Laser-Induced Breakdown Spectroscopy (LIBS) for tropical soil fertility analysis
Introduction
Worldwide, sensor-based analytical methods have been seen to characterize key soil fertility attributes in a rapid way without producing chemical residues (Viscarra Rossel and Bouma, 2016, Molin and Tavares, 2019). Proximal soil sensing (PSS) methods allows to reduce the number of operations con-ducted in traditional analytical procedures (e.g. digestion and extraction), making the analysis faster, cost-effective and environmentally friendly. Different PSS techniques have been applied for determining soil fertility attributes, e.g., X-ray fluorescence (XRF), mid infrared and visible and near infrared (vis-NIR) diffuse reflectance spectroscopies, ion selective electrodes (ISE), are few to mention among others (Silva and Molin, 2018, Munnaf et al., 2019, Demattê et al., 2019, Tavares et al., 2020a). LIBS spectra allow a broad characterization of soil elementary constitution (e.g., N, P, K, Ca, Mg, Al, and Fe) (He et al., 2018, Nicolodelli et al., 2019), and satisfactory determinations of the total contents of these elements in soil samples are commonly found in the literature (Hussain et al., 2007, Riebe et al., 2019, Erler et al., 2020, Xu et al., 2019a). LIBS spectra can also be used as a proxy to infer about soil fertility attributes [e.g., extractable (ex-) nutrients, pH, cation exchange capacity (CEC), organic matter (OM), pH, base saturation (V), and textural attributes] (Senesi, 2020). However, research regarding LIBS applications for soil fertility analysis are still incipient. In temperate soils, the results of the few studies carried out are not always satisfactory. For example, in German agricultural soils, Erler et al. (2020) have achieved good prediction performances for pH (R2 between 0.91 and 0.95), reasonable predictions for humus content (R2 between 0.54 and 0.66), and poor predictions for ex-P (R2 between 0.22 and 0.35). Similarly, in Chinese agricultural soils, studies have shown poor prediction performance for ex-K and ex-P (Xu et al., 2019a), reasonable predictions for OM (R2 = 0.58) (Xu et al., 2019b), and good prediction for CEC, OM, and pH (R2 ≥ 0.81) (Xu et al., 2019a). Xu et al. (2019c) reported satisfactory predictions for pH, OM, and total N (R2 between 0.60 and 0.75), and poor performances (R2 < 0.35) for ex-K and ex-P in Chinese soils. In Brazilian tropical soils, although good prediction performance (R2 ≥ 0.85) have been achieved for textural attributes (Villas-Boas et al., 2016) and pH (Ferreira et al., 2015), evaluation of the prediction accuracy for other key fertility attributes (e.g., CEC, V, OM, and extractable nutrients) has not been addressed in the literature yet.
The selection of an optimal modelling procedure for LIBS spectral data, capable of extracting useful but hidden information is crucial for accurate prediction of soil fertility attributes. Although there is no consensus regarding the best modelling strategy to adopt, recent studies suggested using multivariate techniques for the determination of soils elementary constitution when using LIBS spectra, in particular the partial least squares regression (PLS) (Takahashi and Thornton, 2017, Riebe et al., 2019). In addition, multivariate models are a simple and useful strategy to mitigate the matrix effect commonly present in soil samples (Takahashi and Thornton, 2017, Tavares et al., 2020b). Despite of the benefits, multivariate models using the entire spectrum require high computational capacity, especially due to the high spectral resolution of LIBS data that result in a large number of spectral variables (e.g., thousands data points) (Erler et al., 2020). In this regard, algorithms for the selection of most significant variables, such as deep learning or interval successive projection algorithm in partial least squares (iSPA-PLS) (Gomes et al., 2013, Riebe et al., 2019, Niu et al., 2021), can be a useful technique to reduce the amount of LIBS input variables and simplify data acquisition and processing procedures. To our best knowledge, no previous work reported the performance of iSPA-PLS for the prediction of key soil attributes in tropical soils, in comparison with full-spectra based PLS analyses.
This work aimed at evaluating different modelling approaches, namely, multiple linear regression (MLR) using selected emission lines of LIBS data, PLS using the full LIBS spectra from 200 to 540 nm (with 38,880 variables), and PLS using specific spectral regions selected by the iSPA algorithm for the prediction of key soil fertility attributes in Brazilian tropical soils. The best performing model was compared with univariate linear regression (ULR) in order to assess the influence of the matrix effect on the prediction of extractable nutrients (ex-P, ex-Ca, and ex-Mg).
Section snippets
Soil samples and reference analysis
A set of 102 soil samples, belonging to the soil sample bank of the Precision Agriculture Laboratory (LAP) from Luiz de Queiroz College of Agriculture, University of São Paulo, were used in this study. These samples were collected from 0 to 20 cm depth in two Brazilian agricultural fields (designated as Field 1 and Field 2) and stored after being air-dried and sieved at 2 mm. The analysis results of the LAP’s soil sample bank were used to choose samples with wide ranges of variability of key
Results
Laboratory measured soil properties.
