Code repository for integrated occupancy model for single-visit semi-structured and single-visit time-removal sampling surveys

Multi-species structured surveys following specific sampling design are powerful tools for understanding species’ distributions and trends, but are not equally well-suited for surveying all species. Combining data collected under different protocols or sampling designs has the potential to boost sample size, extend geographic and environmental coverage, and improve model estimates. Datasets comprised of citizen science observations tend to be large, however, these data are not often used in analyses accounting for imperfect detection – volunteer observers typically do not collect the types of data which are most often used to estimate detectability. Here, we introduce a statistical model to combine discrete time-removal samples and continuous time detection/non-detection data to jointly estimate detection rate from single-visit surveys that either record time-to-detection, or total survey duration without time-to-detection. We demonstrate the model using the Integrated Monitoring of Bird Conservation Regions program which follows structured time-removal sampling, coupled with eBird checklist data of known duration to create a single hierarchical occupancy model accounting for imperfect detection. With a case study and simulations of 3 bird species in Utah spanning levels of prevalence and sample size, we demonstrate the potential for analytical improvements in predictive performance and coefficient estimation gained by coupling information gathered under these two survey designs. We find the largest performance gains in low prevalence species coupled with large detection increase with data integration. Furthermore, we find performance gains in a low prevalence species of conservation concern even with the integration of only two additional detections. Augmenting ecological analyses through strategic data integration can facilitate inference on increasingly rare species that may be data-deficient in any one dataset alone, and can improve the precision of ecological estimates for management planning.