Evaluation of algal species distributions and prediction of cyanophyte cell counts using statistical techniques.

Safe drinking water sources are crucial for human health. Consequently, water quality management, including continuous monitoring of water quality and algae at sources, is critical to ensure the availability of safe water for local residents. This study aimed to construct statistical prediction models considering probability distributions relevant to cyanophyte cell counts and compare their prediction performance. In this study, water quality parameters at Juam Lake and Tamjin Lake, representative water sources in the Yeongsan and Seomjin rivers, South Korea, were investigated. We used a water quality monitoring network, algae alert system, and hydraulic and hydrological data measured every 7 days from January 2017 to December 2022 from the Water Environment Information System of the National Institute of Environmental Research. Using data for 2017-2021 as a training set and data for 2022 as a test set, the performances of seven models were compared for predicting cyanophyte cell counts. Environmental factors associated with algae in water sources were observed based on the monitoring data, and a prediction model appropriate for the cyanophyte distribution was generated, which also included the risk of toxicity. The extreme gradient boosting with the random forest model had the best predictive performance for cyanophyte cell counts. The study results are expected to facilitate water quality management in various water systems, including water sources.

Verification of water environment monitoring network representativeness under estuary backwater effects.

The multi-functional weirs constructed as part of the Four Major River Restoration Project in Korea are operated for water level management and may have a backwater effect in estuaries. If the main channel of the Nakdong River flows backward and affects the estuary water, the water quality in the estuaries may not be representative of the tributary water quality. In this study, we confirmed the representativeness of the existing water quality monitoring networks using spatiotemporally disperse electrical conductivity observations, self-organizing maps (SOMs) for monthly pattern analysis, and the LOcally WEighted Scatter plot Smoother (LOWESS) technique for trend analysis. The results show that the Namgang 4-1 site, which is located in the Nam River estuary, is not affected by the Nakdong River, while the Baekcheon (Sunwongyo) site in the Baekcheon estuary is always affected by the Nakdong River. Therefore, it is necessary to relocate the existing monitoring network or establish a new monitoring network for locations affected by mainstream backflow, as is seen in Baekcheon (Sunwongyo). The methods proposed in this study, including spatiotemporally diverse electrical conductivity measurement, dimensionless fluctuation values, SOMs, and LOWESS, can be used to verify the representativeness of water quality measurement networks in other regions.