Rapid urbanization and industrialization have caused water quality issues in urban rivers. Appropriate measures based on water quality monitoring systems and prediction methods are needed for water quality management. While South Korea has operated a water quality monitoring system that measures various environmental factors and has accumulated water quality data, a water quality prediction system is not in place. This study suggests a water quality prediction method based on a long short-term model using water quality and meteorological monitoring data. Additionally, we present a derived input set of the prediction model that can improve the prediction model performance. The prediction model’s performance was evaluated by the coefficient of determination under various conditions, such as the hyperparameters, temporal resolution of input data, and application of upstream and downstream data. As a result, using the temporal resolution of the input data as hourly data improved predictions by an average of 25.6% over three days of the prediction period compared to daily data. Meanwhile, it was analyzed that the hyperparameters and using upstream and downstream data have a minor effect on the model performance. The results of this study underscore the crucial role of the number, duration, and temporal resolution of available monitoring data in water quality management.

In order to perform adequate water quality management, it is important to predict the water quality through measurement and data accumulation of the concentration of contaminants. However, daily measurement of water quality pollutant is unrealistic in practical aspect. In this study, the possibility of daily- or hourly-based water quality prediction through dissolved oxygen (DO) using RNN-LSTM (Recurrent Neural Network-Long Short-term Memory) algorithm, which is well-known for time-series learning, was performed. The research selected Bugok Bridge in Oncheon-stream, Busan, South Korea. Hourly-based DO, temperature, wind speed, relative humidity, rainfall data was collected at the target location and was converted to daily data. To forecast the DO concentration, TensorFlow, a deep learning open source library developed by Google, was utilized. Data of four years (2014-2017) was used for daily learning data and 2018 data was used for verification of the trained model. The performance with the adjusted number of hidden layers, number of repetitions, and the sequence length, as well as the accuracy of the model was analyzed. As a result of this research, it is proven that the performance of the prediction can be improved when weather data and large amount of data is available.

In this study, the catchment hydrology cycle analysis tool (CAT) model was used to conduct a comprehensive hydrological analysis of the water balance of agricultural reservoirs. Data from 2010 to 2017, including precipitation, water level data in the reservoir, groundwater usage, and wastewater discharge, were collected and compiled for the upper reaches of the Hantan River Dam. The current conditions and content curves of the 11 reservoirs within the watershed were investigated and recorded. The results were analyzed by simulating three scenarios: treating the entire watershed as 1 unit, dividing the watershed into 5 sub-watersheds according to the standard watershed criteria, and further subdividing it into 27 watersheds, taking into account the presence of agricultural reservoirs. In cases where watershed information is lacking, it is deemed that subdividing the watershed can enhance efficiency. The highest model efficiency was observed in the 27 sub-basins, particularly when accounting for agricultural reservoirs. This study proposed an efficient method for hydrological analysis of watersheds including ungauged areas.

The use of monitoring in river management is known to be both economical and rational, and the amount of digital information globally is increasing over time. AI research utilizing such data has been widely employed recently in the field of water resources and hydrology, yielding excellent predictive results. In this study, we utilized DO (Dissolved Oxygen) factor and meteorological data collected from the Bugok Bridge site in the Oncheoncheon watershed through an automatic water quality measurement network. We employed the LSTM (Long Short Term Memory) algorithm, a type of deep learning known for its excellent time series learning capabilities, as the learning algorithm. To confirm the potential of the use of big data, we conducted a comparative analysis by performing hourly and daily predictions, and an accuracy analysis by comparing actual and predicted data. For data utilization, missing data from the data collected by the automatic measurement network were linearly interpolated. It was confirmed that the predictive performance for the DO factor was higher using hourly than daily data.

Effective flood management in agricultural fields has become increasingly important due to the rising frequency and intensity of rainfall events driven by climate change. This study investigates the applicability of urban flood analysis models—SWMM (1D) and K-Flood (2D)—to irrigated agricultural fields with artificial drainage systems. A case study was conducted in a rural area near the Sindae drainage station in Cheongju, South Korea, using rainfall data from an extreme weather event in 2017. The models simulated inland flooding and were validated against flood trace maps provided by the Ministry of the Interior and Safety (MOIS). Receiver Operating Characteristic (ROC) analysis showed a true positive rate of 0.565, a false positive rate of 0.21, and an overall accuracy of 0.731, indicating reasonable agreement with observed inundation. Scenario analyses were also conducted to assess the effectiveness of three improvement strategies: reducing the Manning coefficient, increasing pump station capacity, and widening drainage channels. Among them, increasing pump capacity most effectively reduced flood volume, while channel widening had the greatest impact on reducing flood extent. These findings demonstrate the potential of urban flood models for application in agricultural contexts and support data-driven planning for rural flood mitigation.