Estimation of cyanobacteria pigments in the main rivers of South Korea using spatial attention convolutional neural network with hyperspectral imagery

Authors: SEOKMIN, HONG - Ulsan National Institute Of Science And Technology (UNIST); CHO, KYUNGHWA - Ulsan National Institute Of Science And Technology (UNIST); PARK, SANGHYUN - Environmental Research Complex; KANF, TAEGU - Environmental Research Complex; Kim, Moon; NAM, GIBEOM - Environmental Research Complex; PYO, JONGCHWOL - Ulsan National Institute Of Science And Technology (UNIST)

Submitted to: GIScience and Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 1/28/2022
Publication Date: 2/1/2022
Citation: Seokmin, H., Cho, K., Park, S., Kanf, T., Kim, M.S., Nam, G., Pyo, J. 2022. Estimation of cyanobacteria pigments in the main rivers of South Korea using spatial attention convolutional neural network with hyperspectral imagery. GIScience and Remote Sensing. 59(1):547-567. https://doi.org/10.1080/15481603.2022.2037887.
DOI: https://doi.org/10.1080/15481603.2022.2037887

Interpretive Summary: Harmful algal blooms (HABs) from toxic cyanobacteria in fresh water rivers and streams can cause water quality problems that directly damage aquatic ecosystems and indirectly damage, through agricultural and domestic water use, the quality and safety of plant crops as well as the health of animals and humans. Traditional in-situ water quality monitoring and analysis methods are limited in their speed and quantification of spatial water quality identification and distribution. Remote spectral sensing techniques via aircraft and satellite have been used effectively in nearly real-time image-based monitoring for chlorophyll-a (Chl-a) and phycocyanin (PC) concentrations in the water to evaluate HABs, but the conventional analysis algorithms typically used with remote sensing must be optimized for each specific body of water. This study developed a general-use algorithm based on hyperspectral visible/near-infrared imaging with a convolutional neural network (CNN) method to analyze different water bodies for Chl-a and PC concentrations, and compared the results with those from conventional analysis algorithms, for spatially and temporally assessing HABs. Using study sites (reservoirs, weirs, or dike areas) on three major rivers in South Korea, this research acquired hyperspectral images of the sites via drone and aircraft, and demonstrated that the CNN-based general algorithm could perform better in estimating Chl-a and PC concentrations compared to the conventional analysis algorithms. The results show that with the collection of additional data, the CNN-based model has great potential for implementation via drone, aircraft, and satellite imagery to monitor water quality across various bodies of water, for the benefit of agricultural producers as well as industrial or metropolitan water users.

Technical Abstract: Although remote sensing techniques have been used to monitor toxic cyanobacteria with hyperspectral data in inland water, it is difficult to optimize conventional bio-optical algorithms for individual water bodies because of the complex optical properties of various water components. Therefore, this study adopted a spatial attention convolutional neural network (spatial attention CNN) to estimate the chlorophyll-a (Chl-a) and phycocyanin (PC) concentrations in the Geum, Nakdong, and Yeongsan rivers in South Korea in order to evaluate cyanobacteria using remote sensing reflectance data. The CNN model utilized a spatial attention module to analyze the importance of the bands in the reflectance data. Then, the spatial attention CNN model was compared with different bio-optical algorithms for each study area. The spatial attention CNN model was generalized to estimate the pigment concentrations in the target rivers, and the model performance was evaluated by correlation coefficient (R) and root mean squared error (RMSE) between the observed and estimated concentrations of the algal pigments. The spatial attention CNN model, which was generalized to estimate the pigment concentrations in the target rivers, had R values above 0.87 and 0.88 for Chl-a and PC, respectively. However, the optimized band ratio algorithms for Chl-a and PC had R values above 0.83 and 0.70, respectively. Hence, it showed better performance than the conventional bio-optical algorithms. The spatial attention module provided attention weights for visualizing important features in the reflectance data. Specifically, the 600 nm, 650 nm, and near-infrared regions had high attention weights for estimating the concentrations of Chl-a and PC. Based on these findings, this study demonstrated that the spatial attention CNN model has a high potential for good application performance in various water bodies.