SIMULATING ECOLOGICAL CHANGE IN THE STRATIGRAPHIC RECORD: WHEN CAN ABRUPT ENVIRONMENTAL CHANGE BE RESOLVED BY FAUNAL PROXIES?

The vast majority of fossil assemblages are a blend of specimens that did not coexist in life, and thus, represent different environmental conditions. This impacts paleoenvironmental reconstructions because the ecological indices calculated from such mixed samples can obscure short-lived environmental perturbations. We use a simulation approach to measure power to accurately reconstruct abrupt environmental changes using faunal assemblages given variable sampling strategies, intensities of bioturbation, and magnitudes of environmental change. As a case study, we used benthic foraminifera assemblages collected by IODP Expedition 341 at Site U1419 in the Gulf of Alaska. This record has a well constrained, high-resolution age model, and a strong faunal gradient related to oxygenation, which is represented by the first score of a detrended correspondence analysis (DCA-1). To simulate faunal assemblages, we fit kernel density estimates to the abundance of each species as a function of DCA-1, and used these KDEs to simulate possible communities at select DCA-1 values. After verifying that the simulated assemblages recovered DCA-1 values with high fidelity to the generating values, we simulated how processes such as sampling resolution, bioturbation, and completeness of sampling affect the ability to accurately reconstruct environmental shifts over a stratigraphic column. We found that completeness of sampling and sample width has the greatest effect on the ability to accurately detect short-lived shifts in faunal proxies, but brief events could still generally be distinguished from background even if their apparent magnitude was an underestimate of their true magnitude. These simulations also allow researchers to evaluate if an observed DCA-1 value could result from a mixture of communities expected during an event and during background conditions. We further demonstrate how this simulation framework can enable a researcher to create a sampling plan with a high probability of sampling abrupt events from a limited number of pilot samples. While our specific case study is framed around low-oxygen events ~200 years in duration, such simulation methods can be adapted to any faunal abundance distribution and temporal resolution to evaluate the probability of detecting abrupt events given a sampling scenario.