Optimizing habitat management for amphibians: From simple models to complex decisions

Highlights

• Artificial wetland construction is an effective conservation tool for amphibians.

• Identifying the best locations to build wetlands can be extremely challenging.

• We solved the problem by coupling optimization with a metapopulation model.

• Our approach solved a conservation problem with millions of potential solutions.

• Optimization is a valuable tool for prioritizing amphibian conservation actions.

Abstract

Habitat loss, degradation and fragmentation remain leading causes of amphibian declines across the globe. To mitigate these impacts, conservation managers may protect core habitats and pursue habitat creation or enhancement actions, including construction of artificial wetlands, manipulation of wetland hydroperiods, removal of invasive species or restoration of aquatic and riparian vegetation. Yet management budgets are universally tight. When planning such actions, managers face the fundamental and complex problem of choosing where and when to invest limited resources to maximize the likelihood of species persistence. Here, we extend our previous research on this problem, and demonstrate the utility of coupling occupancy models with optimization algorithms to identify preferred habitat management schemes across multiple, disjunct habitat networks. Our real-world case study, completed in close collaboration with conservation managers, focussed on optimal habitat creation schemes for a threatened Australian frog in a rapidly urbanizing region. Our new technique identified clear priorities for investment in wetland construction both among and within seven disjunct habitat networks, solving a spatial prioritization problem that entailed millions of potential solutions and which was otherwise intractable. Such complex, multi-scale spatial prioritization problems are pervasive in amphibian conservation. Coupling occupancy models with spatial optimization algorithms represents a promising avenue to solve these problems and design habitat protection, creation and management schemes that maximize the chance of species persistence.

Introduction

Direct human impacts pose ongoing and pervasive threats to the world's amphibians, and their mitigation must remain a key focus for amphibian conservation (Gallant et al., 2007). Of these threats, ameliorating the destruction, degradation and fragmentation of habitat is particularly pressing, with at least 40% of threatened amphibians being impacted by these processes globally (Stuart et al., 2004). Habitat loss and degradation occurs through numerous mechanisms, including wetland destruction for agricultural and urban development (Cushman, 2006; Hamer and McDonnell, 2008), hydrological transformation due to urban runoff (Canessa and Parris, 2013), degradation of water quality through sedimentation and chemical pollution (Ficken and Byrne, 2013; Hale et al., 2019; Sievers et al., 2019), introduction of invasive predators (Vredenburg, 2004; Hamer and Parris, 2013), over-grazing by livestock (Schmutzer et al., 2008), and the loss or modification of terrestrial habitat upon which many amphibians rely (Harper et al., 2008). Similarly, habitat fragmentation has numerous sources, from the modification of the intervening terrestrial matrix to the construction of barriers to dispersal, such as roads or other infrastructure (Matos et al., 2019 - this issue; Petrovan and Schmidt, 2019 - this issue).

Despite their multiple impacts, habitat loss, degradation and fragmentation are manageable threatening processes. Mitigation is possible through legislative protection of key habitats and the dispersal paths between them, and is often accompanied by restoration of remnant habitats and the creation of artificial habitat (Smith and Sutherland, 2014; Schmidt et al., 2019 - this issue). Such initiatives can be remarkably effective. For example, rapid population recovery of threatened ranid frogs has been achieved by removing non-native fish from breeding habitats (Vredenburg, 2004; Pope, 2008), and constructed wetlands can support diverse amphibian assemblages even in highly-modified landscapes (Knutson et al., 2004; Parris, 2006). Nevertheless, while several authors have addressed the restoration and design of habitat to maximize its suitability for amphibians (reviewed by Smith and Sutherland, 2014), less attention has been paid to identifying where in a given landscape management interventions should be pursued to maximize conservation gains (Matos et al., 2019 - this issue; Schmidt et al., 2019 - this issue).

