Determining the impact of reservoir water transfers on water quality using advanced methods

Abstract

Inter-basin water transfers in reservoir systems, usually driven by purely economic purposes (e.g. water quantity objectives such as water supply and hydropower production), can have complex physical, chemical, hydrological and biological implications both in the upstream and downstream reservoir. Phytoplankton growth is a complex process, usually driven by internal dynamics and mainly dependent on nutrient concentrations, temperature and light availability. When water is transferred from an upstream to a downstream reservoir, phytoplankton cells can potentially be transported, thus causing changes in the phytoplankton community. Determining if a causal relationship exists between the phytoplankton flux from the upstream reservoir and the measured phytoplankton biovolume in the downstream reservoir becomes an important step towards the understanding and management of reservoir systems. In the present study, simple and novel statistical methods (e.g. randomized intervention analysis and Granger causality), based on time series analysis and linear regression, were used to detect trends of phytoplankton biovolume and causal relationships between two interconnected reservoirs, thus providing significant evidence of how phytoplankton biovolume was affected by the magnitude of water transfers. Our analysis was supported by a weekly dataset of phytoplankton biovolume in both the reservoirs and by daily water quantity data (transferred flow rate from the upstream to the downstream reservoir). The studied scheme is the Shoalhaven System, Australia, which was built in the 1970s as a water supply and hydropower generation system. All the applied methods are suitable to be used in any kind of connected water system, provided a long-term dataset is available. Two different time periods, characterized by low and high water transfers between the reservoirs, were identified and a Granger causality statistical test was applied on each of them. A causal relationship was found only during the high transfer period, i.e. when the transferred flow rate exceeded a certain threshold. This result demonstrates that the phytoplankton flux due to water transfers was one of the causes of the observed phytoplankton biovolume in the downstream reservoir. Therefore the increase of the observed biovolume in the downstream reservoir during the high transfer period was due to a combination of processes, i.e. the transport of phytoplankton cells from the upstream reservoir, a seeding effect of the imported phytoplankton and internal processes. The same procedure was applied on specific phytoplankton groups, i.e. diatoms, chlorophytes and cyanobacteria, testing the hypothesis of a causal relationship between measured biovolume in the downstream reservoir and flux from the upstream reservoir via water transfers. It was demonstrated that diatoms was the only phytoplankton group experiencing a causal relationship with the diatom flux from the upstream reservoir. The application of the Granger causality test to ecological time series in the context of reservoir systems revealed to be a valuable statistical contribution to the understanding and management of connected water bodies. Future analyses will concentrate on broader applications of this methodology in the context of the understanding and management of reservoir systems.