A General Model for Designing Networks of Marine Reserves

See allHide authors and affiliations

Science  06 Dec 2002:
Vol. 298, Issue 5600, pp. 1991-1993
DOI: 10.1126/science.1075284


There is debate concerning the most effective conservation of marine biodiversity, especially regarding the appropriate location, size, and connectivity of marine reserves. We describe a means of establishing marine reserve networks by using optimization algorithms and multiple levels of information on biodiversity, ecological processes (spawning, recruitment, and larval connectivity), and socioeconomic factors in the Gulf of California. A network covering 40% of rocky reef habitat can fulfill many conservation goals while reducing social conflict. This quantitative approach provides a powerful tool for decision-makers tasked with siting marine reserves.

Networks of marine reserves can be an important tool for the conservation of marine biodiversity (1). However, although there is an increasing body of theory about marine reserves (1, 2), there has been almost no practical application of theory on large spatial scales (from hundreds to thousands of km). Some theory suggests that marine reserves should protect more than 20% of the habitat to enhance fisheries (3–6), but there is no agreement on how much habitat should be protected to preserve biodiversity (7), nor on how to maintain ecological links (connectivity) between reserves (8–10).

To address these questions, we designed a network of marine reserves to protect biodiversity and complement fisheries management in the Gulf of California, a tropical marine biodiversity hot spot (11), by collecting basic biodiversity and ecological data from all important rocky coast habitats and applying them to a reserve-siting model based on optimization algorithms that maintain connectivity. The rocky shores of the Gulf of California harbor 10 distinct habitats along ∼1000 km of latitude (12). As a starting point, we set a goal of protecting 20% of each representative habitat and 100% of rare habitats (12) and of the areas with the highest species richness. We also set a goal of maximizing the protection of ecosystem functioning by protecting larval sources (13–16) and nurseries for targeted fish species (16) and by ensuring the connectivity among populations through larval dispersal (10). Existing marine protected areas on the rocky coasts of the Gulf of California are negligible with regard to conservation at the regional scale; there is only one no-take area (Cabo Pulmo Marine National Park, 7111 ha) covering ∼0.2% of the coastal area.

The biodiversity patterns of reef fishes in the Gulf of California showed clear gradients in species richness along latitude: The number of species decreased as the latitude increased (Fig. 1) (17). We used a canonical correspondence analysis to identify the main axes of variation in species abundance among habitats and sites. Latitude and depth explained 66% of the variation in the fish assemblages, indicating the existence of three main zoogeographic regions for reef fishes in the Gulf of California (17). Although the focus was on reef fish, we also addressed plant and invertebrate biodiversity, using habitat as a surrogate (18), and estimated the area of each habitat type around every island and along each section of coast (17).

Figure 1

(A) Map of the Gulf of California with location of the study area (rocky shores) and sampling sites. (B) Gradients of species richness of reef fishes on shallow rocky bottoms (boulders and walls, 5 to 20 m). Other habitats showed similar patterns, with decreasing species richness and increasing latitude.

To determine the existence and location of fish larval sources, we interviewed local fishers, conducted diving surveys from 1998 to 2000, and identified the location of spawning aggregations for seven commercial species (15). We focused on these large fishes because they are the only rocky-habitat species that spawn at specific locations and are targeted by fishers at spawning (15). Larval sources of noncommercial fishes, invertebrates, and algae exist throughout the habitat and are not restricted to a few specific locations. Hence, we assume that the protection afforded by a reserve network for commercial species will ensure sufficient larval production for nonthreatened species. We also identified the habitat requirements for recruitment of vulnerable fish species (16, 19).

We divided the rocky coasts of the Gulf of California into 69 planning units, for which we obtained information about biodiversity and ecological processes (20). Every planning unit had data on reef fish species richness, the presence of spawning aggregations and nurseries of commercial fishes, and the total area of each habitat. We used a model based on optimization algorithms to select a number of planning units that would fulfill the above conservation goals while minimizing the number of reserves (17) and would ensure connectivity among them. The distance between reserves in a network must be determined on the basis of larval dispersal patterns (21), although there is much uncertainty about dispersal patterns (9, 22–24). Assuming that a reserve network should consider mainly the connectivity between vulnerable species populations, we determined that the distance between adjacent reserves in the Gulf of California should not exceed 100 km (25). The selection model was replicated in each of the three zoogeographic regions.

