Original ArticleIdentification of influencing factors of A. catenella bloom using machine learning and numerical simulation
Introduction
Harmful algal blooms (HABs) tend to harm human health and the economy in affected areas (Anderson et al., 2012). As the geographical scale, frequency, and period of HABs have increased globally in recent decades, they have become a serious threat to aquatic ecosystems (Gobler et al., 2017; Hoagland et al., 2002; Weiher and Sen, 2006). In particular, Asian countries have frequently experienced HABs with respect to Alexandrium species (Han et al., 2016). Blooms of A. catenella, which is a dinoflagellate species known to cause paralytic shellfish poisoning (PSP), has been observed regularly in spring (Han et al., 1992; Ishikawa et al., 2014; Kim et al., 2002). In particular, the southeastern coastal areas of Korea have been reported as hotspots of toxic species (Han et al., 1992; Ishikawa et al., 2014; Kim et al., 2002; Shin et al., 2017).
Although numerous studies have been conducted to understand the mechanism and control of Alexandrium blooms, bloom dynamics have not been fully understood due to the complexities of physical, chemical, and biological environments (Anderson et al., 2012). McGillicuddy Jr (2010) demonstrated that the physical, chemical, and biological factors heterogeneously triggered the initiation, development, and ultimate demise of HABs. In particular, HABs are often associated with weather conditions (Brandenburg et al., 2017). Changes in hydrodynamics have been reported to cause rapid and massive development of HABs (Pettersson and Pozdnyakov, 2013). In addition, there also existed the interactive relationship between the environmental variables such as the water velocity affects the dispersion rate of nutrients and temperature stimulates the growth of cyst (Fischer, 1976; Pospelova et al., 2008). The typhoon and anthropogenic source can also change the trend of bloom (Ding et al., 2012; Gobler et al., 2017; Heisler et al., 2008; Zhu et al., 2014). Therefore, these complex process of HAB might impede management of it.
Given the limited resources, the modeling and machine learning approach could be an alternative to understanding and controlling harmful algae (Pyo et al., 2019; Ruardij et al., 2005; Yoshioka and Yaegashi, 2018). The numerical model of the ocean, along with its physicochemical background, has been used to simulate salinity, water temperature, and water elevation (Murtugudde and Busalacchi, 1998; Thompson et al., 2006). Salinity, water temperature and water elevation, and nutrients have been found to influence the outbreak of algal blooms (Armi et al., 2011; Bearon et al., 2006; Ralston et al., 2014). In particular, the numerical model can be useful to identify the temporal-spatial distribution of algae (Ralston et al., 2014). Several studies have been conducted to simulate HABs in the ocean. Pinto et al. (2016) used a particle-tracking model to predict abundance of HAB species. He et al. (2008) developed a numerical model to predict A. catenella (former A. fundynes) bloom in the western Gulf of Maine. Previous modeling studies have focused on the quantification of algae. In those studies, the numerical models were useful for generalizing the overall trends of algae populations. However, those models are limited in their ability to explain specific situations such as unexpected outbreaks and extinction of blooms due to the complex blooming mechanisms.
The limitations of the numerical models can be addressed by using a data-driven model, namely machine learning. Machine learning is capable of handling big data and can identify the most important and consequential features from a wide variety of variables (e.g., categorical and continuous variables) (Szegedy et al., 2015; Young et al., 2018). Machine learning (Chen and Manning, 2014; Litjens et al., 2017; Veselý et al., 2013; Young et al., 2018) has been applied in numerous fields such as image recognition, speech analysis, and biology. Among machine learning models, the DT model is suitable for recognizing influencing factors from the data because it can interpret a trained model by examining the tree of the model (Safavian and Landgrebe, 1991; Song and Ying, 2015). Kamikawaji et al. (2016) detected errors in water temperature and salinity observations using the DT model. Liu et al. (2015) estimated sea surface salinity in coastal waters using random forest. Although the model using machine learning showed the acceptable performance, most studies mainly focus on the simulations of water temperature, salinity, and water depth. In addition, few studies have applied machine learning to HABs.
Herein, we identify the factors influencing the bloom of Alexandrium species using a physical and machine learning model. A. catenella was selected as the target species; it is known to produce PSP and it frequently occurs during spring in South Korea (Ki et al., 2004; KI and HAN, 2007). The numerical model generated the temporal-spatial distribution of the ocean's environmental factors in Geoje Island, South Korea, which had frequently experienced outbreaks of A. catenella bloom. Machine learning estimated the extent of the bloom and analyzed the factors affecting it. The aims of this study were 1) to conduct temporal and spatial monitoring of A. catenella, 2) analyze the physical properties of the ocean using a numerical model, 3) estimate the extent of the A. catenella bloom using the DT model, and 4) identify the factors influencing the bloom of A. catenella by analyzing the tree model.
