Our modelling approach involved the following components.
1. The output from an existing hydrodynamic model covering Scottish waters and the compilation of a climatological flow-field to represent "average" conditions.
2. Proposed MPA locations as "source" and "target" areas for the dispersal of individual species (at the relevant life stage for dispersal).
3. Species life cycles divided into categories, based on common biological characteristics that may influence dispersal patterns, such as the duration of larval phase/settlement window and season of spawning.
4. Simplistic Individual-Based Models that allow the characterisation at individual level of the origin, destination and trajectory of particles representing PMF larvae, and could also be used to simulate the interplay between physical transport and biological characteristics such as development, mortality and "behaviour", although such interactions were not taken into account here largely due to lack of reliable relevant biological information.
5. Simulation results processed to quantify connectivity and export/import out of/into proposed MPA locations to assess the most suitably located sites and the replication needed.
With respect to 3, year was split into quarters to consider approximate spawning windows, whilst PLD was split into 11 daily time intervals for the 18 PMF species considered. The lack of information on even the most basic behavioural attributes in most cases, such as vertical distribution in the water column, meant that species behaviour could not be considered. An existing hydrodynamic model covering Scottish waters and the compilation of a climatological flow-field to represent "average" conditions was used to predict larval transport. Proposed MPA locations were considered as "source" and "target" areas for the dispersal of individual species (at the relevant life stage for dispersal).
Input Data for PMF Species
A summary of the spawning times and pelagic larval duration ( PLD) of these eighteen benthic species that have been important in the possible Scottish MPA network selection is given in Table 2, together with the key supporting literature. As most of these species spawn over a few months, spawning time was considered by season. Some cnidarian settle within ten days of release, whilst the larvae of spiny lobster may drift in the plankton for up to six months. Many bivalves have a one or two month larval duration. Consequently, PLD categories were derived on the basis of these reported ranges. Spawning locations of PMF species were based on the feature under consideration being identified in site descriptions ( http://www.scotland.gov.uk/Topics/marine/marine-environment/mpanetwork/MPAParliamentReport). Due to the choice of species, connectivity among some of the proposed MPAs was not considered. Figure 2 shows the distribution of proposed MPAs and the location of MPAs important in the present study (shaded green).
Figure 2: Location of possible Nature Conservation MPAs. Shaded areas refer to MPAs identified as important to benthic invertebrate PMF species, which have been proposed to the Scottish Government. Stipled MPA location refers to a proposed MPA still under review that contains benthic invertebrate PMF species.
Area code descriptions follow overleaf.
|FSC||Faroe-Shetland Channel||I & II|
|FSS||Faroe-Shetland Sponge Belt||I, II & V|
|CFL||Central Fladen (core)||II|
|ECC||East Caithness Cliffs||II|
|EGM||East of Gannet and Montrose Fields||II|
|FOF||Firth of Forth Banks Complex||II|
|FTH||Fetlar to Haroldswick||II|
|MTB||Mousa to Boddam||II|
|NSP||Norwegian boundary sediment plain||II|
|WYR||Wyre and Rousay Sounds||II|
|WSK||Windsock||II & III|
|CSS||Clyde Sea sill||III|
|DLA||Lochs Duich, Long & Alsh||III|
|EPL||Eye Peninsula to Butt of Lewis||III|
|GLE||Gairloch and Wester Loch Ewe||III|
|LFG||Upper Loch Fyne and Loch Goil||III|
|NWS||North-west sea lochs and Summer Isles||III|
|SEB||Shiant East Bank||III|
|SJU||Loch Sunart to the Sound of Jura||III|
|STM||Skye to Mull||III|
|BHT||The Barra Fan and Hebrides Terrace Seamount||III & V|
|GSH||Geikie Slide and Hebridean Slope||III & V|
|SSH||South-west Sula Sgeir Slide and Hebridean Slope||III & V|
|RBS||Rosemary Bank Seamount||V|
Spawning times and pelagic larval duration of the benthic PMF. Spawning time key: W=winter, S=spring, Su=summer, A=autumn, NK=not known (all seasons assumed). *reference relates to a similar species, as no other published information is available.
