Research article
Predicting spatiotemporal patterns of road mortality for medium-large mammals

https://doi.org/10.1016/j.jenvman.2019.109320Get rights and content

Highlights

  • Used a roadkill dataset, 2 years of surveys on 800 km, in Mato Grosso do Sul, Brazil.

  • Regressed the spatiotemporal patterns of 7 sp. with a set of environmental variables.

  • Intrinsic spatial and temporal roadkill risks were the most important predictors.

  • We successfully predicted the spatiotemporal patterns for all species.

  • This approach allows defining specific spatiotemporal tailored management actions.

Abstract

We modelled the spatiotemporal patterns of road mortality for seven medium-large mammals, using a roadkill dataset from Mato Grosso do Sul, Brazil (800 km of roads surveyed every two weeks, for two years). We related roadkill presence-absence along the road sections (1000 m) and across the survey dates with a collection of environmental variables, including land cover, forest cover, distance to rivers, temperature, precipitation and vegetation productivity. We further included two variables aiming to reflect the intrinsic spatial and temporal roadkill risk. Environmental variables were obtained through remote sensing and weather stations, allowing the estimate of the roadkill risk for the entire surveyed roads and survey periods. Overall, the models could explain a small fraction of the spatiotemporal patterns of roadkills (<0.23), probably due to species being habitat generalists, but still had reasonable discrimination power (AUC averaging 0.70 ± 0.07). The intrinsic spatial and temporal roadkill risk were the most important variables, followed by land cover, climate and NDVI. We show that identifying spatiotemporal roadkill patterns may provide valuable information to define specific management actions focused on road sections and time periods, in complement to permanent road mitigation measures. Our approach thus offers a new insight into the understanding of road effects and how to plan and strategize monitoring and mitigation.

Introduction

Transportation infrastructures, particularly roads, are the most common human-made features in the globe, spreading across nearly its entire surface (Ibisch et al., 2016). Despite their value to human living and development, they may be responsible for significant and deleterious impacts on biodiversity, of which roadkill is amongst the most visible and significant ones (Bennett, 2017; Forman and Alexander, 1998; Van der Ree et al., 2015). When roadkill rates are not offset by an increase in per-capita recruitment, the growth rate of populations can be seriously affected, ultimately leading to local extinctions or to a strong population depletion in the surrounding areas (Borda-de-Água et al., 2014, 2011; Silva et al., 2010). For this reason, strategies to reduce wildlife road mortality are becoming a significant component of many wildlife conservation efforts (e.g. www.giantanteater.org and www.tapirconservation.org.br).

The success of conservation and mitigation management strategies may greatly depend on the knowledge of the temporal and spatial patterns of roadkill risk, and its relation with key environmental drivers, in order to place or perform mitigation actions where and when they are most required (Lesbarrères and Fahrig, 2012). Whilst most research studies on roadkill effects have addressed separately the spatial and the temporal patterns of mortality (Crawford et al., 2014; Cureton and Deaton, 2012; Garrah et al., 2015; Santos et al., 2017), we claim that focusing simultaneously on spatial and temporal patterns may provide important and complementary information for management and mitigation of the transportation infrastructures (e.g. Santos et al., 2018). For example, most of the reported roadkill of turtle species occurred in specific road sections and during the breeding season (Beaudry et al., 2009; Crawford et al., 2014; Cureton and Deaton, 2012). Hence, it may be more cost-effective to apply specific actions in such well-defined road segments with higher mortality and during those periods of higher likelihood of collision, i.e. when animals tend to move more and/or longer distances.

In this study, we aimed to assess how the locations and occasions of collisions relate to a collection of environmental variables in order to model the spatiotemporal patterns of roadkill risk. We were particularly interested in freely available environmental variables that could be easily obtained and used on the broader landscape scale to assess and predict the roadkill risk along different roads, and therefore to better assist in road and landscape planning and management actions. These variables included habitat information from remote sensing observations and climatic information from weather stations. The use of these environmental variables was meant to explain the variation in spatiotemporal patterns of roadkills, assuming that the number of casualties along the road and year have a direct relation with species' occupancy and movement activity in neighboring areas (Barrientos and Bolonio, 2009; D'Amico et al., 2015).

