Improving the estimations of transect length and width for underwater visual surveys of targets on or near the seabed

Abstract

Demersal marine organisms may be surveyed visually using cameras deployed on a submersible vehicle, e.g. a remotely operated vehicle (ROV). For estimating animal densities along visual transects, the sampled area may be calculated from the products of transect distance and width. In this study, distance was measured relative to submerged pipeline features of known length using: 1) speed estimates from a Doppler velocity log (DVL); and 2) position estimates from an ultra-short baseline (USBL) acoustic system. DVL estimates of distance were within 0.8 to 1.0% of known feature lengths ranging from 6.01 to 1520.53 m, respectively. USBL estimates of distance were less accurate, ranging from 3.2 to 5.0% over the same distances, and positively biased compared to DVL estimates. Also, transect-width estimates were compared using: 1) calibrated laser metrics; and 2) measurement of camera orientation and optical properties. Both methods produced similar estimates of transect width, and thus area (r2 = 0.81), but width varied with camera altitude and orientation, and increased with seabed relief. Therefore, an assumption of constant transect width may bias estimates of sampled area, and estimates of animal density and abundance, particularly when surveying over seabeds with variable relief. To minimize these biases, practical methods are described to quantify transect distance and width throughout surveys using standard survey equipment and analytical methods.