?Mixed? occupancy designs: When do additional single-visit data improve the inferences from standard multi-visit models?

Estimating occupancy while accounting for imperfect detection typically requires repeated surveys at sampling units. How-ever, mixed sampling designs are very common, where only a subset of sites is visited repeatedly, while the remainder are vis-ited only once, providing single-visit (SV) data. It is unclear whether SV data contribute to parameter estimates. Consequently, they have often been discarded in occupancy analyses. We conducted two simulation studies to understand the degree to which SV data contribute information to the estimation of occupancy and detection probability. In Simulation 1, we simulated detec-tion/non-detection data under different scenarios of repeated sampling and varying magnitudes of occupancy and detection probabilities. In Simulation 2, we included continuous covariates, to see whether these could enhance the information content of SV data. To each simulated data set, we fitted models containing between 0 and 5000 SV sites and compared the standard errors of the occupancy and detection estimates. We found that SV data always contributed some information to the estimation of both occupancy and detection in a mixed design. Their relative contribution was greatest when > 2 visits were conducted at the repeated-visit sites, and for species with higher detection probabilities. These results suggest that SV data are valuable when combined with repeated-visit data and lead to more precise estimates than when repeated-visit data are used alone. Includ-ing suitable continuous covariates into the analysis of the simulated data increased the contribution of SV data even more. This suggests that, in a mixed design, occupancy estimation could be optimized by measuring and modelling continuous covariates that explain at least some heterogeneity in occupancy and detection amongst sites. Thus, we recommend that for mixed-design data all the available information be used in a joint model to obtain the most precise detection-corrected occupancy estimates.(c) 2023 The Author(s). Published by Elsevier GmbH on behalf of Gesellschaft fur Okologie. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/)