- Split View
-
Views
-
Cite
Cite
Juliette Delavenne, Kristian Metcalfe, Robert J. Smith, Sandrine Vaz, Corinne S. Martin, Ludovic Dupuis, Franck Coppin, Andre Carpentier, Systematic conservation planning in the eastern English Channel: comparing the Marxan and Zonation decision-support tools, ICES Journal of Marine Science, Volume 69, Issue 1, January 2012, Pages 75–83, https://doi.org/10.1093/icesjms/fsr180
- Share Icon Share
Abstract
The systematic conservation approach is now commonly used for the design of efficient marine protected area (MPA) networks, and identifying these priority areas often involves using specific conservation-planning software. Several such software programmes have been developed in recent years, each differing in the underlying algorithms used. Here, an investigation is made into whether the choice of software influences the location of priority areas by comparing outputs from Marxan and Zonation, two widely used conservation-planning, decision-support tools. Using biological and socio-economic data from the eastern English Channel, outputs are compared and it is shown that the two software packages identified similar sets of priority areas, although the relatively wide distribution of habitat types and species considered offered much flexibility. Moreover, the similarity increased with increasing spatial constraint, especially when using real-world cost data, suggesting that the choice of cost metric has a greater influence on conservation-planning analyses than the choice of software. However, Marxan generally produced more efficient results and Zonation produced results with greater connectivity, so the most appropriate software package will depend on the overall goals of the MPA planning process.Delavenne, J., Metcalfe, K., Smith, R. J., Vaz, S., Martin, C. S., Dupuis, L., Coppin, F., and Carpentier, A. 2012. Systematic conservation planning in the eastern English Channel: comparing the Marxan and Zonation decision-support tools. – ICES Journal of Marine Science, 69: 75–83.
Introduction
The 1992 Convention on Biological Diversity set the ambitious target of establishing, by 2012, a global system of marine protected areas (MPAs) covering 10% of all marine ecological regions, comprising both multiple-use and strictly protected areas. MPAs are increasingly seen as crucial instruments for conserving biodiversity and maintaining fish stocks (Leathwick et al., 2008), and there is evidence of their potential benefits in the management of fisheries (Halpern and Warner, 2002; Gell and Roberts, 2003). In its strategy for the marine environment (EC, 2008), the European Commission (EC) is promoting the idea of marine spatial planning (MSP) providing a framework to improve decision-making and delivering an ecosystem-based approach to the management of marine activities. MSP is also expected to provide a more transparent process of conflict resolution in situations where there are many demands for the use of marine resources and sea space.
This context has led to renewed interest in developing methods for designing efficient MPA networks (Smith et al., 2009). In particular, it is recognized widely that conservation planners need to account for opportunity costs and potential biodiversity loss when designing MPA systems. This has led to the widespread adoption of the systematic conservation-planning approach (Margules and Pressey, 2000; Margules and Sarkar, 2007), which is a target-driven process that aims to identify networks of priority areas for ensuring the representation and long-term persistence of biodiversity (Margules et al., 2002; Leslie et al., 2003). Setting targets helps increase transparency and measure progress, but it also allows socio-economic data to be included in the planning process without influencing or endangering conservation goals. Thus, MPA networks can be designed so that they meet targets, while also minimizing impacts on stakeholders and increasing the likelihood of their successful implementation (Knight et al., 2006).
Systematic conservation planning generally involves (i) producing a list of important species and habitat types, known collectively as conservation features, (ii) setting targets for each of these conservation features, (iii) dividing the planning region into a series of planning units, (iv) calculating the amount of each feature found in each planning unit, (v) assigning a cost value to each planning unit, and (vi) using computer software to identify priority areas for conserving biodiversity, reducing fragmentation levels, and minimizing planning unit costs (Moilanen et al., 2009a). Several conservation-planning software packages have been produced, some of which have been used to design MPA networks (e.g. Leslie et al., 2003; Fernandes et al., 2005; Klein et al., 2008a; Leathwick et al., 2008). However, this has created some uncertainty among practitioners as to whether the location of the identified priority areas varies with the software used. Here, we investigate this issue by comparing results from Marxan and Zonation, two of the most widely used conservation-planning, decision-support tools (Moilanen et al., 2009a).
