Toward a generalized framework for the analysis of historical data used in 'resurvey' studies

Background/Question/Methods

Long-term, large-scale environmental change necessitates empirical studies that span time frames from decades to centuries. As this length of time often precludes planned experiments, an increasingly popular option in ecology is the “resurvey” study, or the revisitation of past research sites. The unplanned nature of resurveys, however, brings challenges and potential inferential pitfalls to the contemporary researcher. An important emerging issue is how researchers can use historical survey data—which may greatly differ in methodological properties from contemporary data—to make analytically robust comparisons. Previous work has largely acknowledged the problems of such comparisons while employing a diverse set of strategies, statistical or otherwise, to account for errors. While no single analytical framework will satisfy the needs of all researchers, the increasing popularity and scientific value of resurvey studies demands a generalized accounting of inferential problems as well as common strategies for overcoming them.

Results/Conclusions

This talk will outline how statistical methods can be used to account for bias derived from issues such as imperfect detection, changing taxonomies, inability to re-find survey sites, differing survey methodologies, and varying levels of survey effort. Among various options, emphasis will be placed on the flexible nature of occupancy models to allow the inclusion, characterization, and estimation of multiple sources of uncertainty deriving from the resurvey process. Examples will be drawn from a centennial resurvey of birds in the Sierra Nevada of California, illustrating how distribution limits, site occupancy, and community metrics can be estimated from historical data while accounting for survey-based uncertainty. While such analytical tools show great potential in resurvey studies, their prospective success relies on the standards by which both historical data and contemporary data are collected. Thus, I will conclude by highlighting the survey attributes—including temporally replicated samples—that facilitate occupancy comparisons. Incorporating these features into survey methodologies will aid the measurement of occurrence change whether investigators are resurveying the past or setting baselines for the future.