Conservation planning with dynamic threats: The role of spatial design and priority setting for species’ persistence

https://doi.org/10.1016/j.biocon.2009.12.018Get rights and content

Abstract

Conservation actions frequently need to be scheduled because both funding and implementation capacity are limited. Two approaches to scheduling are possible. Maximizing gain (MaxGain) which attempts to maximize representation with protected areas, or minimizing loss (MinLoss) which attempts to minimize total loss both inside and outside protected areas. Conservation planners also choose between setting priorities based solely on biodiversity pattern and considering surrogates for biodiversity processes such as connectivity. We address both biodiversity processes and habitat loss in a scheduling framework by comparing four different prioritization strategies defined by MaxGain and MinLoss applied to biodiversity patterns and processes to solve the dynamic area selection problem with variable area cost. We compared each strategy by estimating predicted species’ occurrences within a landscape after 20 years of incremental reservation and loss of habitat. By incorporating species-specific responses to fragmentation, we found that you could improve the performance of conservation strategies. MinLoss was the best approach for conserving both biodiversity pattern and process. However, due to the spatial autocorrelation of habitat loss, reserves selected with this approach tended to become more isolated through time; losing up to 40% of occurrences of edge-sensitive species. Additionally, because of the positive correlation between threats and land cost, reserve networks designed with this approach contained smaller and fewer reserves compared with networks designed with a MaxGain approach. We suggest a possible way to account for the negative effect of fragmentation by considering both local and neighbourhood vulnerability to habitat loss.

Introduction

Protected areas are the foundation of many conservation strategies and can be effective tools in maintaining biodiversity. However, the global network of protected areas still has extensive gaps in representing rare and endangered species (Rodrigues et al., 2004a). Of the 11,633 species of terrestrial vertebrates analysed by Rodrigues et al., 74% were poorly represented in protected areas. The picture was even darker for threatened and critically endangered species with 89% and 92% of these species analysed being poorly represented. An important reason for these gaps is that protected areas are often biased towards areas of low productivity and accessibility, also dubbed the “land nobody wants” (Pressey, 1994).

To redress this bias, an approach to priority setting has been proposed that combines biodiversity value and vulnerability to threats (Myers et al., 2000, Pressey and Taffs, 2001, Rodrigues et al., 2004b). This approach (hereafter MinLoss) aims to minimize the expected loss of biodiversity and prioritize areas for conservation that are both important for biodiversity and likely to be lost without intervention. MinLoss considers areas that are not vulnerable to threats as protected de facto, therefore effectively contributing to the objective of minimizing biodiversity loss. In contrast, a more opportunistic approach, maximizing gain (hereafter MaxGain), prioritizes areas for conservation based only on biodiversity value and assumes that unreserved areas will not contribute to biodiversity persistence because they will eventually be lost. An implicit assumption of this approach is that threats to biodiversity are homogeneous and static. Threats to biodiversity, however, are spatially variable, and, in many cases, dynamic. Dealing with dynamic threats requires planners to first predict spatial and temporal changes in threats (Wilson et al., 2005) and then to devise responses (conservation actions). It is usually unrealistic to assume that conservation actions can be implemented all at once or that there are no obstacles to implementation arising from limits on funds, availability, feasibility of interventions and so on (Meir et al., 2004). For this reason, managers are required to schedule conservation actions (Pressey and Taffs, 2001). Scheduling is the coordination of actions over time and space depending on the urgency for intervention, the spatial options for protecting features, the availability of funds, and other factors. Scheduling calls for the formulation of the dynamic area selection problem in which protection and loss are incremental, parallel processes (Costello and Polasky, 2004).

Comparisons of MaxGain and MinLoss in solving the dynamic area selection problem have shown that MinLoss loss generally outperforms MaxGain in retaining biodiversity features. One exception to this occurs when there is low spatial variability in vulnerability to threats, in this case the second assumption made by MaxGain is valid, i.e. vulnerability is homogeneous and the two approaches effectively converge. A second exception occurs when there is considerable uncertainty in future conservation funding or implementation opportunity (Costello and Polasky, 2004, McBride et al., 2007, Wilson et al., 2006).

