The application of absolute reflection thresholds in active fire monitoring

Forty-seven NOAA AVHRR scenes covering parts of West Africa are examined and active fires are classified using the contextual algorithm for three sub-windows. One sub-window covers the grass- and shrublands of southern Mauritania. The transition zone from grasslands to savannas in Eastern Senegal is covered in the second sub-window, while the woody savannas of northern Guinea are covered by the last sub-window. In the contextual algorithm active fires are classified based on a combination of emission and reflection properties. The application of absolute channel 2 reflectance thresholds was originally included in the algorithm in order to neglect pixels with a high reflection, which could cause an artificially high response of the sensor channel 3. However, by neglecting pixels with a high channel 2 reflection, also areas with early fires in still green surrounding are neglected due to the high reflection of near-infrared radiation from green vegetation. It is found that the application of channel 2 thresholds seriously limits the amounts of fires in the central sub-window in the beginning of the early dry season. This area is well known to be an area of intense burning activity early in the dry season. Thus, approximately 40 percent of the detected active fires are neglected if a channel 2 threshold is applied. This pattern is not repeated for the northern and southern sub-windows, since channel 2 reflections are constantly too low or too high to influence the result.