Causal impact analysis of enhanced phosphorus effluent standard on river water quality

The effect of environmental policies on water quality is an important factor in evaluating a project's impact and economic feasibility. The aim of this study was to evaluate the causal relationship between strengthening the total phosphorus (TP) concentration standard in wastewater treatment plant (WWTP) effluents and river water quality in the Geum River watershed (Rep. of Korea). Data, including precipitation, streamflow, and water quality, were collected for 2005-2020 and temporally divided into "before" and "after" the event of strengthening standards (in 2012). Further, the data were spatially divided into influence sites (impact) affected by the event and control sites (control) unaffected by the event. A median difference test for a before/after and control/impact (BACI) analysis and a causal impact analysis (CIA) based on a Bayesian structural time-series model were performed to evaluate the changes in water quality after the event. The BACI test showed that the TP concentration was reduced significantly (p-value<0.05) at all impact sites after the event, whereas the difference was not significant at the control sites. In contrast, other water quality variables, except TP, showed different statistical significance depending on the site. The CIA was performed by controlling the rainfall, flow rate, suspended solids, water temperature, biochemical oxygen demand, and chemical oxygen demand as covariates, which were selected based on a directed acyclic graph and bi-variable correlation analysis. The CIA results showed that the TP concentration was reduced significantly (p-value<0.05) at all impact sites, except for the control sites after the event, which is consistent with the BACI results. The causal impact of environmental management policies was previously difficult to evaluate by objectively targeting the natural systems because of the confounding bias. Our study demonstrated that strengthening the TP concentration standard from WWTPs majorly contributed to reducing TP in the receiving river, even when confounding factors, such as fluctuations in non-point source pollution loads caused by rainfall and runoff, were excluded. The statistical approaches used in this study can be valuable and practical methods for such evaluations if sufficient prior and posterior data and appropriate covariates are obtained.