The descriptive statistics of soil fertility attributes for the calibration and validation datasets are shown in Fig. 2. The Kennard Stone algorithm ensuring comparable range and standard deviation (SD) for the calibration and validation sets (Fig. 2) is fundamental for an effective evaluation of the models' predictive performance (Mourad et al., 2005, Mouazen et al., 2006). It is also important to mention that the selected samples have wide variability
Predicting key soil fertility attributes using LIBS
The LIBS spectra obtained from the pelletized soil samples allowed satisfactory prediction results (0.59 ≤ R2 ≤ 0.94) using multivariate models developed for eight out of nine fertility attributes, although pH was the only attribute that showed poor prediction performance (R2 ≤ 0.31). Soil fertility attributes are indirectly predicted via elemental analysis sensors due to relationship existing between such attributes and the elemental constitution of the soil samples. Textural attributes are
Conclusion
LIBS proved to be efficient for predicting fertility attributes in tropical soils. The LIBS spectra obtained from the pelletized soil samples associated with the multivariate models achieved satisfactory predictions (RPD > 1.40) for eight out of the nine key soil fertility attributes, with pH being the only exception that showed poor predictive performance (RPD ≤ 1.12). The best prediction performances were obtained for CEC, ex-Ca, and ex-Mg, which had RPD always greater than 2.0, regardless of
Funding acknowledgements
T.R.T. was funded by São Paulo Research Foundation (FAPESP) [grant number 2017/21969-0 and 2020/16670-9]. Soil fertility tests were funded by National Council for Scientific and Technological Development (CNPq) – “Edital de Chamada Universal” [grant number 458180/2014–9]. L.C.N. was funded by CNPq [grant number 381261/2020–4]. LIBS facilities were funded by FAPESP (FAPESP 04/15965–2) [grant number 2015–19121–8]. Authors also acknowledge the financial support received from the Research
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
References (46)
- et al.
Prediction of soil fertility via portable X-ray fluorescence (pXRF) spectrometry and soil texture in the Brazilian Coastal Plains
Geoderma
(2020) - et al.
Soil analytical quality control by traditional and spectroscopy techniques: Constructing the future of a hybrid laboratory for low environmental impact
Geoderma
(2019) - et al.
Laser-induced breakdown spectroscopy: Extending its application to soil pH measurements. Spectrochim.
Acta B.
(2015) - et al.
Sample treatment and preparation for laser-induced breakdown spectroscopy
Spectrochim. Acta Part B
(2016) - et al.
Elemental analysis of Cerrado agricultural soils via portable X-ray fluorescence spectrometry: Inferences for soil fertility assessment
Geoderma
(2019) - et al.
Recent advances and future trends in LIBS applications to agricultural materials and their food derivatives: an overview of developments in the last decade (2010–2019). Part I. Soils and fertilizers
Trends Anal. Chem.
(2019) - et al.
Simultaneous optimization by neuro-genetic approach for analysis of plant materials by laser induced breakdown spectroscopy
Spectrochim. Acta Part B
(2009) - et al.
Laser-induced breakdown spectroscopy (LIBS) to measure quantitatively soil carbon with emphasis on soil organic carbon. A review
Anal. Chim. Acta
(2016) - et al.
Quantitative methods for compensation of matrix effects and self-absorption in Laser Induced Breakdown Spectroscopy signals of solids
Spectrochim. Acta B
(2017) - et al.
Soil sensing: A new paradigm for agriculture
Agric. Syst.
(2016)
Laser-induced breakdown spectroscopy to determine soil texture: a fast analytical technique
Geoderma
Detection of soil organic matter from laser-induced breakdown spectroscopy (LIBS) and mid-infrared spectroscopy (FTIR-ATR) coupled with multivariate techniques
Geoderma
Improved intact soil-core carbon determination applying regression shrinkage and variable selection techniques to complete spectrum laser-induced breakdown spectroscopy (LIBS)
Appl. Spectrosc.
Comparing vis–NIRS, LIBS, and combined vis–NIRS‐LIBS for intact soil core soil carbon measurement
Soil Sci. Soc. Am. J.
Near-infrared reflectance spectroscopy–principal components regression analyses of soil properties
Soil Sci. Soc. Am. J.
Guidelines for calibration in analytical chemistry. Part I. Fundamentals and single component calibration (IUPAC Recommendations 1998)
Pure Appl. Chem.
Method 3051A microwave assisted acid digestion of sediments, sludges, soils, and oils. Z. Für
Anal. Chem.
Soil Nutrient Detection for Precision Agriculture Using Handheld Laser-Induced Breakdown Spectroscopy (LIBS) and Multivariate Regression Methods (PLS, Lasso and GPR)
Sensors
Intemperismo de rochas e minerais
The successive projections algorithm for interval selection in PLS
Microchem. J.
Quantitative analysis of nutrient elements in soil using single and double-pulse laser-induced breakdown spectroscopy
Sensors
Measurement of nutrients in green house soil with laser induced breakdown spectroscopy
Environ. Monit. Assess.
Computer aided design of experiments
Technometrics
Cited by (23)
Optimization of pXRF instrumentation conditions and multivariate modeling in soil fertility attributes determination
2024, Spectrochimica Acta - Part B Atomic SpectroscopySensing technologies for characterizing and monitoring soil functions: A review
2023, Advances in AgronomyCitation Excerpt :Their results suggested that the LIBS methodology rapidly and efficiently measures soil carbon with excellent detection limits (∼ 300 mg kg), precision (4–5%) and accuracy (3–14%). Tavares et al. (2022) proved that LIBS efficiently predicted fertility attributes in tropical soils. They achieved an RPD > 1.40 for eight out of the nine properties studied, being pH the one that showed the worst result.
Comparative elemental analysis of soil of wheat, corn, rice, and okra cropped field using CF-LIBS
2022, OptikCitation Excerpt :In this technique, a very energy rapid laser pulse is focused on the sample to generate plasma, and the ablated material breaks down into atomic species and excited ionic species [8]. LIBS can detect several elements, has a quick reaction time, and requires little/no target material preparation [9–13]. LIBS is a type of atomic emission spectroscopy in which a high-powered laser pulse contacts the sample surface to create a plasma, which subsequently generates a unique fingerprint for the elemental composition of the sample utilizing its characteristic lines. [14]