Interpolation of animal tracking data in a fluid environment

Abstract

SUMMARY Interpolation of geolocation or Argos tracking data is a necessity for habitat use analyses of marine vertebrates. In a fluid marine environment,characterized by curvilinear structures, linearly interpolated track data are not realistic. Based on these two facts, we interpolated tracking data from albatrosses, penguins, boobies, sea lions, fur seals and elephant seals using six mathematical algorithms. Given their popularity in mathematical computing,we chose Bézier, hermite and cubic splines, in addition to a commonly used linear algorithm to interpolate data. Performance of interpolation methods was compared with different temporal resolutions representative of the less-precise geolocation and the more-precise Argos tracking techniques. Parameters from interpolated sub-sampled tracks were compared with those obtained from intact tracks. Average accuracy of the interpolated location was not affected by the interpolation method and was always within the precision of the tracking technique used. However, depending on the species tested, some curvilinear interpolation algorithms produced greater occurrences of more accurate locations, compared with the linear interpolation method. Total track lengths were consistently underestimated but were always more accurate using curvilinear interpolation than linear interpolation. Curvilinear algorithms are safe to use because accuracy, shape and length of the tracks are either not different or are slightly enhanced and because analyses always remain conservative. The choice of the curvilinear algorithm does not affect the resulting track dramatically so it should not preclude their use. We thus recommend using curvilinear interpolation techniques because of the more realistic fluid movements of animals. We also provide some guidelines for choosing an algorithm that is most likely to maximize track quality for different types of marine vertebrates.

Demographic implications of alternative foraging strategies in juvenile loggerhead turtles Caretta caretta of the North Pacific Ocean

“…In order to obtain data that were equally spaced temporally, we interpolated consecutive ARGOS hits every 12 h following the methods of Tremblay et al (2006). We determined multiindividual utilization distributions (UDs) with a Gaussian kernel density analysis of the positions from tracks from each habitat.…”

Trophic cascades in the western Ross Sea, Antarctica: revisited

“…Satellite positions were filtered using the 'Argosfilter' package for R (Freitas 2012), which first removes all records with invalid locations (class Z), then all locations that require unrealistic swimming speeds (maximum 2.3 m s −1 ), and finally removes all spikes with angles <15 and 25° if their extension is higher than 2500 and (Freitas et al 2008). Because raw Argos positions are biased by satellite orbital parameters and the penguins' latitudinal positions (Georges et al 1997), we used a linear interpolation algorithm (Tremblay et al 2006) to create a temporally uniform distribution of locations every 15 min along each foraging track. Because of the differences in temporal resolution between the dive (1 s) and location data (15 min, after interpolation), we used temporal proximity to assign an approximate geographic location to each dive.…”

Behavioural mapping of a pelagic seabird: combining multiple sensors and a hidden Markov model reveals the distribution of at-sea behaviour

“…Because of some irregularity in GPS signal acquisition, occasional missing locations and the need to calculate speed within discrete intervals, we found it necessary to interpolate regular locations prior to implementing the HMM. We used a curvilinear interpolation, which typically produces more realistic estimates of animal movements than linear interpolation [26] and is less prone to underestimating path length and speed. Our tests indicated that it performed well in recreating GPS data recorded at a higher resolution (see the electronic supplementary material).…”