Hybrid CNN–LSTM Framework for Detection of Heavy Metal Contamination in Agricultural Water

Contamination of agricultural water by heavy metals has become a major environmental and public health concern due to rapid industrialization, excessive fertilizer usage, and improper wastewater disposal practices. Continuous use of contaminated irrigation water negatively affects soil fertility, crop productivity, and food safety, leading to severe ecological and human health risks. Traditional laboratory-based water quality assessment methods are often time-consuming, expensive, and unsuitable for real-time monitoring applications. To address these limitations, this paper proposes a hybrid Convolutional Neural Network–Long Short-Term Memory (CNN–LSTM) framework for intelligent detection and prediction of heavy metal contamination in agricultural water systems.

The proposed model integrates the feature extraction capability of CNN with the temporal learning strength of LSTM to analyse multidimensional water quality parameters such as pH, turbidity, electrical conductivity, dissolved oxygen, temperature, and concentrations of heavy metals including lead (Pb), cadmium (Cd), arsenic (As), and mercury (Hg). Sensor-based and historical water quality datasets are utilized for training and validation of the framework. Data pre-processing techniques including normalization, missing value handling, and noise reduction are applied to improve e model efficiency and prediction accuracy.

Experimental results demonstrate that the hybrid CNN–LSTM model achieves superior performance compared with conventional machine learning techniques in terms of accuracy, precision, recall, and prediction stability. The proposed system enables early detection of contamination patterns and supports real-time agricultural water quality monitoring for sustainable irrigation management. Furthermore, the framework can be integrated with IoT-enabled smart farming systems for automated environmental assessment and decision support.