Overcoming radiotelemetry bias in habitat-selection studies

Abstract

For many species, determination of habitat selection is based on habitat-use data obtained through radiotelemetry. Recent papers pertaining to study techniques have largely ignored the effect of habitat-dependent bias in the performance of radiotelemetry systems. Such biases cannot be overcome by increasing radiotelemetry precision, excluding data, or increasing sample sizes, as the biases are centred around data that are missing or that contain habitat-dependent errors in location. The problem is best addressed at the data-analysis stage through the use of geographic information systems. We used Monte Carlo simulations to assess the effect of habitat-dependent bias in radiotelemetry studies on the assessment of habitat selection. We looked at the effects of habitat-patch size, level of telemetry signal inhibition, level of habitat co-occurrence, and selection pattern. We demonstrated that regarding use as the composition of habitat types within a circular area around each telemetry location can help to overcome the inaccurate assessment of habitat-selection patterns that biased data produce. The size of the circular area best able to overcome the bias is related to habitat patch size and to the level of association between two or more habitat types. Furthermore, we argue that the characteristics of habitat mosaics selected by animals can and should be studied in this way.