Using AI to Monitor Groundwater for Contaminants

DZone 's Guide to

Using AI to Monitor Groundwater for Contaminants

Read on to get an insight about AI-Based system that provides real-time monitoring of pollutants using low-cost sensors.

· AI Zone ·
Free Resource

When groundwater becomes contaminated, it can have a number of extremely serious implications, both for humans and the wider environment. To avoid contamination from happening usually requires long-term monitoring, which can be costly to undertake.

A team from Berkeley Labs has developed an AI-based system that provides real-time monitoring of pollutants using low-cost sensors. Their work, which was documented in a recently published paper, promises to allow officials to monitor large, complex, and long-term events and their impact on contaminant levels.

"Conventional methods of monitoring involve taking water samples every year or every quarter and analyzing them in the lab," the team says. "If there are anomalies or an extreme event, you could miss the changes that might increase contaminant concentrations or potential health risk. Our methodology allows continuous monitoring in situ using proxy measurements, so we can track plume movement in real time."

Automating Analysis

Instead, the team proposed to utilize Machine Learning methods to analyze the data autonomously. In this way, they believe they can achieve an earlier warning of any sudden changes in contaminant levels and intervene accordingly.

There has been a gradual move away from intensive methods of cleanup due to the high environmental cost of such approaches. This means that monitoring has to be performed over a longer timeframe, but this is in itself costly. What's more, existing methods of long-term monitoring often fail to take into account the abruptness of change in weather and how this might influence plume behavior. The new method uses sensors to track the various qualities of the water that have been previously identified as reliable indicators of contaminant levels. The data collected by the sensors was fed into a mathematical algorithm called a Kalman filter to enable them to estimate the concentration of the contaminant.

Put to the Test

20 years worth of historical data from a test site was used to test the model, and it proved reliable in predicting plume behavior over a long timeframe. The team is confident that these early results showcase the promise of their approach. Not only does it reduce the cost involved in monitoring, but they also believe it can be achieved with less frequent sampling and analysis, which could reduce costs even further.

The team is confident that the system can be used for both surface and underground water, whilst also proving capable of tracking a wide range of metals, radionuclides and other organic compounds commonly found in groundwater.

"There are so many different types of sensors available now, and sensor networking and rapid statistical analysis is straightforward," the team says. "We can put together all types of in-situ sensors and estimate the target contaminant concentration using this framework for data integration in real-time."

They continue: "Improved monitoring techniques are essential to protect public health and the ecology. People feel safe if it's properly monitored. Our technique is a way to monitor such sustainable remediation-effectively and cheaply."

ai, environmental monitoring, machine learning, monitoring, real-time monitoring

Published at DZone with permission of Adi Gaskell . See the original article here.

Opinions expressed by DZone contributors are their own.

{{ parent.title || parent.header.title}}

{{ parent.tldr }}

{{ parent.urlSource.name }}