Abstract
1. Human activities have led to ecological regime shifts, first revealed at the community level in ecosystems. A regime shift in a biological community is a sudden change in the relative contributions of several taxa, resulting in a post-shift state that remains stable over the long term with a structure that is outside the boundaries of the ‘normal’ pre-shift variability. Most methods for regime shift detection are based on univariate statistics (e.g. commercial fish catch data, sea surface temperature anomalies). Multivariate methods suitable for identifying change in multi-species communities can be used to identify regime shifts in communities. 2. In this paper, I use a 37-year record (1972–2008) of phytoplankton in the Bay of Quinte (northeastern Lake Ontario) to demonstrate the use of several largely independent data analysis methods that are shown here to concur in their output. Among the most powerful procedures is an approach that models the anomalies around long-term Grand Mean and reference-point community structures that were compared to annual structures using Bray–Curtis community similarity coefficients. CUSUM plots of model residuals, segmented regression analysis and other tests are all useful to identify the location of break-points in records of anomalies. Follow-up significance testing was performed separately with permutation tests. Improved sensitivity of these techniques when applied to highly seasonal data was demonstrated after extraction of seasonal components as periodic functions. 3. Statistically significant shifts in the Bay of Quinte phytoplankton were detected in the year following an approximate 50% reduction in point-source phosphorus loading in early 1978 and again immediately after the establishment of invasive dreissenid mussels in the mid-1990s. Associated with this second intervention was an increased representation by species of the potentially toxic Cyanoprokaryote Microcystis, and dramatic declines in some diatom species, with significant implications for human use and food web function. 4. This paper provides a ‘tool box’ of methods (most freely available on the WWW) for those needing to distinguish between true shifts and normal inter-annual variability in biological communities. Ability to measure statistically significant change in communities can lead to enhanced understanding of cause–effect relations and to enhanced capabilities for prediction of change.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have