Marxan uses a minimum-set approach to identify portfolios of planning units that achieve conservation targets at near-minimal cost. It does this by first defining the cost of a portfolio as an objective function made up of: (i) the combined cost of the planning units in the portfolio, which can be a measure of any aspect of the planning unit, such as its area, the risk of being affected by anthropogenic impacts, or the opportunity costs resulting from protection; (ii) a penalty for each unmet target; and (iii) a spatial constraint cost reflecting the portfolio's fragmentation level (Ball and Possingham, 2000; Possingham et al., 2000). The spatial constraint is based on the boundary length of the portfolio, because fragmented portfolios have more of this exposed edge. Reducing this fragmentation involves adding more planning units to the portfolio, producing more viable, but less efficient, results (Ball and Possingham, 2000).
In contrast, Zonation uses a maximum-cover approach that aims to maximize the conservation benefits for a fixed cost specified by the user by first calculating the marginal loss for each of the cells in the planning region (Moilanen et al., 2005, 2009b). It then removes cells one at a time based on maximizing the overall conservation value of the remaining area to produce a conservation-value map based on the hierarchical ranking of the landscape. This conservation-value map then forms the basis of further analyses, and Zonation has a range of options for incorporating connectivity and viability into the prioritization process (Arponen et al., 2006; Cabeza and Moilanen, 2006; Moilanen et al., 2009b).
This means the approaches that underpin Marxan and Zonation are fundamentally different, with Marxan seeking to minimize costs while meeting specified targets, and Zonation seeking to maximize biodiversity benefits given a specified cost. However, the Zonation outputs can be modified to identify priority areas for meeting specified targets, and this is why marine conservation planners have used both software packages to identify MPA networks based on a target-setting approach (e.g. Klein et al., 2008b; Leathwick et al., 2008). Given these differences, one might expect Marxan and Zonation to identify different sets of priority areas, which could create confusion and doubt about the value of both software packages. Alternatively, it might be assumed that results should be similar because areas needed to meet targets will always be selected, and this was found in earlier work that compared outputs between Marxan and C-Plan, another reserve-system-design tool (Carwardine et al., 2007). Further, one might expect similar results using real-world cost data in the analyses, such as information on opportunity costs or threats. This is because these data have a specific spatial pattern within the planning region, so the same low-cost areas containing important biodiversity tend to be selected (Richardson et al., 2006; Nhancale and Smith, 2011). Therefore, data from the eastern English Channel are here used to investigate whether priority-area and conservation-value maps produced by Marxan and Zonation differ and whether the difference is sensitive to the conservation target and type of cost metric used in the analysis.
Material and methods
The English Channel is a shallow epi-continental sea located in the temperate Northeast Atlantic, covering ∼77 000 km2 and separating the south coast of the UK from the north coast of France (Dauvin, 2008). The Channel has two distinct parts, western and eastern, with markedly different oceanographic characteristics and which can be regarded as different ecosystems (Vaz et al., 2007; Coggan and Diesing, 2010). The eastern English Channel (Figure 1) is a biogeographical transition zone between the warm, temperate Atlantic oceanic system and the boreal North Sea, and the area encompasses a wider range of ecological conditions than other European seas (Dauvin, 2008; Carpentier et al., 2009). The area is shallow (<50 m) and strongly influenced by the River Seine. The eastern English Channel is not only important from an ecological perspective, but is of considerable economic value for fisheries, maritime traffic, marine aggregate extraction, and other sectors (Martin et al., 2009).
Mapping physical data
Five environmental parameters were selected to describe the range of ecosystems found in the eastern English Channel: depth, temperature, sediment type, salinity, and bed-shear stress. Depth combined bathymetry and mean sea level. Bathymetric data were derived from SHOM (Service Hydrographique et Océanographique de la Marine) hydrographic charts, and the mean sea level (at mid-tide) was estimated using a hydrodynamic model. Temperature and salinity data were measured in situ during Ifremer's Channel Ground Fish Survey (CGFS) from 1997 to 2006 and were used to estimate anomalies (observed surface temperature or salinity minus the mean for the area surveyed) and bottom–surface differences. Seabed shear-stress estimates were obtained from a two-dimensional hydrodynamic model originally developed for the Irish Sea, but extended to cover the northwest European shelf (Carpentier et al., 2009). Seabed sediment types were extracted from a sediment map of the English Channel (Larsonneur et al., 1982), in which the original 29 sediment classes were aggregated into the following five broader classes: (i) fine sand, (ii) coarse sand, (iii) fine heterogeneous sandy gravel, (iv) coarse heterogeneous sandy gravel, and (v) pebbles. This sediment-type map was used because previous studies in the eastern English Channel showed that benthic invertebrate communities (San-Vicente Añorve, 1995), and fish, cephalopod, and macroinvertebrate species assemblages (Vaz et al., 2007) were related to substratum type.
