Using Computer Software to Design Marine Reserve Networks: Planners Discuss Their Use of MARXAN

The process of designing a network of marine reserves can involve an extraordinary amount of labor and data. This is particularly the case when planners seek an “optimal” network design – one that provides the best balance of biodiversity and socioeconomic considerations. Such a design requires accounting for multiple species, habitats, oceanographic factors, and resource uses across a wide geographic area. The data and computations involved can overwhelm planners without the aid of computers and special software to handle the challenge.

MARXAN, a software program to help design optimal reserve networks, is gaining fans in the MPA-planning community, thanks to its flexibility and capacity for processing large amounts of information. It was instrumental, for example, in designing the new zoning plan for the Great Barrier Reef Marine Park, which resulted in the world’s largest marine reserve network (MPA News 5:10). This week, MPA News examines MARXAN, its strengths and weaknesses, and how reserve planners are using it around the world.

How the software works

MARXAN supports decision-making on reserve design. It is not meant to replace decision-making. By incorporating data on species, habitats, and other biodiversity features, the software can identify networks of reserve sites that would meet biodiversity targets while minimizing costs to resource users, such as fishermen. It is then up to planners to decide which of the possible networks would be preferable, or to modify the networks by addressing specific stakeholder concerns or incorporating other data.

MARXAN works by performing an algorithm – or series of computations – called “simulated annealing”. This technique finds optimal combinations of sites that best satisfy the planning objective, subject to various constraints. (Simulated annealing has several other real-world applications, including the arrangement of telephone networks.) Although other software programs provide similar services for reserve design, MARXAN is generally considered to be the most powerful. It accommodates the most datasets and offers the greatest flexibility for entering a variety of decision-making factors.

Applying simulated annealing tools to marine systems is different then applying them to terrestrial ones, says Hugh Possingham, an ecologist at the University of Queensland, Australia, who helped develop the underlying basis for MARXAN in 1998. (The software design comprised the Ph.D. thesis of graduate student Ian Ball.) The main difference, Possingham suggests, is that on land, private land ownership and irreversible habitat change are more common factors than in the sea. When terrestrial sites targeted for protection are privately owned, it takes time for the government to procure them for the network; conversely, any delays in designation increase the likelihood those habitats will experience irreversible change. As a result, computer-generated plans for terrestrial networks can fall out of date rapidly, even within a year, due to changes in habitat. The resulting networks, if still based on the original plan, are less than optimal. Possingham describes this phenomenon in a paper he co-authored in the journal Ecology Letters in July 2004 (7:615-622), available online at http://www.nceas.ucsb.edu/collab/2135/docs/ele6241.pdf.

“The Ecology Letters paper is not so relevant to marine systems,” says Possingham. “It is primarily relevant to systems where sites are being destroyed and where sites only become available at the whim of the owner. In many marine systems, sites are not being irreversibly destroyed, aside from perhaps in some areas with dynamite fishing, for example. And with government control of most waters, planners do not have to wait for sites to become available. In the case of the Great Barrier Reef Marine Park, for example, the government was able to implement its re-zoning plan in one fell swoop.”

Applications of MARXAN

Three cases in which MARXAN was used to plan marine reserve networks are discussed below:

Great Barrier Reef Marine Park, Australia

Leanne Fernandes, who managed the program to design a re-zoning plan for the Great Barrier Reef Marine Park, needed a tool that would be powerful enough to compute 16,000 planning units (each ranging in size from 10-30 km²) and 20 or more datasets. “We wanted a system that would optimize for a solution, but given that we calculated there might be 10¹⁵ or so options, a normal optimization program wasn’t going to do it,” she says. Planners also wanted a tool that would account for “representativeness” as a target – not just biodiversity hotspots or irreplaceable areas with unique species – and would account for the cost of designating particular sites.

At the time, no systems available met all of GBRMPA’s needs, including an early version of MARXAN. So, with the help of Possingham and Ball, GBRMPA revised the MARXAN software, making it more appropriate for use in the marine environment and the Great Barrier Reef in particular. Using the software requires expertise, says Fernandes. “In my opinion, you need to be highly competent in GIS analysis and preferably programming with an idea about optimization algorithms,” she says. “We had the technical expertise in-house and – with the forbearance of Dr. Ball, who had to revise the code of the program on occasion for us to overcome glitches – we were able to make it work for us.”

Fernandes says MARXAN was helpful in delivering a solid beginning point – a network of reserves within the marine park – from which to launch the process of re-zoning the park. “But the tool did more than that,” she says. “It highlighted data gaps and inadequacies, displayed the extreme importance of a good reporting tool, and ensured a structured approach to setting objectives, using data, understanding data limitations, etc.”

