Figures
Abstract
The aims of this study were to determine the extent and distribution of an OSPAR priority habitat under current baseline ocean temperatures; to illustrate the prospect for habitat loss under a changing ocean temperature scenario; and to demonstrate the potential application of predictive habitat mapping in “future-proofing” conservation and biodiversity management.
Maxent modelling and GIS environmental envelope analysis of the biogenic bed forming species, Modiolus modiolus was carried out. The Maxent model was tested and validated using 75%/25% training/test occurrence records and validated against two sampling biases (the whole study area and a 20km buffer). The model was compared to the envelope analysis and the area under the receiver operating characteristic curve (Area Under the curve; AUC) was evaluated.
The performance of the Maxent model was rated as ‘good’ to ‘excellent’ on all replicated runs and low variation in the runs was recorded from the AUC values. The extent of “most suitable”, “less suitable” and “unsuitable” habitat was calculated for the baseline year (2009) and the projected increased ocean temperature scenarios (2030, 2050, 2080 and 2100). A loss of 100% of “most suitable” habitat was reported by 2080.
Maintaining a suitable level of protection of marine habitats/species of conservation importance may require management of the decline and migration rather than maintenance of present extent. Methods applied in this study provide the initial application of a plausible “conservation management tool”.
Citation: Gormley KSG, Porter JS, Bell MC, Hull AD, Sanderson WG (2013) Predictive Habitat Modelling as a Tool to Assess the Change in Distribution and Extent of an OSPAR Priority Habitat under an Increased Ocean Temperature Scenario: Consequences for Marine Protected Area Networks and Management. PLoS ONE 8(7): e68263. https://doi.org/10.1371/journal.pone.0068263
Editor: Howard I. Browman, Institute of Marine Research, Norway
Received: January 25, 2013; Accepted: May 27, 2013; Published: July 22, 2013
Copyright: © 2013 Gormley et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: Heriot Watt University funded the PhD research project. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
It is widely accepted that the natural distribution patterns of organisms are primarily driven by their environmental requirements [1]; and that climate change is potentially having an impact on natural distribution patterns through range expansion, contraction and migration [2,3]. The effect which climate change has on geographic distribution is often assessed in terms of potential envelopes/spatial niches shifting in altitude, longitude or latitude; and this influence could, in turn, threaten biodiversity and the conservation of many species [3-5].
Priority marine habitats (determined as ‘threatened and/or declining species and habitats’ under the OSPAR Convention for the Protection of the Marine Environment of the north-east Atlantic 1992) are considered to be of greatest marine nature conservation importance within the North-East Atlantic and are being used to prioritise marine biodiversity conservation and protection under Annex V of the OSPAR Convention 1992. The maintenance of priority habitats will also contribute to the achievement of ‘Good Environmental Status’ (GES) under the European Union (EU) Marine Strategy Framework Directive (MSFD; 2008/56/EC; see also 6). Appropriate area-based management strategies, including a network of Marine Protected Areas (MPAs), are being considered under the MSFD with these and other habitats in mind [7].
Data on the distribution of marine species and habitats are often limited, mainly because of the complexity and costs of surveying and sampling extensive sea areas. For example, habitat maps based on survey data and ground truthing currently cover just 10% of the UK continental shelf [8]. The use of predictive species distribution modelling might therefore provide a suitable tool to fill knowledge gaps, but it may be subject to the issue of over-prediction of range when studying individual species [9]. Ross and Howell [9] acknowledged that a more robust approach might be to apply predictive modelling methods to a habitat formed by a species, rather than to the indicator species itself. This principle has been adopted in the present study.
The objective of this study is to explore the use of a predictive Species Distribution Model (SDM) and a Geographical Information System (GIS) based Environmental Envelope Analysis (EEA) method to create modelled habitat maps for a priority habitat: the biogenic horse mussel reefs formed by the bivalve mollusc Modiolus modiolus (Linnaeus, 1758).
Although M. modiolus is a widespread and common species, actual horse mussel beds are limited in their distribution [10] and often represent biodiversity ‘hotspots’ e.g. [11], some of which have been, or are in the process of being selected for Marine Protected Area status [12-14]. M. modiolus is an Arctic-Boreal species, with a distribution range covering the seas around Scandinavia (including Skagerrak and Kattegat) and Iceland south towards the Bay of Biscay [15-17]. M. modiolus is known to inhabit the subtidal and lower intertidal region of the northern Atlantic and Pacific oceans [15], often in water depths between 5 and 50m; however some individuals have been found at a depth of 280m [15,18]. Dense aggregations/beds reach their southerly limit around the British Isles, in the Irish Sea. This suggests that their occurrence around the British Isles may be vulnerable to a long-term rise in water temperature [16,18].
M. modiolus beds are thought to play an important role in benthic productivity and seabed stabilisation. The beds contribute to high biodiversity and may provide refugia and feeding opportunities to other marine organisms [10,19,20]. Although maps of bed distribution have been created, there is still a considerable amount of uncertainty as to the true extent of these beds within the OSPAR region [21].
The primary goal of this study is to use publicly available datasets to test the modelling approaches for a M. modiolus habitat case example, to see whether it may provide a new tool to inform the MSFD spatial management process for key habitats. The models will be applied to determine the extent of habitat suitable for M. modiolus beds under current baseline conditions; predict habitat loss under an increased ocean temperature scenario; and demonstrate the application of a predictive habitat mapping tool for “future-proofing” spatial planning for habitats and biodiversity management planning.
Methods
Modiolus modiolus Occurrence Data
The M. modiolus bed occurrence records were extracted from the 2011 OSPAR priority habitats dataset [22] and corrected based on areas of uncertainty published by Rees [21]. The data were supplemented with occurrence records collected during more recent UK surveys [13,14,23]. A total of 215 occurrence records were obtained (Figure 1). As a result of the limited geographical coverage of some of the environmental layers, 82 records were excluded because they did not coincide with the environmental layers.
Projection: WGS 1984 UTM 31N.
