You are here:
AUTOMATED PRODUCTION OF SEAGRASS MAPS FROM SIDESCAN SONAR IMAGERY: ACCURACY, VARIABILITY AND PATCH RESOLUTION
Citation:
DEWITT, T. H., P. LATTIN, P. J. CLINTON, H. A. STECHER, III, C. FOLGER, AND S. STRICKLAND. AUTOMATED PRODUCTION OF SEAGRASS MAPS FROM SIDESCAN SONAR IMAGERY: ACCURACY, VARIABILITY AND PATCH RESOLUTION. Presented at Estuarine Research Federation 2007, Providence, RI, November 04 - 08, 2007.
Impact/Purpose:
research presentation
Description:
Maps of seagrass beds are useful for monitoring estuarine condition, managing habitats, and modeling estuarine processes. We recently developed inexpensive methods for collecting and classifying sidescan sonar (SSS) imagery for seagrass presence in turbid waters as shallow as 1-m. We used a supervised image classification approach based on a maximum likelihood classification algorithm. Variances and covariances of signatures derived from local SSS image texture statistics and depth class were used to classify images. Underwater video data were used to train the image classification and to test the thematic accuracy of the resulting seagrass maps. Replicate surveys were conducted over study sites in Yaquina R. estuary (OR) to determine the among-survey variability in the estimation of seagrass area, spatial distribution, and thematic accuracy. Ground-based surveys to measure patch size-distribution and shoot density were used to determine the minimum patch size and shoot density that could be resolved with the SSS-based seagrass maps. Spatial accuracy of the seagrass maps was determined by comparing the measured and imaged positions of artificial sonar reflectors. The results of these assessments of the variability, accuracy , and effective resolution of SSS-based seagrass maps provided the foundat