Depth and seabed shear-stress maps were classified into five types based on quantile values. Temperature and salinity exhibited less variation, so the same approach was used to divide these into three types. This classification system produced maps that contained an equal area of each type, so that each broad range of the physical environment would be represented in the final portfolios. However, this is a preliminary approach and there is a need for a better classification that takes into account the temporal dynamics and biodiversity value of the different types of physical phenomena.
Mapping biological data
Two types of biological distribution data were used in the analyses: a habitat map based on benthic invertebrate communities to represent broader biodiversity, and eight species-distribution maps to represent fine-scale biodiversity patterns (Noss, 1990). We selected these eight additional species because they are economically and ecologically important and ensured the representation of species with offshore and inshore spatial distributions (Table 1). The study aimed to compare outputs from the different software packages, so it was not deemed necessary to include a large number of species in the exercise.
Common name . | Latin name . | Development stage . |
---|---|---|
Herring | Clupeus harengus | <1-year old and >1-year old |
Cod | Gadus morhua | All ages |
Tope | Galeorhinus galeus | All ages |
Veined squid | Loligo forbesi | All ages |
Plaice | Pleuronectes platessa | <1-year old and >1-year old |
Spider crab | Maja brachydactyla | All ages |
Lesser-spotted dogfish | Scyliorhinus canicula | All ages |
Spurdog, spiny dogfish | Squalus acanthius | All ages |
Common name . | Latin name . | Development stage . |
---|---|---|
Herring | Clupeus harengus | <1-year old and >1-year old |
Cod | Gadus morhua | All ages |
Tope | Galeorhinus galeus | All ages |
Veined squid | Loligo forbesi | All ages |
Plaice | Pleuronectes platessa | <1-year old and >1-year old |
Spider crab | Maja brachydactyla | All ages |
Lesser-spotted dogfish | Scyliorhinus canicula | All ages |
Spurdog, spiny dogfish | Squalus acanthius | All ages |
Common name . | Latin name . | Development stage . |
---|---|---|
Herring | Clupeus harengus | <1-year old and >1-year old |
Cod | Gadus morhua | All ages |
Tope | Galeorhinus galeus | All ages |
Veined squid | Loligo forbesi | All ages |
Plaice | Pleuronectes platessa | <1-year old and >1-year old |
Spider crab | Maja brachydactyla | All ages |
Lesser-spotted dogfish | Scyliorhinus canicula | All ages |
Spurdog, spiny dogfish | Squalus acanthius | All ages |
Common name . | Latin name . | Development stage . |
---|---|---|
Herring | Clupeus harengus | <1-year old and >1-year old |
Cod | Gadus morhua | All ages |
Tope | Galeorhinus galeus | All ages |
Veined squid | Loligo forbesi | All ages |
Plaice | Pleuronectes platessa | <1-year old and >1-year old |
Spider crab | Maja brachydactyla | All ages |
Lesser-spotted dogfish | Scyliorhinus canicula | All ages |
Spurdog, spiny dogfish | Squalus acanthius | All ages |
Designing the conservation planning system
The planning unit theme was produced by creating a series of 5629 16-km2 grid squares using the repeating-shapes extension in ArcView 3.2; the area of each conservation feature found in each planning unit was then calculated. As some of the planning units overlapped the coastline, less of their area fell within the planning region, so for those overlapping planning units, we calculated their area within the planning region by clipping them with the coastline boundary and using these area values as the basis of the planning unit costs. Hence, planning-units that only contained a small amount of the English Channel tended to contain less of each conservation feature, but also had a lower cost.
