A post-classifier for mangrove mapping using ecological data

Abstract

A global decline in tropical mangrove forests is one of the most serious problems of the world's coastal ecosystems. This problem results in an increasing demand of detailed mangrove maps at the species level for monitoring mangrove ecosystems and their diversity. Consequently, this research is the first to investigate the unexplored potential of exploiting mangrove–environment relationships for improving the quality of the final mangrove map at the species level. The relationships between mangroves and the surrounding environmental gradient were incorporated into the mapping process via a typical Bayesian probability model. The Bayesian model functioned as a post-classifier to improve the quality of an already-produced mangrove map. The environmental gradient in use was a GIS layer of soil pH data. Despite the remaining confusion between R. mucronata and S. caseolaris, it was found that the investment of the integration of soil pH into the mapping process paid off as it significantly increased the mapping accuracy from 76.04% to 88.21%. It is anticipated that the methodology presented in this study may be used as a guideline for producing mangrove maps at a finer level.