Effects of imperfect detectability on inferences from monitoring

Principal Investigator: David E. Andersen.

Co-Principal Investigator: Doug Johnson.

Student: Beth Rigby (Ph.D., Natural Resources Science and Management – Wildlife Ecology and Management).

Status: Research Work Order established and Ph.D. graduate student selected. Field component of the project completed in 2012. Model development completed in 2014 and final report near completion. 

The value of bird monitoring has come under increasing scrutiny recently due to concerns about imperfect detectability. In particular, because the probability of detecting a bird in the area surveyed often is less than one, counts are indices of abundance, rather than actual estimates of abundance. Often the issue is cast in the equation,

E(C) = pN,

where E(C) is the expected count of some species made during a survey, N is the true number of that species in the surveyed area at the time of the survey, and p—detectability—is the proportion of the true number that is recorded. In recent years many authors have emphasized that variation in C reflects not only variation in N but also variation in p. From that fact, they caution against drawing inferences about population changes from indices.

A variety of methods that attempt to adjust counts for imperfect detectability have been advanced, including distance sampling, multiple-observer sampling, and time-to-detection sampling. Surveys, especially of breeding birds, are constrained by limited time—both during the season and within a day, suitable weather conditions, the number and skills of observers, access to sites, and other factors. Gathering the information necessary to employ these techniques can take additional effort and thereby reduce the number of sites that can be surveyed. Furthermore, the suitability of available adjustment methods in multispecies surveys has been questioned.

A critical question is the extent to which additional effort to employ these techniques is rewarded by improved results. Clearly, estimates of abundance will be affected by adjustments for detectability; what is not known are the consequences on estimates of population trajectories. The objective of the proposed study is to evaluate the influence of imperfect detectability on inferences about population changes. Results from this study could suggest that certain situations involving detectability seriously compromise conclusions drawn from a survey; in that case, appropriate adjustments may be strongly commended. Conversely, other types of variation in detectability might induce only inconsequential errors in survey results; then detectability adjustments may not be warranted.

Final Report: 2014 Final Report – Dectability in Bird Surveys (.pdf)