根据2004年的MODIS数据和森林火灾地面调查数据, 以福建省为例, 来探讨典型南方丘陵山区森林火灾发生后火灾面积大小评估的技术方法。在对火区250 m分辨率MODIS数据特征分析的基础上, 通过MODIS近红外通道反射率和归一化植被指数在火灾发生前后的散点图, 结合火区的假彩色合成影像, 利用ENVI软件的ROI处理功能建立火灾面积评估技术方法。对35起森林火灾样本进行评估, 过火面积总体估算误差为15 hm2, 结果表明, 在南方丘陵山区, 应用MODIS数据开展森林火灾面积评估, 能较好地满足业务需求。
Based on the MODIS data which have 250 meters resolution and the ground investigation data of forest fire in 2004, the study taking Fujian Province as an example analyzes and discusses the technological method about how to assess the forest fire acreages when there is a forest fire in hills regions in the southeast China.According to a number of forest fires cases happened in 2004, 35 typical cases are selected, the corresponding MODIS data characteristics are calculated and analyzed before and after the forest fires. The results show that the reflectivity of the near infrared channel (CH2) and the value of normalized difference vegetation index (NDVI) of MODIS data before the forest fire are obviously lower than after it, whatever are the data such as the minimum value of CH2 and NDVI, the maximum value or the average value. Moreover, the average reflectivity of CH2 declines by 6.65 points, and the average value of NDVI decreases by 0.23 points. The reflectivity of the visible channel (CH1) of MODIS data before the forest fire is much lower than that after it, and the ratio of samples to the total samples is 71%, but on the opposite, the ratio of samples with higher values reaches 29%. From the analysis, the reflectivity of CH2 channel of MODIS data and the NDVI value have different characteristics obviously before and after the forest fire on the mountain areas in Southern China, which can be used as an index to evaluate the size of the fire areas after the forest fire.On the basis of analyzing the properties of MODIS data to detect the forest fire areas, the method of comparing the dispersed graph between reflectivity of the MODIS near-infrared channels and NDVI before and after the forest fires is adopted in the study, combining with the pseudo color composite image in the fire regions, some means to evaluate the forest fire acreages with the help of using ROI function to deal with MODIS data in ENVI software are found.Finally, the study shows that the overall estimate error of burned acreages is 15 hectares in the evaluated 35 forest fire sample cases in 2004. At the same time, the calculated forest fire areas by the method of analyzing the reflectivity of the near-infrared channel (CH2) of MODIS data are showed and they are close to the actual results by ground investigation, and the areas of forest fire calculated by the means of adopting the value of NDVI are near the areas of the destroyed forest from ground investigation. Moreover, the results indicate it is significant that to evaluate the forest fire areas by the means of applying the MODIS data in hills of Southern China can be used in daily operational works and can meet the needs of monitoring the forest fire.