Radio frequency (RF) signals are a cornerstone of modern technology, enabling wireless communication, data transfer and sensing applications across a wide range of fields and tasks.
A Cornell Tech-led research group is in the early stages of developing a portable, inexpensive device that uses radio frequency signals and machine learning for another important job: measuring lead contamination levels in soil.
The lab of Rajalakshmi Nandakumar, assistant professor at the Jacobs Technion-Cornell Institute at Cornell Tech, has developed SoilScanner, which sends radio waves of different frequencies from an RF transmitter, through a soil sample to an RF receiver, which reveals the effect the soil - and how much lead is in it - has on the signal.
Nandakumar is senior author and Yixuan Gao, a doctoral candidate in computer science, is lead author of "Feasibility of Radio Frequency Based Wireless Sensing of Lead Contamination in Soil," which won a best-paper award at the International Conference on Embedded Wireless Systems and Networks (EWSN '24), held Dec. 10-13 in Abu Dhabi, United Arab Emirates.
Other contributors were Tanvir Ahmed, a doctoral student in information science and a member of Nandakumar's lab; Zhongqi (Joshua) Cheng, professor in the Department of Earth and Environmental Sciences at Brooklyn College of the City University of New York; and Mikhail Mohammed, a 2023 Brooklyn College graduate who now works for the Environmental Protection Agency (EPA).
"In recent years, especially during COVID, a lot of us got excited about having our own backyard garden, or spending more time at home," said Nandakumar, who's also a member of the Department of Information Science in the Cornell Ann S. Bowers College of Computing and Information Science.
"But if you look at instructions for how to grow tomatoes, no one actually tells you that you have to check your soil for lead," she said. "It's all about pH levels. A lot of us, even though we interact very often with soils, are totally unaware of possible lead contamination."
Gao said the group was motivated by a map of lead contamination in New York City that Cheng's Urban Soils Lab (USL) had produced over the course of several years of testing for hundreds of soil samples throughout the five boroughs. The testing revealed dangerously high levels of lead in many locations, most notably in northern Brooklyn.
About 45% of the soil samples tested by USL had lead levels above 400 parts per million (ppm), the previous EPA recommended screening level (revised a year ago to 200 ppm for residential soils). "This means there is a significant risk when gardening in these urban soils," Gao said.
Testing for lead in soil generally involves either sending samples to a lab for analysis, which relies upon harsh chemicals and can be expensive, or using a portable X-ray fluorescence device, which is cost-prohibitive for many communities.
The group's device is simple: Dirt is placed in a one-liter plastic container between the transmitter and receiving antenna. The transmitter sends short bursts of single-tone RF signals at low and high frequencies (700 to 1,000 megahertz and 2.3 to 2.5 gigahertz) through the soil sample.
The receiver's power spectrum reading is passed to a machine-learning model for further analysis.
SoilScanner was calibrated and tested on two different sets of samples: 23 lab-prepared samples of dirt, spiked with lead; and 22 field samples of varying compositions. It was able to detect lead contamination in natural field samples with 72% overall accuracy, and became even more accurate as the ppm value rose. It had a zero-error rate when lead levels were greater than 500 ppm.
Nandakumar's lab will continue to work with the USL, which Cheng founded, to use SoilScanner on samples sent in from across the country, Gao said. The team is also working on an even smaller, less costly, battery-operated version of the device, using Wi-Fi and RFID chipsets such as those found in smartphones and computers.
"The goal of this project," Nandakumar said, "is to make this technology cheaper and more accessible - both for people to test their own soils and for governments to plan for remediation."
This research was supported by the Cornell Institute for Digital Agriculture.
