Author: Bassam Waheed
Waheed, Bassam, 2025 Mangrove Biomass from Space: A South Australian case study using optical, RADAR and LiDAR datasets., Flinders University, College of Science and Engineering
Terms of Use: This electronic version is (or will be) made publicly available by Flinders University in accordance with its open access policy for student theses. Copyright in this thesis remains with the author. You may use this material for uses permitted under the Copyright Act 1968. If you are the owner of any included third party copyright material and/or you believe that any material has been made available without permission of the copyright owner please contact copyright@flinders.edu.au with the details.
ABSTRACT
Mapping and quantitative assessment of above ground biomass (AGB) of mangroves may be possible using multiple remote sensing datasets such as optical, RADAR and LiDAR supported by fieldwork. This study assessed South Australian (SA) mangroves (Avicennia marina) in the St Kilda area near Salisbury. Literature asserts that mangroves absorb four times more carbon than normal forest; thus estimation of AGB is important for planning and decision making in coastal zones. The main objective of this research was to assess the capacity of satellite data to study mangrove extent and AGB in St Kilda and to build a technical framework to integrate multiple remote sensing datasets. A specific research question was what are the optimum methods to study AGB using optical, RADAR and LiDAR? Additionally, does a data fusion approach assist quantise AGB? This comprehensive study of mangroves was undertaken using optical satellites (Pléiades Neo, and Sentinel-2) and RADAR (ALOS-1 PALSAR, ALOS-2 PALSAR, and Sentinel-1) and demonstrated that AGB estimation is possible by identifying correlation between remote sensing variables and biomass determined from field observations using statistical regression. Linear regression for LiDARCHM-mean and mean heights from field work, showed a high correlation between these two variables with a R2 value of 0.87. The average value for AGB estimation from field data was 59.7 tons h-1. An optimum regression model from LiDARCHM, with a R2 value of 0.74, yielded a total biomass of 96,149 tons. Regression models from optical satellite data, Pléiades Neo, provided the R2 value of 0.48 using NDVI and 0.27 using principal component analysis (PC-2). In contrast, Sentinel-2 gave R2 value of 0.40 using red-edge NDVI (NDVIre) and 0.48 using the mangrove vegetation index (MVI). The final AGB estimation from optical imagery yielded a value of 69,024 tons with an optimum model. In assessing mangrove extent, two classification approaches were used for Pléiades Neo; random tree (RT) and ISO cluster. RT, a variant of Random Forests, application resulted in the extent of mangroves (13.09 km2 & 15.78 km2) for pan-sharpened and multispectral respectively and similarly, ISO cluster computed (14.35 km2 & 14.99 km2) for both sets of imagery. The RT model achieved a user accuracy of 0.98 with Kappa value of 0.86 for Pléiades Neo, using a combination of six spectral bands, vegetation index (NDVI) and a CHM. Three RADAR datasets were used to calculate AGB estimation using backscatter coefficients, RADAR indices, GLCM matrix, PCA techniques and fieldmean biomass by applying statistical regression. The regression results showed R2 values for fieldmean biomass against backscatter (HV=0.18, VH=0.14, VH=0.12), PCA (0.27, 0.09, 0.16), GLCM (0.23, 0.15, 0.17) and RVI (0.29, 0.09, 0.12) for ALOS-1 PALSAR, ALOS-2 PALSAR, and Sentinel-1 satellites respectively. Additionally, a pixel based fusion approach for ALOS-2 PALSAR and Sentinel-1 resulted in R2 between PC-2 and fieldmean biomass values of 0.27 at 4m. The satellite remote sensing approach made this study non-intrusive, potentially repeatable and regular, and showed excellent capability to map mangrove extent, especially when multiple datasets were used.
Keywords: Above Ground Biomass, AGB, Machine Learning, Biomass, Mangrove Biomass, Decompositoin
Subject: Environmental Science thesis
Thesis type: Masters
Completed: 2025
School: College of Science and Engineering
Supervisor: David Bruce