Method Research Of Forest Fire Warning Based On SAR Images

In the western mountains of China, people want to monitor and forecast forest fire which is not an easy thing. Because the mountains have harsh natural environment, high altitude, rough road and the traffic is inconvenient. What’s more, the vegetation covers in the forest of the mountains are high and lots of fallen leaves lead the forest fire occurred easily. So, using suited data and method in fire warning is particularly important. The using of multi-source remote sensing technology can save more manpower, material and time, because the technology is macroscopic and timely. In particular, microwave remote sensing, with its all-time, all-weather characteristics and penetrate clouds and fog and in the rainy time extensive applications. However, factors resulting in forest fires are more difficult to get all of the relevant factors parameters and fires warning is a challenge. How to accurate monitoring of fires and fires warning in only a small number of surface parameters, is a technical difficulty.This study wants to solve that problem. Using SAR data, the experimental data and some ground meteorological data, via information extracting algorithms and fires warning modeling, we can warn the forest fires in cloudy, rainy, mountainous terrain complex areas. The contents and results of this paper are as follows:(1) By analyzing the relationship between complex terrain area and the radar echo, DEM terrain correction method is used to eliminate errors caused by terrain distortion, including layover, shadow and other distortions.(2) Using SVM method and SAR image multipolar features, based on analysis of the complex terrain area, the influences of the bare soil, water, buildings, but vegetation are removable.(3) Different data sources and different parameter inversion algorithms are used to quantitative retrieval vegetation biomass, vegetation moisture, soil water content under the vegetation cover respectively, and assessing the accuracy via the ground measured data and multi-source remote sensing data.(4) In order to researching the process of forest fires(before-fire warning, in-fire monitoring, and after-fire forest recovery), using ground microwave scatterometer system to measure simulate fires, three phases of the scattering characteristics are obtained and explaining the phenomenon by theoretical knowledge.(5) Different surface parameters quantitative retrieving via the SAR image, temperature and other factors, are combined with forest fire danger index to calculate the fire probability and get the forest fire warning model. What’s more, using of the correlation between soil moisture and fires to correct the forest fire danger index. Finally, the forest vegetation fire danger map and incentives of fire are created according to the model. The MODIS fire data is used to assess the accuracy of the fire warning model and validation results show that the fire warning model constructed in this paper has a good effect.