Improved MSG-SEVIRI images cloud masking and evaluation of its impact on the fire detection methods

One of the most important factors responsible of the firedetection algorithms fail is represented by the inaccurate cloud detection methods. In fact, the cloud-contaminated pixels are often associated with false fire pixel because of the brightness temperature increase in the mid-infrared channel. On the other hand an incorrect cloud masking could hide a real fire pixel, especially at the borders of clouds. Together with, the SEVIRI images EUMETSAT provides its own cloud mask (CLM product). This mask is computed by making full use of the MSG-SEVIRI spectral channels. Among the 12 channels, only channels 8 (IR 9.7) and 12 (HRV) are not included in the cloud detection and analysis. Due to the particular application for which CRPSM is using SEVIRI images, detection of fire at its early stage (sizes lower than 0.1 ha), a high sensitivity to changes in the radiance measured by the sensor in channel 4 (3.9 μm) is required. Since the presence of a cloud covering only a fraction of the pixel (∼4x4 km at mid latitude) can produce an increase in the estimated brightness temperature, in such channel, capable to provoke a false alarm we decided to use also channel 12 in the cloud detection algorithm. Thus, in order to improve the cloud masks provided by EUMETSAT a new methodology has been introduced. The approach, is firstly based on the application of the HRV channel during daytime. This paper aims to describe the cloud detection method and to present the results of the comparison with the CLM-EUMETSAT product as well as to assess the impact of the new process in the fire detection method. ©2008 IEEE.