Nonetheless, she adds, MARXAN was only a beginning point. “Prior to release of the draft zoning plan, much information was added that was not amenable to application in the software, such as information about important uses and values that were in people’s heads, not in databases,” she says, citing areas of importance to recreational fishers as an example. “This information was integrated prior to finalization of the draft zoning plan. Then GIS, as well as roundtable planning teams and a database of analyses of public submissions, were all tapped to revise the draft plan based on the next round of submissions. The textual information often contained spatial references that were relatively easily interpreted by the planning teams working with the submissions database and GIS, but more difficult to transform into a format usable in MARXAN.”

New Zealand

In 2002, New Zealand’s Department of Conservation (DOC) and National Institute of Water and Atmospheric Research (NIWA) used MARXAN on a trial basis. They tested its possible future application for designing a network of MPAs throughout the nation’s 3.8 million-km² Exclusive Economic Zone (EEZ). Kathy Walls, senior marine conservation officer for the Department of Conservation, and scientist Mark Weatherhead of NIWA worked together on the project. “This EEZ-wide approach presented some exciting opportunities to test the application of MARXAN over a large marine area,” says Walls.

Weatherhead says the greatest challenge was the volume of input data, which led to very long run times for the computer – over 40 hours on a powerful computer. “Because the computation time was too long, we coarsened the data for the purposes of the trial by stepping up to 100-km² planning units from 1-km² units, which reduced the computation time substantially,” he says. He and Walls concluded that the 100-km² dataset was adequate for investigating different management options at the EEZ-wide scale, and that the 1-km² dataset should be used to produce final outputs. They decided MARXAN was a promising tool to assist with selection of MPAs, and now hope to continue development of it with possible application at a finer scale in the nearshore.

Walls and Weatherhead say the outputs of MARXAN are only as good as the data that go in. “So we need to know the data limitations,” says Walls. “We also need to be aware of the gaps in our knowledge that may be of importance. For data that have a high degree of temporal or spatial variability, you would probably want to look at them a little more abstractly, or look at long-term average conditions.” Ultimately, setting up MPAs is a sociopolitical decision, she says. “MARXAN is just one of many sources of information feeding into that decision-making process.”

British Columbia, Canada

The Living Oceans Society, a Canadian NGO, has used MARXAN to design a potential network of marine reserves in the waters of British Columbia, along the Pacific coast of Canada. Jeff Ardron, who managed the process for the organization, says MARXAN is not perfect. “I have found it finicky to run correctly, and although it was designed for large amounts of data and large areas, certain functions can grind the program almost to a full halt,” he says. “But then again, we were looking at about 32,000 different planning units and 93 layers of data. This is far beyond the abilities of any other program, and certainly not something a planner can intuit. Having said that, when MARXAN is running smoothly, it is a thing of beauty.”

Notably, the Living Oceans Society was not interested in the “best solutions” that MARXAN produced, in light of unknown factors or assumptions that could potentially render such solutions infeasible, says Ardron. “Rather, we were interested in emergent patterns over the course of hundreds or thousands of runs under a variety of modeled reserve sizes and fragmentations,” he says. “From these aggregated runs, we identified areas as having high ‘conservation utility’. That is, under a variety of conditions, certain areas appeared again and again as being useful, perhaps even essential, in reserve design. These we viewed as the logical place to begin marine reserve planning.”

Since 2000, the Living Oceans Society has provided its findings to the provincial and federal governments, but despite several false starts, integrated marine planning has yet to begin. “Am I happy with MARXAN? Generally, yes,” says Ardron. “Am I happy with marine planning in BC? No comment.” Reports on his organization’s MARXAN work are available online at http://www.livingoceans.org/library/index.shtml.

For more information:

Hugh Possingham, Ecology Centre, University of Queensland, St Lucia, QLD 4072, Australia. Tel: +61 7 3365 9766; E-mail: hpossingham@zen.uq.edu.au

Leanne Fernandes, GBRMPA, PO Box 1379, Townsville, QLD 4810, Australia. Tel: +61 7 4750 0779; E-mail: leannef@gbrmpa.gov.au

Kathy Walls, Marine Conservation Unit, Department of Conservation, PO Box 10420, Wellington, New Zealand. E-mail: kwalls@doc.govt.nz

Mark Weatherhead, National Institute of Water &Atmospheric Research Ltd., New Zealand. E-mail: m.weatherhead@niwa.cri.nz

Jeff Ardron, Living Oceans Society, PO Box 755, Salt Spring Island, BC V8K 2W3 Canada. Tel: +1 250 653 9219; E-mail: jardron@livingoceans.org

---