|Species||Spawning time||Settlement window (d)||Reference Number|
|min||max||spawning time||'settlement window'|
|White cluster anemone||AW||1||10||70||70, 71|
|Northern feather star||NK||1||10||-||5*, 44*|
|Pink soft coral/sea fingers||Su||1||10||30, 38||7*, 9*, 13*, 14*, 20*, 28*, 74*|
|Northern sea fan||SuW||1||10||34||86|
|Heart cockle||A||1||30||61||64*, 56*, 52*, 68*|
|Gravel sea cucumber||SSu||1||30||40||36*, 3, 54*|
|Native oyster||Su||10||30||41, 45, 87, 89||1, 8, 41, 82|
|Small brackish water snail||SSu||20||30||5, 23||23|
|Horse mussel||AWSSu||30||40||10, 11, 15, 16, 75||59*, 67, 81|
|Fan mussel||SSuA||30||50||51*, 60*, 77||60*|
|Flame shell||Su||20||60||73*, 79||46, 83|
|Heart sea urchin||SuA||40||60||12, 21, 22,||12, 43*, 50*, 58*|
|Tall sea pen||AW||28||65||18, 19, 37||25*, 35*, 47*, 69*, 84*|
|Burrowing sea anemone||SSuA||28||90||63||25*, 35*, 47*, 69*, 84*|
|Crayfish, spiny lobster||Su||60||180||2, 39||2, 39, 55, 76*|
Year-specific daily 3-dimensional flow-fields for an area between 50-65° N latitude and 15° W - 15° E longitude were obtained by running the SNAC model (Logemann et al., 2004) for 16 years, between 1995-2010, forced with air pressure data from the European Centre for Medium-Range Weather Forecasts ( ECMWF) Operational Data set. Daily 16-year averages were then calculated for each hydrodynamic model grid node. M 2 tidal velocities were superimposed onto residual currents at each node. The spatial resolution of the model was 0.125° latitude by 0.250° longitude, corresponding approximately to < 15 km in our model domain, with 11 fixed (Z) vertical layers.
Due to the lack of detailed biological information for most PMF species, further simplifying assumptions were made within the bio-physical model. Particles representing larvae of the invertebrate species of interest were released at 5 km regular spacing, within the MPAs under consideration. Particles were only released from start positions considered "wet" (i.e. water depth > 0m), based on the model bathymetry. As a consequence, five MPAs were excluded from the simulations because they were too close to the coast to be resolved by the model. Rosemary Bank and other deep water locations west of 15° W were also disregarded, but this is unlikely to affect the outcome of our study because PMF species were largely restricted to the European continental shelf. One hundred particles were released from each start location within each MPA, making up a total of just under 290,000 particles per simulation. The simultaneous release of multiple particles from each point was necessary for numerical stability to account for the stochastic effect of horizontal diffusion. Spawning times were assigned to seasons (spring, summer, autumn and winter) and represented by single particle releases at the mid-point of each season (calendar days 80, 172, 264 and 355, respectively). Particles were kept at a constant depth (25 m) throughout the simulations, which were run for a total of 180 days. The tracking time-step was one hour and particle positions were stored at daily intervals. The particle tracking methodology has been described in detail by Gallego and Heath (2003) and Heath and Gallego (1998, 2000).
Analysis of Simulation Results
The simulations described above were common to all PMF species, so the stored model results were queried off-line on the basis of simplified biological information (Table 2), to extract data applicable to individual species. The criteria used were spawning season (one or more seasons, depending on information in the literature; when the timing of spawning was unknown, all seasons were selected), approximate settlement time window (≤ PLD) and origin MPA. Origin MPA was identified on the basis of species presence data in the 2011 GEMS database and species identified as an important feature in the selection guidelines for each MPA. We assumed a uniform distribution over the whole MPA area of each species present.
Based on the above criteria, the outcome of the bio-physical simulations was queried for each species, to extract the tracks that originated in the relevant MPA(s) in the appropriate season(s). The final particle positions at the end of the settling period were recorded, as an indication of the export potential of that species to other (protected and non-protected) areas. The presence of particles on any MPA (including their origin MPA) during the settlement period was also recorded, as an indication of connectivity between MPAs. These results were displayed as maps. In addition, the relative connectivity between source and sink MPAs, as percentage of all particles released at each source MPA was shown as colour matrices for each species (where multiple spawning seasons occurred, we only present cumulative connectivity plots over all relevant seasons). We also produced Tables for each species with summary statistics (mean and standard deviation) of the dispersion distance from each origin MPA, assuming a straight line between the origin and destination at the end of the PLD. Note that several MPAs could be visited by the same particle during the settlement period (we measured potential contact and made no assumptions about how a pelagic larva would decide to settle and finalise its pelagic stage sometime within its settlement window). Also, we quantified contacts with all MPAs along the drift track, regardless of whether a species had been recorded on a given MPA, as a nil record does not necessarily preclude the presence or potential presence of the species on that area ( MPA suitability for any given species was not examined here). Finally, as particle positions were only stored daily for post-processing, it is possible that we missed particles over MPAs between position recording intervals. Note that, as the focus of the study was to investigate connectivity between and export potential from MPAs, we did not consider the potential export of larvae from non-protected areas.
Analysis of General Patterns
A Generalised Additive Model ( GAM) was fitted to mean transport distance estimates (response variable) derived for each species by MPA and quarter (season) combination. Maximum pelagic larval duration and distance to shore were considered as continuous explanatory variables and OSPAR region and quarter were treated as factors. As the effect of explanatory variables may not have been linear, these terms were treated as splines within the GAM. A gamma response distribution coupled with a log-link function was chosen due to the increasing variance in the response variable with the explanatory covariates. A minimum adequate model was derived by removing terms from the full models successively, comparing successive models with an ANOVA with an F statistic.