Elsevier

Animal Behaviour

Volume 85, Issue 1, January 2013, Pages 269-280
Animal Behaviour

Commentary
Using conditional circular kernel density functions to test hypotheses on animal circadian activity

https://doi.org/10.1016/j.anbehav.2012.09.033Get rights and content

Highlights

► We describe conditional circular kernels to test hypotheses in animal activity. ► The models allow control of the smoothing parameter and density isopleth. ► We apply our models to camera-trap data on large mammals from the Pantanal. ► The models adequately describe the timing of main activity peaks. ► Activity range and activity overlap prove useful in comparing the species.

Section snippets

Circular kernel density estimation

Nonparametric kernel density estimation is a widely used technique for nonparametric estimation of densities based on a data set (Silverman 1986; Wand & Jones 2005). Kernel density estimators for circular/directional data sets are discussed in several studies, such as those of Hall et al. (1987), Bai et al. (1988), Prayag & Gore (1990), Hendriks (1990), Kim (1998) and Klemelä (2000). Because we are interested in estimation on a circle, that is, circular data, we can restrict our attention to

Testing the Effect of the Parameters on the Activity Range and Overlap

The analytical methods described here were applied to data obtained through camera-trap monitoring conducted at the Nhumirim Research Station (18°59′S, 56°37′W), in the Pantanal wetlands of Brazil. The study was approved under license number ICMBio-21560-1 issued by Brazil's federal environmental agency. We established a 50-trap-station grid, with camera traps (Tigrinus Equipment Research Ltd, Timbó, Santa Catarina, Brazil) systematically placed 1.5–2 km apart, covering 80 km2. The entire grid

Results

The graphic representation of the activity patterns fitted using our circular kernel approach described the raw data distribution in accordance with our previous expectation, both from field experience and from the exploratory analysis of the raw data (Fig. 1c, d; Appendix, Fig. A2). The estimated activity range conditioned to any given isopleth was smaller for higher κ values (overfitted functions) for all species (Fig. 1a, b; Appendix, Fig. A3). However, this tendency was stronger for species

Case Study: Activity Ranges and Overlap among the Invasive Feral Hog and Two Peccary Species

The feral pig, Sus scrofa, has been introduced worldwide and is considered to be one of the most harmful invasive species in the world, causing damage to populations, communities and ecosystems (Lowe et al. 2000). As it was introduced in the Pantanal approximately 200 years ago and shares several ecological, behavioural (Kiltie 1982; Desbiez et al. 2009), morphological and physiological (Bodmer 1991; Sicuro & Oliveira 2002; Elston et al. 2005) features with the South American peccary species,

Activity range and activity overlap models

The major advantages of using kernel functions to fit activity data are related to their continuous and circular structure. These factors allow us to avoid problems associated with the arbitrariness of the timescale categorization and definition of the origin (usually noon or midnight in linear approaches). Additionally, as nonparametric functions, they can adequately fit either a bimodal or multimodal pattern, which are both widespread among animals.

For our purposes, the automated smoothing

First page preview

First page preview
Click to open first page preview

References (48)

  • M. Di Bitetti et al.

    Time partitioning favors the coexistence of sympatric crab-eating foxes (Cerdocyon thous) and pampas foxes (Lycalopex gymnocercus)

    Journal of Mammalogy

    (2009)
  • D. Donoho et al.

    The ‘automatic’ robustness of minimum distance estimators

    Annals of Statistics

    (1988)
  • J. Elston et al.

    Digestive efficiency of collared peccaries and wild pigs

    Southwestern Naturalist

    (2005)
  • T. Gabor et al.

    Renal morphology of sympatric suiforms: implications for competition

    Journal of Mammalogy

    (1997)
  • H. Gómez et al.

    Dry season activity periods of some Amazonian mammals

    Studies on Neotropical Fauna and Environment

    (2005)
  • P. Hall et al.

    Kernel density estimation with spherical data

    Biometrika

    (1987)
  • H. Hendriks

    Nonparametric estimation of a probability density on a Riemannian manifold using Fourier expansions

    Annals of Statistics

    (1990)
  • L. Ilse et al.

    Resource partitioning in sympatric populations of collared peccaries and feral hogs in southern Texas

    Journal of Mammalogy

    (1995)
  • A.d.A. Jácomo et al.

    Niche separation between the maned wolf (Chrysocyon brachyurus), the crab-eating fox (Dusicyon thous) and the hoary fox (Dusicyon vetulus) in central Brazil

    Journal of Zoology

    (2004)
  • S. Jammalamadaka et al.

    Topics in Circular Statistics. Vol. 5. Multivariate Analysis

    (2001)
  • R. Kiltie

    Bite force as a basis for niche differentiation between rain forest peccaries (Tayassu tajacu and T. pecari)

    Biotropica

    (1982)
  • P. Kim

    Deconvolution density estimation on SO(N)

    Annals of Statistics

    (1998)
  • N. Kronfeld-Schor et al.

    Partitioning of time as an ecological resource

    Annual Review of Ecology, Evolution and Systematics

    (2003)
  • M. Linkie et al.

    Assessing tiger–prey interactions in Sumatran rainforests

    Journal of Zoology

    (2011)
  • Cited by (66)

    • Diel timing of migration is not plastic in a migratory ungulate

      2022, Animal Behaviour
      Citation Excerpt :

      After we converted speed into detection probabilities, we fitted nonparametric kernel density curves to activity patterns (Cunningham et al., 2019; Ridout & Linkie, 2009). After visually inspecting multiple possible bandwidths, we set the bandwidth to five for all populations and seasons because it appeared to appropriately capture nuanced variation in activity without overfitting (Oliveira-Santos et al., 2013; Ridout & Linkie, 2009). We derived activity curves at the level of individual, population and season.

    • Species differences in temporal response to urbanization alters predator-prey and human overlap in northern Utah

      2022, Global Ecology and Conservation
      Citation Excerpt :

      Moderate changes were considered to be when one of the two significance criteria above was met. Finally, we used the methodology described by (Oliveira-Santos et al., 2013) to calculate and describe the 50% isopleth for each study site’s activity distribution, which we hereafter refer to as ‘core activity.’. To assess whether species-specific temporal activity differed across study sites, we first grouped each species’ detections into either day (representing detections captured between 0800 and 1859 h), night (representing detections captured between 2200 and 0459 h), or crepuscular (representing detections captured between both 0500–0759 h and 1900–2159 h) detections.

    • Circular-linear-linear probabilistic model based on vine copulas: An application to the joint distribution of wind direction, wind speed, and air temperature

      2021, Journal of Wind Engineering and Industrial Aerodynamics
      Citation Excerpt :

      In the present study, KDE is employed for the goodness-of-fit test of the multivariate parametric joint distribution model built using vine copulas. At present, KDE of univariate data has already become a mature technique, such as KDE of linear variables (Ouarda et al., 2015; Qin et al., 2011; Wang et al., 2016; Xu et al., 2015), and KDE of circular variables (Oliveira-Santos et al., 2013; Oliveira et al., 2012; Taylor, 2008; Tsuruta and Sagae, 2018). However, only a few studies have reported KDE of the C-L dataset (García-Portugués et al., 2013; Han et al., 2018), and there are barely any discussions on KDE of the C-L-L dataset.

    View all citing articles on Scopus
    View full text