Elsevier

Forest Ecology and Management

Volume 397, 1 August 2017, Pages 108-116
Forest Ecology and Management

Robust predictive performance of indicator species despite different co-occurrence patterns of birds in natural and managed boreal forests

https://doi.org/10.1016/j.foreco.2017.04.034Get rights and content

Highlights

  • Species co-occurrence patterns differed between natural and managed boreal forests.

  • We identified indicator species of avian diversity using multivariate statistics.

  • Indicators selected using data across disturbance regimes had robust performance.

  • Biodiversity monitoring can benefit from carefully selected biodiversity surrogates.

Abstract

Indicator species are widely used biodiversity surrogates that allow the assessment of biodiversity without the expensive and time-consuming construction of species inventories. The selection of indicator species often relies on species co-occurrence patterns, which may be altered by anthropogenic disturbance such as forest harvesting, imposing a unique challenge to their application in managed forest. Here, we studied boreal bird communities in natural forests originating from wildfire and managed forests originating from clearcutting. We aimed to (1) compare species co-occurrence patterns in natural and clearcut forest stands, (2) select indicator species based on species co-occurrence patterns to predict avian diversity, and (3) evaluate the predictive performance of indicator species under both natural and clearcutting disturbance regimes using the same training data set and an independent testing data set. We found that species co-occurrence patterns differ substantially between natural and clearcut stands, suggesting that forest harvesting alters species-environment relationships and/or interspecific interactions. Consequently, we selected different sets of species as indicators of avian diversity based on data from natural or clearcut stands. However, according to internal and external evaluation, selecting indicator species using data from both natural and clearcut stands produced surrogates that predicted avian diversity accurately and precisely in both types of forests. Our results suggest that, despite forest harvesting altering species co-occurrence patterns, a comprehensive understanding of species co-occurrence patterns across natural and managed forests can be used to develop robust biodiversity surrogates. Our study shows that small sets of indicator species can represent the biodiversity of a wide range of species in ecosystems undergoing anthropogenic disturbance, which has important implications for the application of biodiversity surrogates for conservation.

Introduction

The success of decision-making in biodiversity conservation and natural resource management relies upon a detailed understanding of spatial and temporal patterns of biodiversity (Heywood and Watson, 1995, Van Jaarsveld et al., 1998). However, thorough biodiversity assessments can be extremely expensive and time-consuming (Knight et al., 2006, Pressey et al., 2000), and conservation biologists frequently use biodiversity surrogates such as indicator species to assess biodiversity without detailed species inventories (Leal et al., 2010, Pearson, 1994, Thomson et al., 2007). Indicator species, which are small sets of species whose presence or absence are correlated with the biodiversity of a larger group of species (Fleishman et al., 2005), have been successfully used to predict biodiversity of numerous taxa, including plants, invertebrates, and vertebrates, and across a variety of ecosystems (Azeria et al., 2009, Lindenmayer and Likens, 2011). One of the most important applications of indicator species is to assess the impacts of anthropogenic disturbance, such as forest harvesting, on biodiversity (Drever et al., 2008, Lindenmayer et al., 2000).

The selection of indicator species often relies on knowledge of species co-occurrence patterns (Azeria et al., 2009), which may be altered by anthropogenic disturbance. Habitats under anthropogenic disturbance tend to harbor different species communities compared to corresponding habitats in their natural state, even many years following disturbance events (Edwards et al., 2014, Kneitel and Chase, 2004, Zhao et al., 2013). Anthropogenic disturbance can also disrupt patterns of interspecific associations (Azeria et al., 2011, Sarà et al., 2006). As a result, the performance of indicator species identified under natural conditions can be poor in ecosystems under anthropogenic disturbance due to alterations in environmental filtering, interspecific interactions, and, ultimately, species co-occurrence (Lindenmayer and Likens, 2011). This is a critical issue because the reliability of indicator species may be lowest where there is the greatest need for accurate assessments of anthropogenic impacts on biodiversity, thus imposing a unique challenge to the applications of indicator species. Surprisingly, few studies have investigated how the performance of indicator species in predicting biodiversity changes with anthropogenic disturbance (Lawton et al., 1998, Schulze et al., 2004).

