Elsevier

Forest Ecology and Management

Volume 432, 15 January 2019, Pages 942-948
Forest Ecology and Management

Assessing habitat quality when forest attributes have opposing effects on abundance and detectability: A case study on Darwin’s frogs

https://doi.org/10.1016/j.foreco.2018.10.022Get rights and content

Highlights

  • We found certain forest attributes had opposite effects on frog abundance and detectability.

  • Abundance was negatively correlated with daily microclimatic fluctuation.

  • Abundance was positively correlated with stand basal area.

  • However, raw counts were not associated with stand basal area.

Abstract

Forest management can be used to increase the local abundance of species of conservation concern. To achieve this goal, managers must be sure that the relationships between the targeted forest attributes and the focal species abundance are based on robust data and inference. This is a critical issue as the same forest attributes could have opposing effects on species abundance and the detectability of individuals, impairing our ability to detect useful habitat quality surrogates and to provide correct forest management recommendations. Using spatially stratified capture-recapture models (a.k.a. multinomial N-mixture models), we evaluated the effects of stand-level forest attributes on detection probability and local abundance for the endangered Southern Darwin’s frog (Rhinoderma darwinii), a forest-specialist and fully terrestrial amphibian endemic to the South American temperate forest. Our results show that an increase of stand basal area and a decrease of daily microclimatic fluctuation (i.e. an increase in structural complexity) were positively associated with the local abundance of R. darwinii. These stand-level forest attributes also explained the among-population variation in detection probability, although the relationships were opposite to those for abundance. Consequently, an analysis of raw frog counts (i.e. not adjusted for imperfect detection) did not reveal all the factors associated with local abundance. Our results provide further support to previous claims that raw counts of individuals should not be used, generally, as a proxy of abundance in species inhabiting forest ecosystems and elsewhere. More importantly, the opposite effect of forest attributes on abundance and detectability observed in our study highlights the need to use methods that quantify species-habitat relationships in a robust way and which take habitat-specific imperfect detection into account.

Introduction

Anthropogenic land use change is one of the main threats to terrestrial biodiversity (Newbold et al., 2015). Understanding species-habitat relationships is vital to inform the protection and management of imperilled species and their habitats (Martin et al., 2017). Habitat is a species-specific concept referring to the resources and conditions present in an area that allow occupancy, survival and reproduction of individuals (Hall et al., 1997). Yet, from a practical perspective, it is neither possible nor efficient to measure all the resources and conditions that comprise the habitat of a given species. Furthermore, the concept of habitat is not restricted to a binary classification of the environment into habitat and non-habitat but considers a gradient of habitat quality (Fischer and Lindenmayer, 2007). Consequently, conservation biologists interested in managing habitat to increase the local abundance of species of conservation concern, have traditionally focused on habitat quality surrogates (e.g. vegetation types, cover of herbaceous understory, etc.) that may exhibit strong correlations with local abundance or vital rates (e.g., survival and recruitment) of focal species (Franklin et al., 2000, Oldham et al., 2000, Indermaur and Schmidt, 2011, Romano et al., 2017, Unglaub et al., 2018). In forested ecosystems, structural attributes of forest stands seem to be particularly good habitat quality surrogates for forest-specialist animals (Ross et al., 2000, Payer and Harrison, 2003, Watson et al., 2004, Berg et al., 2012, Ibarra et al., 2014, Romano et al., 2017).

In the context of habitat protection and management, it is vital that the relationships between the proxies for habitat quality and the focal species abundance are based on robust data and inference (Kroll et al., 2009, Welsh et al., 2009, Otto et al., 2013, Romano et al., 2017). For instance, the inadequacy of relative abundance (i.e., simple counts) as a proxy of true abundance has been largely acknowledged: due to imperfect detection, true abundance can be underestimated to an unknown degree following this approach (Preston, 1979, Anderson, 2001, Schmidt, 2004, Kéry and Schmidt, 2008). Indeed, the probability of counting an individual that is present and exposed to sampling (i.e. detection probability) often shows a spatial and/or temporal variation associated with observational and environmental covariates (MacKenzie and Kendall, 2002, Pollock et al., 2002, Kéry, 2008, Tanadini and Schmidt, 2011). Worryingly, if covariation between detection probability and habitat quality exists, the confounding effect of imperfect detection could hamper our capability to detect any association between habitat attributes and true abundance (Bailey et al., 2004, Kéry, 2008).

