Predicting water quality in the Great Barrier Reef Catchments: Learning from long-term water quality monitoring data

© 2019 Shuci LiuRiverine water quality deterioration has become a global issue, affecting freshwater and near-shore marine ecosystems. To protect aquatic environmental health in these systems, effective water quality management strategies are critically important. Riverine water quality needs to be evaluated and managed at the catchment scale. Management decisions can be greatly supported and informed by understanding of: (1) the processes influencing water quality constituent concentrations; and (2) the key factors affecting these processes and by predictive capability. As such, catchment water quality modelling tools are important to gain this knowledge and predict water quality dynamics under different climatic and land use scenarios. A key challenge in modelling stream water quality dynamics is that there is the lack of comprehensive understanding of the role and importance of the various factors that drive differences in water quality across space and time, specifically: 1) the importance of catchment characteristics on the spatial variability in water quality; and 2) the importance of variations in environmental variables including weather and vegetation cover on the temporal variability in water quality in different landscapes. These issues have jointly led to a lack of predictive power of water quality models for multiple catchments over large regions. This research aims to improve current understanding of how riverine water quality varies spatially and temporally and the factors that drive this and to develop improved data-oriented analytical and predictive frameworks to achieve this. In this research, I investigated long-term, multiple-site water quality monitoring datasets to identify the dominant factors affecting pollutant processes and thus to inform water quality management. This research focused on the Great Barrier Reef (GBR) catchments in sub-tropical and tropical north-eastern Australia. The GBR has high environmental, societal and economical values both nationally and globally. This thesis contains three results chapters to achieve the overarching research objective. Chapter 4 addresses the spatial pattern of water quality and its linkage with heterogeneity in catchment landscape characteristics, using multivariate statistical analyses. By comparing spatial differences in long-term averaged water quality and catchment landscape characteristics, two groups of catchments were identified. This grouping reflected the main differences in water quality: group one sites had lower time-averaged concentrations (except for NOX), which were situated in wet areas with high sugar cane production and conservation zones, while group two sites had relatively higher average concentrations, were drier and with considerable grazing land use. Chapter 5 reports on an investigation into the relationships between spatial variability in water quality and catchment characteristics, using a multi-model statistical modelling approach. Specifically, the aim was to identify key catchment characteristics influencing average event-mean concentrations and to develop a spatial model predicting this. Results indicated that natural catchment characteristics were important factors for determining the variation of sediments and particulate nutrients over space, while anthropogenic characteristics (i.e., land use) were more important factors for dissolved nutrients species. The models developed were able to predict average event-mean concentrations well, with Nash-Sutcliffe efficiency (NSE) ranging from 0.64 to 0.98. Chapter 6 presents an investigation into the key factors affecting temporal changes in water quality, as summarized by variations in constituent event-mean concentrations. A Bayesian hierarchical modelling coupled with a Bayesian model averaging approach was used to evaluate the temporal change in stream water quality. The modelling results demonstrated that the key temporal controls varied between the two clusters of sites identified in Chapter 4, as well as among different constituents. Overall, catchment ground cover condition and soil moisture prior to runoff events were of greatest importance in determining the temporal change in event-mean concentration and event discharge characteristics were also important. Results from the three chapters were brought together to assess the degree to which the result has the potential to support the current water quality management and inform future design of water quality management in the GBR catchments. The spatial modelling results support the current working hypotheses underpinning water quality management in the GBR catchments, that cattle grazing, and sugarcane land uses are management priorities to reduce sediment and dissolved inorganic nitrogen, respectively. Based on the temporal modelling results, the future management practices could focus on maintaining good vegetation cover, especially for large and grazed catchments before the wet season. Finally, the water quality modelling framework used in this research provides a potential tool to evaluate catchment water quality, and what can be done to improve it, under future changes to climate, land uses and land management