Mapping and Monitoring Mangrove Forest Baselines Across the Globe

Authors Organisations
Type

Student thesis: Doctoral ThesisDoctor of Philosophy

Original languageEnglish
Awarding Institution
Supervisors/Advisors
Award date10 Mar 2016
Links
Show download statistics
View graph of relations

Abstract

Mangrove forests are woody vegetation located amongst the coastal wetlands of the tropics that hold importance for local populations, carbon sequestration and biodiversity. Despite this, mangrove forests are in decline as a consequence of both direct and indirect anthropogenic activities, such as aquaculture practices,coastal development and climate change. Through the interpretation of time-series colour composite radar imagery, mangrove forests were observed to have been anthropogenically disturbed and to have experienced changes in extent caused by natural processes. This revealed that as much as 40% of the world’s mangrove forests occur within a region at risk of further loss and degradation.In light of this observation, this study has developed a method for automatically mapping and monitoring mangrove extent using time-series Japanese (JAXA) ALOS PALSAR (2007-2010) and JERS-1 (1996) radar imagery at a variety of locations across the tropics. Random Forests was used to classify mangrove forest extent for the year 2010 at 16 study sites across the tropics, representing a broad range of mangrove habitats and forest types. Existing mangrove extent maps were refined using a Bayesian Maximum Likelihood approach and were used to generate training data for the Random Forests algorithm. Changes in mangrove forest extent were detected using a novel map-to-image approach developed in this study. The technique detected change in mangrove extent in an automated fashion for the period 2007-2010 using ALOS PALSAR imagery and 1996-2010 using JERS-1 imagery. An area in excess of 2.5 million ha of mangrove forest extent was classified with the baseline and changes in extent mapped with an accuracy >90%. Limitations pertaining to image registration, classification error and the size of the minimum mapping unit were identified as sources of error in the baseline and change detection mapping. The results of this work are applicable at the global level and can be scaled to update the existing map of global mangrove forest extent and implement a mangrove monitoring system.