Among the scientific and practical challenges to effective scheduling of limited conservation resources is the need to promote the persistence of biodiversity processes. Biodiversity processes, such as ecological and evolutionary dynamics are fundamental in maintaining and generating biodiversity (Balmford et al., 1998). Despite this, few studies have attempted to combine attention to biodiversity processes with dynamic threats (Pressey et al., 2007). Cabeza and Moilanen (2003) assessed an indicative reserve system based only on biodiversity pattern and the assumption of static threats. By accounting for population dynamics and habitat loss outside the reserves, they showed that some species would decline and disappear from the system. Cabeza (2003) and Van Teffelen et al. (2006) asserted that the impact of habitat loss and fragmentation on metapopulation dynamics might be reduced if reserve selection were based on species models that incorporated connectivity measures as predictor variables. Carroll et al. (2003) and Noss et al. (2002) integrated a spatially explicit population model and a reserve selection algorithm to identify priorities for mammalian carnivores in the Rocky Mountains. To measure priority for reservation, they expressed irreplaceability and vulnerability (sensu Margules and Pressey, 2000) respectively as the population growth rate and its expected decrease without conservation intervention. Williams et al. (2005) developed an approach to selecting reserves that accounted for range adjustments in response to climate change by designing a set of reserves that would provide connectivity over space and time for species with different dispersal abilities. Although connectivity or spatial population dynamics are receiving increasing attention in reserve design, we are aware of only two studies, that have considered both connectivity and threat within a scheduling approach (Harrison et al., 2008, Sabbadin et al., 2007). This is probably due to the complexity of the problem.

The few systematic conservation planning exercises that have addressed scheduling with respect to biodiversity processes and dynamic threats have failed to consider an important issue. They have not considered the variable cost of conservation action, assuming instead that costs were uniform. Conservation costs are rarely uniform across any region, and considering them can increase the cost-efficiency and feasibility of conservation (Naidoo et al., 2006). Moreover, land value is a major conservation cost and is often positively correlated with vulnerability to habitat loss because value is related to potential profits from extraction. Targeting vulnerable areas of low cost-efficiency can therefore preclude the protection of large, intact areas with higher cost-efficiency (Newburn et al., 2006, Spring et al., 2007). The implications of such choices only become obvious when variable costs are considered.

Here, for the first time, we address both biodiversity processes and dynamic threats by testing four different strategies defined by maximizing gain and minimizing loss applied to both biodiversity patterns and processes (species-specific responses to habitat fragmentation) to solve the dynamic area selection problem with variable area cost. We use models of predicted probability of occurrence to approximate persistence, on the assumption that the predicted probability of occurrence of a species at time t is equivalent to the probability of persistence from now until time t. Probability of occurrence has been used previously as a surrogate for probability of persistence because both are dependent on the same factors related to habitat quality (Araujo and Williams, 2000). This surrogacy gains credibility when occupancy models incorporate neighbourhood covariates such as the proportion of suitable habitat in a defined radius. These models are also likely to relate to processes relevant to population viability such as edge avoidance, spatial population dynamics and lowered persistence of local populations in small habitat fragments (Araujo et al., 2002, Moilanen and Wintle, 2007). Species’ persistence depends, of course, on extrinsic factors such as habitat loss (Araujo and Williams, 2000). We account for this by using a land-use change model to simulate loss of native vegetation. We use the results to answer the following questions:

  • (1)

    Can information about species-specific fragmentation effects be used to schedule conservation actions and improve conservation outcomes?

  • (2)

    Is minimizing loss better than maximizing gain when reservation cost and species-specific influences of fragmentation are incorporated into conservation planning?

