Short communication
Uncertainty analysis favours selection of spatially aggregated reserve networks

https://doi.org/10.1016/j.biocon.2005.11.006Get rights and content

Abstract

It has been widely argued that habitat fragmentation is bad for (meta)population persistence and that a high level of fragmentation is a similarly undesirable characteristic for a reserve network. However, modelling the effects of fragmentation for many species is very difficult due to high data demands and uncertainty concerning its effect on particular species. Hence, several reserve selection methods employ qualitative heuristics such as boundary length penalties that aggregate reserve network structures. This aggregation usually comes at a cost because low quality habitats will be included for the sake of increased connectivity. Here a biologically justified method for designing aggregated reserve networks based on a technique called distribution smoothing is investigated. As with the boundary length penalty, its use incurs an apparent biological cost. However, taking a step further, potential negative effects of fragmentation on individual species are evaluated using a decision-theoretic uncertainty analysis approach. This analysis shows that the aggregated reserve network (based on smoothed distributions) is likely to be biologically more valuable than a more fragmented one (based on habitat model predictions). The method is illustrated with a reserve design case study in the Hunter Valley of south-eastern Australia. The uncertainty analysis method, based on information-gap decision theory, provides a systematic framework for making robust decisions under severe uncertainty, making it particularly well adapted to reserve design problems.

Introduction

The primary goal of reserve planning is to increase the probability of the long-term persistence of biodiversity (Vane-Wright, 1996, Pimm and Lawton, 1998, Margules and Pressey, 2000, Cabeza and Moilanen, 2001, Araújo and Williams, 2001, Polasky and Solow, 2001). It is widely recognized that the spatial pattern of reserved habitat may influence the biological value of reserves through its influence on spatial population dynamics. A basic tenet of spatial (meta)population theory is that dispersal of individuals between sites and colonization of empty habitat are influenced by connectivity (distance) and thus aggregated networks are predicted to maintain species better than fragmented ones (Hanski, 1998). This effect has been demonstrated but it is difficult to incorporate in reserve designs when several species are involved due to uncertainty about the effects of fragmentation on individual species, the lack of data for quantifying such effects, and the computational demands of incorporating individual fragmentation effects in design algorithms.

Heuristic methods have been devised that increase aggregation in reserve networks, based on the assumption that aggregation is good, especially if it can be achieved with low extra cost. A common way of aggregating reserve networks is the boundary length penalty (Possingham et al., 2000, McDonnell et al., 2002.; Nalle et al., 2002, Önal and Briers, 2002, Fischer and Church, 2003, Cabeza et al., 2004a, Cabeza et al., 2004b). This method is qualitative in the sense that the biological value of the reserve network is not influenced by the degree of fragmentation. Instead, aggregation is induced via a qualitative penalty given for the boundary length of the reserve.

Aggregation involves trade-offs. There is usually a notional biological cost to increase aggregation because it is usually necessary to include lower quality habitats in order to increase connectivity. However, most studies using the boundary length penalty (cited above) have found that a major decrease in the boundary length of the reserve network can be achieved with a small biological or financial cost. The choice of the most appropriate value for a penalty is heuristic, guided by the trade-off between biological value and reserve aggregation.

Here we investigate aggregated reserve structures obtained by another method – distribution smoothing (Moilanen et al., 2005). In this technique, the distribution of the species is smoothed using a kernel with its width set by an estimate of dispersal distances for the species. The smoothing effectively identifies important semi-continuous regions where the species has overall high levels of occurrence, although not necessarily in every grid cell.

It is shown that if there is uncertainty about the biological value of habitat close to the edge of a reserve then aggregated reserves also provide more robust conservation outcomes in terms of biological value. Biologically, it can be expected that negative edge effects (increased disturbance or predation, invasive species, and changes in abiotic conditions; Debinski and Holt, 2000, Gaston et al., 2002) and metapopulation dynamics (Hanski, 1998) would lead to decreased biological value in cells close to the edge of the reserve network. However, this can be difficult to quantify for many species due to the complexity and high data demands of spatial population modelling. Therefore, an uncertainty analysis approach to assess potential effects of fragmentation on conservation outcomes was used here.

Decision analytical methods (Drechsler, 2000, Harwood, 2000, du Ray et al., 2005, Westphal and Possingham, 2003, Wilson et al., 2005) are well suited for application in conservation planning where resources are limited, tradeoffs between different goals are common, and many uncertain factors may plague biological data and analyses. The aim of this paper is to investigate a method for incorporating uncertainty about species responses to fragmentation into reserve design. Info-gap decision theory (Ben-Haim, 2001) was developed for decisions in the face of severe uncertainty. The objective of info-gap analyses is to identify decisions that achieve a desired outcome or aspiration with the maximum possible level of robustness to uncertainty. In this study, info-gap decision theory is used for quantifying the robustness of reserve design options to uncertainty in species responses to fragmentation. Using a case study in the Hunter Valley region of south-eastern Australia, the potential of uncertainty analysis to assist in the design of aggregated reserve networks is investigated.

Section snippets

An uncertainty analysis on the effects of fragmentation

Before going to the details of the uncertainty analysis, it is necessary to clarify the meaning of linear and nonlinear reserve selection models (Moilanen, 2005). The analysis that follows below is most relevant for linear reserve selection models. In such models the structure and quality of the landscape may influence the initial conservation value (here probability of occurrence) of sites (grid cells). However, the structure of the selected reserve network does not have an effect on

Results

Fig. 1 demonstrates the difference in a probability of occurrence map (Fig. 1(a)), a smoothed map (Fig. 1(b)) and reserve network structures arising from the use of original (Fig. 1(c)) and smoothed distributions (Fig. 1(d)) for many species. There is a marked difference in the aggregation levels of the networks. For example, the scattering of tiny areas in the upper middle area are included in the fragmented solution due to high predicted occurrence levels for the sugar glider. These areas do

Discussion

The present analysis describes a quantitative argument that can be used for evaluating uncertain effects of fragmentation on the biological value of a reserve network. Two reserve network structures are compared, a fragmented one and an aggregated one, which apparently has slightly lower biological value because connectivity has been purchased via the inclusion of biologically suboptimal cells into the reserve. The aggregated solution could have been obtained using any reserve selection

Acknowledgements

We thank Professor Mark Burgman and Astrid van Teeffelen for helpful comments on the manuscript. We also thank Professor Yakov Ben-Haim and an anonymous reviewer for constructive comments. This study was funded by the Academy of Finland project 1206883 to AM, the Finnish Centre of Excellence Programme 2000–2005, Grant no. 44887, and Australian Research Council Linkage Project LP0347473 to BW. We thank professor Mark Burgman, the University of Melbourne and the Australian Mathematical Sciences

References (33)

  • D.T. Fischer et al.

    Clustering and compactness in reserve site selection: an extension of the biodiversity management area selection model

    Forest Sci.

    (2003)
  • K.J. Gaston et al.

    Persistence and vulnerability, retaining biodiversity in the landscape and in the protected areas

    J. Biosci.

    (2002)
  • I. Hanski

    A practical model of metapopulation dynamics

    J. Anim. Ecol.

    (1994)
  • I. Hanski

    Metapopulation dynamics

    Nature

    (1998)
  • C.R. Margules et al.

    Systematic conservation planning

    Nature

    (2000)
  • M.D. McDonnell et al.

    Mathematical methods for spatially cohesive reserve design

    Environ. Model. Assessment

    (2002)
  • Cited by (87)

    View all citing articles on Scopus
    1

    Tel.: +61 3 8344 4572; fax: +61 3 9347 5460.

    View full text