Environmental Data
Data on environmental variables of potential biological relevance to M. modiolus were obtained from publically available sources (Table 1) then assigned to a 0.005o grid using ArcMap 9.3 Geographical Information System (GIS) software. Temperature, depth, substratum, water movement and salinity were chosen based on the M. modiolus environmental requirements as outlined by Holt et al. [18], but water quality and suspended sediment were not available for inclusion in this model.
| Variable | Source |
|---|---|
| Bathymetry: depth (m) | GEBCO_08 30-second arc Bathymetry resolution [63] |
| Slope: percentage gradient of the seafloor (%) | Adapted in ArcGIS 9.3 from: GEBCO_08 30-second arc Bathymetry resolution [63] |
| Sea Bottom Temperature: climatological annual mean sea bottom temperature (oC). Adapted from NOAA depth interval data | NOAA, World Ocean Atlas [24] |
| Bottom Salinity: climatological annual mean sea bottom salinity (PSS). Adapted from NOAA depth interval data | NOAA, World Ocean Atlas [64] |
| Landscape: seabed landscape features [Broad patterns in seabed character, such as seabed morphology determined by major geological and hydrographic processes] | UKSeaMap/MESH webGIS [65] http://www.searchmesh.net/ (“Marine Landscapes” layer on interactive map) |
| Current Speed: average spring current speed (ms-1) | Atlas of UK marine renewable energy resources [66] Supplemented by: Current speed data on UKHO Navigation Charts [67] and BODC oceanographic data [68] |
Table 1. Environmental variables and data sources.
Increased Ocean Temperature Scenario
Increased ocean temperature scenarios were established for the following epochs: 2009 (Figure 1), 2030, 2050, 2080 and 2100 based on Locarnini et al. [24] and the International Panel on Climate Change (IPCC) scenario planning methodology [25]. Predictions were based on the IPCC climate change scenario A1B in which a 4oC increase in ocean surface temperature would occur by 2100 [25]. A linear increase in ocean bottom temperature was calculated between 2009 and 2100, therefore increases of 0.92oC, 1.80oC and 3.12oC were expected for 2030, 2050 and 2080 respectively.
Model scenarios assumed a uniform increase in temperature over the entire spatial domain and throughout the water column.
Environmental Envelope Analysis
Initial baseline species distribution analysis was carried out through the creation of an environmental envelope for M. modiolus bed populations in ArcMap 9.3. The M. modiolus bed occurrence records were grouped into populations based on their location and proximity to each other. Populations were selected if the occurrence records were within 10km of each other, excluding areas of obvious boundaries, e.g. land or sealochs etc. Within this 10km population grouping, the individual occurrence records were given a 1km buffer which would represent bed extent within that particular population. Environmental layers were plotted in vector format and overlaid with the population records. The "preferred range" of environmental attributes was characterised in terms of the interquartile ranges of the environmental variable values over the occurrence locations.
The "preferred range" for the landscape was calculated based on qualitative data (therefore the interquartile range calculation within ArcMap was not suitable). The area of overlap for each population and landscape type was calculated. The percentage of each landscape type inhabited by a population was calculated (landscape range), ranked, then the median and maximum of these percentages (landscape range) was determined. The "preferred" landscape types were determined as representing those that were inhabited by the majority of the populations (≥ the median of the landscape range).
Areas where "preferred range" attributes occurred for all overlying environmental layers were classed as the environmental envelope for horse mussel beds.
Species Distribution Model
Maxent is a predictive method that models the geographic distribution of species using presence-only data. Probability of occurrence is modelled in relation to environmental variables under the assumption that the species distribution will follow the property of maximum entropy [26-28]. Maxent has been used in a number of comparative studies examining the effectiveness of species distribution modelling (SDM) in the marine environment [28-30] and is considered to be reliable in this context [29].
Model Validation
The model predictions were tested using the ‘Area Under the Curve’ (AUC) produced by Maxent. The area under the Receiver Operating Characteristic (ROC) curve is a widely used test statistic which measures model performance [28]. The AUC varies between 0 and 1, with values above 0.9 indicating excellent prediction, between 0.7 and 0.9 indicating good prediction, below 0.7 indicating poor prediction, and below 0.5 no better than random [29].
Owing to the lack of independent test datasets, models were assessed by 2-fold cross validation on ten replicate runs [28]. The occurrence dataset was randomly split in ArcMap 9.3 using the Hawths Analysis Tools for ArcGIS extension [31] each containing a randomly selected 75% of records for model training and the corresponding 25% for model testing. A further model cross-validation was run using the full occurrence dataset randomly split into a 90% training/10% test dataset internally using the Maxent random test setting.
No absence data were available and therefore 10,000 randomly chosen pseudo-absence/background points were run. Selecting the background points from the whole study area may artificially inflate the AUC value, especially if the geographic area is particularly large or the area of suitable habitat is small in relation to the whole study area [28]. During model evaluation, models were tested using background points selected from within a 20km buffer of the known occurrence locations (bias model) and compared with models run with background points selected from the whole study area (global model).
It was considered that the landscape layer might artificially influence the distribution of suitable habitat within the model, therefore, jack-knife contributions of each variable were measured to test the contribution of each variable to the model.
The tested models were visually inspected and compared to the environmental envelope analysis, and occurrence data. This enabled the assessment of model plausibility with respect to the known distribution and areas of suitable habitat outside known occurrence range (over-prediction) [28].
Probability of Habitat Distribution
The probability of occurrence values (0 to 1) estimated in the Maxent model training and projection runs were separated into 10 bands and the area (Km2) covered by each band was calculated.
The 10 probability bands were further separated into 3 categories for MPA region assessment:
- i) 0.5–1.0 representing a situation where M. modiolus beds may be more likely to occur (“most suitable habitat”);
- ii) 0.1-0.49 representing a situation where M. modiolus beds are less likely to occur (“less suitable habitat”); and
- iii) 0.0–0.09 representing a situation where M. modiolus beds are highly likely not to occur (“unsuitable”).
In this study, MPA Regions are defined as designated regions of search for potential MPAs within UK waters (200nm limit).
Results
Environmental Envelope Analysis
The environmental envelope analysis method was applied to the M. modiolus bed population locations (Figure 1) and is a simple summarisation of potential suitable habitat for M. modiolus beds within UK waters. Table 2 outlines the environmental envelope calculated for M. modiolus beds.
| Environmental Layer | Preferred Range |
|---|---|
| Temperature: | 9 to 10 oC |
| Landscape: | Sealoch |
| Shallow coarse sediment plain - moderate tide stress | |
| Shallow coarse sediment plain - weak tide stress | |
| Shelf coarse sediment plain - moderate tide stress | |
| Shallow sand plain | |
| Embayment | |
| Shallow mixed sediment plain - weak tide stress | |
| Shallow mud plain | |
| Shelf coarse sediment plain - strong tide stress | |
| Photic rock | |
| Bathymetry: | -20 to 0 m |
| Current Speed: | 0.5 to 1.115 m/s |
| Slope: | 0 to 0.345% |
| Salinity: | 34 to 35 ppt |
Table 2. The selected Environmental Envelope for Modiolus modiolus (Linnaeus, 1758) beds.
Figure 2 illustrates the environmental envelope for M. modiolus beds and represents areas of suitable M. modiolus bed habitat generated by the envelope analysis.
Projection: WGS 1984 UTM 31N.
