Page 21 - A U4SSC deliverable - Accelerating city transformation using frontier technologies
P. 21
In focus – AI for water quality monitoring, Shandong province
Despite nearly 70 per cent of the Earth’s surface being covered with water, only three per cent of its fresh
water is available for agriculture and human consumption. Consequently, the recycling and recovery of
water from wastewater is essential for optimising its utilisation for drinking, industry, renewable-energy
generation and agriculture. Rivers are an important source of fresh water for irrigation, especially in
agrarian communities. However, the quality of rivers could be threatened by point sources and non-
point sources through which pollutants enter the water cycle. Therefore, prior to using river water for
irrigation, it is important to identify the characteristics of water pollutants and trace their origins . To
reduce the burden on freshwater resources, wastewater (in other words, reclaimed water) can also be
used effectively for irrigation. However, reclaimed water irrigation also holds potential pollution risks
with the added peril of pollutants percolating into groundwater water reserves.
In a study conducted by a group of universities, including: Beijing Normal University (China), Asian
Institute of Technology (Thailand), and Shandong Academy of Environmental Planning (China), an AI-
based system referred to as the integrated long short-term memory network (LSTM), that leverages
Apriori cross-correlation and association rules, was implemented in the Shandong Province to identify
the characteristics of water pollutants and trace industrial point sources. For the purpose of this project,
existing water quality monitoring data from Shandong Province were used to develop a water quality
cross-correlation map. This map formed the basis for identifying highly correlated pollutants (affecting
water quality), which were then tracked using Apriori to determine the pollutant, its source, as well
as common sites of contamination. Moving beyond binary classification, in order to understand the
complex relationships between point sources of pollutants and water quality, monitoring indicators
were identified by the neural network using the LSTM algorithm. Throughout the project, the LSTM
algorithm provided high-prediction accuracy and was also able to trace potential industrial point
sources of pollutants that would affect water quality in the future.
Accordingly, LSTM was considered to be a reliable system to detect changes in water quality and
identify abnormal fluctuations in point source pollutants. Based on the findings of this project, the
application of LSTM can be utilised for the effective, large-scale AI-based monitoring of water resources
to uncover toxic materials present in irrigation water (obtained from fresh water sources, as well as from
reclaimed water). These features, coupled with the potential of LSTM to predict other possible sources
of pollution, would be very pertinent in tackling the challenges faced by the agricultural industry.
Therefore, this system could enable environmental concerns relating to the contamination of irrigation
sources to be overcome, which would, in turn, support the sustainable yield of crops and the protection
of the livelihoods of individuals engaged in this sector. 43
Accelerating city transformation using frontier technologies | A U4SSC deliverable 11