Nine Marxan and nine Zonation analyses were carried out using three different planning-unit cost metrics and three different sets of targets. Cost metric 1 was area cost, which was based on the surface area of each planning unit, so that all the planning units had the same cost value apart from those located at the edges of the planning region. Cost metric 2 was accessibility cost, which was also based on surface area, but values were reduced by 50% in planning units considered more likely to be suitable for inclusion in an MPA network based on current human activity patterns. Hence, lower values were given to planning units falling within the 3-nautical-mile zone where trawling is restricted, and within shipping lanes and ferry routes, where fishing pressure is reduced (Figure 1). Cost metric 3 was fishing cost, which was based on the fishing profitability of the planning unit, using official data from the French Maritime Fisheries and Aquaculture Office (undertaken by the Ifremer-Halieutic information system) that was then modified to weight the costs by the distance to the nearest French port. French vessels dominate fisheries in the eastern English Channel (Martin et al., 2009). The number of vessels vary, but in 2005, for example, 641 French boats and 49 English boats of >10 m were recorded (Carpentier et al., 2009), so this cost variable was only based on data from the French Maritime Fisheries and Aquaculture Office and French ports.
In each analysis, the same percentage target was used for all habitat types and species, but in the different analyses, we used targets of 10, 30, and 50%. The 10% target has been applied commonly in the literature, but it has also been criticized for not being ecologically relevant (Pressey et al., 2003), and the 30% target is currently recommended by the IUCN (IUCN, 2003) and has also been used in earlier studies (Klein et al., 2008b). The maximum target of 50% has a stronger ecological basis, but is rarely used in conservation planning because it is assumed to be too politically contentious (Soule and Sanjayan, 1998). However, English and French MPA agencies have developed or are developing their own targets (e.g. JNCC and Natural England, 2010).
Marxan and Zonation analyses
As described above, Marxan and Zonation use different approaches for identifying priority areas and measuring conservation value, so we needed to select methods and outputs that were most comparable. In terms of methodology, this involved selecting the following options in Zonation: (i) the target-based, cell-removal rule to produce the priority-area map, so that Zonation sequentially removes the lowest-value planning unit from its conservation-value map, as long as that planning unit is not needed to meet the targets for the different features (Moilanen, 2005); and (ii) the boundary-length-penalty (BLP) option, which most closely resembles the BLM factor in Marxan (Moilanen, 2007; Moilanen and Wintle, 2007).
Identifying suitable outputs was relatively straightforward, but it is important to understand the differences in the software packages. A Marxan analysis involves running the software a number of times and producing a near-optimal, but often different, portfolio at the end of each run. It then identifies the best portfolio as that with the lowest cost and produces a selection-frequency output by counting the number of times each planning unit appeared in the different portfolios (Ball et al., 2009). In this analysis, we used the best output as Marxan's priority-area map and the selection-frequency output as Marxan's conservation-value map. Hence, Marxan's priority-area map can change between different analyses, and the extent of its near-optimality tends to increase with the number of runs used. Similarly, Marxan's conservation-value map can vary between analyses, although these differences tend to be much smaller because each output is based on a number of runs.
In contrast, Zonation produces the same conservation-value map for a given set of inputs, based on the same hierarchical-ranking output, and also produces the same priority-area map for meeting the specified targets. Despite these differences, conservation practitioners use the outputs in similar ways: both priority-area maps show areas that are needed to meet the specified targets, and both conservation-value maps show the relative importance of each planning unit for meeting the conservation objectives.
Nine analyses were undertaken with Marxan and Zonation to run assessments based on the three different planning-unit cost metrics (Table 2) and the three different targets: 10, 30, and 50%. The Marxan analyses involved running the software 100 times, each run consisting of a million iterations. After conducting a sensitivity analysis, we used a BLM value of 5 in all three Marxan analyses, because this best balanced efficiency and portfolio-fragmentation levels (Possingham et al., 2000; Carpentier et al., 2009), and used a target-penalty factor of 100 000 for each conservation feature to ensure that Marxan identified portfolios that met all the targets. The Zonation analyses used the target-based removal rule to identify portfolios that best met the targets, and the BLP value was selected to ensure the lowest boundary length/area value.
