Elsevier

Biological Conservation

Volume 143, Issue 11, November 2010, Pages 2525-2531
Biological Conservation

An approach for ensuring minimum protected area size in systematic conservation planning

https://doi.org/10.1016/j.biocon.2010.06.019Get rights and content

Abstract

One of the most efficient approaches for designing protected area (PA) networks is to use systematic conservation planning software. A number of software packages are available and all of them include a spatial cost or constraint component in their prioritisation algorithms, which allow the user to determine the level of fragmentation of the final PA system. Many conservation planners want to set minimum PA size thresholds, as small PAs are less viable and more expensive to manage, but this can only be achieved with existing software packages by repeatedly reducing the fragmentation levels of the PA system until every PA meets the threshold. Such an approach is inefficient because it increases the size of every PA, not just the smaller ones. Here we describe MinPatch, a software package developed to overcome this problem by manipulating outputs from the Marxan conservation planning software, so that every PA meets the user-defined size threshold. We then investigate the impacts of this approach with a dataset from the Maputaland Centre of Endemism, and find that using MinPatch to meet the PA thresholds is a much more efficient approach than using Marxan alone. We also show that setting a minimum PA threshold can have important effects on where new PAs are located when compared with Marxan outputs. Based on these results, we recommend that conservation planners use MinPatch whenever they want each PA in a network to meet a minimum size threshold.

Introduction

Many species and habitats are poorly represented by current protected area (PA) systems (Rodrigues et al., 2004, Jenkins and Joppa, 2009), so most countries have pledged to develop representative PA systems as part of their Convention on Biological Diversity commitments (CBD, 2004). One commonly used approach for expanding and modifying PA systems is systematic conservation planning (Moilanen et al., 2009), which involves: (i) developing a list of important conservation features; (ii) setting targets for how much of each of these features should be conserved, and; (iii) undertaking a conservation assessment to identify new priority areas that best meet the targets in combination with any existing PAs. A number of conservation planning software packages have been developed to help with this prioritisation process, with the most commonly used being Marxan, Zonation, ConsNet and C-Plan (Moilanen et al., 2009). Each program uses unique prioritisation algorithms but all of these approaches involve dividing the planning region into a series of user-defined planning units. The software then identifies groups of planning units, referred to as “portfolios” hereafter, which achieve the targets whilst meeting other specified constraints, such as minimising the combined cost of the planning units. A range of planning unit cost metrics can be used and the efficiency of each portfolio is measured in terms of how well this cost is minimised.

These software packages are also designed to consider the spatial pattern of their results, as networks of many small and isolated PAs are less ecologically viable and more expensive to manage (Cabeza and Moilanen, 2001, Balmford et al., 2003). Previously, the only way to ensure that assessments identified viable PAs was to increase the size of the planning units, but this produces inefficient results because larger planning units tend to contain superfluous habitats that are not needed to meet the targets (Pressey and Logan, 1998). Current software packages overcome this problem by including spatial cost or constraint components in their prioritisation algorithms, which allow the user to determine the level of fragmentation of the final PA system. Thus, analyses can be based on smaller planning units, as long as the software is set to identify clusters of planning units that would make viable PAs. Typically, this involves running the software a number of times and adjusting the emphasis placed on this clustering until the user is satisfied that the resultant reserve system balance the conflicting needs for efficiency and viability.

Including this spatial component has ensured that conservation planning software produces results with more real-world relevance (Smith et al., 2009). However, the current approach has one major disadvantage that results from the spatial cost or constraint being applied to the whole portfolio. In general, conservation practitioners want to ensure that each PA is above a minimum size, for the reasons of viability and expense mentioned above, although this PA size threshold is generally context specific and based on expert opinion or quantitative studies. At present, the only way to ensure that every PA within a portfolio meets that size threshold is to gradually increase the importance of the spatial component until all the identified PAs are sufficiently large. Unfortunately, this can be a far from efficient approach, as this increases the size of every PA, not just those that are smaller than the threshold. This is an important limitation, as efficiency is a key aspect of systematic conservation planning (Ando et al., 1998, Moilanen, 2008). Here we describe a software package that overcomes this problem by modifying outputs from an existing conservation planning software package, allowing users to specify directly the minimum size of PAs contained within every portfolio.