Road mortality in Brazil is known to be a major concern for numerous species (Ascensão et al., 2017; Cunha et al., 2010; González-Suárez et al., 2018). Medium-large mammals are particularly relevant as some species are highly vulnerable to roadkill given their general low reproductive rates and low densities, associated with large area requirements that often lead to higher road crossing rates and eventually higher collision rates (Rytwinski and Fahrig, 2012). On the other hand, given their high body-mass, they represent a serious threat when involved in animal-vehicle collisions, potentially resulting in significant costs and human fatalities (Huijser et al., 2013). Here, we modelled the spatiotemporal patterns of roadkill of medium-large mammals across the state of Mato Grosso do Sul, Brazil. We expected favorable habitat areas to be related to higher density of individuals and therefore to higher concentration of casualties; and time periods of longer and more frequent displacements (e.g. mating, dispersal) to be related to temporal peaks of mortality (D'Amico et al., 2015; Grilo et al., 2009). The combination of both habitat and climate data should therefore allow to model and predict the spatiotemporal patterns of road mortality. We further expected those species having a more generalist niche breadth to show a less defined and more variable spatiotemporal roadkill patterns, when compared to more specialized species.

Section snippets

Study area and data collection

We used the roadkill datasets collected within the research projects “Anteaters and Highways” (www.giantanteater.org) and “Lowland Tapir Conservation Initiative” (www.tapirconservation.org.br) in the state of Mato Grosso do Sul (MS), Brazil. Therein, researchers conducted two distinct year-round roadkill surveys, on a regular basis, with intervals of 11–17 days (mean 14 days), comprising 37 and 38 surveys in the two years, respectively. The surveys spanned from April 2013 to March 2014, and

Spatial information

Land cover was obtained from remote sensing via the MODIS MCD12Q1 product (Friedl and Sulla-Menashe, 2015). This product provides global land cover types at yearly intervals (2001–2017), with 500 m resolution, derived from six different classification schemes, from which we used the Annual International Geosphere-Biosphere Programme (IGBP) classification (Friedl and Sulla-Menashe, 2015). The land cover classes are derived using supervised classifications of MODIS Terra and Aqua reflectance

Model building, performance and validation

For each of the seven focal species we performed binomial logistic regressions (GLM) relating the roadkill presence-absence in the road sections (1000 m) across the survey dates, with the collection of environmental variables (land cover classes, forest cover, distance to rivers, temperature, precipitation, and NDVI). We further included two additional variables aiming to reflect the intrinsic roadkill risk, by using the total number of roadkills (per dataset) at each date and road section,

Results

Higher values of variance explained were found for Dasypus novemcinctus (0.23 ± 0.05), followed by Hydrochoerus hydrochaeris (0.12 ± 0.04), Euphractus sexcinctus (0.08 ± 0.03), Tamandua tetradactyla (0.05 ± 0.03), Tapirus terrestris (0.05 ± 0.04), Cerdocyon thous (0.04 ± 0.02) and Myrmecophaga tridactyla (0.03 ± 0.02) (Fig. 2, panel A). The models had a reasonable discrimination power, with the AUC for the training data averaging 0.70 ± 0.07 across species and replicates; and the AUC for

Discussion

In this study, we modelled and predicted the spatiotemporal patterns of road mortality for seven medium-large mammals, all of which have different habitat and food specializations, suggesting that our approach can be generalized to other vertebrates. In general, higher roadkill risk aggregations spanned for some kilometers and days, suggesting that tailored mitigation measures focusing simultaneously on certain road sections and periods could be beneficial. Importantly, the modeling procedure

Funding

FA was funded through a post-doctoral grant from Fundação para a Ciência e Tecnologia (FCT, SFRH/BPD/115968/2016). The Anteaters &Highways Project is funded by the Fondation Segré, Houston Zoo, Nashville Zoo, Naples Zoo at Caribbean Gardens as well as several other Zoos and institutions listed at http://www.giantanteater.org/supporters.html. The Lowland Tapir Conservation Initiative (LTCI) – Instituto de Pesquisas Ecológicas (IPÊ) has the institutional and financial support from a multitude of

References (47)

  • R.A.L. Santos et al.

    Predicting wildlife road-crossing probability from roadkill data using occupancy-detection models

    Sci. Total Environ.

    (2018)
  • J.P. Silva et al.

    Estimating the influence of overhead transmission power lines and landscape context on the density of little bustard Tetrax tetrax breeding populations

    Ecol. Model.

    (2010)
  • B. Zimbres et al.