Metric . | Marxan BLM value . | Zonation BLP value . | Cost layer . |
---|---|---|---|
Area | 5.0 | 0.5 | Area of planning unit |
Accessibility | 5.0 | 10.0 | Area of planning unit, but reduced by 50% for inshore areas and shipping lanes |
Fishing | 5.0 | 0.0 | Profitability of French fishers |
Metric . | Marxan BLM value . | Zonation BLP value . | Cost layer . |
---|---|---|---|
Area | 5.0 | 0.5 | Area of planning unit |
Accessibility | 5.0 | 10.0 | Area of planning unit, but reduced by 50% for inshore areas and shipping lanes |
Fishing | 5.0 | 0.0 | Profitability of French fishers |
Metric . | Marxan BLM value . | Zonation BLP value . | Cost layer . |
---|---|---|---|
Area | 5.0 | 0.5 | Area of planning unit |
Accessibility | 5.0 | 10.0 | Area of planning unit, but reduced by 50% for inshore areas and shipping lanes |
Fishing | 5.0 | 0.0 | Profitability of French fishers |
Metric . | Marxan BLM value . | Zonation BLP value . | Cost layer . |
---|---|---|---|
Area | 5.0 | 0.5 | Area of planning unit |
Accessibility | 5.0 | 10.0 | Area of planning unit, but reduced by 50% for inshore areas and shipping lanes |
Fishing | 5.0 | 0.0 | Profitability of French fishers |
The Marxan and Zonation conservation-value maps were used to measure the impact of using different planning-unit cost metrics. This was done by first using a quantile classification in ArcGIS to convert both outputs into maps divided into ten classes of equal area based on their measure of conservation value. Hence, each planning unit was given a ranking value from 1 to 10 for both software outputs, and we then used Spearman's rank tests to determine the similarity of the outputs, although we did not record the significance values for these tests because the data were influenced by spatial autocorrelation (Balmford et al., 2001; Nhancale and Smith, 2011). We also investigated the priority-area maps produced by the two different software packages and tested for differences in total area, number of patches, and median patch size using a Wilcoxon signed-ranks test. Finally, we tested whether the priority areas selected by Zonation had higher Marxan conservation-value scores using the Mann–Whitney tests.
Results
The conservation-value maps produced by Marxan and Zonation were strongly influenced by the cost metric, with similar areas being identified as important (Figure 2). However, important areas were widely scattered using the area metric, more likely to be found around the coast and the shipping lanes in the Dover Strait using the accessibility metric, and more likely on the English side of the planning region using the fishing metric. Using higher targets tended to increase the number of planning units with high conservation-value scores (Figure 2). Zonation outputs generally consisted of more-rectangular patches of planning units, whereas the important areas in the Marxan outputs had less regular boundaries (Figure 3, Table 3). The conservation values of the planning units calculated by Marxan and Zonation were correlated and varied with cost metric (Table 4). The results also broadly showed that correlations were better with increasing conservation targets and using the fishing-cost metric.
Target (%) . | Cost metric . | Number of patches . | Median patch area (km2) . | Total area of portfolio (km2) . | |||
---|---|---|---|---|---|---|---|
. | . | Marxan . | Zonation . | Marxan . | Zonation . | Marxan . | Zonation . |
10 | Area | 8 | 6 | 1 315.8 | 3 200.0 | 11 821.4 | 20 047.9 |
10 | Accessibility | 11 | 6 | 384.0 | 3 200.0 | 13 380.6 | 20 047.9 |
10 | Fishing | 8 | 34 | 509.4 | 40.0 | 16 334.7 | 19 167.9 |
30 | Area | 8 | 8 | 857.3 | 432.0 | 27 855.0 | 25 514.3 |
30 | Accessibility | 12 | 3 | 1 038.5 | 6 015.4 | 26 493.6 | 28 695.8 |
30 | Fishing | 19 | 27 | 4 141.5 | 32.0 | 28 810.9 | 31 613.9 |
50 | Area | 9 | 6 | 15.4 | 40.0 | 45 333.4 | 54 863.9 |
50 | Accessibility | 6 | 4 | 336.0 | 40.0 | 47 879.8 | 67 583.8 |
