Protected areas are an important conservation measure. However, there are controversial findings regarding whether closed areas are beneficial for species and habitat conservation as well as for harvesting. Species dispersal is acknowledged as a key factor for the design and impacts of protected areas. A series of agent-based models using random diffusion to model fish dispersal were run before and after habitat protection. All results were normalized without the protected habitat in each scenario to detect the relative difference after protecting an area, all else being equal. Model outputs were compared with published data regarding the impacts over time of MPAs on fish biomass. In addition, data on species' dispersal potential in terms of kilometres per year are compared with model outputs. Results show that fish landings of species with short dispersal rates will take longer to reach the levels from before the Marine Protected Areas (MPAs) were established than landings of species with long dispersal rates. Further, the establishment of an MPA generates a higher relative population source within the MPA for species with low dispersal abilities than for species with high dispersal abilities. Results derived here show that there exists a feasible win-win scenario that maximizes both fish biomass and fish catches.
Habitat protection is a complex issue which has only recently achieved high public visibility (UN, 2010). In marine environments it covers many aspects, such as conservation of juvenile fish habitats, protection of corals, and development of marine recreational parks or dive sites. Fishing is often seen as a destructive force, and habitat destruction by fishing practices has to be considered in any comprehensive management plan (Jones et al., 2011). Habitat protection can be total or partial. Total closures are often associated with Marine Protected Areas (MPAs) and the designation of certain areas for alternate uses such as recreation.
Closing an area affects several stakeholders. Closed areas are of interest to biologists, conservation scientists, land use planners, but also to fishermen and the fishing industry in general as well as the tourism industry (Ami et al., 2005; Rees et al., 2010b). While there are cases where closed areas are beneficial for species and habitat conservation (Jones et al., 2011; Seytre and Francour, 2014), there are also studies that question the benefits of closures from an economic perspective regarding fish landings (Gårdmark et al., 2006; Jones et al., 2011). This in turn has implications for both food security (Pauly et al., 2005) and economic impacts on fisheries (Eide et al., 2003, 2011; Jentoft and Eide, 2011). Thus a win-win scenario in terms of both increased fish biomass and increased fish landings after establishing an MPA is ideal (Rees et al., 2010a) but questionable.
The design of MPAs involves specifying the total surface area to be protected, the distribution in space of that area, and its connectivity (Moustakas and Silvert, 2011). That leaves a fairly wide range of choices: there is controversy about whether single large reserves are more effective than several smaller ones of the same total area, whether edge effects diminish their efficacy, and whether closely spaced reserves are more effective than distantly spaced ones (Moustakas and Silvert, 2011). It is acknowledged that dispersal is a key factor in designing MPAs (Coleman, 2013; Underwood et al., 2013). The reasons behind dispersal being a key factor (Lewis et al., 2013) are that (i) MPAs should be large enough so that adults can stay long enough inside them, but how large is large enough is clearly related to dispersal potential; (ii) MPAs should be close enough so that larvae can move between them, but how close is close enough is also related to dispersal potential.
Assuming dispersal to be an important factor in determining the ability of species to reach the protected areas, the impacts on species with different dispersal abilities may vary in time from the establishment of an MPA (Claudet et al., 2008; Silvert and Moustakas, 2011) for various reasons related to species growth rates or the ability of species to reach or remain within the MPA. Here, assuming all other factors that influence the efficacy of MPAs remain equal, the impacts of MPA(s) on biomass inside and outside the reserves, as well as on landings over time to species with different dispersal abilities, are investigated. In an effort to provide the relative differences in fish biomass and fish landings with and without MPAs, agent-based simulation modelling is used (Moustakas and Evans, 2015; Moustakas and Silvert, 2011) to model migration (Schönfisch and Kinder, 2002) via diffusion (Augustijn et al., 2016). Model outputs of each simulation scenario after the establishment of an MPA were normalized by model outputs of the same scenario prior to the establishment of an MPA in order to detect relative changes before and after closing an area.
A simulation model is used to predict the efficacy of MPAs as a function of species' dispersal potential and different catch rates across two different MPA spatial design scenarios. All results presented here (regarding fish biomass and annual catch) were normalized to 100 % in the steady-state situation without the MPA in each scenario. Thus, results presented here are presented as dimensionless numbers. Clearly, results from field studies are expected to differ in their values but in comparison with field data the shape of the curves should be at least similar. The model assumes that fish move around at random (Blackwell, 1997). Such a modelling attempt can serve as a null model (Silvert and Moustakas, 2011) and potentially as a minimal model for pattern formation (Petrovskii and Malchow, 1999). This is a conservative (and often an unrealistic) approach as many species exhibit directed dispersal by seasonal migration between feeding and spawning areas. However there are also species that exhibit such dispersal behaviour such as littoral fish species (estuarine fish, intertidal fish, coral reef fish), and the fishery that is mainly involved with this type of fishing is trawl and recreation fishing (Mant et al., 2006). In addition, habitat-dependent species like coral reef species (e.g. clownfish, anemonefish, and damselfish) are also characterized by this type of movement. The active fishery that is predominantly linked to this type of fish is artisan fishing (Campbell and Pardede, 2006).