This method indicates that the west of Scotland, Strangford Lough in Northern Ireland, Wales, Orkney and Shetland are the most suitable areas for M. modiolus beds, with more scattered areas around the Isle of Man and the east coast of England.
When the envelope analysis was applied to the projected climate change scenarios, results indicated that there would be a decrease of potentially suitable habitat by 2050 (58% loss by 2030; and 98% loss by 2050) and complete loss of suitable M. modiolus bed habitat by 2080.
The envelope analysis was re-run for the baseline model, excluding the landscape environmental layer (to test for environmental variable bias) and a small increase in suitable habitat was noted, however, results still showed the same distribution pattern as before, with a slight increased presence around the coast of Wales, east England and south west Scotland. This comparison shows that the landscape layer did not have a disproportionate effect on the baseline model outcome.
Species Distribution Model
Model Selection.
The Maxent model was trained using cross-validation of 2 externally selected sub-sets of the 2009 baseline data and further trained for an internally selected sub-set within Maxent’s automated validation test. The training AUC values, shown in Table 3, ranged from 0.92 to 0.99 with little variation shown over the 10 replicates (maximum difference from 0 to 0.006). The test AUC values ranged from 0.86 to 0.98 and showed slightly higher variation over the replicated runs (maximum difference 0.008 to 0.047).
| Average AUC Test Statistic | ||||
|---|---|---|---|---|
| Training | Testing | |||
| Model (Training/Test) | Bias | Global | Bias | Global |
| Set 1 (75/25%) | 0.92 ±0.003 | 0.98 ±0.001 | 0.86 ±0.051 | 0.97 ±0.023 |
| Set 2 (75/25%) | 0.94 ±0.003 | 0.99 ±0.001 | 0.90 ±0.047 | 0.97 ±0.043 |
| All (90/10%) | 0.93 ±0.006 | 0.99 ±0.000 | 0.92 ±0.039 | 0.98 ±0.008 |
| Final model (90/10%) | 0.93 | 0.99 | 0.88 | 0.97 |
Table 3. Threshold-independent area under the curve (AUC) indices for Modiolus modiolus (Linnaeus, 1758) habitat model.
Test statistic values decreased when calculated using pseudo-absences restricted to 20 km from occurrence records.
A final model was run for each of the sampling scenarios using the full occurrence records and a 90%/10% training/test ratio run on a single replicate. The AUC values for the final model ranged from 0.93 to 0.99 for model training and 0.88 to 0.97 from model testing and generally equalled the average of the cross-validated models, indicating little variation between the overall model test statistics. Overall, the environmental variable with the highest gain when used in isolation was landscape, which therefore appears to have the most useful information by itself when determining the location of suitable habitat. In contrast, when Bathymetry was omitted the jack knife analysis showed the lowest gain, indicating that the Bathymetry variable has the most information not present in the other variables, when determining location of suitable habitat. The AUC values remained above 0.96 for each model run following omission of each environmental variable in turn; this indicates ‘excellent’ model performance.
Pseudo-absence selection models were compared. The models where pseudo-absences were chosen from within 20km of the known occurrence records predicted suitable habitat to occur to the west of Scotland and Northern Ireland, but with the highest probability of suitable habitat occurring on the North Norfolk sandbanks. There were also areas of low probability predicted in the English Channel and a lack of suitable habitat predicted around Orkney (Figure 3). In comparison, in the models where pseudo-absences were selected from the whole study area, the highest probability of suitable habitat occurring was observed predominantly to the West of Scotland, Shetland and Northern Ireland. The area around the Norfolk sandbanks showed lower levels of probability (Figure 4).
Model sampling bias: 20km. Projection: WGS 1984 UTM 31N.
Model sampling bias: Global. Projection: WGS 1984 UTM 31N.
The “most suitable” habitat output (probability ≥0.5) for 2009 from Maxent were compared with the environmental envelope.
The output of the Environmental Envelope Analysis (EEA) showed a 50% overlap of the "most suitable" habitat predicted by the Maxent global sampling model; with an overlap of 22% of the "less suitable" habitat (Table 4) and <1% of the “unsuitable” habitat.
| Method/Model | Area (Km2) | Percentage of Maxent overlapped by envelope | Combined overlap (excluding "unsuitable" habitat) | Percentage "over prediction" (model vs envelope) | |
|---|---|---|---|---|---|
| Envelope Analysis | 7,009 | n/a | n/a | n/a | |
| Global model | "Most Suitable" | 2,191 | 50% | 26% | 58% |
| "Less Suitable" | 14,390 | 22% | 26% | 58% | |
| Bias model | "Most Suitable" | 6,471 | 55% | 16% | 81% |
| "Less Suitable" | 29,659 | 8% | 16% | 81% | |
Table 4. Comparison of Environmental Envelope Analysis and Maxent model outputs. Overlap area calculations.
Model Projections.
The selected baseline model was projected against the predicted 2030, 2050, 2080 and 2100 conditions. Figure 5 illustrates that the percentage of sea area suitable for M. modiolus beds decreases rapidly over the 4 projected epochs with a 100% loss of M. modiolus bed habitat predicted by 2100. The 10 probability bands were separated into 3 categories for ease of examination and discussion: “most suitable” (MS), “less suitable” (LS) and “unsuitable” (US) habitat. Calculated areas indicated a 100% loss of “most suitable” habitat by 2080 (Figure 5). Figure 6 illustrates the rapidity of habitat loss of the epochs. The steepest decline of potential habitat occurs in bands 0.1 to 0.39 between 2050 and 2080, and band 0.8 to 0.89 between 2030 and 2050. The modelled projections are illustrated in Figure 7. The extent of predicted distribution as represented by the shading, decreases significantly over the 4 epochs.
Illustration of speed of habitat loss.
Projection: WGS 1984 UTM 31N.