Section snippets
Study site
Our study implemented temporal and spatial monitoring to investigate the occurrence of A. catenella. The study area is located on the eastern coastal waters of Geoje Island in South Korea (Fig. 1). A spatial survey was conducted in the entire region near Geoje Island. Horizontal distributions of A. catenella were surveyed at 34 stations in March 2017 and 34 stations in April 2017; moreover, 24 stations were surveyed in March 2018 (Fig. 3). Temporal sampling was conducted intensively at a fixed
Intensive monitoring of A. catenella
An extreme bloom of A. catenella was recorded in late March of 1989 and in early May of 1997 in the Chinhae Bay of the southeastern coastal waters in South Korea, having a density greater than 10,000 cells L−1 (Lee, 2005). In this study, the maximum density in 2017 was 44,012 cells L−1 and the temporal variation of A. catenella in 2017 showed a similar seasonality of species occurrence, indicating a spring outbreak of A. catenella bloom in mid-April (Fig. 4a). However, there was absence of this
Conclusions
This study investigated the factors influencing the bloom of A. catenella. Intensive and spatial monitoring was conducted to investigate the temporal variation and spatial distribution of the species, respectively. Additionally, a numerical model was built to analyze the physical factors of the ocean that contributed to the bloom of A. catenella. Using the monitoring and simulation results, this study identified the hierarchical structure of the bloom. The major findings of this study are as
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgement
This research was supported by the Basic Core Technology Development Program for the Oceans and the Polar Regions of the National Research Foundation (NRF) funded by the Ministry of Science, ICT & Future Planning (NRF-2016M1A5A1027457).
References (110)
- et al.
Detecting the red tide algal blooms from satellite ocean color observations in optically complex Northeast-Asia Coastal waters
Remote Sens Environ
(2006) - et al.
Alexandrium fundyense cyst dynamics in the Gulf of Maine
Deep Sea Res. Part II: Topical Stud. Oceanography
(2005) - et al.
Insights into saline intrusion and freshwater resources in coastal karstic aquifers using a lumped Rainfall–Discharge–Salinity model (the Port-Miou brackish spring, SE France)
J Hydrol (Amst)
(2016) - et al.
Combined physical, chemical and biological factors shape Alexandrium ostenfeldii blooms in the Netherlands
Harmful Algae
(2017) - et al.
A neural-network-based classification scheme for sorting sources and ages of fecal contamination in water
Water Res.
(2002) - et al.
Water residence time in Chesapeake Bay for 1980–2012
J. Marine Syst.
(2016) - et al.
Comparison between neural networks and multiple logistic regression to predict acute coronary syndrome in the emergency room
Artif Intell Med
(2006) - et al.
Paralytic shellfish toxin production by the dinoflagellate Alexandrium pacificum (Chinhae Bay, Korea) in axenic, nutrient-limited chemostat cultures and nutrient-enriched batch cultures
Mar. Pollut. Bull.
(2016) - et al.
Eutrophication and harmful algal blooms: a scientific consensus
Harmful Algae
(2008) - et al.
Exploring currents and heat within the North Sea using a numerical model
Journal of Marine Systems
(2009)
The growth and cyst formation of a toxic dinoflagellate, Alexandrium tamarense, at low water temperatures in northeastern Japan
J. Exp. Mar. Biol. Ecol.
Salinity intrusion characteristics analysis using EFDC model in the downstream of Geum River
J. Environ. Sci.
Relationship between phytoplankton bloom and wind stress in the sub-polar frontal area of the Japan/East Sea
J. Marine Syst.
Tracking Alexandrium catenella from seed-bed to bloom on the southern coast of Korea
Harmful Algae
Application of machine learning methods to spatial interpolation of environmental variables
Environ. Model Software
A survey on deep learning in medical image analysis
Med Image Anal
Multivariable integration method for estimating sea surface salinity in coastal waters from in situ data and remotely sensed data using random forest algorithm
Comput Geosci
Application of an unstructured 3D finite volume numerical model to flows and salinity dynamics in the San Francisco Bay-Delta
Estuar Coast Shelf Sci
Models of harmful algal blooms: conceptual, empirical, and numerical approaches
J Marine Syst.
Monitoring of the toxic dinoflagellate Alexandrium catenella in Osaka Bay, Japan using a massively parallel sequencing (MPS)-based technique
Harmful Algae
Modeling the transport pathways of harmful algal blooms in the Iberian coast
Harmful Algae
Distribution of dinoflagellate cysts in surface sediments from the northeastern Pacific Ocean (43–25 N) in relation to sea-surface temperature, salinity, productivity and coastal upwelling
Mar. Micropaleontol.
