Optimized rule-based logistic model tree algorithm for mapping mangrove species using ALOS PALSAR imagery and GIS in the tropical region

Abstract

The main objective of this study is to map the spatial distribution of mangrove species and assess their changes from 2010 to 2015 in Hai Phong City of Vietnam located on the tropical region using the ALOS PALSAR data and an optimized rule-based decision tree technique. For this purpose, ALOS PALSAR imagery for the above period and GIS data were collected and used, and then, spatial distributions of mangrove species were derived using logistic model tree (LMT) classifier. The LMT is current state-of-the-art machine learning method that has not been used for mapping of mangrove species. The results showed that incorporation of ALOS PALSAR imagery and GIS in the LMT algorithm provides satisfactory overall accuracy and kappa coefficient. The ALOS-2 PALSAR for 2015 achieved better overall accuracy, with an increment of 3.6% in mapping mangrove species than that of the ALOS PALSAR for 2010. The ALOS-2 PALSAR-derived model yielded the overall accuracy of 83.8% and the kappa coefficient of 0.81, compared with those of the ALOS PALSAR-derived model, 80.2% and 0.78, respectively. The results of classification for 2010 and 2015 were significantly different using the McNemar test. This research demonstrates the potential use of ALOS PALSAR together with machine learning techniques in monitoring mangrove species in tropical areas.