50 | Fishing | 9 | 15 | 384.0 | 16.0 | 21 007.8 | 55 823.9 |
Target (%) . | Cost metric . | Number of patches . | Median patch area (km2) . | Total area of portfolio (km2) . | |||
---|---|---|---|---|---|---|---|
. | . | Marxan . | Zonation . | Marxan . | Zonation . | Marxan . | Zonation . |
10 | Area | 8 | 6 | 1 315.8 | 3 200.0 | 11 821.4 | 20 047.9 |
10 | Accessibility | 11 | 6 | 384.0 | 3 200.0 | 13 380.6 | 20 047.9 |
10 | Fishing | 8 | 34 | 509.4 | 40.0 | 16 334.7 | 19 167.9 |
30 | Area | 8 | 8 | 857.3 | 432.0 | 27 855.0 | 25 514.3 |
30 | Accessibility | 12 | 3 | 1 038.5 | 6 015.4 | 26 493.6 | 28 695.8 |
30 | Fishing | 19 | 27 | 4 141.5 | 32.0 | 28 810.9 | 31 613.9 |
50 | Area | 9 | 6 | 15.4 | 40.0 | 45 333.4 | 54 863.9 |
50 | Accessibility | 6 | 4 | 336.0 | 40.0 | 47 879.8 | 67 583.8 |
50 | Fishing | 9 | 15 | 384.0 | 16.0 | 21 007.8 | 55 823.9 |
Target (%) . | Cost metric . | Number of patches . | Median patch area (km2) . | Total area of portfolio (km2) . | |||
---|---|---|---|---|---|---|---|
. | . | Marxan . | Zonation . | Marxan . | Zonation . | Marxan . | Zonation . |
10 | Area | 8 | 6 | 1 315.8 | 3 200.0 | 11 821.4 | 20 047.9 |
10 | Accessibility | 11 | 6 | 384.0 | 3 200.0 | 13 380.6 | 20 047.9 |
10 | Fishing | 8 | 34 | 509.4 | 40.0 | 16 334.7 | 19 167.9 |
30 | Area | 8 | 8 | 857.3 | 432.0 | 27 855.0 | 25 514.3 |
30 | Accessibility | 12 | 3 | 1 038.5 | 6 015.4 | 26 493.6 | 28 695.8 |
30 | Fishing | 19 | 27 | 4 141.5 | 32.0 | 28 810.9 | 31 613.9 |
50 | Area | 9 | 6 | 15.4 | 40.0 | 45 333.4 | 54 863.9 |
50 | Accessibility | 6 | 4 | 336.0 | 40.0 | 47 879.8 | 67 583.8 |
50 | Fishing | 9 | 15 | 384.0 | 16.0 | 21 007.8 | 55 823.9 |
Target (%) . | Cost metric . | Number of patches . | Median patch area (km2) . | Total area of portfolio (km2) . | |||
---|---|---|---|---|---|---|---|
. | . | Marxan . | Zonation . | Marxan . | Zonation . | Marxan . | Zonation . |
10 | Area | 8 | 6 | 1 315.8 | 3 200.0 | 11 821.4 | 20 047.9 |
10 | Accessibility | 11 | 6 | 384.0 | 3 200.0 | 13 380.6 | 20 047.9 |
10 | Fishing | 8 | 34 | 509.4 | 40.0 | 16 334.7 | 19 167.9 |
30 | Area | 8 | 8 | 857.3 | 432.0 | 27 855.0 | 25 514.3 |
30 | Accessibility | 12 | 3 | 1 038.5 | 6 015.4 | 26 493.6 | 28 695.8 |
30 | Fishing | 19 | 27 | 4 141.5 | 32.0 | 28 810.9 | 31 613.9 |
50 | Area | 9 | 6 | 15.4 | 40.0 | 45 333.4 | 54 863.9 |
50 | Accessibility | 6 | 4 | 336.0 | 40.0 | 47 879.8 | 67 583.8 |
50 | Fishing | 9 | 15 | 384.0 | 16.0 | 21 007.8 | 55 823.9 |
Metric . | 10% . | 30% . | 50% . |
---|---|---|---|
Area | 0.270 | 0.284 | 0.554 |
Accessibility | 0.249 | 0.394 | 0.133 |
Fishing | 0.720 | 0.830 | 0.788 |
Metric . | 10% . | 30% . | 50% . |
---|---|---|---|
Area | 0.270 | 0.284 | 0.554 |
Accessibility | 0.249 | 0.394 | 0.133 |
Fishing | 0.720 | 0.830 | 0.788 |
Metric . | 10% . | 30% . | 50% . |
---|---|---|---|
Area | 0.270 | 0.284 | 0.554 |
Accessibility | 0.249 | 0.394 | 0.133 |
Fishing | 0.720 | 0.830 | 0.788 |
Metric . | 10% . | 30% . | 50% . |
---|---|---|---|
Area | 0.270 | 0.284 | 0.554 |
Accessibility | 0.249 | 0.394 | 0.133 |
Fishing | 0.720 | 0.830 | 0.788 |
In general, the planning units identified as part of the Zonation priority-area maps had higher Marxan selection-frequency scores than planning units that were not selected by Zonation, with the exception of the 10% targets and accessibility cost-metric analysis (Table 5). Marxan generally produced smaller priority-area systems than Zonation (n = 9, Z = −2.429, p = 0.015), but there was no pattern with median patch size or the number of patches. There was a linear relationship between priority-area extent and targets, so priority-area extent ranged between 11 821 km2 for the 10% target- and area-cost metric Marxan analysis and 67 583 km2 for the 50% target- and accessibility-cost metric Zonation analysis (Figure 4), but there was no obvious trend with the number of patches and median patch size (Table 3).
