This study was carried out in the English Channel (Fig. 1), a cold-temperate epicontinental sea separating the south coast of the UK from the north coast of France (Delavenne et al., 2012). The English Channel constitutes a biogeographical transition zone between the warm temperate Atlantic oceanic system, and the boreal North Sea and Baltic Sea continental systems of northern Europe, encompassing a wide range of ecological conditions (Coggan & Diesing, 2011; Delavenne et al., 2012). The study region focused on the eastern English Channel (EEC), which is delimited by the Dover Strait to the east and Cotentin Peninsula to the west and is a key area for tourism, shipping, energy production and aggregate extraction (Carpentier et al., 2009). In addition, it supports an important commercial fishery, as well as key nursery, spawning areas and migratory routes linked to specific environmental characteristics (Martin et al., 2009).
Figure 1. EUNIS levels 3 and 4 habitat map for the eastern English Channel showing the location of the 1314 sampling points. See Table S1 for a key to EUNIS habitat codes, levels and descriptions.
Download figure to PowerPoint
There are several ongoing MPA designation projects in this section of the English Channel. Both France and the UK have implemented MPAs as part of their EU Birds and Habitats Directive commitments, and France is currently developing an MPA network in the ‘Three Estuaries region’ (Bay of Somme, Authie and Canche; Fig. 1). In addition, the EEC is the focus of the Balanced Seas project (http://www.balancedseas.org/), which is one of the four regional MCZ projects, which seeks to identify and recommend MPAs for the inshore and offshore waters of south-east England (JNCC & Natural England, 2010). Balanced Seas uses habitat targets based on the SAR that were developed at a national level from biodiversity data collected in English waters (JNCC & Natural England, 2010).
We used a broad-scale habitat map in this analysis, which is based on the European Nature Information System (EUNIS) habitat classification hierarchy developed by the European Environment Agency (EEA, 2006; Coggan & Diesing, 2011). Figure 1 shows the distribution of each EUNIS habitat class that was modelled using physical and environmental data, including depth, substratum and energy levels. Rock habitats were modelled to level 3 in the EUNIS hierarchy, whilst sediment habitats were modelled to level 4 (Coggan & Diesing, 2011). The EUNIS level 3 habitats are broken down into three habitat types and coded as follows: infralittoral rock (A3.x), circalittoral rock (A4.x) and sublittoral coarse sediment (A5.x), which was further divided into its finer-scale EUNIS level 4 habitats (A5.xx).
Biodiversity survey data
Given the importance of macrobenthic diversity in the EEC (Vaz et al., 2007; Carpentier et al., 2009), the increasing emphasis on their conservation (Sanvicente-Anorve et al., 2002; Vincent et al., 2004) and the large amount of benthic sampling that has taken place (e.g. Desroy et al., 2003; Dauvin et al., 2004; Carpentier et al., 2009), we developed targets using presence/absence data from macrobenthic surveys carried out between 1985 and 2007, providing data from 1314 sampling points (Fig. 1). These surveys used a range of sampling protocols and gear sizes (0.1–0.5 m2), with samples predominantly collected using a Hamon grab, with the exception of 16 stations in the Ridens that used a van Veen grab. The sampling strategy in the study area was predominantly regularly spaced; however, there was more intensive sampling in surveys from the east of the Isle of Wight, in the Ridens and in coastal areas such as between Dieppe and Calais, the Bay of Veys and the Bay of Seine (Fig. 1).
Calculating habitat targets
We calculated habitat targets following the SAR-based approach developed by Desmet & Cowling (2004), which treats the SAR as a power function. Whilst concerns about using this particular approach in conservation planning have been expressed in the literature (see Smith, 2010 for a detailed review), we employed it in our study because (i) we specifically sought to investigate the uncertainties around this existing approach; (ii) the power function has been shown to perform well for macrobenthic datasets containing between 42 and 1300 samples (Azovsky, 2011).
This approach involves transforming the power function (equation (1)) to estimate the proportion of habitat area required to represent a given percentages of species (equation (2)):
Here, S′ and A′ denote the proportion of species and habitat area respectively (Desmet & Cowling, 2004; Rondinini & Chiozza, 2010), and z describes the slope of the power function, which is the rate of species accumulation with increase in area (Lomolino, 2000; Tjorve & Tjorve, 2008). The constant c is a scaling factor that relates to the size (area) of an individual sampling unit and can be ignored when comparing proportions or percentages of species and area (Desmet & Cowling, 2004; Rondinini & Chiozza, 2010). Thus, it is possible to calculate habitat targets by (i) determining the z-value of the SAR for a given habitat; (ii) using the z-value to calculate the proportion of area required to represent a given percentage of species; and (iii) multiplying this proportion by the total habitat area.
