Quantifying telemetry collar bias when age is unknown: A simulation study with a long-lived ungulate

Radiotelemetry collars are frequently used to estimate demographic parameters of animals, such as annual survival and parturition rates. If animals are collared for multiple years and statistical adjustments are not made, these estimates can be biased by an unrepresentative age structure and individual variability of collared animals. To quantify the effects of different factors on the magnitude of these potential biases, we created a computer simulation of the female portion of a barren-ground caribou (Rangifer tarandus granti) herd and then randomly assigned collars to individuals within the simulated population. Under our default model, based on the Western Arctic Herd monitoring program, caribou were collared randomly from all females aged 2 years and over, and they remained collared for a mean of 7 years. Our simulations revealed that survival rates were underestimated by approximately 3.4% and parturition rates were overestimated by approximately 3.3%. The magnitude of these biases increased when individuals remained collared for longer periods. Increased individual variability in the population resulted in only small increases in survival and parturition rates. Because the magnitude of the bias increased steadily during the first years of the study, we found a substantial risk of incorrectly identifying a significant decline in survival in the first 7 years after marking. Including the number of years individual animals have been collared as a covariate in analyses can reduce the biases in demographic parameters and should be considered for inclusion in analyses when animal age is unknown. Actual survival rate estimates from telemetry data for the Western Arctic Herd were generally consistent with the results of these simulations. These potential biases should be considered when interpreting demographic parameters from multi-year collaring studies.