The model follows previous modelling attempts in which a full description is
provided (Moustakas and Silvert, 2011; Moustakas et al., 2006), modified
accordingly here so that dispersal is random. The model is run on a square
grid with 100
Population growth occurs at each time step with a constant (time- and
space-independent) growth rate
Fish movement is random with an equal probability of diffusing to the eight
adjacent neighbouring cells. The probability of migrating to one of the
eight neighbouring cells is multiplied by
For each time step
There are no periodic boundary conditions meaning that fish located in the four corner cells of the simulation grid may move only to their three neighbouring cells.
The spatial design of MPA(s) included two different scenarios: a single large
and two small MPAs totalling the surface of the single large MPA, and in each
case the same total surface area was protected. The total protected surface
area spanned from 1 % up to 20 % of the simulation surface area. In
all cases mortalities
In order to examine relative differences with and without MPAs, each
simulation scenario is replicated with a common parameter space,
The simulation scenarios examined here (parameter space) include fish
dispersal coefficients
In order to constrain model outputs with data (Moustakas and Evans, 2015), published data regarding fish biomass of fish species pre- and post-MPA establishment were used from the California Channel Islands, USA, including five fish species (see next paragraph for details regarding species) (Karpov et al., 2012) for model validation. The data included species-specific biomass data before and after MPA establishment (Karpov et al., 2012), allowing comparisons of impacts over time, as well as within and outside the protected area after the MPA was established from 2003 to 2008, allowing comparisons inside and outside the protected area after MPA establishment. Further, the data set also provides statistics on landings of commercial species before and 3 years after the establishment of MPAs.
The species-specific landings post-/pre-MPA establishment were regressed
against their dispersal potential. Dispersal potential of each species was
retrieved from the following published studies:
In order to link model predictions with marine species dispersal potential ,
thus predict the time impacts on landings of different species groups,
analysis on (adult) marine taxa dispersal data was conducted. The data
derived from a meta-analysis of 1897 publications (Moustakas and Karakassis,
2005, 2009): Within this data set a search regarding dispersal rate of
species was conducted. From the 1897 publications, only the ones that
explicitly mentioned dispersal rates per species and length of the study so
that dispersal can be normalized as km yr
An empirical cumulative density function (ECDF) was used to evaluate the
dispersal range of each species (in kilometres) against the percentage of
species in the data set that have a dispersal potential less than or equal to
that value. The ECDF
Model outputs showed that recovery of landings (in comparison to the levels
of pre-MPA establishment) was faster for species with high dispersal rates
than for those with low dispersal rates. This applies to both single large
and multiple small MPA spatial designs for mortalities (
Statistical analyses of fish density data post- and pre-MPA establishment
showed that landings of commercial fish species in post-MPA establishment
divided by the landings of the same species pre-MPA establishment regressed
against the dispersal potential of each species. This showed that in the case
of the five commercial fish species that were examined, the relative change
in landings post-normalized by pre-MPA establishment was more pronounced in
species with longer dispersal rates (Fig. 3a;
Highest dispersal potential is exhibited amongst the phyla of Gadirormes,
Crustaceans, Perciformes, Echinodermata, Mollusca, and Pleuronectiformes
(Fig. 4a). With the exceptions of Gadirormes and Pleuronectiformes, phyla
with high dispersal rates have a high variation of dispersal rates between
individual species within the phylum (Fig. 4a). The majority of phyla
examined have dispersal rates of less than 1 km (Fig. 4a). From the species
considered here, 48 % have dispersal rates of < 1 km, while
90 % have dispersal rates of < 200 km (Fig. 4b). Overall,
dispersal rates between species was very high as indicated by differences
between 95 % confidence intervals of the mean
Model outputs derived here depict the relative time needed for fish landings to reach levels from before the establishment of an MPA. The method – normalizing outputs after a change in the system has been introduced by model outputs prior to the change – may serve as a valuable null model tool in ecology and biological sciences in order to investigate the relative effects of a key parameter (here, dispersal on the impacts of MPAs on both fish biomass and landings). Models are used when experiments are costly, require significant labour effort, ethics, and effects of spatial or temporal scales associated. Cellular automata and agent-based models are useful tools for addressing such issues (Bastardie et al., 2013; Convertino et al., 2015; DeAngelis and Yurek, 2015; Eide, 2012, 2014; Moustakas and Silvert, 2011; Moustakas et al., 2006).
Model outputs derived here showed that fish catches are more likely to recover faster at the original levels pre-MPA(s) establishment and above. Statistical analysis of normalized post-/pre-MPA establishment data exhibited a monotonic pattern and faster recovery of landings of long dispersers – data were available for 5 species and 5 years after closures. Previous spatially explicit studies on population recovery after disturbance have indicated that long dispersers recover more homogeneously than short dispersers (Johnson et al., 2001; Reed et al., 2000), and to that end model outputs are in agreement with this. For additional discussion on the interplay between highly mobile fish and the efficacy of MPAs, see also Breen et al. (2015).