Boreal forests are regularly harvested under sustainable forest management strategies, such as ecosystem-based management, with the goal of maintaining a balance between economic interests and biodiversity conservation. Despite the fact that ecosystem-based management aims to develop forestry practices that best emulate natural disturbance regimes such as wildfire (Bergeron et al., 2002), the practice of clearcutting is still widely used. Clearcutting strongly modifies the structure of boreal forest ecosystems by converting old-growth stands to young, even-aged stands (Cyr et al., 2009, Imbeau et al., 2001) and removing most standing trees and snags, which are key structural elements in post-fire conditions (Hutto, 2006). Consequently, managed forests often harbor contrasting species communities to natural forests, even many years following harvest (Azeria et al., 2011, Zhao et al., 2013). The potential for contrasting species co-occurrence patterns in managed and natural boreal forests means that indicator species could be ineffective for assessing the long term impacts of forestry on biodiversity.

Our overarching goal was to examine the effect of variation in bird species co-occurrence on the predictive performance of indicator species in natural and managed boreal forests. Indicator species may provide a practical and efficient measure for evaluating avian diversity in boreal forests, where poor visibility in dense forests and the cryptic nature of many species can make full inventories of diverse avian communities difficult to achieve (Drapeau et al., 2000, Zhao et al., 2013). We studied boreal bird communities in natural forests (originating from natural disturbances such as wildfire, disease and blow-downs) dominated by old-growth stands, and managed forests (originating from clearcutting for the timber industry) dominated by early successional stands. Our objectives were to (1) compare species co-occurrence patterns in natural and clearcut stands, (2) select indicator species based on co-occurrence patterns to predict species richness, and (3) evaluate the predictive performance of indicator species under both natural and clearcutting disturbance regimes using the same training data set and an independent testing data set.

Section snippets

Study area

The study was conducted in the boreal forest of the Côte-Nord region of Québec, Canada (49°-52°N, 65°-70°W) (Zhao et al., 2013). The climate in the region is humid, with a mean annual temperature of −2.5 to 0.0 °C and total annual precipitation of 1000–1400 mm, with 35% of this falling as snow (Grondin et al., 1996). Due to its high precipitation the area has a long fire cycle (average length >270 years), resulting in landscapes dominated by uneven-aged, old-growth stands (Bouchard et al., 2008).

Species co-occurrence classification

We recorded 70 bird species across the 185 sites in the training data set, of which 56 occurred in more than 1% of sites. Of these 56 species, 25 (44.6%) were mature forest species, 10 (17.9%) were young forest species, 10 (17.9%) were shrubland species and 11 (19.6%) were generalists (Table S1). Hierarchical clustering analysis based on sites in both stand types identified a set of five empirical groups (Fig. S1a). We recorded 47 bird species in more than 2% of the sites in natural stands, of

Discussion

Our study demonstrates that examining species co-occurrence patterns in natural and managed forests can produce important insights into the performance of biodiversity surrogates. First, we reveal that species co-occurrence patterns are not always consistent with species’ known primary habitat associations, and are different in natural and clearcut stands, suggesting that species have responded to novel environmental conditions created by forest harvesting differently. Second, due to the

Acknowledgments

We thank our industrial partners for providing access to forest management areas. We are grateful to all field assistants for their efforts in data collection. Two anonymous reviewers provided valuable comments. This project was funded by the Natural Sciences and Engineering Research Council of Canada-Laval University Industrial Research Chair in Silviculture and Wildlife.

References (50)

  • Y. Bergeron et al.

    Natural fire regime: a guide for sustainable management of the Canadian boreal forest

    Silva fennica

    (2002)
  • M. Bouchard et al.

    Fire return intervals and tree species succession in the North Shore region of eastern Quebec

    Can. J. For. Res.

    (2008)
  • D. Boucher et al.

    Développement d un outil de classification de la structure des peuplements et comparaison de deux territoires de la pessière à mousses du Québec

    Forest. Chronicle

    (2003)
  • W.K. Cornwell et al.