This work focused on modifying outputs from Marxan because it is the most widely used conservation planning software package. Marxan uses a simulated annealing approach, which involves running an algorithm many times and identifying a different but near-optimal portfolio each time (Ball and Possingham, 2000). Marxan then identifies the “best” solution, which is the portfolio with the lowest cost, and a “selection frequency” output, which counts the number of times each planning unit appeared in the different portfolios. The combined cost for portfolios that meet all the specified targets is calculated as the total cost of all the planning units plus the boundary cost. The boundary cost is the length of the boundary of the entire systems and it acts as a spatial constraint because fragmented portfolios have more edge. Thus, selecting for less edge will favour portfolios containing larger clusters of planning units. This boundary cost is the product of multiplying total edge length by a user-specified boundary length modifier (BLM) value. Thus, the user can adjust portfolio fragmentation levels by adjusting the BLM value: a higher BLM value increases the relative importance of the boundary cost compared to the planning unit costs, and so produces less fragmented but more extensive portfolios (Ball et al., 2009).

Marxan could be adapted to identify portfolios containing PAs that met a user-defined size threshold, simply by adding a new component to the portfolio cost that penalised portfolios containing smaller PAs. However, measuring cluster size is a relatively time consuming process and would have a very large impact on processing time, as simulated annealing involves recalculating the combined portfolio cost many thousands of times per run. Instead, we have developed an approach that takes the portfolios produced by Marxan and modifies them so that each PA meets a user-defined size threshold. In this article we describe software that uses this new approach and test its effectiveness on an existing dataset from southern Africa. More specifically, we investigate how this new approach affects portfolio efficiency and the spatial pattern of the identified PA systems.

Section snippets

Methods

MinPatch for Marxan is a software package written in the Python programming language (Lutz, 2009) and consists of the following stages: (1) it imports a Marxan portfolio and identifies each planning unit cluster, referred to as “PAs” hereafter. It then removes the PAs in this Marxan portfolio that are smaller than a user-defined threshold; (2) it adds entirely new planning unit clusters, referred to as “patches” hereafter, that form the basis of new PAs of the minimum size until every

Results

The median area of the portfolios identified by Marxan ranged between 7750 km2 for the BLM = 0.125 Marxan analysis and 8601 km2 for the BLM = 16 Marxan analysis (Fig. 1; Table 1). Using MinPatch increased the median portfolio area for all of the eight analyses, so that they ranged between 8110 km2 for the BLM = 0.125 MinPatch analysis and 8761 km2 for the BLM = 16 MinPatch analysis (Table 1). This meant that using MinPatch increased the median portfolio size from between 1.77% for the BLM = 16 analyses and

Discussion

Conservation planning software is commonly used in systematic assessments to help develop PA systems. The information they provide is only part of the decision making process and the eventual PA system can differ significantly from the initial priority area maps (Knight et al., 2009). However, there are great advantages in producing outputs that reflect the real-world issues faced by conservation practitioners as closely as possible, as it reduces the amount of effort needed to convert the

Acknowledgements

We would like thank to Nigel Leader-Williams for his help and National Directorate of Conservation Areas, Ministry of Tourism, Mozambique and AEDA for funding this work. DS and HPP were partly funded by a Commonwealth Environmental Research Facility grant and Australia Research Council.

References (22)

  • A. Balmford et al.

    Global variation in terrestrial conservation costs, conservation benefits, and unmet conservation needs

    Proceedings of the National Academy of Sciences of the United States of America

    (2003)
  • Cited by (0)

    View full text