Detecting and explaining long‐term changes in river water quality in south‐eastern Australia

Abstract

AbstractUnderstanding the temporal changes in river water quality is important for catchment water quality management. This study aims to detect and attribute long‐term trends and abrupt changes in river water quality. We used 26 years of water quality data (1994–2020) collected from 102 river monitoring sites across Victoria, south‐eastern Australia. We analysed six water quality constituents that are of key concerns for Australian catchment management, namely: electrical conductivity (EC), total suspended solids, nitrate‐nitrite, total Kjeldahl nitrogen, total phosphorous and filtered reactive phosphorus. To detect trends and abrupt changes in water quality at each site, a Bayesian ensemble modelling approach was applied, namely, the Bayesian estimator of abrupt change, seasonal change, and trend (BEAST). To explain water quality trends, we then built multivariate regressions to link water quality with streamflow and seasonality, and then compared alternative model structures with and without a change in the regression relationships informed by the changes detected by BEAST. Among the six constituents studied, EC shows the most distinct systematic trends, with 21 sites having a significant increase followed by a non‐significant trend; within the 21 sites, 14 had a significant change point in EC around Year 2010. The regression analyses between water quality and streamflow suggested that the observed systematic change in EC could be largely related to reduced streamflow during the Millennium drought, which greatly impacted the climate and hydrology of south‐eastern Australia over the first decade of 2000. The results of this study can help inform the design of effective mitigation strategies and avoid further degradation of water quality across Victoria. Besides, our trend analysis and attribution approaches are applicable to water quality time series in other regions for robust trend analysis and change point detection.