In the second part of the MODIS algorithm, it is critical to determine valid neighboring pixels for every potential fire pixel, which will be used to derive the background parameters designed to set the remaining dynamic thresholds. The separation of fire pixels and non-fire background pixels becomes ambiguous with increasing background temperature caused by the presence of undetected background fires, seasonal change and certain surface types. This can directly affect the performance of the contextual algorithms. Giglio et al. (1999) excluded the eight pixels surrounding the potential fire pixel from the processing window in order to take out the fire contaminated background pixels. This algorithm showed a higher sensitivity to small, cool fires compared with the algorithms of Justice et al. (1996) and Flasse and Ceccato (1996).