Jae-Ki Shin

Rivers act as natural conduits for the transport of plastic debris from terrestrial sources to marine environments. Accurately quantifying plastic debris in surface waters is essential for comprehensive environmental impact assessments. However, research on the detection of plastic debris in surface waters remains limited, particularly regarding real-time monitoring in natural environments following heavy rainfall events. This study aims to develop a real-time visual recognition model for floating plastic debris detection using deep learning with multi-class classification. A YOLOv8 algorithm was trained using field video data to automatically detect and count four types of floating plastic debris such as common plastics, plastic bottles, plastic film and vinyl, and fragmented plastics. Among the various YOLOv8 algorithms, YOLOv8-nano was selected to evaluate its practical applicability in real-time detection and portability. The results showed that the trained YOLOv8 model achieved an overall F1-score of 0.982 in the validation step and 0.980 in the testing step. Detection performance yielded mAP scores of 0.992 (IoU = 0.5) and 0.714 (IoU = 0.5:0.05:0.95). These findings demonstrate the model’s robust classification and detection capabilities, underscoring its potential for assessing plastic debris discharge and informing effective management strategies. Tracking and counting performance in an unknown video was limited, with only 6 of 32 observed debris items detected at the counting line. Improving tracking labels and refining data collection are recommended to enhance precision for applications in freshwater pollution monitoring.

Small-scale reservoirs located in river estuaries are a significant water resource supporting agricultural and industrial activities; however, they face annual challenges of eutrophication and algal bloom occurrences due to excessive nutrient accumulation and watershed characteristics. Efficient management of algal blooms necessitates a comprehensive analysis of their spatiotemporal distribution characteristics. Therefore, this study aims to develop a chlorophyll-a (Chl-a) estimation model based on high-resolution satellite remote sensing data from Sentinel-2 multispectral sensors and multiple linear regression. The multiple linear regression (MLR) models were constructed using multiple reflectance-based variables that were collected over 2 years (2021–2022) in an estuarine reservoir. A total of 21 significant input variables were selected by backward elimination from the 2–4 band algorithms as employed in previous Chl-a estimation studies, along with the Sentinel-2 B1-B8A wavelength ratio. The developed algorithm exhibited a coefficient of determination of 0.65. Spatiotemporal variations in Chl-a concentration generated by the algorithm reflected the movement of high Chl-a concentration zones within the body of water. Through this analysis, it turned out that Sentinel-2-based spectral images were applicable to a small-scale reservoir which is relatively long and narrow, and the algorithm estimated changes in concentration levels over the seasons, revealing the dynamic nature of Chl-a distributions. The model developed in this study is expected to support effective algal bloom management and water quality improvement in a small-scale reservoir or similar complex water quality water bodies.

Many countries have attempted to mitigate and manage issues related to harmful algal blooms (HABs) by monitoring and predicting their occurrence. The infrequency and duration of HABs occurrence pose the challenge of data imbalance when constructing machine learning models for their prediction. Furthermore, the appropriate selection of input variables is a significant issue because of the complexities between the input and output variables. Therefore, the objective of this study was to improve the predictive performance of HABs using feature selection and data resampling. Data resampling was used to address the imbalance in the minority class data. Two machine learning models were constructed to predict algal alert levels using 10 years of meteorological, hydrodynamic, and water quality data. The improvement in model accuracy due to changes in resampling methods was more noticeable than the improvement in model accuracy due to changes in feature selection methods. Models constructed using combinations of original and synthetic data across all resampling methods demonstrated higher prediction performance for the caution level (L-1) and warning level (L-2) than models constructed using the original data. In particular, the optimal artificial neural network and random forest models constructed using combinations of original and synthetic data showed significantly improved prediction accuracy for L-1 and L-2, representing the transition from normal to bloom formation states in the training and testing steps. The test results of the optimal RF model using the original data indicated prediction accuracies of 98.8% for L0, 50.0% for L1, and 50.0% for L2. In contrast, the optimal random forest model using the Synthetic Minority Oversampling Technique–Edited Nearest Neighbor (ENN) sampling method achieved accuracies of 85.0% for L0, 85.7% for L1, and 100% for L2. Therefore, applying synthetic data can address the imbalance in the observed data and improve the detection performance of machine learning models. Reliable predictions using improved models can support the design of management practices to mitigate HABs in reservoirs and ultimately ensure safe and clean water resources.

Understanding water quality events in a multiple-impoundment series is important but seldom presented comprehensively. Therefore, this study was conducted to systematically understand the explosion event of geosmin (GSM) in the North Han River (Chuncheon, Soyang, Euiam, and Cheongpyeong Reservoirs) and Han River (Paldang Reservoir), which consists of a cascade reservoir series, the largest drinking water source system in South Korea. We investigated the spatiotemporal relationship of harmful cyanobacterial blooms in the upstream reservoir (Euiam) with the water quality incident event caused by the GSM in the downstream reservoir (Paldang) from January to December 2011. The harmful cyanobacterial bloom occurred during August–September under a high water temperature (>20 °C) after a heavy-rainfall-based flood runoff event. The high chlorophyll-a (Chl-a) concentration in the upper Euiam Reservoir was prolonged for two months with a maximum concentration of 1150.5 mg m⁻³, in which the filamentous Dolichospermum circinale Kütz dominated the algal community at a rate of >99%. These parameters remarkably decreased (17.3 mg Chl-a m⁻³) in October 2011 when the water temperature decreased (5 °C) and soluble reactive phosphorus was depleted. However, high and unprecedented GSM concentrations, with a maximum value of 1640 ng L⁻¹, were detected in the downstream reservoirs (Cheongpyeong and Paldang); the level was 11 times higher than the value (10 ng L⁻¹) recommended by the World Health Organization. The concentrations of GSM gradually decreased and had an adverse effect on the drinking water quality until the end of December 2011. Our study indicated that the time lag between the summer–fall cyanobacterial outbreak in the upstream reservoir and winter GSM explosion events in the downstream reservoirs could be attributed to the transport and release of GSM through the effluent from hydroelectric power generation in this multiple-reservoir system. Therefore, we suggest that a structural understanding of the reservoir cascade be considered during water quality management of drinking water sources to avoid such incidents in the future.

In the recent era of AI, instance segmentation has significantly advanced boundary and object detection especially in diverse fields (e.g., biological and environmental research). Despite its progress, edge detection amid adjacent objects (e.g., organism cells) still remains intractable. This is because homogeneous and heterogeneous objects are prone to being mingled in a single image. To cope with this challenge, we propose the weighted Mask R‐CNN designed to effectively separate overlapped objects in virtue of extra weights to adjacent boundaries. For numerical study, a range of experiments are performed with applications to simulated data and real data (e.g., Microcystis, one of the most common algae genera and cell membrane images). It is noticeable that the weighted Mask R‐CNN outperforms the standard Mask R‐CNN, given that the analytic experiments show on average 92.5% of precision and 96.4% of recall in algae data and 94.5% of precision and 98.6% of recall in cell membrane data. Consequently, we found that a majority of sample boundaries in real and simulated data are precisely segmented in the midst of object mixtures.