REAL-TIME SOIL CONTAMINANT MONITORING USING MICROBIAL BIOSENSORS AND SUPPORT VECTOR MACHINE

Abstract

Continuous‍‌ real-time monitoring of soil contamination is imperative for environmental safeguarding. Traditional laboratory-based analyses, however, are slow, expensive, and not viable for continuous on-site field work. In this work, we propose a hybrid soil monitoring strategy that combines microbial biosensors with a Support Vector Machine (SVM) classifier for quick detection of contaminants. Specifically modified Escherichia coli bacteria are engineered to react specifically to lead (Pb²⁺), mercury (Hg²⁺), and organophosphate compounds. They do this by generating electrical bio-signals with voltages ranging from 0.2 V to 1.5 V at contaminant concentrations above 50 ppm. The bio-signals are converted to digital form using a 12-bit ADC at a 1 kHz sampling rate, and then filtered, normalized, and statistical features extracted. A dataset containing 3,200 measurement points gathered over three months is used, of which 1,200 are used for both training and validation of the SVM. Using a radial basis function kernel, the system under discussion achieves 96.8% classification accuracy and 96.4% F1-score across the five soil condition categories. Sensor drift is compensated for through online model updating; thus, the system is reliable and capable of continuous soil contamination monitoring in a field environment.