Fig. 3

Illustration of the study design. We varied the autocorrelation range when simulating spatial covariates from 5 to 20 and tested for different missingness scenarios (ranging from 0% to 50% missing locations). To investigate how increasing forgiveness (i.e., the willingness to include steps with duration above the regular step duration) influenced model results, we varied its value from 1 (regular step selection) to 5 (considering steps that are five times the regular step duration). Finally, we tested five different methods to account for potential biases introduced by including irregular steps. This gave 3 x 6 x 5 x 5 = 450 combinations, each of which we replicated 100 times. We assumed step lengths (sl) to follow a gamma distribution, whereas turning angles (ta) followed a von Mises distribution