Can space replace time in occupancy studies when the method of detection requires retaining individuals?

Models that estimate occupancy probabilities in the presence of imperfect detection are gaining popularity in fisheries and wildlife studies.  The data for these models result from repetitive sampling over time or space within the study area.  Sampling over space rather than time may be necessary when sites are remote or difficult to access repeatedly.  Occupancy models require that the repeated samples provide independent observations of detection or nondetection.  If spatial replicate samples are used and animals are retained during detection sampling, as in most fisheries studies, samples may not be effectively independent because retained individuals cannot be detected in other samples.  This problem causes occupancy and detection estimates to be biased and is most likely to occur when local population abundance is low.  We conducted a field study on endangered juvenile catostomids in Upper Klamath Lake, Oregon, to compare results from occupancy models based on temporal and spatial replicate study designs. Occupancy estimates were imprecise due to small sample sizes, but appeared to be biased high when based on spatial replicates. Accordingly, detection probabilities for spatial replicates were lower than for temporal replicates. Our results support previous simulation studies that caution against substituting space for time in occupancy studies.