Application of AVHRR data and mixture modelling to estimate forest and range burning in West Africa: The Gambia.

Abstract

Burning in savannas significantly impacts the physical environment and has recently been recognized for the significant role it could play in climate change by releasing large amounts of greenhouse gases (Menaut et al., 1991). This is particularly true for West African savannas which cover approximately 3 x 10⁶ km², one of the largest contiguous extents of savannas in the world. Recently it has been suggested that 2.4 - 4.2 gt C (gigatons of carbon) are released by savanna burning annually (Hall and Scurlock, 1990). Even with this impact our knowledge of tropical vegetation systems is limited (Sadar et aID, 1990). One method of monitoring vegetation over large regions is using satellite remote sensing. The Advanced Very High Resolution Radiometer (AVHRR) is well suited for the task due to its high temporal resolution (daily), large area coverage (2700 km swath width) and spectral channel availability (nearinfrared band useful for monitoring vegetation). In this study AVHRR data for four burn seasons were analyzed to determine the origin, spread and extent of burning for The Gambia. The accuracy of the assessment was verified using coincident higher spatial resolution (30 m) Thematic Mapper (TM) data and field assessments. The burn extents calculated using a traditionally classified AVHRR image were within 11% of those calculated by the TM. Further a linear mixture model was applied to the AVHRR imagery to determine if burn estimates could be improved. It was found that by thresholding mixture model results, derived using AVHRR bands 1, 2 and 3, acceptable burn estimates were produced. The mixture model estimates were consistently higher than the TM estimations, particularly for agricultural areas.