By taking advantage of recently developed spatially stratified capture-recapture (CR) models, a special class of so-called N-mixture models (Royle et al., 2007, Kéry and Royle, 2010, Kéry and Royle, 2016), we evaluated the parallel effects of forest stand attributes on local abundance and detection probability in the endangered Southern Darwin’s frog (Rhinoderma darwinii). This forest-specialist and fully terrestrial amphibian is endemic to the South American temperate forest (Valenzuela-Sánchez et al., 2015). This ecosystem, with its high levels of species endemism and threats, comprises one of the 35 global biodiversity hotspots (Mittermeier et al., 2011). Despite its uniqueness and ecological importance, around half of the South American temperate forest has been lost due to anthropogenic activity (Tecklin et al., 2010). The major threats to R. darwinii are habitat loss and the fungal disease, chytridiomycosis (Soto-Azat et al., 2013a, Soto-Azat et al., 2013b, Valenzuela-Sánchez et al., 2017). Additionally, based on evidence from other forest-specialist animals, where a positive correlation between stand structural complexity and species abundance has been often found (e.g. Ross et al., 2000, Payer and Harrison, 2003, Watson et al., 2004, Berg et al., 2012, Ibarra et al., 2014), we hypothesise that a reduction of forest structural complexity could lead to habitat degradation for this frog. Habitat degradation could reduce individual vital rates and population abundance and viability. For instance, a reduction of structural complexity could increase microclimatic variability (Spies and Franklin, 1988, Chen et al., 1999), which can negatively affect several organismal processes in amphibians (Raffel et al., 2013, Nowakowski et al., 2018). In particular, we expect to find a positive correlation between R. darwinii local abundance and stand-level forest structural attributes such as stand basal area. We test this prediction by using CR data from 10 local populations inhabiting forest stands of differing structural complexity. Additionally, we provide empirical evidence showing that commonly measured forest attributes can have opposing effects on local abundance and detection probability; a possibility that is not always recognized by conservation scientists and managers. Failure to recognize and account for this can impair both the ability to detect useful habitat quality surrogates and to provide correct forest management recommendations.

Section snippets

Study area

We conducted our field study during January 2013 near Neltume, in the Reserva Biológica Huilo Huilo, southern Chile (39°49′48″S, 72°03′14″W). In this part of the South American temperate forest, the vegetation types vary according to altitude (Caviedes and Ibarra, 2017). We selected 10 forest stands with known presence of R. darwinii, all located at a similar altitude (range = 444–623 m a.s.l) in the transition zone between the Valdivian forest and the Nothofagus dombeyi forest types (sensu

Results

We made a total of 351 captures of 179 different frogs (Table 2). Of these, 98 (54.7%) were recaptured at least once across capture occasions.

Most of the detection models with p being modelled as a function of one or more covariates exhibited a higher support than the M0 model (Appendix B). The ΔAIC between the M0 model and the AIC best detection model was 34.32. The top ranked model, which was retained as a basis for the modelling of N, included the additive effect of three site-level

Discussion

Our results support the hypothesis that forest structural attributes are an important component of the habitat of this forest-specialist frog. Namely, an increase of stand basal area and a decrease of daily microclimatic fluctuation (i.e. an increase in structural complexity; see below) were positively associated with the local abundance of R. darwinii (Fig. 1).