MPA Region Assessment
The area of MS, LS and US habitat within each MPA region was calculated over the 5 epochs and these data are summarised in Table 5. The results show that there are some MPA regions that are potentially more important to M. modiolus beds than others. The area and percentage loss of “most suitable” habitat within each MPA region is summarised in Table 6. The results (Tables 5 and 6) show that the West of Scotland (Territorial) MPA region is the most important region in terms of predicted habitat. The Net Gain, North Scotland (Territorial), South West Scotland (Territorial) and Northern Ireland are also important regions. Most significantly, the West of Scotland (Territorial) region loses 56% of its “most suitable” habitat by 2030 and 100% is lost by 2050.
| MPA Region | Area (Km2) | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2009 | 2030 | 2050 | 2080 | 2100 | Total area of region | |||||||||||
| MS | LS | US | MS | LS | US | MS | LS | US | MS | LS | US | MS | LS | US | ||
| East Scotland | 0 | 0 | 98761 | 0 | 0.00 | 99204 | 0 | 0 | 98961 | 0 | 0 | 99242 | 0 | 0 | 99210 | 101511 |
| East Scotland (Territorial) | 20 | 923 | 12391 | 11 | 817 | 12537 | 4.83 | 151 | 13167 | 0 | 0 | 13386 | 0 | 0 | 13380 | 13656 |
| Balanced Seas | 0 | 6 | 15013 | 0 | 0 | 15282 | 0 | 0 | 15294 | 0 | 0 | 15273 | 0 | 0 | 15202 | 17846 |
| Finding Sanctuary | 3 | 931 | 77999 | 0 | 2 | 79059 | 0 | 0 | 78739 | 0 | 0 | 79391 | 0 | 0 | 79028 | 95979 |
| Irish Sea Conservation Zone | 10 | 1493 | 15485 | 0 | 9 | 16687 | 0 | 0 | 16833 | 0 | 0 | 16730 | 0 | 0 | 16700 | 17551 |
| Isle of Man | 7 | 601 | 3159 | 0 | 6 | 3774 | 0 | 0 | 3780 | 0 | 0 | 3780 | 0 | 0 | 3780 | 4609 |
| MCZ Project Wales | 24 | 2283 | 13627 | 0 | 8 | 15917 | 0 | 0 | 15918 | 0 | 0 | 15951 | 0 | 0 | 15946 | 16375 |
| Net Gain | 583 | 7031 | 103228 | 81 | 3300 | 107776 | 6.50 | 213 | 111024 | 0 | 0 | 110931 | 0 | 0 | 111117 | 113204 |
| North Scotland | 0 | 0 | 22742 | 0 | 0 | 23247 | 0 | 0 | 23185 | 0 | 0 | 23130 | 0 | 0 | 23194 | 29967 |
| North Scotland (Territorial) | 324 | 1540 | 21361 | 136 | 1229 | 21662 | 32 | 232 | 22708 | 0 | 0 | 22996 | 0 | 0 | 23018 | 23860 |
| Northern Ireland | 210 | 683 | 3197 | 0 | 153 | 4090 | 0 | 21 | 4196 | 0 | 0 | 4160 | 0 | 0 | 4234 | 9071 |
| South West Scotland (Territorial) | 273 | 766 | 6538 | 37 | 258 | 7319 | 0 | 0 | 7598 | 0 | 0 | 7644 | 0 | 0 | 7645 | 7996 |
| West Scotland | 0 | 93 | 16698 | 0 | 0 | 17165 | 0 | 0 | 16650 | 0 | 0 | 17156 | 0 | 0 | 16942 | 27701 |
| West Scotland (Territorial) | 1345 | 2311 | 36913 | 590 | 1074 | 39231 | 0 | 382 | 40435 | 0 | 0 | 40698 | 0 | 0 | 40742 | 43640 |
Table 5. The area of “most suitable” (ms), “less suitable” (LS) and “unsuitable” (US) habitat within each MPA region.
| MPA Region | 2009 | 2030 | 2050 | 2080 | 2100 | ||||
|---|---|---|---|---|---|---|---|---|---|
| Area | Area | % Loss | Area | % Loss | Area | % Loss | Area | % Loss | |
| East Scotland | 0.00 | 0.00 | n/a | 0.00 | n/a | 0.00 | n/a | 0.00 | n/a |
| East Scotland (Territorial) | 19.91 | 10.71 | 46 | 4.83 | 76 | 0.00 | 100 | 0.00 | 100 |
| Balanced Seas | 0.00 | 0.00 | n/a | 0.00 | n/a | 0.00 | n/a | 0.00 | n/a |
| Finding Sanctuary | 3.20 | 0.00 | 100 | 0.00 | 100 | 0.00 | 100 | 0.00 | 100 |
| Irish Sea Conservation Zone | 10.28 | 0.00 | 100 | 0.00 | 100 | 0.00 | 100 | 0.00 | 100 |
| Isle of Man | 7.29 | 0.00 | 100 | 0.00 | 100 | 0.00 | 100 | 0.00 | 100 |
| MCZ Project Wales | 24.11 | 0.00 | 100 | 0.00 | 100 | 0.00 | 100 | 0.00 | 100 |
| Net Gain | 582.81 | 80.86 | 86 | 6.50 | 99 | 0.00 | 100 | 0.00 | 100 |
| North Scotland | 0.00 | 0.00 | n/a | 0.00 | n/a | 0.00 | n/a | 0.00 | n/a |
| North Scotland (Territorial) | 323.64 | 136.15 | 58 | 31.72 | 90 | 0.00 | 100 | 0.00 | 100 |
| Northern Ireland | 210.26 | 0.00 | 100 | 0.00 | 100 | 0.00 | 100 | 0.00 | 100 |
| South West Scotland (Territorial) | 273.09 | 36.97 | 86 | 0.00 | 100 | 0.00 | 100 | 0.00 | 100 |
| West Scotland | 0.00 | 0.00 | n/a | 0.00 | n/a | 0.00 | n/a | 0.00 | n/a |
| West Scotland (Territorial) | 1345.32 | 590.39 | 56 | 0.00 | 100 | 0.00 | 100 | 0.00 | 100 |
Table 6. The area and percentage loss of “most suitable” habitat within each MPA region.
Discussion
The aim of this study was to model the ecological niche and bioclimatic envelope of M. modiolus beds within UK waters as a baseline for subsequent increased ocean temperature projections, and to demonstrate its application as a tool for future management of habitats. Species Distribution Modelling techniques have previously been applied in the marine environment to a range of motile species [28,29,32-35]; but, with the possible exception of Ross and Howell’s 2012 [9] study on deep sea organisms, this is the first study the authors are aware of that deals with marine habitat forming species of high conservation management interest, under an increasing ocean temperature scenario. In a terrestrial setting bioclimatic envelope models provide perhaps the best available guide for conservation managers and policy makers [2,4,36-39] and have been considered as first approximations of the magnitude and broad patterns of future impacts [2]. In this context, terrestrial conservation protection has appeared inadequate under future climate change scenarios [36]. For example, Carvalho et al. [36] concluded that protected areas covered 10% of the current distribution of all Iberian herptiles; and that to maintain this coverage the protected area network would have to be increased by 1-2% by 2080.
