In the last ten years, with the help of satellite remote sensing, we build up a huge database of fire points in China. The
remote sensing data that we used to do the fire monitoring include NOAA, FY-1, FY-3 and MODIS. In this paper, we
present a new model for fire forecast base on the former database and NCEP reanalysis data of last ten years. As we
know, the reason of land surface fire can be divided to two groups: subsurface property and meteorological factors. Both
of them are very complicated. For subsurface property, there are many factors that relational to wild fire, such as land
surface type and combustible material. For meteorological factors, they also strongly impact to the fire occur. There are
four factors of meteorological should be pay attention in the fire forecast, they are wind speed, precipitation, temperature
and humidity. For the former two groups of reasons of fire's taken place, we build a two-part model to do the fire
forecast. For the first part, corresponding to the subsurface factors, we used the ten years fire points monitoring database
to describe it. We do the statistics on the database by five days (overlapping, 366 periods totally) and 0.5625 degree grid
(according to NCEP). In each grid and each period of days, the average number of fire points describes the fire status
corresponding to the average meteorological conditions and subsurface condition at that grid and at that time period. For
the second part, firstly, we average the four meteorological factors into five days periods and 0.5625 degrees grids;
secondly we evaluate the different of the four factors from the average value in the target day (forecast day).
Field experiments with man-made fires in a forest were conducted to verify fire warning products from satellite remote sensing techniques and to select more effective channels for producing these products. Pine branches and trunks as well as other woods were burned at a designated place in a pine-dominated forest to simulate wild forest fires when a satellite was passing over the sky. Infrared spectral irradiances, visible spectrum, brightness, and temperature were measured concurrently with satellite data at the ground using a medium and near-infrared MOMEM MR154 FT-Spectroradiometer, an infrared thermal imager, and a visible and near-infrared spectroradiometer (ASD FR). The measurements showed two emission peaks in middle infrared band that corresponded exceptionally to the combustion strength. One of the spikes at 4.17 μm reflected the CO emission peak. The other peak spanned through the wavelengths of 4.34-4.76 μm, which exhibited a much stronger response to the fire than the commonly used channel 3.5-4.0 μm for fire monitoring in remote sensing. The results suggest that the wave band 4.34-4.76 μm is probably more sensitive and more effective than the common-used channel for wild fire monitoring using satellite remote sensing techniques. However, the peak of this wavelength band drifted during the burning process, which should be taken into account in channel selection. This band is suitable to determine forest fires. Further studies are needed to use it for retrieving fire strength quantitatively.
Meteorological satellite are widely used to monitor forest and grassland fire in China and other developing countries in recent years as its AVHRR sensor with middle infrared channel are very sensitive to hot spot. However, some new issues recently in this application need be considered: the forest department want more information from satellite remote sensing, like the condition of a fire, not only location of fire spot; from NOAA-K, the mid-infrared channel (3.7µm) closed at daytime, that give much difficulty to detect forest fire by using meteorological satellite; EOS/MODIS sensor has more channel and higher resolution in ground surface monitoring. This paper introduces some improved method for detecting and evaluating the conditions of forest and grassland fire by using multiple channel data from meteorological and environment satellite data, These methods includes:
a, Evaluating the sub-pixel size and temperature of fire spot by using multiple channel data of AVHRR in polar meteorological satellite in various condition, including. The way of multiple channel combination are: mid- infrared (3.7µm) and thermal infrared (11µm) channel; near infrared (1.6µm) and thermal infrared (11µm) channel; two long infrared (11µm and 12µm) channel. And also introduce the means of presentation for evaluation result.
b. Detecting the hot spot in the condition without mid- infrared channel data, includes: using two thermal infrared (11µm and 12µm)channel data from AVHRR; using near infrared (1.6µm) and thermal infrared (11µm) channel data from AVHRR.
c. Some features in fire monitoring by using multiple channel data of EOS/MODIS.
Conference Committee Involvement (1)
Disaster Forewarning Diagnostic Methods and Management