2007 - Sussex Sea Fisheries District Committee (SFDC), Sussex coastline, towed underwater video footage for habitat classification dataset https://portal.medin.org.uk/portal/start.php?tpc=010_752620811acf337a9c52bc3cd63b36fe Management through spatial planning is seen as a tool to facilitate competing resource demands in the marine environment. The process of spatial resource management requires the availability of relevant seabed descriptions. Currently habitat descriptions at a scale relevant to the inshore zone are not available to inshore managers. The Sussex Sea Fisheries District Committee has been progressing towards a policy of marine spatial resource management and planning and in recent years has been developing a series of Geographical Information Systems data sets which support this objective. The Sussex District encompasses several aggregate extraction sites and the Committee are called upon to provide information in the aggregate extraction consenting process. Single beam acoustic ground discrimination systems (AGDS) based on standard echo sounders are used for commercial and research applications to describe the seabed. In contrast to multibeam and sidescan systems AGDS are relatively simple, low in cost and can be attached to the hull of any vessel with minimal operator supervision once installed. The Sussex Sea Fisheries Committeeâ??s Fishery Patrol Vessel (FPV) â??Watchfulâ?? is equipped with an AGDS (QTCâ?¢ View 5) manufactured by the Quester Tangent Corporation of Canada. The QTCâ?¢ system is linked to a Simradâ?¢ 50 Khz single beam echo sounder and output data is processed and logged to an onboard PC using proprietary software â??Impactâ?? develop by QTCâ?¢. Subject to suitable operating conditions the system is normally operated during routine enforcement patrols. In the summer of 2007 FPV â??Watchfulâ?? provided a survey platform to sample 270 inshore field sites. The field sites were selected to coincide with locations where acoustic data had been collected. A Bowtechâ?¢ underwater camera system was modified for towed use with a custom designed video sledge. The system was deployed along a transect at each of the field sites to collect c. 100 m of video data. In excess of 13 hours of video survey was interpreted by a marine biologist. This data set, which represents c. 3 km of seabed imagery, was categorised using course resolution benthic morphology descriptors which might be detected by the QTCâ?¢ View system and could also be relevant to the identification of essential fish habitats within the Sussex Sea Fisheries District. In addition the sites were categorised according to the EUNIS marine habitat classification system to ensure the comparability of the data with other regional and international data sets. All video footage was transferred from DAT and archived to DVD. Selected video and stills information is hosted on the Committeeâ??s website to maximise rapid and wide dissemination (see http://www.sussexsfc.gov.uk/project_malsf02.htm). The multivariate statistical package Primerâ?¢ (Clarke, 2001) was used to cluster the categorised data interpreted from the video survey into broad habitat classes providing a custom designed habitat classification scheme with a course level of descriptive resolution that emphasised morphological seabed features rather than community structure. The geographical positions from the video survey sites were each attributed a class based on this classification scheme. Half of the data for each habitat class was used in a supervised classification process to develop a habitat classification. This was undertaken in the image processing software Idrisiâ?¢. The images used in the classification were created from the vector based acoustic data in the gridding software Surferâ?¢ (by averaging the acoustic data within a 20 m search radius of each grid node with a node spacing of 20 m). After classification in Idrisiâ?¢ the data was then reconverted to vector format and QTC Clamsâ?¢ was used to interpolate the class values using a categorical interpolator to create a final habitat map for interpretation purposes. The second half of the data interpreted from the video survey was then used to test the validity of the final habitat classification. An error matrix was produced and accuracy statistics were calculated for each class in the habitat map and for the map overall. The accuracy of the resulting map was low (Tau Coefficient = 0.28); the cause of this low accuracy was investigated and a re-survey carried out. A repeat survey of a proportion of the field sites from the video survey was undertaken. Time constraints and weather conditions meant 96 of the original 270 field sites were re-surveyed; this provided sufficient data to re-evaluate the system. During the analysis of this new data set the field data clustering, supervised classification and accuracy assessment was repeated several times and each time the constituents of the classification scheme were altered to assess the ability of the QTCâ?¢ AGDS system to detect different morphological seabed features. The classification schemes developed were as follows: â?¢ Substrate Only â?¢ Substrate and Sediment Features â?¢ Substrate and Lifeforms â?¢ Substrate Sediment Features and Lifeforms. In addition a final classification was undertaken using the EUNIS classification scheme. The EUNIS classification scheme produced the lowest classification accuracy (Tau = 0.16), the substrate classification produced the highest (Tau = 30%). The addition of the sediment features descriptors reduced the overall accuracy and the addition of the Lifeforms descriptors produced a similar result to the Substrate based classification (Tau = 29%). Overall the accuracy of the system was found to be poor. It is suggested that the low accuracy is a result of the wide variety of environmental conditions under which the survey was conducted. The frequent occurrence of sloped bedrock, angular boulders and dunes, waves and ripples is likely to have confused the acoustic record by in some cases reflecting sound away from the transducer. The variety 6 of depths surveyed (0-60 m) is another potential source of error. The water off the Sussex coast is often sediment dominated and in deep water this can cause significant absorption if reflected sound leading to a broad variation in the acoustic signature for the same seabed type. The system could be more effective if a standard survey design is used in constant environmental conditions; furthermore it is likely to provide more accurate results over smaller areas of seabed with low levels of depth variation. The influence of the classification system design on the thematic accuracy of the final map classifications demonstrates the importance of using a system which incorporates morphological features which are more readily detected by acoustic systems. The video system produced the most useful information from the project. The database and interpretation of the video record provided information at a higher resolution than expected; this alone provides an extensive and valuable additional source of information on the distribution of habitats off the Sussex coast. Given the value in the video database obtained in this study and the Sea Fisheries Committeeâ??s capacity to collect this type of information, in future a good bathymetric model of the seabed combined with information on seabed geology may be a reliable proxy for determining seabed habitats. An interactive seabed map, with video extracts and full copies of this report are available. 20071231 20071231