Environmental Envelope Analysis
The Environmental Envelope Analysis (EEA) provided a relatively quick and simple method for analysing the potential distribution of the M. modiolus habitat and was performed in order to validate the Maxent model method. The EEA greatly improves the visualisation and analysis of potential projected conditions in support of conservation planning without the requirement for specialised modelling knowledge; and methods such as this demonstrate the possibilities of generating new knowledge from existing data sets. It was important that all environmental variable layers used were freely and publically available in order to demonstrate the immediate applicability of such modelling tools to inform contemporary policy and management decision making for the marine environment.
The envelope analysis, however, will only take into account areas where all the individual "preferred ranges" overlap, a concept that is corrected for within the Maxent model. In addition, the EEA does not lend itself sufficiently to model testing and statistical analysis, therefore it would not necessarily provide robust evidence, unless run alongside another model. It does however provide a robust representation of shifting habitats in a timely and cost-effective manner.
The EEA method developed within this study is, as far as the authors are aware, a new use of the method for the selection of an environmental envelope based on the interquartile range analysed within a GIS setting. Two other proposed methods were also investigated [40,41], but these methods were judged to be unsuitable for the data used within this particular study. These methods were either based on descriptive data and on species that inhabitat a very particular niche [41], or suggested too wide an envelope (minimum to maximum ranges) [40].
The envelope analysis utilised, predicts that the habitat will retreat northwards as sea temperature increase, with more limited extent of distribution in the Irish Sea and Shetland regions compared to the current known bed occurrence records (Figure 1). These results would suggest that although this type of analysis is useful for simple visualisation and summarisation of suitable habitat areas, more refinement of environmental layers is required for detailed application.
Species Distribution Model
The Maxent model outputs in this study provide an overview of potentially suitable M. modiolus bed habitat. Despite the present model being built on environmental variables with coarse resolution, species with a narrow ecological niche can show high accuracy of predicted distribution under modelled conditions compared to those with a broader niche [29]. In addition, the global model which was utilised in this study closely resembled the output of the comparative environmental envelope analysis. Overall, therefore, the baseline trained model (global model) can be interpreted as a good predicted range, with projections showing that the M. modiolus beds lose their ability to fulfil that range by 2100. Under these modelled conditions M. modiolus beds in the UK will be increasingly vulnerable.
Details of climate change scenarios in the marine environment are poorly understood. The extent to which environmental changes (e.g. alterations to hydrodynamics and sediment dynamics) might occur alongside temperature increases is not well studied. Other environmental variables such as salinity and acidity were excluded in the present study because there was a lack of information [42], or conflicting literature on the potential levels and direction of change in these variables (e.g. salinity increasing [43,44], salinity decreasing [45]; salinity decreasing at high latitudes and increasing at low latitudes [46,47]). Under the climate change scenario A1B [25] ocean pH is predicted to decrease to 7.9 from a baseline of 8.1 in 2007. However, no environmental data on the variability of pH of the seawater around the UK was readily available to allow this scenario to be defined in terms of spatial variation. Depth was excluded from the "climate change" scenario based on the quality of the bathymetry data used. The sea level rise predicted under the climate change scenario A1B indicates an increase of up to 0.5 m by 2100.
The assumptions made on increased ocean temperature at depth in the present study are supported by research conducted by Levitus et al. [48]. This research suggested warming of the upper 300m of the world’s oceans between 1948 and 1998, particularly the Indian and Atlantic Oceans. However, it is unclear as to what magnitude ocean warming at depth will occur in the future; and variations in the speed of climate change between UK regions are unknown [49].
An issue with SDM techniques for sessile organisms like M. modiolus is that SDM, including Maxent, base predicted distributions on an ecological niche theory, and do not give consideration to propagule dispersal [50], dispersal vectors and propagule establishment [51]. Although knowledge of larval dispersal may not necessarily refine habitat suitability models in definite terms, it may lead to an enhanced understanding of model predictions or contribute to model accuracy.
Presently, little information is available on genetic connectivity of the beds. Holt et al. [18] and Comely [52] suggest recruitment from outside the area for beds off the Lleyn Peninsula and the Isle of Man; and self-sustaining populations occurring in Strangford Lough and the Scottish sealochs based on perceptions of wide dispersal from and to highly tidal areas, and low dispersal from and to sealochs with high water residence times.
M. modiolus are thought to spawn in a relatively narrow temperature window (7-10oC) [16] suggesting that, although the model shows a reduction of potentially suitable M. modiolus habitat, recruitment may be the mechanisms by which reefs cease to be viable. Established reefs may therefore persist beyond the prediction of the present study, but their reproduction may be hindered; and local adaption to the changing climate may occur over time.
M. modiolus are relatively long lived, with a life-span of approximately 20-100 years [53] giving some indication of the lag-time before senescence might be detected. There is, as yet, no evidence of reefs that are senescing. Many beds studied in the 1950s still exist [23,54-56] and reefs in North Wales are thought to have persisted for approximately 150 years [12], with evidence that these beds are still recruiting [56]. Studies have recorded an overall decline in the extent of M. modiolus beds in the period between 1950 and 1990 [53].
The trained model output illustrated that the most suitable baseline areas occurred in west Scotland, Northern Ireland (Strangford Lough) and Shetland, with less suitable habitat occurring in the Irish Sea and Orkney. Patches of suitability around the east coast of England (Norfolk coast) appear misleading because beds are not known to occur there (Figure 7). It is possible that the model is predicting the existence of suitable environmental conditions for M. modiolus beds in in this area, but other unaccounted factors (e.g. connectivity, fishing impacts, or turbidity etc.) could be preventing actual bed presence. Limitations of knowledge, low numbers of targeted surveys or decline of beds in this area are also possible explanations. For example, the Southern North Sea, the Western Channel/Celtic Sea and Irish Sea are known to have the highest intensity of trawling and dredging pressure in the UK [8]: an anthropogenic pressure thought to impact these biogenic habitats (e.g. Strain [57]). Furthermore, the North Norfolk Sandbanks and Saturn Reef are designated MPAs (Special Area of Conservation; SAC) for Sabellaria spinulosa Leuckart, 1849 beds, a tube dwelling polychaete, which require silty, turbid conditions to build their tubes and reefs [18]. In this study the model may therefore be interpreting the suitability of areas for biogenic reefs and may not be refined enough to distinguish the environmental envelope for functionally similar species structures. S. spinulosa requires suspended sediment to build their tubes, M. modiolus does not, and may be sensitive to smothering and/or lack of suitable suspended food.