Organic-walled dinoflagellate cyst production, composition and flux from 1996 to 1998 in the central Strait of Georgia (BC, Canada): a sediment trap study
Mar. Micropaleontol.
A convolutional neural network regression for quantifying cyanobacteria using hyperspectral imagery
Remote Sens Environ
Applications of machine learning to ecological modelling
Ecol Modell
Modeling the bloom dynamics of the polymorphic phytoplankter Phaeocystis globosa: impact of grazers and viruses
Harmful Algae
ANN constitutive model for high strain-rate deformation of Al 7075-T6
J. Materials Process Technol.
Which species, Alexandrium catenella (Group I) or A. pacificum (Group IV), is really responsible for past paralytic shellfish poisoning outbreaks in Jinhae-Masan Bay, Korea?
Harmful Algae
Complexity in the eutrophication–harmful algal bloom relationship, with comment on the importance of grazing
Harmful Algae
Modeling complex nonlinear responses of shallow lakes to fish and hydrology using artificial neural networks
Ecol Modell
In situ and satellite observations of a harmful algal bloom and water condition at the Pearl River estuary in late autumn 1998
Harmful Algae
Health effects of chronic pesticide exposure: cancer and neurotoxicity
Annu Rev Publ Health
Permutation importance: a corrected feature importance measure
Bioinformatics
Progress in understanding harmful algal blooms: paradigm shifts and new technologies for research, monitoring, and management
Ann Rev Mar Sci
Alexandrium catenella and Alexandrium tamarense in the North Lake of Tunis: bloom characteristics and the occurrence of paralytic shellfish toxin
Afr. J. Aquat. Sci.
Effects of salinity structure on swimming behavior and harmful algal bloom formation in Heterosigma akashiwo, a toxic raphidophyte
Mar. Ecol. Prog. Ser.
Physiological responses of a Southern Ocean diatom to complex future ocean conditions
Nat Clim Chang
An empirical comparison of supervised learning algorithms
A study on paralytic shellfish poison of sea Mussel, Mytilus edulis-food poisoning accident in Gamchun Bay, Pusan, Korea, 1986
Korean J. Fisheries Aquat. Sci.
Log-transformation and its implications for data analysis
Shanghai Arch Psychiatry
An unstructured grid, finite-volume, three-dimensional, primitive equations ocean model: application to coastal ocean and estuaries
J. Atmospheric and Oceanic Technol.
A fast and accurate dependency parser using neural networks
Unusual salinity conditions in the Yangtze Estuary in 2006: impacts of an extreme drought or of the Three Gorges Dam?
Ambio
Fundamentals of Estuarine Physical Oceanography
Effects of typhoon Morakot on a large shallow lake ecosystem
Lake Taihu, China. Ecohydrology
Comparative study of selenium requirements of three phytoplankton species: gymnodinium catenatum, Alexandrium minutum (Dinophyta) and Chaetoceros cf. tenuissimus (Bacillariophyta
J. Plankton Res.
Asymptotic equivalence between cross-validations and Akaike information criteria in mixed-effects models
J. Data Sci.
Chesapeake Bay nitrogen fluxes derived from a land-estuarine ocean biogeochemical modeling system: model description, evaluation, and nitrogen budgets
J. Geophys. Res. Biogeosci.
Mixing and dispersion in estuaries
Annu Rev Fluid Mech
Estimation of prediction error by using K-fold cross-validation
Stat Comput
Cited by (8)
Predicting coastal harmful algal blooms using integrated data-driven analysis of environmental factors
2024, Science of the Total EnvironmentFormation mechanism and environmental drivers of Alexandrium catenella bloom events in the coastal waters of Qinhuangdao, China
2022, Environmental PollutionCitation Excerpt :In the present study, the temperature increased from 4 to 11 °C between the start and end of the bloom event, which is relatively lower than the temperature ranges of 7–15 °C (Condie et al., 2019) and 10–13 °C (Álvarez et al., 2019) reported in previous studies. Also, this increase is significantly lower than the ideal temperature of 17 °C estimated for the growth of A. catenella via machine learning and numerical simulations (Baek et al., 2021). This indicates the uniqueness of the algal strain in the region.
Comprehensive Insights into Harmful Algal Blooms: A Review of Chemical, Physical, Biological, and Climatological Influencers with Predictive Modeling Approaches
2024, Journal of Environmental Engineering (United States)First insight into H3K4me3 modification in the rapid growth of Alexandrium pacificum (dinoflagellates)
2022, Frontiers in Marine Science