Source–sink theory has been applied to the spatial design and impacts of
MPAs (Andrello et al., 2013; Seijo and Caddy, 2008). Results derived here
exhibited that MPAs are increasingly acting as population sources as species'
dispersal range decreases. Species with shorter dispersal rates are likely to
also be smaller in size and/or body mass (Alimov, 2003; Williams, 1999), thus
they benefit more simply from the fact that in all scenarios MPAs had an
equal total size. Clearly, home range areas of short dispersers will be
smaller than those of long-distance dispersers (the model does not account
for individual body length or mass). However, given that species with short
dispersal potential have more restricted distributions (Bradbury et al.,
2008; Curini-Galletti et al., 2012), overall it seems reasonable to expect
that protecting the habitats of short dispersers will create larger
population buffers within the protected area than when protecting the
habitats of long dispersers. Data of movement of lingcod (
According to the results derived here, the abundance of species of phyla with very low dispersal rates such as Porifera, Rhodophyta, Bryozoa, and Anthophyta will be considerably higher within the MPA than outside. The majority of these species are not commercial (and would not be targeted by fishers) but a “blind” fishing method such as trawling would affect them (González-Irusta et al., 2013; Heery and Cope, 2014). Further, several of the short dispersing species are habitat-forming species (Lilley and Schiel, 2006). It should be noted, however, that these conclusions are based upon a fairly large data set (Moustakas and Karakassis, 2005, 2009), but this data set is not exhaustive.
In general the variables used in this work have no units, as they are
normalized. However when comparisons with real fish species are made, since
real
This study shows that in the explored parameter space a win-win scenario in
terms of fish biomass and increase in landings after some years of closing an
area is feasible, but it does not show what the actual parameter space
leading to this result is. It only shows that this is mathematically
possible. Despite the fact that the results presented here are unitless
(ratio), the sensitivity to the scale of analysis has not been accounted for
(Gautestad 2013) in terms of multiscale modelling (Duan et al. 2014). A ratio
is scale-free, but the actual processes as they are defined here are not.
There are several scales involved:
The time interval of simulation (10 years) may seem short, because the effect of MPAs is usually visible after long time intervals (Claudet et al., 2008) and the lifespan of some species may exceed this time. Moreover, Figs. 1a and 2a suggest that with a longer time interval more curves could reach the 100 % target. However, in general there are several behavioural changes in fishers after establishing an MPA (e Costa et al., 2013). While it would be interesting to know whether landings attain the levels observed before the implementation of MPAs (convex curves for high dispersal distance) and how long this will take, other acting processes such as increased fishing pressure (García-Rubies et al., 2013) or phenotypic evolution (Diaz Pauli and Heino, 2014; Moustakas and Evans, 2013) also occur, thus long-term outputs are unlikely to be realistic. Thus, the model was only run long enough to discern some variability between species' dispersal abilities.
Additionally in the simulation grid, the corner cells get inputs only from their three neighbouring cells, giving a lower growth at the edge of the area as no periodic boundary conditions were used. For a view on scaling issues in gridded models and model structure with scenario boundary conditions, see discussion in (Millington et al., 2011; Moustakas and Evans, 2015).
There are very large differences in the dispersal potential of species as
indicated by differences between mean (
Introducing MPAs may lead to a temporary decline of landings, owing to stronger fishing effort outside the protected areas to compensate for lack of fishing inside MPAs. However, over time the source–sink effect – due to a gradual many-fold increase in fish abundance inside the MPAs – may not only gradually make landings from the unprotected fishing areas rise again but even overshoot the pre-MPA level. This result was achieved under overfishing, a 25 % of total mortalities (natural and fishing mortalities) higher than the growth rate as it often happens in reality (Daskalov, 2002; Jackson et al., 2001). Thus, a win-win result is achieved (Rees et al., 2010a): fish and the local ecosystem are protected and can thrive inside protected areas, and the fishing industry will benefit from a net gain after a temporary decline while waiting for the MPA population(s) to increase sufficiently, so that it can become a strong provider of dispersing individuals (Rees et al., 2010a). This win-win scenario needs time (Rees et al., 2010a; Russ and Alcala, 2004), and in general an integration of science and stakeholder-based methods may facilitate such scenarios (Gall and Rodwell, 2016; Ruiz-Frau et al., 2015).
Fast recovery or even overshoot of landings relative to pre-MPA level
basically depend – under the given model design – on two main aspects:
dispersal rate
At present MPAs generally cover much less than 20 % of fishing areas;
consequently this policy need revision in order to achieve the net fishing
gain over time. Other studies suggested that the yield from the harvest
effort is strongly affected by the fraction of area protected from harvesting
and that maximum yield is independent of the size of the protected area
unless the fraction is > 0.56 (Kaitala et al., 2004) The
dependence on
Comments of two anonymous reviewers and the handling editor Ronald Brandl considerably improved an earlier manuscript draft. This paper is dedicated to William (Bill) Silvert with whom I very much would have liked to write the paper. Edited by: R. Brand Reviewed by: A. Eide and one anonymous referee