    A trait-based test for habitat filtering: convex hull volume

    Ecology

    (2006)
  • D. Cyr et al.

    Forest management is driving the eastern North American boreal forest outside its natural range of variability

    Front. Ecol. Environ.

    (2009)
  • B. Diedenhofen et al.

    Cocor: a comprehensive solution for the statistical comparison of correlations

    PLoS ONE

    (2015)
  • P. Drapeau et al.

    Landscape-scale disturbances and changes in bird communities of boreal mixed-wood forests

    Ecol. Monogr.

    (2000)
  • D.P. Edwards et al.

    Selective-logging and oil palm: multitaxon impacts, biodiversity indicators, and trade-offs for conservation planning

    Ecol. Appl.

    (2014)
  • E. Fleishman et al.

    Using indicator species to predict species richness of multiple taxonomic groups

    Conserv. Biol.

    (2005)
  • N.J. Gotelli

    Null model analysis of species co-occurrence patterns

    Ecology

    (2000)
  • P. Grondin et al.

    Écologie forestière

  • V.H. Heywood et al.

    Global Biodiversity Assessment

    (1995)
  • S. Holm

    A simple sequentially rejective multiple test procedure

    Scand. J. Stat.

    (1979)
  • R.L. Hutto

    Toward meaningful snag-management guidelines for postfire salvage logging in North American Conifer Forests

    Conserv. Biol.

    (2006)
  • L. Imbeau et al.

    Long-term effects of forestry on birds of the eastern Canadian boreal forests: a comparison with Fennoscandia

    Conserv. Biol.

    (2001)
  • Cited by (3)

    • On the efficiency of indicator species for broad-scale monitoring of bird diversity across climate conditions

      2022, Ecological Indicators
      Citation Excerpt :

      By contrast, 40% of species were indicators in a single region. Our 1,750,000 km2 study area certainly exceeds the spatial domain over which indicators are typically used for biodiversity monitoring (e.g., Azeria et al., 2009; Pakkala et al., 2014; Zhao et al., 2017), with few exceptions (Hess et al., 2006). We grouped sites with similar short-term climate conditions to reduce the risk of selecting indicators that would be most representative of areas where most observations have been collected (e.g., lower latitudes in our case).

    • Effects of environmental changes on interspecific interactions of three sympatric pheasants – A study based on long-term monitoring data

      2022, Ecological Indicators
      Citation Excerpt :

      The understanding of interspecific interactions enables us to better predict how ecological communities would respond to habitat modification or climate change (Bar-Massada et al. 2017; Clark et al. 2018; Harris and Warton, 2015; Kass et al. 2020). Many studies documented that interspecific interactions can be influenced by the environment (Blois et al. 2013; Laliberté and Tylianakis, 2010; Milazzo et al. 2013; Pires et al. 2016; Rodewald et al. 2015; Tylianakis and Morris 2017; Tylianakis et al. 2007; Veech 2006; Zhao et al. 2017), while both negative and positive species interactions are included in ecological theory as important driving forces of community organization and dynamics (Bertness and Callaway 1994; He and Bertness 2014; Qi et al. 2018). Negative interspecific interactions mainly are interspecific competitions (García-Callejas et al. 2018).

    • Are data-mining techniques useful for selecting ecological indicators in biodiverse regions? Bridges between market basket analysis and indicator value analysis from a case study in the neotropics

      2020, Ecological Indicators
      Citation Excerpt :

      According to the De Cáceres et al., (2012), as well as Bachand et al. (2014), combinations of species may not only provide the means to analyse an ecosystem when no single species are found to be indicators, but also detect functional shifts in ecosystems. Carignan and Villard (2002) also argue that no single species can provide sufficient information to characterize an entire ecosystem, while Zhao et al. (2017) demonstrate that sets of co-occuring species can be used as robust biodiversity surrogates. Furthermore, a greater positive predictive power can be attained with the use of species combinations as well as greater coverage of the ecosystem under certain circumstances, even though sensitivity loss may occur in some cases (De Cáceres et al., 2012).

    1

    Qing Zhao and Tom H.E. Mason have equal contributions.

    View full text