RachelMelrose-108x132
Rachel Melrose
Role: 
PhD Candidate
Contact details:
Office: 

Off Campus

 

Mapping flooding and wetlands in arid Australia using satellite radar and optical remote sensing


Rivers and wetlands are under considerable threat around the world through human-induced climate change, unsustainable extraction, pollution, invasive species and overharvesting of resources. Such complex problems demand development of tools to improve wetland management and recognize their ecological and ecosystem service values. Understanding the complex inter-relationships between freshwater ecosystems and human uses necessitates analyses of water resources at large spatial and long temporal scales. Classification of satellite imagery provides a quantifiable tool for rapid and extensive mapping of water at these scales. Landsat optical imagery is routinely used to map wetlands across Australia, but cannot penetrate cloud cover or vegetation to detect inundation. Contrastingly, synthetic aperture radar (SAR) imagery overcomes these shortcomings, and has been used to map tropical and boreal wetlands, but remains largely untested for semi-arid and arid areas. I aimed to test the effectiveness of using L-band SAR to map flooding and arid wetland types in the Paroo and Warrego River floodplains within the Murray-Darling Basin, Australia. Data from the Japanese Advanced Land Observing Satellite (ALOS) Phase Arrayed L-band SAR (PALSAR) sensor, was provided through the Kyoto & Carbon (K&C) Initiative, with the objective to develop and validate SAR products to meet requirements of the RAMSAR convention on wetlands. I showed effective mapping of open water by thresholding a SAR L-HH image. Using two SAR images, I successfully developed inundation mapping methods in arid land cover types using segmentation and change detection to high accuracy. I used SAR data mining in a classification and regression tree analysis for differentiation of arid wetland types with a high accuracy based on extensive field validation. I then characterised the hydrology of the largest floodplain wetland in the system, Yantabulla, by obtaining historic river flow records (1993-2011) and mapping floodwater using a combination of ALOS PALSAR, cloud-free Landsat, and aerial surveys; supporting the use of interoperable data. Flood extent was measured as a percentage of the total wetland area per image, and modelled in relation to upstream flows. The demonstrated superiority and effectiveness of SAR in this study makes it an indispensable tool for rapid and extensive mapping of water in arid wetlands for ongoing monitoring and conservation.

SupervisorsProfessor Richard Kigsford and Professor Tony Milne