A Bayesian Hierarchical Model To Assess Eutrophication Risk In West Lake, Hangzhou

Models are often fit to data sets composed of observations from multiple "sampling units", or called "subsystems" compared to the whole system. These subsystems may share some commonality, but each of them is unique in behaviour more or less, so that assumption of the same set of model parameter values across all subsystems may not always be valid. This problem can be overcome by adopting a hierarchical approach. Under the hierarchical structure, each subsystem has its own set of parameter values, but some commonality in values is assumed across the whole system, which can be structured by an underlying population distribution. We applied this hierarchical method to assessing eutrophication risk in West Lake, Hangzhou, China, a typical shallow eutrophication lake, artificially divided into 5 subsystems. The procedure followed was developing a hierarchical model relating eutrophication response — the level of Chlorophyll a (Chl-a) ~ to multiple routinely monitoring variables, and then introducing a new variable, named "the probability of standard violation", indicating the expected frequency of Chl-a standard exceedance. For comparison, both global and subsystem-specific parameters were estimated using Bayes Theorem. Results showed that the hierarchical model was more realistic than the global model. Furthermore, in Bayesian perspective, predictions expressed as probabilities, rather than a single value, involving more uncertainty information, can be essential to environmental management and decision-making.