In the forests inhabited by R. darwinii, stand structural complexity is determined by forest successional stage and the level of

Conclusions

Our results provide further support to previous claims that raw counts of individuals should not be used, generally, as a proxy of abundance in species inhabiting forest ecosystems (Kroll et al., 2009, Otto et al., 2013) and other ecosystems (Anderson, 2001, Kellner and Swihart, 2014). More importantly, the opposite effect of forest attributes on abundance and detectability observed in our study highlights the need to use methods that quantify species-habitat relationships in a robust way and

Acknowledgments

This study was carried out in accordance with the Chilean law (Agriculture and Livestock Service permit No. 7489/2012). Thanks to Zoo Leipzig and European Association of Zoos and Aquaria for financial support. During data analyses and manuscript writing, AV-S was supported by FONDECYT de postdoctorado (grant no. 3180107).

Declarations of interest

None.

References (54)

  • H.H. Welsh et al.

    How reliable are amphibian population metrics? A response to Kroll et al.

    Biol. Conserv.

    (2009)
  • D.R. Anderson

    The need to get the basics right in wildlife field studies

    Wildl. Soc. Bull.

    (2001)
  • N.D. Berg et al.

    Influence of forest structure on the abundance of snowshoe hares in western Wyoming

    J. Wildl. Manage.

    (2012)
  • L.L. Bailey et al.

    Estimating site occupancy and species detection probability parameters for terrestrial salamanders

    Ecol. Appl.

    (2004)
  • K.P. Burnham et al.

    Model Selection and Multimodel Inference—A Practical Information-Theoretic Approach

    (2002)
  • J. Caviedes et al.

    Influence of anthropogenic disturbances on stand structural complexity in Andean temperate forests: implications for managing key habitat for biodiversity

    PLoS One

    (2017)
  • J. Chen et al.

    Microclimate in forest ecosystem and landscape ecology: variations in local climate can be used to monitor and compare the effects of different management regimes

    Bioscience

    (1999)
  • P.F. Doherty et al.

    Comparison of model building and selection strategies

    J. Ornithol.

    (2012)
  • J. Fischer et al.

    Landscape modification and habitat fragmentation: a synthesis

    Glob. Ecol. Biogeogr.

    (2007)
  • I. Fiske et al.

    unmarked: an R package for fitting hierarchical models of wildlife occurrence and abundance

    J. Stat. Softw.

    (2011)
  • A.B. Franklin et al.

    Climate, habitat quality, and fitness in northern spotted owl populations in Northwestern California

    Ecol. Monogr.

    (2000)
  • L.S. Hall et al.

    The habitat concept and a plea for standard terminology

    Wildl. Soc. Bull.

    (1997)
  • L. Indermaur et al.

    Quantitative recommendations for amphibian terrestrial habitat conservation derived from habitat selection behavior

    Ecol. Appl.

    (2011)
  • K.F. Kellner et al.

    Accounting for imperfect detection in ecology: a quantitative review

    PLoS One

    (2014)
  • M. Kéry

    Extinction rate estimates for plant populations in revisitation studies: importance of detectability

    Conserv. Biol.

    (2004)
  • M. Kéry

    Estimating abundance from bird counts: binomial mixture models uncover complex covariate relationships

    Auk

    (2008)
  • M. Kéry et al.

    Imperfect detection and its consequences for monitoring for conservation

    Commun. Ecol.

    (2008)
  • Cited by (6)

    • Beyond the pond: Terrestrial habitat use by frogs in a changing climate

      2020, Biological Conservation
      Citation Excerpt :

      We assumed that detection was constant but acknowledge that perfect detection may not be possible. For example, frogs may be harder to detect during dry periods (Valenzuela-Sánchez et al., 2019). In previous research conducted at the MFGOWE, data collected in each reserve have been analysed separately (e.g. Manning et al., 2013; Evans et al., 2019).

    • Slow natal dispersal across a homogeneous landscape suggests the use of mixed movement behaviours during dispersal in the Darwin's frog

      2019, Animal Behaviour
      Citation Excerpt :

      HUI1 and HUI2 are within old-growth forest stands of similar composition and structure. Both plots exhibit a high stand basal area (>100 m2/ha), a low daily microclimatic fluctuation and high humidity (Valenzuela-Sánchez et al., 2019). Both forest stands are currently protected from logging activity because they are located within a private protected area.

    View full text