Efficient monitoring is essential for endangered species conservation and recovery programs. A key component of monitoring is using methods that account for imperfect detection. Occupancy modeling is an analytical method that uses detection probabilities to account for imperfect detection, however these methods are rarely used in entomology studies. Evidence for declines in several bumble bees (Bombus spp.) and the recent listing of Bombus affinis, the rusty patched bumble bee, to the federal endangered species list in 2017, demand that better methods be used for monitoring bumble bee populations. Due to a lack of standardized sampling and the use of methods that do not account for imperfect detection, reliable benchmark data for bumble bees is largely lacking. We had two objectives for this study. Our first objective was to estimate occupancy and detection probabilities for Bombus species, particularly B. affinis. Our second objective was to determine which site (area impervious surface within 1 km radius) and survey (area of blooming flowers) variables had the greatest impact on detection probabilities. We performed bumble bee surveys at 94 sites along roadsides in the metro counties of Minnesota in 2018 following a conditional sampling design. We used single season occupancy modeling to analyze our results.
Results/Conclusions
We analyzed the results for five species and species groups within the Bombus genus: B. affinis, B. griseocollis, B. rufocinctus, B. fervidus/B. borealis, and B. vagans/B. sandersonii. We found that predicted occupancy (Ѱ) for each species was: B. affinis Ѱ=0.0364 (sd=0.1759), B. griseocollis Ѱ=0.7389 (sd=0.3477), B. rufocinctus Ѱ=0.3824 (sd=0.4070), B. fervidus/B. borealis Ѱ=0.2596 (sd=0.4002), and B. vagans/B. sandersonii Ѱ=0.6400 (sd=0.4007). Our detection probability (p) estimates were similar across species, with an average of p=0.3497 (0.2590-0.4246). Occupancy differed greatly among species. However, detection probabilities were similar for each species, indicating that surveyors have the same chance of detecting each Bombus species if it is present at a site, including B. affinis. This is important for monitoring programs for the endangered species. We did not find consistent evidence that site and survey variables affected occupancy or detection. In three of the five species groups, the top models did not include any site or survey variables. However, the top model for B. griseocollis included floral area and the top model for B. vagans included floral area and impervious surface. This method was useful for determining occupancy and detection of Bombus species and should be incorporated into long-term bumble bee monitoring programs.