|PYO, JONGCHEOL - Ulsan National Institute Of Science And Technology (UNIST)|
|HONG, SEOKMIN - Ulsan National Institute Of Science And Technology (UNIST)|
|KWON, YONGSUNG - Ulsan National Institute Of Science And Technology (UNIST)|
|CHO, KYUNGHWA - Ulsan National Institute Of Science And Technology (UNIST)|
Submitted to: Journal of Hazardous Materials
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/10/2020
Publication Date: 6/18/2020
Citation: Pyo, J., Hong, S., Kwon, Y., Kim, M.S., Cho, K. 2020. Estimation of heavy metals using deep neural network with visible and infrared spectroscopy of soil. Journal of Hazardous Materials. https://doi.org/10.1016/j.scitotenv.2020.140162.
Interpretive Summary: The presence of heavy metal elements such as arsenic, copper, and lead in soil can have toxic effects on the health of organisms that live in the contaminated soil, plants grown in the soil, and organisms that eat those plants. Effective detection and quantification of heavy metals in soil is important for efforts to mitigate, minimize, or prevent the effects of contamination. This study investigated the use of visible and near-infrared spectroscopy (VNIRS) with data analysis methods involving convolutional neural networks, convolutional autoencoder, artificial neural networks, random forest regression, and principal component analysis to estimate concentrations of arsenic, copper, and lead. Of the multiple prediction models tested, the convolutional neural network model with convolutional autoencoder, produced the highest accuracies for the heavy metal concentrations in soil. The use of VNIRS reflectance measurements was thus demonstrated as a feasible method for rapid and nondestructive analysis of soil for heavy metal detection. Further development of this method will benefit the food and agricultural industries as well as other fields for which environmental contamination may be a critical factor
Technical Abstract: Heavy metal contamination in soil disturbs the chemical, biological, and physical soil conditions and adversely affects the health of living organisms. Visible and near-infrared spectroscopy (VNIRS) shows a potential feasibility estimating heavy metal elements in soil. Moreover, deep learning models have been shown to successfully deal with complex multi-dimensional and multivariate nonlinear data. Thus, this study implemented a deep learning method on reflectance spectra of soil samples to estimate heavy metal concentrations. A convolutional neural network (CNN) was adopted to estimate arsenic (As), copper (Cu), and lead (Pb) concentrations using measured soil reflectance. In addition, a convolutional autoencoder was utilized as a joint method with the CNN for dimensionality reduction of the reflectance spectra. Furthermore, artificial neural network (ANN) and random forest regression (RFR) models were built for heavy metal estimation. Principal component analysis was utilized for dimensionality reduction of the ANN and RFR models. Among these models, the CNN model with convolutional autoencoder showed the highest accuracies for As, Cu, and Pb estimates having R2 values of 0.86, 0.74, and 0.82, respectively. The convolutional autoencoder disentangled the relevant features of multiple heavy metal elements and delivered robust features to the CNN model, thereby generating relatively accurate estimates.