Coral reef change detection using Landsats 5 and 7:: A case study using carysfort reef in the Florida Keys

Satellite remote sensing is increasingly used to map and monitor coral reefs. Satellite data allow for the relatively quick assessment of benthic habitats and can be used for change-detection studies. Landsats 5 and 7, with the Thematic Mapper (TM) and Enhanced Thematic Mapper Plus (ETM+) sensors, respectively, provide the longest time series of satellite observations available for coastal researchers. From 1984 to the present, TM (for U.S.) and ETM+ (worldwide), images are abundant for change detection analysis over coral reefs, especially since the inception of the National Aeronautics and Space Administration (NASA)/United States Geological Survey (USGS) Long Term Acquisition Plan (LTAP). The LTAP was implemented to repeatedly and systematically gather Landsat 7/ETM+ data for specific remote sensing of niche communities, i.e., coral reefs. To properly use these data, a minimum set of corrections needs to be made: noise reduction correction, atmospheric correction, normalization, and ground-truthing. A time series was produced for Carysfort Reef in the Florida Keys from four Landsat 5 and one Landsat 7 images spanning 1984 to 1999. A Mahalanobis distance classification was undertaken to identify four benthic classes: "coral-dominated", "sand", "algae", and "substrate". The results were compared to an in situ database, which included transect and monitoring station data, as well as an aerial photograph. In situ data revealed that Carysfort Reef underwent an extreme loss in coral cover between 1983 and 1999, from 50% coral cover to 4% coral cover, representing a loss of 92% of coral cover. Over a similar time period, Landsat data estimated similar losses, from 33% coral cover to 7% coral cover, which represents a 79% loss in coral cover or "coral-dominated" area. Landsat also provided the progression and locations of "coral-dominated" loss for all of Carysfort Reef, which is not possible using in situ transect or monitoring station data. The statistics developed for classification can easily be used for future studies. Variation in sensor sensitivity was found between Landsat 5 and Landsat 7 imagery causing some spatial and quantitative differences between two images acquired in 1999. This study was performed using methods that have never been done before and proves the utility of satellite remote sensing for small-scale change detection studies. This study opens the possibility of a large-scale study, which could include an entire reef tract. This study also suggests that with the combination of other more expensive, high-resolution sensors, i.e. Ikonos and aerial photography, could allow for a multi-scale analysis and confirmation of Landsat findings. The use of remote sensing as a tool for the mapping and monitoring of coral reefs needs to be more greatly exploited.