algorithm Of Monitoring Forest Fire Points And Estimating Burn Scar Area Based On Modis Data

Forest fire is one of the major disasters in the world, which spreads and expands in the forest open system with freedom. It will not only bring economic loss, but also negatively influences the ecological balance and damages the environment of human. Therefore, it is quite importance to monitor forest fire effectively. Remote sensing technology, with high spatial and temporal resolution, provides broad application prospect in monitoring ecological environment and natural disasters. MODIS date is quite suitable for monitoring forest fire due to EOS/MODIS sensor was especially designed for requirements of fire monitoring and optimized the sensitive bands.Satellite technology can monitor forest fire the nearly real-time, but false fire points and braised burning points restrict the monitoring accuracy. Taking a case study of Wuning country of Jiangxi Province forest fire monitoring was researched by constructing the fire points index model and add the background information based on MODIS as the data source. The traditional method uses the threshold of characteristic parameters such as bright temperature to identify fire points. The size of background window and characteristic parameter values of background pixels can influence monitoring precision greatly when using fixed size of background window to calculate the temperature of potential points, even too small or too big size will affect the monitoring efficiency. Therefore, the paper improved the fixed size of background window into variable background window, set up different sizes of background windows, then extracted local maximum temperature from each size of background window. In addition, too low or too high background temperature can also affect precision. The characteristics of the parameters of MODIS data were analyzed. The improved algorithm can separate the hot fire spots from the background and seek out of the cool fire spots by reasonable thresholds of variance between-class based on variance between-class of brightness temperature. Also, the paper fixed absolute threshold and vegetation index. Integrated the remote sensing information enhancement technology into the process of fire point monitoring, above adjustment and improvement methods are combined to remove interference factors effectively and improve monitoring efficiently. Based on the analysis of the band characteristics, Global Environment Monitoring Index was chosen as characteristic parameters to identify burn scar. The recognition of burn scar was processed by seeds point spread algorithm which often used by computer technique based on the confirmed fire points. Then mixed pixels were decomposed to estimate area of burn scar land. Main conclusions are as follows:(Ⅰ) The fire points index model can outstand point information, eliminate exactly non-fire points and determine the area of suspect fire points.(Ⅱ) The fixed variable background window can eliminate the impact of the fire intensity and the changes of fire point position. The improved algorithm was verified to have good adaptability and can meet the requirements of monitoring fire by application.(Ⅲ) The algorithm based on variance between-class of brightness temperature had been used to monitor the forest fires at Wuning country. The results show that the algorithm is more satisfactory. It can be adapted in different environment. The smolder at low temperature and high temperature fire points can be monitored.(IV) The fixed of the absolute threshold can monitor weak fire points existed in cloudy or low temperature area; the fixed of the vegetation index can remove the interference of high temperature.(V) Using the recognition model to the actual date, accurate recognition of burn scar can reversely verify the accuracy of the improved method. The mixed pixels were decomposed to estimate area of burn scar land. Compared with the reference data, the estimation results are accurate.