Monitoring mangrove forest changes using remote sensing and GIS data With decision-tree learning

This paper presents a decision-tree method for identifying mangroves in the Pearl River Estuary using multi-temporal Landsat TM data and ancillary GIS data. Remote sensing can be used to obtain mangrove distribution information. However, serious confusion in mangrove classification using conventional methods can develop because some types of land cover (e.g., agricultural land and forests) have similar spectral behaviors and distribution features to mangroves. This paper develops a decision-tree learning method for integrating Landsat TM data and ancillary GIS data (e.g., DEM and proximity variables) to solve this problem. The analysis has demonstrated that this approach can produce superior mangrove classification results to using only imagery or ancillary data. Three temporal maps of mangroves in the Pearl River Estuary were obtained using this decision-tree method. Monitoring results indicated a rapid decline of mangrove forest area in recent decades because of intensified human activities.