Section snippets

Study region and species

The Lower Hunter Central Coast (LHCC) region in central-eastern New South Wales includes seven local government areas (Fig. 1a). These local governments have established a Regional Environmental Management Strategy to integrate biodiversity information and coordinate approaches to nature conservation, producing detailed vegetation and fauna survey and mapping (NSW National Parks and Wildlife Service, 2000, Wintle et al., 2004). For our analyses, we used a ∼600-km2 subregion of the LHCC

Correlation between vulnerability, species abundance and cost

Across all planning units, the combined abundance of all three species was strongly negatively correlated with vulnerability (ρ = −0.58 for the neighbourhood model, ρ = −0.48 for the local model; both p < 0.001). The neighbourhood model predicted that approximately 49% of the expected occurrences of yellow-bellied glider, 34% of those of the squirrel glider, and 76% of those of the sooty owl were in existing protected areas or on public land not vulnerable to clearing. Similar values came from the

Discussion

We simulated the effect of incremental, interacting reservation and land conversion for 20 years for three species with a variety of reserve selection strategies and differing rates of reservation and loss of forest. Unlike previous studies, we considered the effect of fragmentation on the predicted occurrence of the target species by including neighbourhood context measures as predictors in the species distribution models. Also unlike previous studies, we considered realistic spatial variation

Acknowledgments

We would like to thank the University of Melbourne and the Lower Hunter and Central Coast Regional Environment Management Strategy administration for providing the environmental layers and the species distribution maps. We thank Natalie C. Ban and Carlo Rondinini for helpful comments on an earlier draft of this manuscript. PV was supported by a visiting scholar grant from University La Sapienza, Rome and by an IDP student mobility postgraduate scholarship. BW was supported by the CERF Hub AEDA

References (56)

  • R. Sabbadin et al.

    Dynamic reserve site selection under contagion risk of deforestation

    Ecological Modelling

    (2007)
  • D.A. Spring et al.

    Pre-emptive conservation versus “fire-fighting”: a decision theoretic approach

    Biological Conservation

    (2007)
  • N. Strange et al.

    Optimal reserve selection in a dynamic world

    Biological Conservation

    (2006)
  • ABARE, 2000. Land Values on Broadacre Farms 1992–1993 to...
  • H. Akaike

    A new look at the statistical model identification

    Automatic Control, IEEE Transactions

    (1974)
  • M.B. Araujo et al.

    Dynamics of extinction and the selection of nature reserve

    Proceedings of the Royal Society B: Biological Sciences

    (2002)
  • P.R. Armsworth et al.

    Land market feedbacks can undermine biodiversity conservation

    Proceedings of the National Academy of Sciences

    (2006)
  • Australian Bureau of Statistics, 2006. National Regional Profile: Hunter (Statistical Division), 2001 Census...
  • Ball, I.R., Possingham, H.P., 2000. Marxan (version 1.8.6): marine reserve design using spatially explicit annealing. A...
  • Balmford, A., Mace, G.M., Ginsberg, J.R.,998. The challenges to conservation in a changing world: putting processes on...
  • M. Cabeza

    Habitat loss and connectivity of reserve networks in probability approaches to reserve design

    Ecology Letters

    (2003)
  • M. Cabeza et al.

    Site-selection algorithms and habitat loss

    Conservation Biology

    (2003)
  • C. Carroll et al.

    Use of population viability analysis and reserve selection algorithms in regional conservation plans

    Ecological Applications

    (2003)
  • Emery, K.A., 1988. Rural Land Capability Mapping Scale 1:100,000. Soil Conservation Service of New South...
  • E.T. Game et al.

    Marxan user manual for Marxan version 1.8.10. University of Queensland

    (2008)
  • R.L. Goldingay et al.

    Home-range estimates and habitat of the yellow-bellied glider (Petaurus australis) at Waratah Creek, New South Wales

    Wildlife Research

    (1993)
  • J.A. Hanley et al.

    The meaning and use of the area under a receiver operating characteristic (ROC) curve

    Radiology

    (1982)
  • T. Hastie et al.

    Generalized Additive Models

    (1990)
  • Cited by (0)

    View full text