MPA Region Assessment
The area of the current SACs that encompass M. modiolus beds (Loch Creran and Lochs Duich, Long and Alsh beds, west Scotland; the Lleyn Peninsula and Sarnau, north Wales; Sanday, Orkney; Strangford Lough, Northern Ireland) cover 141Km2 of the predicted distribution of “most suitable” habitat in 2009; 15Km2 in 2030 and zero in 2050 to 2100. This represents 8% protection of the predicted “most suitable” habitat range in 2009 and this drops to 0.9% by 2030; and 0% by 2050. Protection is therefore limited, and will dwindle in contrast to the Convention on Biological Diversity target: “By 2020, at least ….. 10% of coastal and marine areas, especially areas of particular importance for biodiversity and ecosystem services, are conserved through … …. representative and well-connected systems of protected areas and other effective area-based conservation measures….” Although, this statement is not species specific, the IUCN’s Vth World Parks Congress, 2003, suggested that 20-30% of each habitat should be protected within MPAs by 2012 [9,18,58-60].
Micheli et al. [61] concluded that the protection afforded to species in marine reserves supports population resistance to large scale environmental impacts. This is achieved through greater larval production and recruitment; large adult body size; absence of fishing related mortality and larval spill-over; maintained reproductive output; and recoverability. A network of marine protected areas may therefore be the most effective tool in mitigating the negative impacts of climate change on marine ecosystems and their associated livelihoods [61].
In addition to designated protected areas, consideration also needs to be given to potential dispersal corridors [37] to accommodate movement of conservation interest species/habitats within a changing climate, potentially safeguarding these areas through conservation easement [3].
Pan-European perspectives
Presently, UK Good Environmental Status (GES) targets under the MSFD for rock and biogenic beds are drawn from the Habitats Directive [62] i.e. that the “Area is stable or increasing and not smaller than the baseline value” (EU Habitats & Species Directive, Council Directive 92/43EEC). This is in keeping with one of the key aims of the MSFD to “Protect and preserve the marine environment prevent its deterioration or, where practicable, restore marine ecosystems”. However, one of the key MSFD characteristics of Biodiversity (Descriptor 1) is that “The quality and occurrence of habitats and the distribution and abundance of species are in line with prevailing physiographic, geographic and climatic conditions”, a characteristic that is being interpreted as accommodating climate change [6,62]. The implication of the present study is that, in the short term, maintaining nationally “stable or increasing” areas of some protected habitats may not be achievable within the next 40 years without significant restorative and facilitated migration work. For habitats like these, the connectivity of an MPA network will be of paramount importance, especially for those that have already suffered historic loss and fragmentation. It is also possible that within a life time, maintaining “stable” areas may not be achievable at all within a national or regional context.
The amount of habitat loss that would be tolerated within the assessment of GES under the MSFD is yet to be defined for many target species/habitats and methods such as the one demonstrated within this study, could, with further refinement enable more plausible definition of targets.
Conclusions
Paradoxically, the achievement of GES within ‘prevailing climatic conditions’ may require European Atlantic nations to manage the decline and migration of some of their marine habitats of biodiversity conservation importance rather than maintain their present extent. This concept is relatively novel to marine conservation management and not currently represented within national or international Marine Spatial Planning; nor in the conservation objectives or management plans of MPAs.
Author Contributions
Conceived and designed the experiments: KG MB. Performed the experiments: KG. Analyzed the data: KG WS MB. Wrote the manuscript: KG WS JP AH.
References
- 1. Pearson RG, Dawson TP, Berry PM, Harrison PA (2002) SPECIES: A Spatial Evaluation of Climate Impact on the Envelope of Species. Ecol Modell 154: 289-300. doi:https://doi.org/10.1016/S0304-3800(02)00056-X.
- 2. Pearson RG, Dawson T (2003) Predicting the impacts of climate change on the distribution of species: are bioclimate envelope models useful. Glob Ecol Biogeogr 12: 361-371. doi:https://doi.org/10.1046/j.1466-822X.2003.00042.x.
- 3. Thomas CD, Gillingham PK, Bradbury RB, Roy DB, Anderson BJ et al. (2012) Protected areas facilitate species’ range expansions. Proc Natl Acad Sci U S A 109: 14063-14068. doi:https://doi.org/10.1073/pnas.1210251109. PubMed: 22893689.
- 4. del Barrio G, Harrison PA, Berry PM, Butt N, Sanjuan ME et al. (2006) Integrating multiple modelling approaches to predict the potential impacts of climate change on species’ distributions in contrasting regions: comparison and implications for policy. Environmental Science &. Policy 9: 129-147.
- 5. Thuiller W (2003) BIOMOD - optimizing predictions of species distributions and projecting potential future shifts under global change. Glob Change Biol 9: 1353-1362. doi:https://doi.org/10.1046/j.1365-2486.2003.00666.x.
- 6.
Moffat C, Aish A, Hawkridge JM, Miles H, Mitchell PI et al. (2011) Advice on United Kingdom biodiversity indicators and targets for the Marine Strategy Framework Directive. Healthy and Biologically Diverse Sea Evidence Group Report to the Department for Environment, Food and Rural Affairs. 210 p.
- 7.
SNH (2011) Priority Marine Features. http://www.snh.gov.uk/protecting-scotlands-nature/safeguarding-biodiversity/priority-marine-features/. Accessed: 2011.
- 8. Aish A, Ashworth J, Barrio Froján B, Benjamins S, Bolam S, et al. (2010) Charting Progress 2 Healthy and Biological Diverse Seas Feeder Report: Section 3.1: Marine Habitats. Published by Department for Environment Food and Rural Affairs on behalf of UKMMAS. In: UKMMAS (2010) Charting Progress 2 Healthy and Biological Diverse Seas Feeder Report (Eds. Frost, M & Hawkridge, J). pp. p68-271.
- 9. Ross RE, Howell KL (2012) Use of predictive habitat modelling to assess the distribution and extent of the current protection of 'listed' deep-sea habitats. Diversity and Distributions Early View. On Line: 1-13.
- 10. Elsäßer B, Fariñas-Franco JM, Wilson CD, Kregting L, Roberts D (2013) Identifying optimal sites for natural recovery and restoration of impacted biogenic habitats in a special area of conservation using hydrodynamic and habitat suitability modelling. J Sea Res 77: 11-21. doi:https://doi.org/10.1016/j.seares.2012.12.006.
- 11. Rees EIS, Sanderson WG, Mackie ASY, Holt RHF (2008) Small-scale variation within a Modiolus modiolus (Mollusca: Bivalvia) reef in the Irish Sea. III. Crevice, sediment infauna and epifauna from targeted cores J Mar Biol Assoc UK 88: 151-156.
- 12. Lindenbaum C, Bennell JD, Rees EIS, McClean D, Cook W et al. (2008) Small-scale variation within a Modiolus modiolus (Mollusca: Bivalvia) reef in the Irish Sea: I. Seabed mapping and reef morphology. J Mar Biol Assoc UK 88: 133-141.
- 13.
