Understanding the relationships between the size of mangrove vegetation features and the optimum image pixel size required to map these features is essential to support effective mapping and monitoring activities in this environment. Currently mangroves are under pressure from anthropogenic and natural disturbances, and up-to-date and accurate spatial information is required to support their management. Addressing ecological problems at the correct spatial scale is essential in mangrove environments. There is a lack of knowledge on the types and biophysical properties of mangroves, which can be mapped at different image spatial resolutions. This thesis integrated the spatial and temporal dimension of remote sensing data into a spatio-temporal continuum of mangrove ecology and developed guidelines for multi-scale image-based mangrove mapping. Three objectives were addressed to achieve the aim: (1) characterising mangrove spatial structure to produce an optimum pixel resolution scheme for image-based mangrove mapping; (2) assessing the applicability of the scheme to selected images for mangrove composition and leaf area index (LAI) mapping; and (3) developing guidelines for multi-scale mangrove mapping. The research sites were located in Moreton Bay, Australia, and Karimunjawa Island, Indonesia. Landsat TM, ALOS AVNIR-2, and WorldView-2 images were used for both sites; with additional LiDAR data and a very high-spatial resolution aerial photograph for Moreton Bay.
After two introductory chapters, chapter three focused on the development of a method for estimating the optimum pixel size to map different sizes of mangrove features accurately. The extent of dominant mangrove structural features including tree/shrub crowns, canopy gaps and vegetation formation or community, could be detected using semi-variogram analysis applied to image datasets with different spatial resolutions. The findings showed a gradual loss of mangrove information detail with increasing pixel size. Specific mangrove features could be optimally mapped from a specific pixel size and spectral bands or indices. A pixel size of ≤ 2 m was suitable for mapping canopy and inter-canopy-related features within mangrove vegetation features (such as shrub crown, canopy gaps and single tree crowns), while a pixel size of ≥ 4 m was appropriate for mapping mangrove vegetation formation, communities and larger mangrove features. An optimum pixel resolution scheme was produced for mangrove mapping that served as a basis for an inversion approach to map mangrove features using remote sensing image datasets.
Chapters 4 and 5 focused on the application of the optimum pixel size scheme to the selected images with different spatial resolutions, to map mangrove composition and LAI, respectively. Object-based image analysis successfully produced mangrove composition maps at discrete spatial scales. The findings suggested that the accuracy of the maps was a result of the interaction between the image spatial resolution, the scale of the targeted objects and the number of land cover classes on the map. This task confirmed that the conceptual spatial and temporal hierarchical organisation of mangroves provided an essential aid for effective multi-scale mangrove composition mapping. For LAI mapping, the effect of different image pixel sizes and mapping approaches (i.e. object- and pixel-based) to estimate LAI was investigated. The results suggested that the optimum pixel size to estimate LAI correlated with the dominant object size in the area of interest and the field plot size. The object-based approach significantly increased the LAI accuracy as opposed to the pixel-based approach; with the optimum segmentation size corresponding to the size of the dominant objects in the image scene.
Chapter six synthesised the findings from chapters 3, 4, and 5 and developed guidelines for multi-scale image-based mangrove mapping. Through these guidelines, the relationships between remote sensing and mangrove ecology could be shown explicitly; and at the same time, they provided an effective and efficient way to select the best image datasets and mapping techniques to map mangrove feature(s) at a relevant spatial and temporal scale. These targeted mangrove features can be used as a basic mapping unit for other applications, such as LAI and biomass estimation, carbon storage calculation, species distribution and so on.
This thesis has successfully integrated the field of remote sensing with mangrove ecology and developed rigorous and robust guidelines that provide a fundamental basis for multi-scale image-based mangrove mapping. It also signifies the operational use of remote sensing data for multi-scale mangrove mapping to produce science- and management-ready environmental information at relevant spatial and temporal scales. In a practical context, this guideline will help mangrove scientists and managers select the appropriate image datasets for mapping, measuring and monitoring mangrove environments. To ensure the wider applicability of the guidelines, the methods presented in this thesis need to be tested at other mangrove sites with different environmental settings, using a wider range of image datasets and processing techniques.