Spectral Constrasts Associated With Forest Types In Tropical Areas As Seen On AVHRR Data

Coarse resolution satellite data are increasingly used for monitoring the tropical forest cover at regional to global scales. Given the nature of the data, new approaches have to be devised for classifying forest types. They rely mainly upon three types of discriminants: those based upon spectral constrasts, temporal changes and, to a lesser degree textural patterns. This paper examines the problem of identifying and characterizing spectral contrasts between the evergreen rain forest, the seasonal mixed decidous forest and various forms of secondary or degraded formations in various parts of Southeast Asia. For that purpose, time series of NOAA AVHRR Local Area Coverage (LAC) and Hight Resolution Picture Transmission (HRPT) of the 1989-1990 period have been processed. The analysis is based mainly upon radiometric transects in the five channels (visible, near-infrared and thermal) of the AVHRR across neighbouring vegetation cover types. The problem of establishing a typology of forest-non forest interfaces as seen at the low AVHRR resolution is also discussed and illustrated using a series of examples. The steps of the analysis described in the paper are considered essential to carry out a systematic monitoring of the tropical forest cover over large areas using the AVHRR instrument.