Investigating the role of larval dispersal models in the development of an 'ecologically coherent' network of deep sea marine protected areas

Abstract

There is currently worldwide pressure to establish Marine Protected Area (MPA) networks which are self-sustaining and will persistently protect habitats and species. In order for MPA networks to be effective, the species targeted for conservation must be able to disperse between protected areas and maintain a gene-flow necessary for population sustainability and persistence. This warrants new research on how to quantify and map faunal dispersal to ensure that protection will be effective and sustainable.

Population genetic methods have merit, with the ability to track parentage and gene flow between areas directly. However the costs, quantity of samples, and time required to genetically quantify dispersal for multiple species make these approaches prohibitive as the only method of assessment, especially in relatively inaccessible offshore waters. Dispersal modelling is now becoming more accessible and may fulfil immediate needs in this field (although ground truthing will be necessary in the future).

There have been very few dispersal modelling studies focussed on deep sea or offshore areas, predominantly due to the lack of high resolution hydrodynamic models with sufficient geographic extent away from shore. Current conclusions have been drawn based on shallow water coastal studies, informing offshore MPA network size and spacing. However the differences between these two environments may mean that dispersal abilities are not comparable. Deep water receives less influence from wind and weather, and the scales are vastly different in terms of a) the depth ranges covered, b) the planktonic larval durations (PLDs) of animals, and c) the geographic areas concerned as a consequence.

Global hydrodynamic models with reasonable resolution are now becoming more accessible. With the outputs from these models, and freely available particle simulators, it is becoming more practical to undertake offshore deep water dispersal studies.

This thesis aims to undertake an analysis of these accessible modelling tools within a deep sea context. The guidelines which are currently available to dispersal modellers are yet to encompass the needs of deep water modellers which may require some additional considerations given the extended depth range covered and the different hydrodynamic drivers away from the air/sea interface.

Chapter 1 reviews the larval dispersal process, the factors which may affect dispersal success, and those which should be incorporated into future predictions of dispersal. The current methods for assessing larval dispersal are explored covering genetics, elemental tagging and modelling approaches with an extended look at modelling considerations. Existing marine conservation policy is also touched on in the context of connectivity and larval dispersal.

Chapter 2 is designed to inform future deep sea modellers on how to parameterise and understand a dispersal model. As models appear as a ‘black box’ to the majority of users, sensitivity tests can offer a way of scaling model inputs and tempering expectations from model outputs. A commonly used model pairing (the HYCOM hydrodynamic model and the Connectivity Modeling System) is assessed, using parameters which link to the temporal and spatial scales of mixing in the modelled system: timestep of particle tracer, horizontal and vertical positioning of release points, release frequency of larvae, and temporal range of simulation. All parameters were shown to have a decreased sensitivity with depth, with patterns reflecting local watermass structure. Future studies observing similar hydrodynamic conditions seeking to optimise their model set up would be advised to stratify their model release locations with depth. A means to incorporate all sensitivity test results into optimal input parameters for future studies is demonstrated.

Chapter 3 investigates whether dispersal models provide any advantage over a “sphere of influence” estimate based on average current speeds and PLDs: there is no use pursuing dispersal modelling if the outputs are too erroneous to provide any advantage over a back-of-the-envelope calculation. This chapter examines the outputs of two dispersal models driven by two different hydrodynamic models in order to observe the variability in prediction between models. This model comparison revealed a greater disparity between hydrodynamic model predictions than has been previously understood by ecologists. The two models compared (POLCOMS and HYCOM) may equally be considered as suitable to promote realism in the study region, but slight differences in resolution and numerical error handling resulted in dispersal predictions from which opposing conclusions can be drawn. This chapter therefore emphasises the necessity for model ground truthing before predictions can be trusted.

Chapter 4 assimilates the findings of the previous chapters and applies their advice to a study of MPA network dispersal connectivity. Using the hydrodynamic model which performed best in chapter 3 (HYCOM), a simulation was undertaken for cold water coral (Lophelia pertusa (Linnaeus 1758)) larval dispersal between already established MPAs in the NE Atlantic. As larval characters have only been observed ex situ, dispersal was simulated using two null models (passive and active vertical migration) and averaged to provide an intermediate prediction. A method for assessing dispersal within MPAs and MPA networks is offered based on the intermediate prediction, as well as a network wide assessment of the difference in dispersal patterns for passive and active larvae. It was found that the existing network performs well at supplying larvae to non-networked sites, but performs poorly at supplying other MPAs. The ‘best’ MPAs were central to the network and facilitated the traverse of regional gaps in suitable habitat. The ‘worst’ MPAs were peripheral to the network and small in size. Network-wide passive and active dispersal matrices had no significant difference between them. However site specific variability in the effect of vertical migration was detected subject to variability in local topographic barriers to dispersal, only some of which could be surmounted with vertical migration.

All chapters aim to inform future deep sea dispersal modellers, and encourage exploration of this tool in other contexts, as well as marine conservation. The thesis cautions against the transplantation of shallow water assumptions to deep water environments, and advocates region specific studies and mandatory ground truthing of predictions. An upcoming study will ground truth the findings of this thesis with both genetic and oceanographic data, allowing the accuracy of study results to be quantified.