The applicability of spectroscopy methods for estimating potentially toxic elements in soils: state-of-the-art and future trends

Abstract

Potentially toxic elements (PTEs) in soils pose severe threats to the environment and human health. It is therefore imperative to have access to simple, rapid, portable and accurate methods for their detection in soils. In this regards, the review introduces recent progresses made in the development and applications of spectroscopic methods for in-situ semi-quantitative and quantitative detection of PTEs in soil and critically compares them to standard analytical methods. The advantages and limitations of these methods are discussed together with recent advances in chemometrics and data mining techniques allowing to extract useful information based on spectral data. Furthermore, the factors influencing soil spectra and data analysis are discussed and recommendations on how to reduce or eliminate their influences are provided. Future research and development needs for spectroscopy techniques are emphasized, and an analytical framework based on technology integration and data fusion is proposed to improve the measurement accuracy of PTEs in soil.

Improved Mapping of Potentially Toxic Elements in Soil via Integration of Multiple Data Sources and Various Geostatistical Methods

“…Therefore, p-XRF should have a better possibility of application if we could reduce the negative effects of the environmental factors and then realize in situ and even on-the-go measurements [100]. Advanced pre-processing algorithms, such as external parameter orthogonalization and direct standardization, can be used to deal with this issue [101,102].…”

Current sensor technologies for in situ and on-line measurement of soil nitrogen for variable rate fertilization: A review

“…Data fusion (DF) by definition is the process of combining information from different sources to provide a potentially robust and complete description of an environment or process of interest (Durrant-Whyte, 2003). Nawar et al (2019) reviewed two major types of DF techniques used in modeling and analysis of proximal sensing data; (1) multiple sensor-based DF, where data from multiple sensors are used as input to a single multivariate or machine learning method, and (2) multi-model fusion, where models are combined to improve the prediction accuracy to optimal levels. This classification is adopted also by Ghassemian (2016) for remote sensing data, who claimed that DF can be performed at different levels, by incorporating the main processes in different stages.…”

“…This classification is adopted also by Ghassemian (2016) for remote sensing data, who claimed that DF can be performed at different levels, by incorporating the main processes in different stages. Although these levels are based on remote sensing data, the same principle can be extrapolated to proximal sensing as presented in Horta et al (2015) and Nawar et al (2019). At the first level, DF is implemented using the raw measurement outputs of the sensors, followed by feature extraction using methods such as naïve Bayes, multivariate statistics, Bayesian network, and majority rule few to mention among others (Hoegh and Leman, 2018) that ends up in classification [Fig.…”

Data Fusion of XRF and Vis-NIR Using Outer Product Analysis, Granger–Ramanathan, and Least Squares for Prediction of Key Soil Attributes

“…PLS is one of the popular and widely-used multivariate prediction methods in soil science [22]. It decomposes the matrix of the dependent variables X into scores T and loadings P such that X = TP T with size of P depending on the number of the latent variables.…”