Mammoths: A Vital Nutrient for Early American Societies

Scientists have uncovered the first direct evidence that ancient Americans relied primarily on mammoth and other large animals for food. Their research sheds new light on both the rapid expansion of humans throughout the Americas and the extinction of large ice age mammals. Scientists have uncovered the first direct evidence that ancient Americans relied primarily
HomeTechnologyHarnessing Machine Learning to Identify High-Risk Groundwater Locations for Enhanced Water Quality...

Harnessing Machine Learning to Identify High-Risk Groundwater Locations for Enhanced Water Quality Monitoring

An interdisciplinary group of researchers has created a machine learning framework designed to predict the presence of inorganic pollutants in groundwater using only a limited number of water quality samples. This innovative tool now enables regulators and public health officials to focus their water quality testing efforts on specific aquifers.

An interdisciplinary group of researchers has created a machine learning framework designed to predict the presence of inorganic pollutants in groundwater using only a limited number of water quality samples. This innovative tool now enables regulators and public health officials to focus their water quality testing efforts on specific aquifers.

This initial research was conducted in Arizona and North Carolina but could be useful in addressing significant groundwater quality gaps in other areas as well.

Groundwater serves as a vital drinking water source for millions, but it often contains pollutants that can pose health risks. Nevertheless, many areas do not have complete datasets on groundwater quality.

“Monitoring water quality can be both time-consuming and costly. The more pollutants you test for, the higher the costs and time commitment,” stated Yaroslava Yingling, co-corresponding author of the study and Kobe Steel Distinguished Professor of Materials Science and Engineering at North Carolina State University.

“As a result, there is a growing interest in determining which groundwater sources should be prioritized for testing, allowing for better use of limited monitoring resources,” Yingling explained. “We understand that naturally occurring pollutants like arsenic or lead usually occur alongside specific other elements due to geological and environmental contexts. This raises an important data-driven question: With only limited water quality data available for a groundwater source, can we forecast the presence and concentrations of other pollutants?”

“Besides identifying potentially hazardous elements, we also aimed to see if we could predict the presence of other elements like phosphorus, which can be beneficial for agriculture but may create environmental challenges elsewhere,” noted Alexey Gulyuk, a co-first author and teaching professor of materials science and engineering at NC State.

To tackle this challenge, the research team utilized an extensive dataset that includes over 140 years of groundwater quality monitoring data from North Carolina and Arizona. This dataset consisted of more than 20 million data entries encompassing over 50 parameters related to water quality.

“We trained a machine learning model using this dataset to predict which pollutants would be present based on the available water quality information,” explained Akhlak Ul Mahmood, a co-first author and a former Ph.D. student at NC State. “So, even if we only have information on a few parameters, the system can still estimate which inorganic pollutants might be present and how concentrated they are likely to be.”

A significant discovery from the study is that the model indicates pollutants may be exceeding drinking water safety standards in more groundwater sources than previously recognized. While field data suggested that 75-80% of sampled locations were within safe levels, the machine learning framework indicates that only 15% to 55% of these sites may indeed be free of risk.

“Consequently, we’ve identified numerous groundwater locations that should be prioritized for further testing,” said Minhazul Islam, a co-first author and a Ph.D. student at Arizona State University. “By pinpointing potential ‘hot spots,’ state agencies and municipalities can allocate resources more effectively to high-risk areas, ensuring focused sampling and efficient water treatment solutions.”

“It’s very promising, and we believe it works effectively,” Gulyuk remarked. “However, the true test lies in applying this model in real-world situations and verifying if the prediction accuracy remains reliable.”

Looking ahead, the researchers plan to refine the model by incorporating training data from various regions across the U.S.; including new data sources such as environmental data layers to address emerging contaminants; and conducting real-world experiments to ensure effective groundwater safety measures on a global scale.

“We see extraordinary potential in this approach,” expressed Paul Westerhoff, co-corresponding author and Regents’ Professor in the School of Sustainable Engineering and the Built Environment at ASU. “By continually enhancing its accuracy and broadening its application, we are establishing a foundation for proactive water safety measures globally.”

“This model also serves as a valuable tool for monitoring phosphorus levels in groundwater, assisting us in identifying and mitigating potential contamination risks more effectively,” stated Jacob Jones, director of the National Science Foundation-funded Science and Technologies for Phosphorus Sustainability (STEPS) Center at NC State, which supported this study. “In the future, adapting this model to support wider phosphorus sustainability efforts could yield substantial benefits, aiding us in managing this essential nutrient across different ecosystems and agricultural systems and promoting more sustainable practices.”

This research was funded by the NSF STEPS Center and the Metals and Metal Mixtures: Cognitive Aging, Remediation and Exposure Sources (MEMCARE) Superfund Research Center at Harvard University, which receives support from the National Institute of Environmental Health Sciences under grant P42ES030990.