Such questions may be answered with the aid of metapopulation theory, with which the dynamics of many aquatic-breeding amphibians are consistent (Alford and Richards, 1999). Under this conceptualization, individual wetlands (or ‘ponds’ as they are often referred to in the northern hemisphere) represent patches of habitat in a broader matrix of less-suitable terrestrial habitat. As dispersal and colonization between wetlands occurs at low frequency, the populations associated with them are largely independent, and subject to the stochastic extinction processes typical of small, isolated populations (Alford and Richards, 1999). In a functioning metapopulation, local extinction events are offset by dispersal and recolonization of vacant patches, and can exhibit long-term quasi-stability (Hanski, 1999). While the generality of metapopulation theory as a model of the population dynamics of aquatic-breeding amphibians has rightly been questioned (Marsh and Trenham, 2001), several well-tested examples exist in the literature (Sinsch, 1992; Sjögren-Gulve, 1994; Carlson and Edenhamn, 2000; Vos et al., 2000; Heard et al., 2012a).

For species displaying recurrent population extinction and colonization, it is possible to construct models of these processes from observational data (MacKenzie et al., 2003). Such models can help to elucidate habitat and management correlates of patch-level extinction and colonization, enabling inferences about likely effects of management at both the patch- and landscape-scale (MacKenzie et al., 2003). They can also be used to predict the viability of population networks, by simulating probabilistic extinction and colonization events among patches, and deriving metrics such as the probability of metapopulation (quasi)extinction over defined time periods (Sjögren-Gulve and Ray, 1996; Hanski, 1999). Recent authors have recognized the utility of such models (known as stochastic patch occupancy models, or ‘SPOMs’) for amphibian conservation, and particularly the question of the scale and arrangement of habitat networks required to ensure the persistence of amphibian metapopulations. For example, Chandler et al. (2015) used models of the extinction and colonization dynamics of Chiricahua leopard frog (Lithobates chiricahuensis) to assess viability of re-introduced populations in the U.S. southwest. Similarly, Green and Bailey (2015) developed a model of occupancy turnover for wood frogs (Lithobates sylvatica) in Maryland U.S.A, and used the resulting SPOM to identify cost-effective habitat management initiatives to maximize metapopulation viability.

SPOMs provide useful estimates of metapopulation viability, but are problematic when a great many potential management options exist, as evaluating efficacy for every possible management scenario can be computationally-intensive or impossible. A promising avenue to solve such problems is to couple SPOMs with optimization algorithms to identify preferred management scenarios. In early examples, Moilanen and Cabeza (2002) used a genetic algorithm to identify the optimal set of habitat patches to protect under a fixed budget for Finnish butterfly metapopulations, while Westphal et al. (2003) used stochastic dynamic programming to identify the optimal habitat restoration scheme for a metapopulation of a threatened passerine in Australia. Similarly, Nicholson et al. (2006) demonstrated the use of simulated annealing to identify optimal reserve designs for vertebrates inhabiting forest remnants in Australia.

Here, we demonstrate the utility of coupling SPOMs with optimization algorithms for designing amphibian habitat management programs, extending our previous work on the use of SPOMs to assess metapopulation viability for these animals (Heard et al., 2013). We used the SPOM parameterized empirically in that study as the basis of our new approach, as the annual probabilities of population extinction and colonization are functions of variables that will be familiar to amphibian ecologists: wetland area, hydroperiod, aquatic vegetation composition and connectivity to surrounding occupied wetlands. Simulated annealing was used to identify optimal (or at least, near-optimal) combinations of wetlands to construct across a rapidly urbanizing region, with the objective of maximizing metapopulation viability for our focal species, as estimated using the SPOM. With an overall budget of up to 150 wetlands to be constructed, the decision space spanned many millions of possible solutions and was impossible to solve by running simulations of all potential habitat creation options (as in Heard et al., 2013). We show that our new approach of coupling SPOMs with optimization algorithms has the potential to solve extremely complex habitat management problems of this nature.