Cost metric . | 10% . | 30% . | 50% . |
---|---|---|---|
Area | 15.29* | 12.97* | 34.83* |
Accessibility | 0.18 | 20.14* | 25.45* |
Fishing | 35.79* | 52.79* | 50.29* |
Cost metric . | 10% . | 30% . | 50% . |
---|---|---|---|
Area | 15.29* | 12.97* | 34.83* |
Accessibility | 0.18 | 20.14* | 25.45* |
Fishing | 35.79* | 52.79* | 50.29* |
*p< 0.001.
Cost metric . | 10% . | 30% . | 50% . |
---|---|---|---|
Area | 15.29* | 12.97* | 34.83* |
Accessibility | 0.18 | 20.14* | 25.45* |
Fishing | 35.79* | 52.79* | 50.29* |
Cost metric . | 10% . | 30% . | 50% . |
---|---|---|---|
Area | 15.29* | 12.97* | 34.83* |
Accessibility | 0.18 | 20.14* | 25.45* |
Fishing | 35.79* | 52.79* | 50.29* |
*p< 0.001.
Discussion
Systematic conservation planning is a widely used approach for designing MPA networks, and most planning assessments rely on computer software to identify priority areas for conservation. These software packages are based on the same principles, but generally use different approaches for measuring conservation value and selecting portfolios of planning units. This has created some confusion among conservation practitioners about which software to use and whether this affects the results. This analysis investigated this issue using the Marxan and Zonation software packages and data from the eastern English Channel. Here, we look at whether it is possible to compare the two software packages, given their underlying differences, then discuss how these results are influenced by the application of different cost metrics and targets in the analysis. Finally, suggestions are provided on how practitioners should collect and use data to minimize the influence of these software packages on their results to help produce more relevant results.
Comparing software packages
Marxan uses the minimum-set approach to identify priority areas for meeting specific targets, whereas Zonation uses the maximum-coverage approach to identify priority areas given a fixed budget. Despite this, the software outputs can be compared because Zonation can adapt its ranked hierarchy output to identify the best areas for meeting targets, which it does by sequentially removing the least important planning units until further removal impacts target attainment. However, this ranked hierarchy output is based on the maximum-cover approach, so it will always be impossible to make exact comparisons between the two packages. Moreover, the comparison is further complicated by the different spatial constraints used by the software packages and the difficulty in determining equivalent BLM and BLP values. A standard approach was used for determining both sets of values, based on balancing the relative planning unit and boundary-length costs, but it is likely that their influence on the results differed.
The conservation features and targets used in the analysis were designed to emphasize any differences in the results from the two software packages, because most conservation features were widely distributed, and the targets were never more than 50% of these distributions. Hence, there was great flexibility in the planning region, with no planning units always being needed to meet certain targets, and many planning units having similar conservation value. In such scenarios, it is likely that spatial constraints would have a relatively large influence on which planning units were selected, so differences in the way that the spatial constraints are used may have produced these effects. This contrasts with previous work comparing results from Marxan and C-Plan, another conservation-planning package, which used higher relative targets and included no spatial constraints, and found conservation-value outputs to be similar (Carwardine et al., 2007).
Despite these differences, the two software packages produced similar results. Although the priority areas identified were not identical, which was expected given the flexibility in the system, there was definite overlap, and the Zonation priority areas had significantly higher Marxan selection-frequency scores in almost all the scenarios that investigated the influence of cost metrics and targets (Table 5). Moreover, the strength of this similarity increased using real-world cost data, such as the accessibility- and fishing-cost metrics. This was because using these cost metrics reduced flexibility so that planning units with similar biodiversity value differed in terms of cost, making low-cost units more important (Smith et al., 2008) and more likely to be selected by both software packages (Table 4). Priority-area extent increased with increasing targets, but this relationship was more linear with Marxan than with Zonation. This may be because Zonation tended to select larger and more connected patches, although some of the Zonation outputs also included a number of small fragments, which masked any difference in patch size and number when comparing Marxan and Zonation. Hence, Marxan tended to produce more efficient priority-area networks and Zonation produced networks with greater levels of connectivity.
