Examining the accuracy of trackways for predicting gait selection and speed of locomotion

Abstract

Background: Using Froude numbers (Fr) and relative stride length (stride length: hip height), trackways have been widely used to determine the speed and gait of an animal. This approach, however, is limited by the ability to estimate hip height accurately and by the lack of information related to the substrate properties when the tracks were made, in particular for extinct fauna. By studying the Svalbard ptarmigan moving on snow, we assessed the accuracy of trackway predictions from a species-specific model and two additional Fr based models by ground truthing data extracted from videos as the tracks were being made.

Results: The species-specific model accounted for more than 60% of the variability in speed for walking and aerial running, but only accounted for 19% when grounded running, likely due to its stabilizing role while moving faster over a changing substrate. The error in speed estimated was 0-35% for all gaits when using the species-specific model, whereas Fr based estimates produced errors up to 55%. The highest errors were associated with the walking gait. The transition between pendular to bouncing gaits fell close to the estimates using relative stride length described for other extant vertebrates. Conversely, the transition from grounded to aerial running appears to be species specific and highly dependent on posture and substrate.

Conclusion: Altogether, this study highlights that using trackways to derive predictions on the locomotor speed and gait, using stride length as the only predictor, are problematic as accurate predictions require information from the animal in question.

Keywords: Arctic; Biomechanics; Gait; Locomotion; Snow; Speed; Track.

Fig. 1

Trackways of the Svalbard Rock Ptarmigan. a length (FL), Stride length (l_stride) and Hip Height (h) are used to estimate speed. h is the distance to the top of the hip perpendicular to ground. This measurement however, is entirely dependent on accurate joint angles of the leg bones. As illustrated, a small alteration in joint angle results in different hip height measurements (h < h’). It has been suggested that four times FL can be used as a proxy of effective hip height however this method is also prone to error when applied to extinct animals where there is no information on the joint angles. Digits are numbered 1–4 for analysis as indicated. b Speed (U) predictions from stride length (l_stride) using the ptarmigan species-specific model. The stars represent the data determined from video recordings and the filled circles represent the predicted value for a given l_stride for 50 birds calculated from trackways that corresponded to the video recordings. Red, green and blue represent walking, grounded running and aerial running gaits, respectively. The coloured area delimited by the dashed lines represents the predictive interval for the lines of best fit (corresponding to the filled circles) described by the following equations: U = 3.20 l_stride – 0.23 (walking), U = 2.34 l_stride + 0.34 (grounded running) and for aerial running is U = 4.29 l_stride – 0.03 (aerial running)

Fig. 2

Error estimation for the ptarmigan species-specific regression models and the Alexander (1976), and Thulborn, and Wade (1984) Fr derived equations. Bars to the left of zero represent estimations of U below the measured U, and vice versa. Note that the error estimates from Alexander (1976) and Thulborn, and Wade (1984) for walking gaits are identical because they use the same equation for this gait. Errors have been binned to 10