Scottish Marine and Freshwater Science Volume 5 Number 10: Updating Fisheries Sensitivity Maps in British Waters

4 Discussion and Final Considerations

This work presents a series of 0 group aggregation maps for thirteen commercial species: cod, haddock, whiting, Norway pout, Herring, Mackerel, Horse Mackerel, Sprat, blue whiting, place, sole, hake and anglerfish. Three of these species, horse mackerel, hake and anglerfish, were not present in the previous nursery maps published in Coull et al. (1998).

For hake and anglerfish, due to insufficient data, the maps presented here show the probability of presence of 0 group fish and not the probability of presence of 0 group aggregations, this representation could be improved in the future. For saithe and ling there were not enough hauls represented in the survey datasets with 0 group fish present so that aggregations of these could be identified. As the juveniles of both these species stay in their inshore nursery habitats until they are 2-3 years of age (Heessen et al. 2006; Rowley, 2008) nursery areas could in the future be modelled using data of age 1 fish.

Much consideration was given to approaches for revising spawning areas. Egg and larvae surveys normally have a sampling design that follows a large grid, with sampling stations being averaged to the centre of very large cells - generally 15 x 15 NM, half of an ICES statistical rectangle ( ICES, 2010b; ICES, 2012b). Data with such a coarse resolution cannot be used for species distribution modelling because, as explained in Section 2.3, this approach relies on the close relationship between the presence/absence data and the environmental layers at any given location. For this reason, it was only possible to apply the species distribution modelling methodology to herring larvae, as the IHLS sampling design provides for specific sampling locations.

Even so, the performances of both MAXENT and Random Forest models of herring larvae were evaluated as poor by the kappa statistic, see Table 9. Furthermore, environmental variables like seabed sediments and distance to gravel, which were expected to be especially relevant for defining herring spawning grounds (Parrish et al., 1959; Bowers, 1969; Holliday (1958) cited in Rankine, 1986), showed to have very limited contribution to the predictive value of the models. Because of these, the herring larvae maps are only presented here as a first approach to updating the spawning areas, and should not be published and used as spawning maps for herring. Further work is necessary in refining the methodology and improving fitness of the environmental layers used.

Nevertheless, the species distribution modelling approach could in the future be attempted for the remaining species using adult maturity stage data from the fisheries surveys and, when available, interpret this in conjunction with a simpler spatial interpolation of the egg and/or larvae data.

It should be stressed once more that the maps presented here are outputs of models based mostly on survey data, and that the spatial extent of the modelled area is limited by the availability of environmental data. For this reason, it was not possible in the present work to produce model outputs for many inshore areas, particularly for the sea lochs and the intricate coastline of the west of Scotland.

It should also be noted that these outputs have a limited temporal perspective, in the sense that they do not show annual variations in species distribution or persistence at a given site. Instead, they present an average scenario for the studied period which, in most cases, extended for a period of over 40 years (1970-2012). The exception to this was the gadoid species (cod, haddock, whiting, saithe, ling and Norway pout) for which only the period 2000-2012 was analysed. Also, 0 group aggregation models include only data from Quarters 3 and 4, due to the lack of Age-Length-Keys ( ALKs) for 0 group fish in Quarters 1 and 2.

Although as much scientific rigour as possible was applied to their production, models only produce outputs as good as the data they are based on. For this reason, these maps should be interpreted with caution and used as an additional tool to complement existing information, and not to replace it.

At this stage, these outputs take no account of anecdotal or industry knowledge and, therefore, their use and interpretation is recommended alongside the Coull et al. (1998) maps. Further value can be added to these mapping processes using stakeholder input, particularly from the fishing industry. This will complement the final outputs by communicating the industry's view of where and when, in their experience, the sensitive sites occur. The maps from this report will be provided as spatial layers and will be available for download through the National Marine Plan interactive ( http://marinescotland.atkinsgeospatial.com/nmpi/) and the Marine Scotland interactive web portals ( http://www.scotland.gov.uk/Topics/marine/science/MSInteractive). The outputs included into these Marine Scotland data sharing portals will be readily updated when new or additional data or knowledge becomes available.

In addition to the 0 group aggregation outputs presented in this report, work is currently being undertaken to model the locations of spawning grounds of some important commercial species. Species Distribution Modelling techniques that use international survey data are also being used for this investigation as in this report.