Long term global archives of moderate spatial resolution satellite imagery are now freely and readily available (i.e. Landsat archive). These data sets are not being fully utilised by science or management communities, particularly in the coastal environment and in Australia, due to a lack of appropriate methods to consistently and accurately produce time-series of environmental information. The future use of these image data sets also relies on the science and management communities becoming aware of the types of information and products time-series data sets can deliver. This thesis addresses these issues, by mapping and investigating change in land cover and seagrass cover in South East Queensland (SEQ), Australia, from 1972-2010, using a Landsat image time-series.
Chapter two explains and demonstrates the development of an object-based method to map both land cover and seagrass cover. Landsat MSS (60 m resolution, 18-day repeat, 1972-1987), TM and ETM+ (30 m resolution, 16-day repeat, 1988-2010) data were used without in-situ calibration data, to produce a land cover and seagrass cover map for every year data were available, producing over 60 individual annual map products across the 38 year archive. The land cover classes included: closed canopy, open canopy and sparse vegetation, grassland, bare ground, agriculture and urban; seagrass cover classes included low, medium and high horizontal-projected-foliage-cover, sand, deep and turbid water. Significant attention is given to the thematic accuracy assessment, since both time-series mapping and object-based mapping present new challenges that are not accounted for in traditional pixel-based thematic accuracy assessment. Multiple measures of accuracy are discussed, however, standard overall accuracy figures were still given, with 80% and 65% for land and seagrass cover respectively, which is consistent with previous studies in the area.
Chapter three focuses on the long term spatial and temporal dynamics of land cover change in SEQ, and shows for the first time, a consistent, long term representation of land cover evolution in SEQ. A range of spatially explicit time-series analysis methods were used, including summary maps, variability measures and trajectory analysis. The time-series analysis methods enabled the investigation of a range of land cover processes and dynamics in South East Queensland. The spatial and temporal dynamics of urban development and associated clearing were investigated, and vegetation dynamics were investigated at a broad scale and through local case studies of forestry/plantation practices, bush fire impacts, mining clearing/revegetation and agricultural practices. The results demonstrate the advantages of the thematic time-series mapping approach over traditional image-pair mapping approaches; as well as the advantage of using thematic time-series, as opposed to time-series of vegetation indices or other continuous image based variables. The potential advantages include: the ability to analyse multiple land cover classes simultaneously, without bias toward vegetation; a reduced influence of chosen start/end points and specific timing of event within time-series; a reduced influence of chosen start/end points and specific timing of event within time-series; a reduced temporal resolution requirement compared to some index-based methods; and potential for an easier interpretation of map products by resource managers.
Chapter four focuses on the long term spatial and temporal dynamics of seagrass change in clear shallow water of the Eastern Banks, Moreton Bay. A monthly time-series from 2008 to 2010 was also used to examine seagrass change at both the inter- and intra-annual time scale. This showed for the first time a long term, high temporal resolution record of the broad scale spatial distribution of seagrass on the Eastern Banks and one of the first globally at this spatial and temporal scale. A range of trend and time-series analysis methods are demonstrated for a time-series of annual maps from 1988-2010 and a time-series of monthly maps during 2008-2010. Significant new insight was presented regarding the inter- and intra-annual dynamics of seagrass persistence over time, seagrass cover level variability, seagrass cover level trajectory, and change in area of seagrass and cover levels over time. Overall we found that there was no significant decline in total seagrass area on the Eastern Banks, but there was a significant decline in seagrass cover level condition. A case study of two smaller communities within the Eastern Banks that experienced a decline in both overall seagrass area and condition are examined in detail, highlighting possible differences in environmental and process drivers. We demonstrate how trend and time-series analysis enabled seagrass distribution to be appropriately assessed in context of its spatial and temporal history and provides the ability to not only quantify change, but also describe the type of change. We also demonstrate the potential use of time-series analysis products to investigate seagrass growth and decline as well as the processes that drive it. This study demonstrates clear benefits over traditional seagrass mapping and monitoring approaches, and provides a proof of concept for the use of trend and time-series analysis of remotely sensed seagrass products to benefit current endeavours in seagrass ecology.
This thesis presents and validates a new mapping approach as a “proof of concept” for operational long term, time-series thematic mapping of coastal environments. The approach presented will enable new applications of the progressively growing global and regional long term archives of earth observation data sets for both science and management applications. This work also provided a fundamental data set for use in south-east Queensland to understand interactions between the terrestrial and aquatic coastal environment, since that is not explicitly addressed in this work. New insights were provided by this work into both land cover and seagrass cover change in SEQ, as this is the first time their broad scale spatial distribution has been investigated with high spatial and temporal resolution.