2015; Sousa et al. 2014). Nitrate contamination has been detected in nearly 60 countries (Fig. 4.3).
Newly born babies (those of less than 6 months of age) are the primary victims of
nitrate pollution. They may suffer from methemoglobinemia, popularly known as
“blue baby syndrome,” if they consume water with a nitrate concentration greater
than 10 mg/L. The adverse human health impacts of nitrate first surfaced in 1945
(Zhou 2015) and are summarized in Table 4.3.
4.3 Application of Artificial Intelligence in Groundwater
Contaminant Prediction
Artificial intelligence (AI), a subset of machine learning (ML), has been used to
identify and resolve an array of challenges across various domains, including
information technology, business, banking, retail, pharmacy, healthcare, insurance,
and life sciences, among others. However, the application of AI to the identification,
prediction, and monitoring of environmental problems is a recent phenomenon.
Environmental research is challenged by a lack of ground data, i.e., collected from
fields through surveys and laboratory analysis; therefore, it depends heavily on
remote sensing images. Geographical information system (GIS) tools are used to
extract information from these remote sensing images, simulate the imaged areas,
and develop prediction models. Machine learning, deep learning, and reinforcement
learning (RL) help simulate human knowledge of a specific domain and assist in
developing intelligent systems that can learn from historical data and predict patterns
in unseen data (Chau 2006). Various basic and advanced AI techniques have been
used to develop prediction models of landslide susceptibility and groundwater
potentiality (Pham et al. 2019; Pham et al. 2018; Phong et al. 2019; Rizeei et al.
2019; Chen et al. 2019; Sameen et al. 2019, 2020), for example. The application of
AI techniques in predicting groundwater contaminants, in contrast, is still at a
primitive level. The handful of studies that are available on the application of
these techniques in predicting groundwater contamination by arsenic, fluoride, and
nitrate are summarized below in Tables 4.4, 4.5, and 4.6, respectively.
4.4 Application of AI in Predicting Arsenic Contamination
of Groundwater
Data-analysis techniques for predicting groundwater contamination of arsenic were
first reported in 2006, when researchers developed a probabilistic statistical model to
predict the probability that arsenic concentrations would exceed the level of 5 μg/L
in the drinking-water wells of the New England of the USA (Ayotte et al. 2006). A
logistic regression model of geologic and anthropogenic sources of arsenic,
82
S. K. Singh et al.
Newly born babies (those of less than 6 months of age) are the primary victims of
nitrate pollution. They may suffer from methemoglobinemia, popularly known as
“blue baby syndrome,” if they consume water with a nitrate concentration greater
than 10 mg/L. The adverse human health impacts of nitrate first surfaced in 1945
(Zhou 2015) and are summarized in Table 4.3.
4.3 Application of Artificial Intelligence in Groundwater
Contaminant Prediction
Artificial intelligence (AI), a subset of machine learning (ML), has been used to
identify and resolve an array of challenges across various domains, including
information technology, business, banking, retail, pharmacy, healthcare, insurance,
and life sciences, among others. However, the application of AI to the identification,
prediction, and monitoring of environmental problems is a recent phenomenon.
Environmental research is challenged by a lack of ground data, i.e., collected from
fields through surveys and laboratory analysis; therefore, it depends heavily on
remote sensing images. Geographical information system (GIS) tools are used to
extract information from these remote sensing images, simulate the imaged areas,
and develop prediction models. Machine learning, deep learning, and reinforcement
learning (RL) help simulate human knowledge of a specific domain and assist in
developing intelligent systems that can learn from historical data and predict patterns
in unseen data (Chau 2006). Various basic and advanced AI techniques have been
used to develop prediction models of landslide susceptibility and groundwater
potentiality (Pham et al. 2019; Pham et al. 2018; Phong et al. 2019; Rizeei et al.
2019; Chen et al. 2019; Sameen et al. 2019, 2020), for example. The application of
AI techniques in predicting groundwater contaminants, in contrast, is still at a
primitive level. The handful of studies that are available on the application of
these techniques in predicting groundwater contamination by arsenic, fluoride, and
nitrate are summarized below in Tables 4.4, 4.5, and 4.6, respectively.
4.4 Application of AI in Predicting Arsenic Contamination
of Groundwater
Data-analysis techniques for predicting groundwater contamination of arsenic were
first reported in 2006, when researchers developed a probabilistic statistical model to
predict the probability that arsenic concentrations would exceed the level of 5 μg/L
in the drinking-water wells of the New England of the USA (Ayotte et al. 2006). A
logistic regression model of geologic and anthropogenic sources of arsenic,
82
S. K. Singh et al.