The biologically optimal network involved 24 planning units in 15 aggregated reserves (Fig. 2). The network includes all rare habitats (corals and sea grasses), between 37 and 56% of abundant habitats (boulders, walls, sand, rodolith beds, and shallow algal beds), ∼85% of less abundant habitats (black coral beds and seamounts), 89% of mangroves, and all spawning aggregations (Fig. 3 and Table 1). The network protects 44% of reef habitats in the planning region. The proportion of habitat types targeted for protection is evenly distributed in the three zoogeographic regions, except for rare habitats (12) (Fig. 3). The maximum distance between adjacent reserves is 89 km (median, 36 km; mean, 40 km) (Fig. 3).

Figure 2

Proposed networks of marine reserves for the Gulf of California. (Left) Biologically optimal network, and (right) network that reduces social conflict by excluding areas where fishing pressure and conservation collide. The arrowheads point to planning units removed (left) and added (right) to the network when considering fishing pressure. Some reserves in this figure are aggregates of smaller planning units.

Figure 3

Proportion (%) of total habitat included in the network for each habitat type and zoogeographic region (A) and frequency distribution of distances (in km) between reserves (B).

Table 1

Conservation goals for the rocky coasts of the Gulf of California and achievements of the proposed networks of marine reserves. A null model was conducted, creating 10,000 networks of 24 planning units each, randomly allocated from the total pool of 69 planning units (20).

View this table:

We ran the reserve-siting model again, including fishing pressure, quantified as the density of small fishing boats (17). This solution reduces social conflicts by minimizing the overlap between reserves and heavily fished areas (17), although having reserves near fisheries can be beneficial to fishing (26). This network includes 17 planning units in 13 aggregated reserves covering 40% of reef habitats (Fig. 2). Taking fishing pressure into account does not significantly decrease the proportion of conservation goals achieved relative to the biologically optimal solution (Table 1), mainly because of the low human population density in the Gulf of California and the existence of large areas where coastal fishing pressure is still relatively low.

The most important benefit of this approach is the objectivity it provides to the process of siting marine reserves. Many reserves have thus far been selected more on the basis of social factors than on the basis of biodiversity needs (2). A null model of randomly placed reserves in the Gulf of California showed that although they can provide enough protection for the most abundant habitats, they fail to protect rare habitats (Table 1). The probability that a randomly designed network will achieve conservation goals for all habitats is only 7 × 10−4. Randomly placed reserves would protect an average of only 30% of fish spawning aggregations, but the probability of protecting all aggregations is virtually zero. The probability of including more than 50% of fish nurseries is only 0.4% (Table 1). Ecological processes and critical habitats are not distributed homogeneously, hence reserve networks must be designed on the basis of spatially explicit quantitative data.

The reserve networks presented here allow for the preservation of biodiversity and complement fisheries management. The persistence of populations in a reserve network depends on the size and distance between individual reserves (6, 21). This network allows for the persistence of populations because individual reserves are sufficiently large (50 km) to ensure more than 90% local retention of algal propagules and more than 45% local retention of fish and invertebrate larvae (25, 27). It does not strictly address connectivity for macroalgae and some invertebrates because algae disperse at distances shorter than 5 km and many invertebrates disperse at distances shorter than 100 km (27). However, the average distance between the reserves is 40 km, ensuring connectivity for most fishes and many invertebrates. Finally, the smallest network protects 40% of the habitat, which is in agreement with theoretical work on the minimum fraction of coastline posited for persistence of populations (21).

The use of explicit socioeconomic variables in addition to biodiversity data is particularly important because in marine systems, where fishing is a major threat, ecological criteria and socioeconomic measures are not independent (28). Moreover, portfolios of solutions can be presented to decision-makers (29, 30), who can then evaluate the costs and benefits of different management options within socioeconomic constraints. Prioritization of the reserves can be carried out with this model, using a stepwise selection that evaluates the contribution of each reserve to the preservation of total biodiversity. In the future, new conservation models that account for soft bottoms, pelagic habitats, marine mammals, sea turtles, coastal lagoons, and additional social factors, including future threats, should be developed to obtain networks of reserves to preserve all marine biodiversity. Meanwhile, this procedure can be applied to any coastal region and offers a constructive approach to integrating the economic, social, and biological concerns of marine biodiversity preservation.

Supporting Online Material

Materials and Methods

Figs. S1 and S2

  • * To whom correspondence should be addressed. E-mail: esala{at}


View Abstract

Stay Connected to Science

Navigate This Article