Using Bayesian change point model to enhance understanding of the shifting nutrients-phytoplankton relationship.

Graphical abstract Highlights • A novel Bayesian change point model (BCPM) was proposed to testify abrupt change. • We uncovered the potential of BCPMs for causality determination. • Changes of nutrients and nutrients-Chla relationship did not occur simultaneously. • Nutrients increase did not drive the relationship change in Yilong Lake. Abstract Possibility of shifting nutrients-phytoplankton relationship in lakes requires methods with the ability to testify abrupt relationship change to facilitate efficient management. Bayesian change point model (BCPM) can handle multiple shifts in coefficients and/or in residual errors and therefore suits this requirement. We employed BCPMs to enhance understanding of the shifting nutrients-Chlorophyll a (Chla) relationship in Yilong Lake, which has undergone a regime shift from clear state to turbid state. We developed four candidate models to simulate nutrients-Chla relationship. Model selection results showed the relationship has changed and only one change point exists. Further research based on the selected model showed that (1) the change point was around the 96th observation, (2) nutrients increase did not drive the relationship change, (3) total phosphorus (TP) plays a more important role than total nitrogen on Chla increase, and (4) nutrients reduction is a better strategy than ecosystem recovery to effectively reduce Chla concentration. Therefore, TP reduction should have the priority in Yilong Lake. BCPM is convenient for model selection, posterior distribution acquisition, and relationship change quantification. It could provide critical information for causality deduction. These characters make it useful and extendable to explore shifting relationships in ecological field. [ABSTRACT FROM AUTHOR]