Historical datasets can be useful tools to aid in understanding the impacts of climate change on natural ecosystems. However, resampling historically sampled sites (snapshot resampling) to detect shifts in ecological communities through time needs to be done with caution. Strong inter-annual variability in ecological systems can complicate our ability to draw meaningful conclusions from such one-time resurveys. Here, we combine long-term empirical datasets with simulations to examine real-world implications of the effects of inter-annual variation for our ability to detect the signature of climate change through the resurvey of one-time, or “snapshot” historical sampling efforts. This is an important question given the urgency of detecting and understanding ecological change, and the potential value of historical data and biological collections. We describe and analyze examples of several common types of ecological datasets to assess how reliable some common kinds of ecological resampling applications tend to be. Specifically, we apply randomization approaches to these long-term datasets to determine how often snapshot resampling studies of each ecological response would yield results that are qualitatively in-line with those derived from the full, continuous dataset.
Results/Conclusions
At its best, snapshot resampling can offer insights into how ecosystems have shifted, and may continue to shift, as a result of global change. At its worst, the improper use of snapshot resampling may lead to spurious conclusions regarding the influence of climate change on ecosystem structure and function. Our analyses revealed that snapshot resampling detected the correct long-term trend in anywhere from 20% to 100% of cases. Simulated results found snapshot resampling to be more accurate for community-level metrics such as diversity as compared to measures of individual species abundance. Further, results were more accurate for communities with more species. In both simulated and empirical data, snapshot resampling was more accurate when the magnitude of change was great, and/or inter-annual variability of the response variable was low. The challenge for researchers using historical datasets lies in distinguishing when snapshot resampling of will appropriately reflect climate-induced shifts as opposed to normal inter-annual variability.