Implications for designing MPA networks
The analysis identified three broad aspects that can help inform marine conservation planners when deciding the types of data to be included in their conservation assessments and the type of software to be used. First, although making direct comparisons between Marxan and Zonation was not straightforward, the results were not much affected by the software package used, and the differences were reduced using real-world cost data. Therefore, conservation planners should select the software they consider most appropriate, based on the aims of the project and the additional functionality of the different packages. Second, the conservation-value scores of most of the planning units used in the analysis were generally low, which was probably the main reason for the differences in the results from Marxan and Zonation. This arose because most of our conservation values were widely distributed and the targets relatively low, which meant that there were many similar planning units and, hence, a great deal of flexibility in those selected. This can be overcome by including some conservation features with more limited distributions into the conservation assessment, rather than relying on broad-scale and modelled habitat- and species-distribution data.
The third main finding echoes that from previous studies, which shows that the type of planning-unit-cost metric plays a notable role in determining the location of the priority areas (Klein et al., 2008a; Ban and Klein, 2009). Using real-world data not only produced more robust results, as described above, but also significantly shifted the location of the areas selected by Marxan and Zonation. Therefore, using the accessibility cost meant that most priorities were around the coast and major shipping lanes, whereas using the fishing cost meant that most priorities were on the English side of the planning region. This result highlights the importance of selecting an appropriate cost metric when developing a conservation-planning system to inform decision-makers in the region. However, this is likely to be challenging given that not only do multiple nations share access to the same resources, but the English Channel is commercially important for fishing, transport, aggregate extraction, and energy production sectors (Martin et al., 2009).
The fishing-cost metric also highlights the problems of using the direct financial value in conservation assessments, because this can overly impact marginalized groups (Adams et al., 2010). In this case, the English fishing fleet consists of fewer and smaller boats, so establishing MPAs in English waters would have less impact on the financial value of the catch. However, establishing more MPAs in English waters would have greater impacts on local economies and societies, so any plans advocating such changes would be politically untenable. The results here confirm evidence from a number of studies that show that the value and success of conservation assessments generally depends more on understanding and reflecting the social conditions found in a planning region (Smith et al., 2009) than on the type of selection algorithm or conservation-planning software used. If MPAs are designed to address both conservation and management issues, they will have to be implemented based on a larger set of criteria than those used here. In addition to better descriptors of the socio-economic context, future analyses need to account for the population dynamics of exploited species, as well as the essential habitats for the completion of their life cycles, such as the location of spawning and nursery grounds. Moreover, it may be necessary to link conservation-planning outputs dynamically to bioeconomic models (Mahévas and Pelletier, 2004) to be able to evaluate the medium- to long-term effect of the proposed MPA network on both the exploited population and fishery viability.
MPAs are now expected to be possible management tools in the context of ecosystem-based management of fisheries (Pauly et al., 2002). Although some findings relating to coral reefs led to recommendations that 20–30% of each marine habitat be closed to exploitation (Hughes et al., 2003; Roberts et al., 2003), there are many types of MPA, with management arrangements ranging from multiple-use to strict protection within no-take zones. In complex systems such as the English Channel or the North Sea, MPA networks will have to be designed with different levels of conservation management (Watts et al., 2009) to permit a full MSP exercise. Finally, and more importantly, developing a coherent MPA network for areas shared among many countries will need to move away from the current national approaches, which are limited to the Exclusive Economic Zone and to work on a scale relevant to the ecoregion. This requires international collaboration and shared access to both biological and socio-economic data, which was the approach adopted here.
Acknowledgements
The work was partly co-financed by the European Union through its European Regional Development Fund (INTERREG 3a CHARM II project 162/039/365, INTERREG 4a CHARM III project) and by a PhD studentship granted by the Syndicat Mixte côte d'Opale (SMCO). We also thank two anonymous reviewers for comments that improved the manuscript.
References
Author notes
Handling editor: Emory Anderson