Hirst NE, Cook RL, James B, Kent FEA, Loxton J et al. (2012) The distribution of Priority Marine Features and MPA search features in Orkney waters: Rousay to Copinsay including Eday Sound and points in Scapa Flow: Scottish Natural Heritage Commissioned Report No. 509.
- 14.
Hirst NE, Clark L, Sanderson WG (2012) The distribution of selected MPA search features and Priority Marine Features off the NE coast of Scotland. Scottish Natural Heritage Commissioned Report No. 500. 133pp.
- 15.
Hayward PJ, Ryland JS (1990) The Marine Fauna of the British Isles and north-west Europe. Oxford University Press. 996pp.
- 16. Brown RA (1984) Geographical variations in the reproduction of the horse mussel, Modiolus modiolus (Mollusca: Bivalvia). J Mar Biol Assoc UK 64: 751-770. doi:https://doi.org/10.1017/S0025315400047214.
- 17. Davenport J, Kjørsvik E (1982). Observations A Norwegian Intertidal Populations Horse Mussel Modiolus Modiolus(L). Journal of Molluscan Studies 48: 370-371.
- 18.
Holt TJ, Rees EI, Hawkins SJ, Seed R (1998) Biogenic Reefs (volume IX). An overview of dynamic and sensitivity characteristics for conservation management of marine SACs. Scottish Association of Marine Science.
- 19. Ragnarsson SA, Burgos JM (2012) Separating the effects of a habitat modifier, Modiolus modiolus and substrate properties on the associated megafauna. J Sea Res 72: 55-63. doi:https://doi.org/10.1016/j.seares.2012.05.011.
- 20. Fariñas-Franco JM, Allcock L, Smyth D, Roberts D (2013) Community convergence and recruitment of keystone species as performance indicators of artificial reefs. J Sea Res 78: 59-74. doi:https://doi.org/10.1016/j.seares.2012.10.008.
- 21.
Rees I (2009) Assessment of Modiolus modiolus beds in the OSPAR area. Prepared on behalf of Joint Nature Conservation Committee.
- 22.
JNCC (2012) Mapping OSPAR priority habitats. http://jncc.defra.gov.uk/page-1583. Accessed: January 2012.
- 23.
Sanderson WGunpubl. Dive logs. Edinburgh: Heriot Watt University.
- 24.
Locarnini RA, Mishonov AV, Antonov TP, Boyer HE, Garcia OK et al. (2010) World Ocean Atlas 2009, Volume 1: Temperature. In: S. Levitus. NOAA, Atlas NESDIS 68, Volume 1, Volume 1. Washington, D.C.: U.S. Government Printing Office. p. 184.
- 25.
IPCC (2007) Climate Change 2007: Synthesis Report. Contrib of Working Groups I, II and III to the Fourth Assessment Report of The Intergovernmental Panel on Climate Change [Core Writing Team, R.K PachauriA. Reisinger. In: IPCC, editor. Geneva, Switzerland. pp. 104.
- 26. Phillips SJ, Anderson RP, Schapire RE (2006) Maximum entropy modelling of species geographic distributions. Ecol Modell 190: 231-259. doi:https://doi.org/10.1016/j.ecolmodel.2005.03.026.
- 27. Phillips SJ, Dubik M (2008) Modelling of species distribution with Maxent: new extensions and a comprehensive evaluation. Ecography 31: 161-175. doi:https://doi.org/10.1111/j.0906-7590.2008.5203.x.
- 28. Jones MC, Dye SR, Pinnegar JK, Warren R, Cheung WWL (2012) Modelling commercial fish distributions: Prediction and assessment using different approaches. Ecol Modell 225: 133-145. doi:https://doi.org/10.1016/j.ecolmodel.2011.11.003.
- 29. Reiss H, Cunze S, Konig K, Neumann H, Kroncke I (2011) Species Distribution modelling of marine benthos: a North Sea case study. Mar Ecol Prog S 442: 71-86. doi:https://doi.org/10.3354/meps09391.
- 30. Ready J, Kaschner K, South A, Eastwood P, Rees T et al. (2010) Predicting the distribution of marine organisms at the global scale. Ecol Modell 221: 467-478. doi:https://doi.org/10.1016/j.ecolmodel.2009.10.025.
- 31.
Beyer HL (2004) Hawth’s Analysis Tools for ArcGIS. Available: http://www.spatialecology.com/htools. Accessed: January 2011.
- 32. Hedger R, McKenzie E, Heath M, Wright P, Scott B et al. (2004) Analysis of the spatial distributions of mature cod (Gadus morhua) and haddock (Melanogrammus aeglefinus) abundance in the North Sea (1980–1999) using generalised additive models. Fish Res 70: 17-25. doi:https://doi.org/10.1016/j.fishres.2004.07.002.
- 33. Macleod CD (2010) Habitat representativeness score (HRS): a novel concept for objectively assessing the suitability of survey coverage for modelling the distribution of marine species. J Mar Biol Assoc UK.
- 34. MacLeod CD, Mandleberg L, Schweder C, Bannon SM, Pierce GJ (2008) A comparison of approaches for modelling the occurrence of marine animals. Hydrobiologia 612: 21-32. doi:https://doi.org/10.1007/s10750-008-9491-0.
- 35. Lambert E, MacLeod CD, Hall K, Brereton T, Dunn TE et al. (2011) Quantifying likely cetacean range shifts in response to global climatice change: implications for conservation strategies in a changing world. Endangered Species Res 15: 205-222. doi:https://doi.org/10.3354/esr00376.
- 36. Carvalho SB, Brito JC, Crespo EG, Watts ME, Possingham HP (2011) Conservation planning under climate change: Toward accounting for uncertainty in predicted species distributions to increase confidence in conservation investments in space and time. Biol Conserv 144: 2020-2030. doi:https://doi.org/10.1016/j.biocon.2011.04.024.
- 37. Rose N-A, Burton PJ (2009) Using bioclimatic envelopes to identify temporal corridors in support of conservation planning in a changing climate. Forest Ecol Manag 258, Supplement: S64-S74. doi:https://doi.org/10.1016/j.foreco.2009.07.053.
- 38. Nativi S, Mazzetti P, Saarenmaa H, Kerr J, Tuama ÉÓ (2009) Biodiversity and climate change use scenarios framework for the GEOSS interoperability pilot process. Ecol Inform 4: 23-33. doi:https://doi.org/10.1016/j.ecoinf.2008.11.002.
- 39. Dawson TP, Jackson ST, House JI, Prentice IC, Mace GM (2011) Beyond Predictions: Biodiversity Conservation in a Changing Climate. Science 332: 53-58. doi:https://doi.org/10.1126/science.1200303. PubMed: 21454781.