We calculated habitat-specific z-values using the formula for calculating the slope of a straight line (equation (3)), because a SAR modelled with a power function appears as a straight line with slope z on a log-log plot (Desmet & Cowling, 2004).
where y2 = log(total number of species in a habitat class); y1 = log(average number of species per sampling point); x2 = log(total area of habitat class); and x1 = log(average area of sampling points). Three of these variables (y1, x2, x1) are derived from habitat-specific inventory data (Desmet & Cowling, 2004; Rondinini & Chiozza, 2010), so all that is needed to calculate z-values is to estimate the total number of species (y2) in a given habitat type (Desmet & Cowling, 2004).
The habitat map shows the distribution of each EUNIS level 3 habitat type and subdivides the sedimentary habitat types further into finer-scale EUNIS level 4 types (Fig. 1). Thus, we assigned sampling points on rocky habitats to their associated level 3 habitat types and sampling points on sedimentary habitats to both their associated parent level 3 habitat types, and their constituent level 4 habitat types (see Fig. S1 and Table S1 in Supporting Information for more information regarding EUNIS level 3 parent habitats for level 4 habitat types in the EEC). We then calculated targets for each of these level 3 and level 4 habitats using EstimateS software (Colwell, 2009) to generate estimates of total species richness (y2) and determine habitat-specific z-values for each of these habitat types.
Although there is no consensus as to which estimator provides the best predictions when estimating total species richness for a habitat type (or region) from field survey data (Brose, 2002; Herzog et al., 2002; Chiarucci et al., 2003; Walther & Moore, 2005), there is general agreement that the Bootstrap estimator is the most conservative (Colwell & Coddington, 1994; Chiarucci et al., 2001, 2003; Hortal et al., 2006). A prediction of total species richness based on this estimator should be considered as a minimum estimate (Desmet & Cowling, 2004; Rondinini, 2011a), which is why this estimator was subsequently applied by the SANBI and MCZ projects to develop national targets for both terrestrial and marine habitats.
To assess the effect that choice of species richness estimator has on the calculation of conservation targets, we compared targets derived using the Bootstrap estimator to those derived using several alternative nonparametric estimators of species richness – ICE, Chao2, Jackknife1 and Jackknife2. Whilst these alternative estimators were investigated by both Desmet & Cowling (2004) and Rondinini (2011a), these authors did not explicitly test their effect on target setting (see Colwell & Coddington, 1994; Gotelli & Colwell, 2001; Hortal et al., 2006; Colwell, 2009 for more details on these estimators and their performance). Our comparison involved calculating each richness estimate based on the mean of 1000 estimates that used 1000 randomizations of sample accumulation order without replacement (Colwell, 2009). We then used these results to (i) calculate the proportion of habitat area required to represent 80% of species, hereafter referred to simply as ‘targets’, for each habitat type with > 5 sampling points – we chose to calculate targets based on representing 80% of species because this was used by the Balanced Seas and the other regional MCZ projects (JNCC & Natural England, 2010); (ii) estimate the number of sampling points required to produce a stable target for each habitat type, and each richness estimator, where a target was defined as stable if it exhibited a standard deviation of < 5% (as used by Desmet & Cowling, 2004); (iii) assess how the targets developed in this study compare with those from the MCZ project in the EEC; and (iv) assess how sensitive each of the estimators was to sample size effects using successively larger numbers of accumulated sampling points, which involved dividing the percentage target for each habitat type based on 100, 200 and 300 sampling points by the percentage target based on 50 sampling points (we then took the mean of each of these habitat results for each estimator to show how relative target size changed with sample size).
Finally, we investigated the effects of using different levels of habitat classification on the extent of the MPA network needed to meet the targets. This involved multiplying each habitat target by the extent of its occurrence in the planning region to provide an area target in km2 and then summing these area targets from EUNIS level 4 habitats belonging to the same ‘parent’ level 3 type, so that the combined level 4 result could be compared with the level 3 result.