2013 Volume 33 Issue 1 Pages 38-47
Forest fire smoke detection is an important use of remote sensing applications and image interpretation of moderate resolution imaging spectroradiometry (MODIS) true-color images is a common technique for this purpose. However, there are some difficulties in the decision process, particularly in discriminating between smoke and clouds. To overcome this problem, we propose a new MODIS false-color composite image method that combines a smoke reflectance index, mid-infrared reflectance and a water index. The false-color image depicts smoke in a reddish color that is easily distinguished from clouds and land surfaces. A Smoke Aerosol Reflectance Index (SARI) and a Water Index (WI) were developed to create the false-color composite image of R : SARI, G : MODIS channel 7, and B : WI. The SARI is derived from MODIS channels 1 (red) and 3 (blue) using the Aerosol Enhancement (AE) function and exponential transformation to correlate linearly with an Aerosol Optical Thickness (AOT) of 0.55μm over land. The Water Index (WI) is a synthesis of four water indices : the thick cloud index of channel 32 brightness temperature, the Aerosol Vapor Index (AVI), the Normalized Difference Water Index (NDWI) and the Normalized Difference Snow Index (NDSI). Some smoke pixels from a false-color image were sampled as a training dataset and overall smoke pixels were successfully detected. In this paper, we report three case studies of forest fires (in Korea, China and Russia, and Australia) and one set of results was compared and validated with those of an existing multi-threshold method. This comparision confirmed that the proposed false-color composite image is useful for detecting smoke plumes more accurately.
