Spatial ecology analysis methods generally assume organism locations can be reliably sampled or that detection probability may be reliably estimated and mitigated. However, investigators designing field protocols for new studies or working with long-term datasets may have no reasonable way to predict detection probability a priori. We introduce the R package “DiagnoseHR,” simulation tools for assessing how variation in detection probability arising from landscape, animal behavior, and methodological processes affects spatial ecology inferences. We demonstrate the utility of simulation methods for spatial analysis method selection using a case study of home range analysis under multiple detection scenarios. We simulated correlated random walks in three landscapes that varied in detection probability and constructed home ranges from locations filtered through a range of sampling protocols.
Results/Conclusions
Home range estimates were less biased by reduced detection probability when sampling effort was increased, but the effects of sampling day distribution were minimal. Like others, we found that kernel density estimates were the least affected by variation in detection probability, while minimum convex polygons were most affected. Our results illustrate the value of quantifying uncertainty in spatial analysis methods and suggest that field biologists working in environments with low detection may wish to weight sample-size greater than concerns about temporal autocorrelation when designing sampling protocols. Using our simulation tools, other investigators may adopt a contingency approach to field data collection planning and method selection to increase data collection and analytical efficiency.