- 40.
Jarnevich CS, Stohlgren TJ, Barnett DT, Kartesz J (2007) A simple framework for an invasive species early warning system for counties abs; San Jose McEnery Convention Center – San Jose. California: Ecological Society of America. p. 1.
- 41. Powell M, Accad A, Shapcott A (2005) Geographic information system (GIS) predictions of past, present habitat distribution and areas for re-introduction of the endangered subtropical rainforest shrub Triunia robusta (Proteaceae) from south-east Queensland Australia. Biol Conserv 123: 165-175. doi:https://doi.org/10.1016/j.biocon.2004.11.004.
- 42. Harley CD, Randall Hughes A, Hultgren KM, Miner BG, Sorte CJ et al. (2006) The impacts of climate change in coastal marine systems. Ecol Lett 9: 228-241. doi:https://doi.org/10.1111/j.1461-0248.2005.00871.x. PubMed: 16958887.
- 43. Meredith MP, King JC (2005) Rapid climate change in the ocean west of the Antarctic Peninsula during the second half of the 20th century. Geophys Res Lett 32: L19604. doi:https://doi.org/10.1029/2005GL024042.
- 44. Dore JE, Lukas R, Sadler DW, Karl DM (2003) Climate-driven changes to the atmospheric CO2 sink in the subtropical North Pacific Ocean. Nature 424: 754-757. doi:https://doi.org/10.1038/nature01885. PubMed: 12917678.
- 45. Manabe S, Stouffer RJ (1995) Simulation of abrupt climate change induced by freshwater input to the North Atlantic Ocean. Nature 378: 165-167. doi:https://doi.org/10.1038/378165a0.
- 46. Jacobs S (2006) Observations of change in the Southern Ocean. Philos Trans A Math Phys Eng Sci 364: 1657-1681. doi:https://doi.org/10.1098/rsta.2006.1794. PubMed: 16782605.
- 47. Curry R, Dickson B, Yashayaev I (2003) A change in the freshwater balance of the Atlantic Ocean over the past four decades. Nature 426: 826-829. doi:https://doi.org/10.1038/nature02206. PubMed: 14685235.
- 48. Levitus S, Antonov JI, Boyer TP, Stephens C (2000) Warming of the World Ocean. Science 287: 2225-2229. doi:https://doi.org/10.1126/science.287.5461.2225.
- 49. Burrows MT, Schoeman DS, Buckley LB, Moore P, Poloczanska ES et al. (2011) The Pace of Shifting Climate in Marine and Terrestrial Ecosystems. Science 334: 652-655. doi:https://doi.org/10.1126/science.1210288. PubMed: 22053045.
- 50. Engler R, Guisan A (2009) MIGCLIM: Predicting plant distribution and dispersal in a changing climate. Divers Distrib 15: 590-601. doi:https://doi.org/10.1111/j.1472-4642.2009.00566.x.
- 51. Mokany K, Richardson AJ, Poloczanska ES, Ferrier S (2010) Uniting marine and terrestrial modelling of biodiversity under climate change. Trends in Ecology &. Evolution 25: 550-551.
- 52.
Comely CA (1978) Modiolus modiolus ( L.) from the Scottish West coast. I. Biology. Ophelia 17: 167 - 193.
- 53.
Rees EIS (2009) Assessment of Modiolus modiolus beds in the OSPAR area. Prepared on behalf of Joint Nature Conservation Committee.
- 54.
Seed R, Brown RA The influence of reproductive cycle, growth and mortality on the population structure in Modiolus modiolus ( L.) Cerastoderma (Cardium) edule (L.) and Mytilus edulis (L.) (Mollusca: Bivalvia). In: (Barnes He, editor; 1975. . Aberdeen University Press. pp. 257-274.
- 55. Seed R, Brown RA (1978) Growth as a strategy for survival in two marine bivalves, Cerastoderma edule and Modiolus modiolus. J Anim Ecol 47: 283-292. doi:https://doi.org/10.2307/3936.
- 56.
Hill J, Pearce B, Georgiou L, Pinnion J, Gallyot J (2010) Meeting the MPA network Principle of Viability: Feature specific recommendations for species and habitats of conservation importance. Natural England Commissioning Reports. p.
- 57. Strain EMA, Allcock AL, Goodwin C, Maggs CA, Picton BE, et al. (2012) The long-term impacts of fisheries on epifaunal assemblage function and structure, in a Special Area of Conservation. Journal of Sea Research 67: 58-68.
- 58.
CBD (2010) 2010 Biodiversity Target: Goals and Sub-Targets. http://www.cbd.int/2010-target/goals-targets.shtml. Accessed: January 2013.
- 59.
IUCN (2003) Recommendations of the Vth IUCN World Parks Congress. Gland: IUCN.
- 60. Metcalfe K, Roberts T, Smith RJ, Harrop SR (2013) Marine conservation science and governance in North–West Europe: Conservation planning and international law and policy. Mar Policy 39: 289-295. doi:https://doi.org/10.1016/j.marpol.2012.12.002.
- 61. Micheli F, Saenz-Arroyo A, Greenley A, Vazquez L, Espinoza Montes JA et al. (2012) Evidence That Marine Reserves Enhance Resilience to Climatic Impacts. PLOS ONE 7: e40832. doi:https://doi.org/10.1371/journal.pone.0040832. PubMed: 22855690.
- 62.
HM Government (2012) Marine Strategy Framework Directive Consultation: UK Initial Assessment and Proposals for Good Environmental Status.
- 63.
GEBCO (2011) The General Bathymetric Chart of the Oceans (GEBCO)_08 Grid, version 20100927. http://www.gebco.net. Accessed: May 2012.
- 64.
Antonov JI, Seidov D, Boyer TP, Locarnini RA, Mishonov AV et al. (2010) World Ocean Atlas 2009, Volume 2: Salinity. In: S. Levitus. NOAA, Atlas NESDIS 69, Volume 2, Volume 2. Washington, D.C.: U.S. Government Printing Office. p. 184.
- 65.
Connor DW, Gilliland PM, Golding N, Robinson P, Todd D et al. (2006) UkSeaMap: the mapping of seabed and water column features of UK seas. Joint Nature Conservation Committee. http://www.searchmesh.net/default.aspx?page=1922. Accessed: January 2012.
- 66. DTI (2004) Atlas of UK marine renewable energy resources: technical report. R. 1106R.1106: 154.
- 67.
Digimap Marine (2012) UK Hydrographic Office Charted Raster Navigation Charts. . http://digimap.edina.ac.uk/charteddownloader/. Accessed: April 2012.
- 68.
BODC (1998) UK Digital Marine, Atlas (UKDMAP), version 3 ed.