Bias in estimating animal travel distance: the effect of sampling frequency

Summary 1. The distance travelled by animals is an important ecological variable that links behaviour, energetics and demography. It is usually measured by summing straight‐line distances between intermittently sampled locations along continuous animal movement paths. The extent to which this approach underestimates travel distance remains a rarely addressed and unsolved problem, largely because true movement paths are rarely, if ever, available for comparison. Here, we use simulated movement paths parameterized with empirical movement data to study how estimates of distance travelled are affected by sampling frequency. 2. We used a novel method to obtain fine‐scale characteristics of animal movement from camera trap videos for a set of tropical forest mammals and used these characteristics to generate detailed movement paths. We then sampled these paths at different frequencies, simulating telemetry studies, and quantified the accuracy of sampled travel distance estimation. 3. For our focal species, typical telemetry studies would underestimate distances travelled by 67–93%, and extremely high sampling frequencies (several fixes per minute) would be required to get tolerably accurate estimates. The form of the relationship between tortuosity, sample frequency, and distance travelled was such that absolute distance cannot accurately be estimated by the infrequent samples used in typical tracking studies. 4. We conclude that the underestimation of distance travelled is a serious but underappreciated problem. Currently, there is no reliable, widely applicable method to obtain approximately unbiased estimates of distance travelled by animals. Further research on this problem is needed.