036e3d1fffd3794369c674bf888dca49 English dataset NatureScot (HQ Inverness) data_supply@nature.scot pointOfContact 2025-06-03 MEDIN 3.1.2 EPSG Geodetic Parameter Registry 2021-01-06 revision OSGB 1936 / British National Grid + ODN height 2002 Scottish Natural Heritage (SNH) Isle of May biota broad-scale mapping survey 2003-12-03 publication GB-SCT-SNH-ME-000077-MRSNH00100000024-MAY NS-000077 NatureScot project ID The aim of this project was to survey the intertidal and subtidal areas of the cSAC and a surrounding 1km wide buffer zone, and to provide a comprehensive broad scale biotope distribution map of the area. Other objectives of the work were to review existing information relating to the intertidal and subtidal regions of the Isle of May and to compile the data and outputs into a Geographic Information System (GIS) project compatible with the SNH ArcView GIS. Scottish Natural Heritage (SNH) Scottish Natural Heritage (SNH), Headquaters Data Manager 01463 725000 data_supply@nature.scot owner NatureScot (HQ Inverness) NatureScot Data Manager data_supply@nature.scot custodian ERT (Scotland) Ltd Marine Environmental Consultancy dassh.enquiries@mba.ac.uk originator Esther Hughes Data Archive for Seabed Species and Habitats (DASSH) Data Manager 01752 633102 dassh.enquiries@mba.ac.uk distributor notPlanned Marine Environmental Data and Information Network MEDIN metadata record availability 2025-06-03 revision NDGO0005 NDGO0005 2025-06-03 revision Habitats and biotopes GEMET - INSPIRE themes, version 1.0 2021-09-28 revision Habitat characterisation Habitat extent Macroalgae and seagrass taxonomy-related counts Zoobenthos taxonomic abundance SeaDataNet Parameter Discovery Vocabulary 2021-09-28 revision otherRestrictions Not for navigational use; SNH copyright data which is available for re-use under government licence terms: http://www.nationalarchives.gov.uk/doc/open-government-licence/version/3/ otherRestrictions no restrictions to public access English 8-bit variable size UCS Transfer Format, based on ISO/IEC 10646 biota environment oceans -2.6268 -2.4716 56.1467 56.2159 SeaVox salt and freshwater body gazetteer 2021-01-06 creation North Sea Charting Progress 2 Sea Areas 2021-01-06 creation Northern North Sea 2002-09-09 2002-09-22 SeaVoX Vertical Co-ordinate Coverages 2021-01-06 revision infralittoral SeaVoX Vertical Co-ordinate Coverages 2021-01-06 revision littoral Documents Published report Database Marine Recorder Geographic Information System shapefiles dataset Commission Regulation (EU) No 1089/2010 of 23 November 2010 implementing Directive 2007/2/EC of the European Parliament and of the Council as regards interoperability of spatial data sets and services 2010-12-08 publication Ordnance Survey (OS) raster data for the Isle of May were provided by SNH, and were used to prepare maps for use in the field at a scale of 1:2,500. At this scale a total of six maps covered the Isle of May survey area. All maps, together with forms for recording biotope details, target notes and photography, were printed onto waterproof paper for use in field recording On return from survey, the identity of the few algal specimens taken was checked (though these were not retained), following which the marine biological records were inspected and updated. The biotope designations were then finalised using the current marine biotopes working classification (Version 97.06; Connor et al. 1997a). The neat maps were again checked with reference to field notes and photographs, and minor adjustments made to polygon boundaries if necessary. The neat maps were then scanned, imported into a Geographic Information System (GIS) and the hand-drawn polygons then digitised. For the final map representation, biotopes were coloured according to the MNCR scheme in Connor et al. (1997a) and also to the Life Forms format (Foster-Smith et al. 2000) for comparison. Note that the current MNCR colour scheme (based on the exposure grade as indicated by the higher code prefix to the biotope code) does not cater for lichen or algal crust biotopes. Such biotopes come under the higher code Littoral Rock, which does not incorporate an exposure label. Therefore in the maps coloured according to the MNCR scheme, the Life Form colours have been applied to these biotopes. The AGDS data were processed according to the JNCC Procedural Guidelines (Foster-Smith et al. 2001). Water depths were corrected to the nearest ten minutes using the nearby port of Anstruther in Fife as the reference port. Data were checked for positional and depth spikes, and any dubious data removed. Data collected when the vessel was drifting have also been removed. The key parameters recorded and available for analysis were as follows: 1. High frequency transducer (200kHz): a. E1 (roughness)* b. E2 (hardness)* c. Depth d. Variability of the above 2. Low frequency transducer (38kHz) a. E1 (roughness)* b. E2 (hardness)* c. Depth* d. Variability of E1* e. Variability of other parameters *Parameters used further in Supervised classification. Variability indices (standard deviation of five consecutive values) were calculated for E1, E2 and depth for both high and low frequency data sets. Interpolation: transforms point data to a continuous surface by calculating new values for a grid of positions covering a rectangular area that encompasses the track point data. The reasons for interpolation are, firstly, to produce a coverage that is easier to view as a map than point data and, secondly, to enable raster-based image processing techniques to be applied to the data. The edited track data were interpolated in Surfer™ using the following parameters: • a grid spacing of 10m; • inverse distance algorithm with a weighting toward the grid centroid of 1.5; • a search and display radius of 500m (to ensure there were no gaps in the coverages); • a four-sector search with a maximum of 32 values per sector; • a smoothing coefficient of 50. Interpolation also smoothes data. AGDS track data are inherently variable, even over a seabed of uniform sediment, with the data values for E1 and E2 