Classifying the Seagrass Zostera Marina L. from Underwater Video: An Assessment of Sampling Variation

摘要

Aquatic vegetation perform vital functions in coastal ecosystems. Large-scale loss or decline of this critical resource, particularly near urbanizing estuaries, has been documented throughout the world. In 2000, the Washington State Department of Natural Resources (WDNR) initiated a long-term monitoring effort to track changes in the abundance and depth distribution of the seagrass Zostera marina L. because this marine plant provides valuable habitat in the Puget Sound and is known to be a sensitive indicator of ecosystem health. Data is acquired remotely using an underwater video sampling technique. Appropriate interpretation of monitoring results requires an understanding of sampling variation. This study investigated the importance of intra- and inter-observer classification variation in the estimates of Z. marina cover from underwater video images. Intra-observer coefficient of variation (CV) ranged from 0.4% to 8.9% and inter-observer CV ranged from 1.4% to 22.2%. Examination for conditions associated with areas of low observer agreement found where Z. marina was extremely sparse and/or patchy, there was higher variability in classifications. An analysis of variance and the fractional components showed no significant difference between the Z. marina estimates by observer (p > 0.05) and CV associated with classifying single video image segments of 11%. Our results suggest that while contribution from video processing can vary widely across transects and sites, video processing error makes up a relatively minor component of overall error in site level Z. marina cover estimates.