    Terrestrial mammal responses to habitat structure and quality of remnant riparian forests in an Amazonian cattle-ranching landscape

    Biol. Conserv.

    (2017)
  • F. Ascensão et al.

    Spatial patterns of road mortality of medium–large mammals in Mato Grosso do Sul, Brazil

    Wildl. Res.

    (2017)
  • A.M. Barbosa et al.

    modEvA: model evaluation and analysis

    (2016)
  • R. Barrientos et al.

    The presence of rabbits adjacent to roads increases polecat road mortality

    Biodivers. Conserv.

    (2009)
  • F. Beaudry et al.

    Identifying hot moments in road-mortality risk for freshwater turtles

    J. Wildl. Manag.

    (2009)
  • V.J. Bennett

    Effects of road density and pattern on the conservation of species and biodiversity

    Curr. Landsc. Ecol. Rep.

    (2017)
  • L. Borda-de-Água et al.

    Spatio-temporal impacts of roads on the persistence of populations: analytic and numerical approaches

    Landsc. Ecol.

    (2011)
  • M.J. Corriale et al.

    Patterns of habitat use and selection by the capybara (Hydrochoerus hydrochaeris): a landscape-scale analysis

    Ecol. Res.

    (2014)
  • B.A. Crawford et al.

    Hot spots and hot moments of diamondback terrapin road-crossing activity

    J. Appl. Ecol.

    (2014)
  • H.F. da Cunha et al.

    Roadkill of wild vertebrates along the GO-060 road between Goiânia and Iporá, Goiás state, Brazil

    Acta Sci. Biol. Sci.

    (2010)
  • G.M. Cunnington et al.

    Culverts alone do not reduce road mortality in Anurans

    Ecoscience

    (2014)
  • Cited by (32)

    • Forecasting seasonal peaks in roadkill patterns for improving road management

      2022, Journal of Environmental Management
      Citation Excerpt :

      In general, well-defined patterns of seasonal mortality were related to climatic variables. This relation probably stems from the fact that the phenology of the species is influenced by climatic conditions, with different seasonal activity, namely movement and abundance (e.g., mating and reproduction), leading to a greater number of crossings and, consequently, of roadkill (Ascensão et al., 2019; Cureton and Deaton, 2012; Garrah et al., 2015). For example, the mortality pattern of yacare suggests a higher likelihood of collisions with this species during summertime and rainy periods.

    • Assessing the impact of roadkill on the persistence of wildlife populations: A case study on the giant anteater

      2022, Perspectives in Ecology and Conservation
      Citation Excerpt :

      In BR262, the traffic volume is ca. 3350 vehicles per day in the west section (Fig. 1B) and 3850 vehicles per day in the east section (Fig. 1C) (DNIT, 2020) (see for further details on study area in Ascensão et al., 2021 and Noonan et al., 2021). Roadkill data has been systematically collected throughout the four road sections (Fig. 1B–E) (Ascensão et al., 2017, 2019b, 2021). Here, we used the dataset from Ascensão et al. (2021).

    • An artificial intelligent framework for prediction of wildlife vehicle collision hotspots based on geographic information systems and multispectral imagery

      2021, Ecological Informatics
      Citation Excerpt :

      This algorithm allows the optimization of multiple parameters efficiently and effectively (Kramer, 2017). Regarding the comparison of the classifiers, the RF ADASYN algorithm provided the best result (0.78 ± 12), improving the result obtained by (Ascensão et al., 2019), being the precedent that most closely approximates the methodology applied in this research. However, comparing results of different methodologies applied in different study areas are particularly difficult.

    • Spatiotemporal persistence of bat roadkill hotspots in response to dynamics of habitat suitability and activity patterns

      2021, Journal of Environmental Management
      Citation Excerpt :

      For instance, bat-vehicle collision patterns are often described annually (Fensome and Mathews, 2016), but substantial inter-annual variations occur and this has been scarcely analyzed. For other mammal species (e.g. mule deer, black bear, lowland tapir) these inter-annual differences are often associated with variations in environmental factors (e.g. land cover or climate), traffic volume or population fluctuations (Seiler and Helldin, 2006; Shilling and Waetjen, 2015; Ascensão et al., 2019). Moreover, some species showed marked seasonality in roadkill occurrences, which were particularly evident in mating and swarming periods (Fensome and Mathews, 2016).

    View all citing articles on Scopus
    View full text