scattered around a mean value. Interpolation can be used to create new values at the grid nodes that are a distance-weighted mean value of the real track data within a set search radius around the node. This results in a more stable surface of E1 and E2 values that is more amenable to analysis, compared to the raw data. The parameters selected and listed above were chosen to ensure that the interpolated values were dependent upon real data from at least two adjacent tracks and with a light weighting towards the data closest to the node. Supervised classification: The raster grid images were imported into Idrisi™ for classification. Supervised classification using the maximum likelihood classifier is generally regarded as the most satisfactory means of interpreting multispectral data, and the different acoustic variables have been considered as analogous to electromagnetic data from satellite or airborne sensors. The maximum likelihood classification process also allows Bayesean prior probabilities to be included based on rules and this feature has been used to improve the classification as detailed below. Not all grid images were used for classification. The low frequency depth data were considered more stable than the high frequency depth data (the former required significantly fewer records to be removed during editing and were less variable than the latter). Also, the along-track variability index of the high frequency data was similar to that for the low frequency data, but with less obvious spatial pattern. Images derived from six acoustic variables were used: high frequency E1, high frequency E2, low frequency E1, low frequency E2, low frequency E1 variability, low frequency depth. The video data were categorised to biotope or, if this fine level of detail was not possible, to higher habitat complex level. A buffer zone of 25m was created around each of the video samples which were then used as ‘training’ sites to create the acoustic signatures from the six variables used (as above). The signatures were then applied using the maximum likelihood classifier. Because of the difficulty of sampling within the kelp zone close to the island, the signature developed for kelp was based on scant data. To rectify this deficiency, a constraint was placed on the classification process such that kelp biotopes were given a higher prior probability of occurrence at depths shallower than 15m than other biotopes. This was justified in that no other biotopes were observed in shallow water. Supervised classification does not work particularly well when there is just a single ground truth record for a particular biotope, because there are insufficient data from which to derive a statistically satisfactory acoustic signature. In these circumstances it is usual to amalgamate the record into a broader biotope category. This may result in an apparent mismatch between the number of biotopes identified from ground truth data and the number classified from the acoustic data. The process was repeated for the habitat classes without the need to constrain or enhance any individual class. Every pixel in the image is matched to the signature and a probability assigned to a pixel of it belonging to each class. Each pixel is then coded to show the class that has scored the highest probability. Note that signatures are statistical and there is the chance that a pixel within a training site will not matching the correct class, particularly where there are large numbers of ground truth samples (as is the case in this survey). Thus, some idea of the success of a classification can be obtained through measures of agreement. A measure of the agreement can be calculated by overlaying the ground truth and classified images and matching predicted to actual on a pixel-by-pixel basis (the ‘error matrix’). The percentage agreement for the habitat and biotope classifications was 82 and 77% respectively, with a Kappa agreement (probability of agreement above chance) of 0.79 and 0.74 respectively. This level of agreement is typical of remote surveys, where perfect agreement is never expected due to a combination of positional error and variability in the acoustic reflectance data. Overall, the classification can be considered successful, particularly in demonstrating broad spatial patterns in habitat and biotope distribution. Analysis of the video data: Video data were recorded using an Access™ database. The video tapes were replayed and at least one frame grab was captured for each sample. The frame grabs were stored in a sub-table linked to the master table. Notes were taken in a systematic way recording main habitat features, conspicuous species or, where this detail was not possible, general growth forms and higher taxonomic categories. The biota were also scored from 1 to 5 to show their relative abundance. The species were recorded in a sub-table linked to the master table so that site/species tables could easily be extracted. The notes were first made on paper and then the tapes were replayed and the notes entered into a database, but with entries for biotope codes and habitat categories left blank. The final stage consisted of reviewing the database, frame grabs and if necessary the tapes, in order to tag the records with biotope codes (according to Connor et al. 1997b) and habitat classes. This staged process to biotope recording and the use of look-up tables for codes and terminology was designed to promote consistency in classification of the records. Sidescan data: The raw sidescan sonar data were processed using EoMap™ and ERDAS Imagine™ software. EoMap was used to strip the positional information from the raw data files, and the sidescan data were then slant range-corrected. This process removes the central strip of water column data and draws the seabed images either side of the centreline together, whilst maintaining positional correctness for data at the outer edges of the images. The data swaths were then imported into ERDAS Imagine, in order to display and geo-reference the images for subsequent viewing in ArcGIS.