Targeted visual surveys improve detection of the western bumblebee in a Glacier study

Combining two methods to observe a species of interest

Teams monitoring pollinator communities may need to track overall diversity while also detecting specific species, with limited time and budgets. A USGS study examined whether adding targeted visual searches for Bombus occidentalis—the western bumblebee, a species described in the study as declining and of conservation concern—to community-wide net surveys could improve its detection without losing information about other bumblebees.

The work was conducted in 2023 at 22 sites in Glacier National Park. Bumblebees of the genus Bombus were surveyed there with nets, and targeted visual searches for the western bumblebee were added when it had not been detected through net capture. The question was not only which method found more individuals, but how to integrate both methods to estimate site occupancy and the probability of detecting the species.

Occupancy here refers to the estimated presence of the species at the study sites; detection refers to the possibility of observing it during surveys. These measures are related but not equivalent: a visit that produces no observation does not, on its own, demonstrate that the species is absent.

What the visual search contributed

When the methods were compared, models integrating net captures and visual searches produced occupancy estimates similar to those obtained using net surveys alone, but achieved a higher cumulative probability of detection. In addition, visual searches required less effort than conducting additional net surveys, according to the study summary.

Some of the observations show why the combination may be useful: 11 of the 66 visits (17%) yielded detections of the western bumblebee only during visual searches. On these visits, the targeted search added detections that had not appeared in the net survey. The result documents an improvement in detection under the evaluated design; it does not mean that visual searches will perform equally well at every location, in every season, or in every monitoring program.

Models and the cost of reducing effort

The team estimated occupancy and detection using two approaches: traditional detection/non-detection (DND) models and time-to-detection (TTD) models. Using the complete datasets, both approaches yielded comparable results. This comparison matters because conclusions about a sampling method also depend on how observations are analyzed.

The researchers also evaluated model performance when sampling effort was reduced through subsampling. In that scenario, they found high bias and lower precision, especially in models based only on net captures and in single-survey TTD models. Therefore, the favorable result from integrating methods does not eliminate the importance of having sufficient data: reducing the number of observations can affect the quality of estimates.

The study supports a practical option for the design evaluated: incorporating targeted visual searches during community-wide visits can increase the detection of a species of interest without increasing the number of visits. The conclusion concerns the probability of detecting the western bumblebee and the methods compared; it does not, on its own, demonstrate an improvement in species conservation or a new population trend.

Scope of the results

The results come from bumblebee surveys at 22 sites in a national park during 2023. The available summary does not detail the conditions of each visit or provide comparisons with other locations or seasons. Therefore, it cannot establish whether the same balance between methods would hold in other contexts.

The National Science Foundation (NSF) pages included among the sources describe the agency’s general support for research and its materials programs; they are not independent corroboration of this study and do not provide evidence about its findings. NSF: research support · NSF: materials research

Main reference: Stein, E., Janousek, W. M., Gustilo, E. S. and Graves, T. (2026). “Integrating targeted surveys during community-wide occupancy assessments improves detection rates for a pollinator species of concern”, Conservation Science and Practice.