Submerged habitats from space: increasing map production capacity with new methods and software.

Abstract

Thematic maps of submerged coastal habitats are essential to ecological studies and marine spatial planning processes. The analysis of multispectral satellite imagery provides a cost effective method for mapping shallow habitats. However, the interaction of light with the water column greatly complicates the production of maps from imagery. A number of water column correction (WCC) methods have been devised to address this complexity, but the remote sensing literature in which these methods are described tend to focus on novelty rather than practical implementation. Consequently, methods are often tailored to specific environmental conditions and described in very technical terms, source code for implementation is rarely provided, and potential users are typically unable to understand or make use of existing WCC methods without expert assistance. This thesis seeks to address this situation by developing new methods, suitable for the widest possible range of environmental conditions, and releasing free and open source software for their implementation. The methods and software presented address the entire process from ground truth data collection and processing with the free and open source software (FOSS) Benthic Photo Survey (BPS), through to depth estimation, image processing, water column correction, and accuracy assessment with the FOSS OpticalRS Python library. BPS expedites and greatly simplifies the processing of field data. The novel empirical depth estimation method presented provides better accuracy than a widely used physics based method and requires fewer image processing steps. The parameter estimation and WCC methods presented improve on previous methods in terms of accuracy, simplicity, and flexibility. Combined with pre-existing FOSS image classification software, these methods and tools form the Marine Optical Remote sEnsing Map and Assessment Production System (MORE-MAPS). To demonstrate the efficacy of MORE-MAPS, broad scale submerged habitats in and around the temperate waters of Cape Rodney to Okakari Point Marine Reserve in northeastern New Zealand were mapped to a depth of 20m with an overall accuracy of 83%, and the results were compared with previous habitat maps of the reserve that were based largely on direct observation and manual photo interpretation methods, and with optical remote sensing studies from temperate regions outside of New Zealand. MORE-MAPS was found to provide mapping accuracy comparable to previous optical remote sensing studies, and to provide better accuracy, greater geographic coverage, and lower cost than the previous mapping efforts within the reserve. Overall, the methods and tools developed in this thesis demonstrate potential to greatly reduce the cost and complexity of habitat mapping and, thereby, increase the capacity for map production in New Zealand and throughout temperate, subtropical, and tropical coastal regions worldwide.