Quantitative comparison of change-detection algorithms for monitoring eelgrass from remotely sensed data

The eelgrass (Zostera marina L.) population in Great Bay, New Hampshire has recently undergone dramatic changes. A reoccurrence of the 1930s wasting disease and decreasing water quality due to pollution led to a reduction in the eelgrass population during the late 1980s. Currently, the eelgrass populations in Great Bay have experienced a remarkable recovery from the decline in the late 1980s. Eelgrass is important in our estuarine ecosystems because it is utilized as habitat by many commercial and non-commercial organisms and is a food source for waterfowl. In order to monitor the eelgrass populations in Great Bay, a change detection analysis was performed to determine the fluctuation in eelgrass meadows over time. Change defection is a technique used to determine the change between two or more time periods of a particular object of study. Change detection is an important process in monitoring and managing natural resources and urban development because it provides quantitative analysis of the spatial distribution in the population of interest. A large number of change-detection techniques have been developed, but little has been done to quantitatively assess the accuracies of these techniques. In this study, post-classification, image differencing, and principal components change-detection techniques were used to determine the change in eelgrass meadows with Landsat Thematic Mapper (TM) data. Low altitude (1,000 m), oblique aerial photography combined with boat surveys were used as reference data. A proposed change-defection error matrix was used to quantitatively assess the accuracy of each change-detection technique. The three differ ent techniques were then compared using standard accuracy assessment procedures. The image differencing change-detection technique performed significantly better than the post-classification and principal components analysis. The overall accuracy of the image differencing change detection was 66 percent with a Khat value of 0.43. This study provided an application of Landsat Thematic Mapper to detect submerged aquatic vegetation and the methodology for comparing change detection techniques using a proposed change detection error matrix and standard accuracy assessment procedures. In addition, this study showed that image differencing was better than the post-classification or principal components techniques for detecting changes in submerged aquatic vegetation.