Assessing ecological monitoring design using a hybrid approach to guide changes to long term monitoring: A zooplankton case study

Ecosystem monitoring frequently focuses on populations, but a community perspective better reflects food web ecology. We asked a common question among natural resource managers – how much sampling is needed to understand the community in my system? We identified two objectives: (1) provide a community perspective and (2) ensure computational effort does not become unwieldy. In the San Francisco Estuary, five decades of zooplankton samples have been collected due to their importance as prey for fishes. We conducted a thorough investigation into the gaps and strengths of these data using a combination of three analytical approaches: representative taxa selections, power analysis, and a simulated reduced sampling. A principal tensor analysis divided the zooplankton community into co-occurring subassemblages and a silhouette analysis allowed us to select representative taxa from each subassemblage. We used univariate statistical analyses for each representative taxon to find gaps or overlaps in sampling that span the zooplankton community. We conducted power analyses on generalized linear mixed effect models and found that statistical power varied depending on the combination of temporal and spatial scale and across the effect sizes we tested (0.1, 0.2, 0.5, and 1). For mesozooplankton, which had greater statistical power, we simulated reductions in sampling effort and determined its impact on abundance estimates using Bayesian models; we observed less redundancy for temporal reductions compared with spatial reductions in sampling. Our approach is generally transferrable to evaluating the sampling effort needed to monitor changes in a biological community over time and space.