Visible Near-infrared Reflectance Spectroscopy Research Articles

Optimized near infrared reflectance model based on preprocessing selection strategy for rapid analysis of 2-acetyl-1-pyrroline content in rice flour

The chemical compound 2-acetyl-1-pyrroline (2-AP) is a major contributor to the fragrance of rice. This report presents the first efforts towards determining the quantities of 2-AP in rice by grating-based visible–near-infrared reflectance spectroscopy using a diverse set of rice varieties. Calibration models generated by preprocessing treatments using Design of Experiment (DoE) and ‘one-variable-at-atime’ (OVAT) methods were compared and optimized. A modified partial least squares model with “2, 3, 1, 1”/SNV and 174 variables yielded a coefficient of determination for calibration values of 0.978. The results showed that a DoE-based preprocessing selection could efficiently and accurately optimize 2-AP calibration models, with advantages over OVAT. The findings of this study provide high-throughput screening technology for high-precision fragrance quality breeding, monitoring and control in the rice industry. The method demonstrated here also has potential for application in many other disciplines, including identification of plant fragrances and quality control of cereal products.

Using visible and near infrared spectroscopy and machine learning for estimating total petroleum hydrocarbons in contaminated soils

Petroleum pollution in soil is very damaging to the areas affected by the accidental release of petroleum hydrocarbons and has destructive impacts on natural resources and environmental health. Therefore, its monitoring and analysis are critical, however, due to the cost and time associated with chemical approaches, finding a quick and cost-effective analytical method is valuable. This study was conducted to evaluate the potential of using visible near infrared (Vis-NIR) spectroscopy to predict total petroleum hydrocarbons (TPH) in polluted soils around the Shadegan ponds, in southern Iran. One hundred soil samples showing various degrees of pollution were randomly collected from topsoil (0–10 cm). The soil samples were analyzed for TPH using Vis-NIR reflectance spectroscopy in the laboratory and then following application of preprocessing transformation, partial least squares PLS regression as well as two machine learning models including random forest (RF), and support vector machine (SVM) were examined. The results showed that the reflectance values at 1725 nm and 2311 nm, respectively, served as indicative TPH reflectance features, exhibiting weaker reflection with rising TPH. Among the preprocessing methods, the baseline correction method indicated the highest performance than the others. According to the evaluation model criteria in the validation dataset, the efficiency of the three selected models was observed in the following order SVM > RF > PLS regression. The SVM model provided the best performance in the validation dataset with r2 = 0.85, root mean of square (RMSEP = 1.59 %, and the ratio of prediction to deviation (RPD) = 2.6. Overall, this study provided strong evidence supporting the considerable potential of Visible-NIR spectroscopy as a rapid and cost-effective technique for estimating TPH levels in oil-contaminated soils, surpassing traditional chemical analytical methods. Applying the mid-infrared spectrum (MIR) in combination with Visible-NIR data is expected to provide more comprehensive and accurate results when assessing soils in polluted areas, thereby enhancing the accuracy and reliability of the results across a diverse range of soil types.

Feasibility of using Vis-NIR spectroscopy and PXRF spectrometry to estimate regional soil cadmium concentration

Conventionally, soil cadmium (Cd) measurements in the laboratory are expensive and time-consuming, involving complex processes of sample preparation and chemical analysis. This study aimed to identify the feasibility of using sensor data of visible near-infrared reflectance (Vis-NIR) spectroscopy and portable X-ray fluorescence spectrometry (PXRF) to estimate regional soil Cd concentration in a time- and cost-saving manner. The sensor data of Vis-NIR and PXRF, and Cd concentrations of 128 surface soils from Yunnan Province, China, were measured. Outer-product analysis (OPA) was used for synthesizing the sensor data and Granger-Ramanathan averaging (GRA) was applied to fuse the model results. Artificial neural network (ANN) models were built using Vis-NIR data, PXRF data, and OPA data, respectively. Results showed that: (1) ANN model based on PXRF data performed better than that based on Vis-NIR data for soil Cd estimation; (2) Fusion methods of both OPA and GRA had higher predictive power (R2) = 0.89, ratios of performance to interquartile range (RPIQ) = 4.14, and lower root mean squared error (RMSE) = 0.06, in ANN model based on OPA fusion; R2 = 0.88, RMSE = 0.06, and RPIQ = 3.53 in GRA model) than those based on either Vis-NIR data or PXRF data. In conclusion, there exists a great potential for the combination of OPA fusion and ANN to estimate soil Cd concentration rapidly and accurately.

Application of visible-near infrared spectroscopy in tandem with multivariate analysis for the rapid evaluation of matcha physicochemical indicators.

Consumer preference for matcha is heavily influenced by its physicochemical properties. The visible-near infrared (Vis-NIR) spectroscopy technology coupled with multivariate analysis was investigated for rapid and non-invasive evaluation of particle size and the ratio of tea polyphenols to free amino acids (P/F ratio) of matcha. The multivariate selection algorithms such as synergy interval (Si), variable combination population analysis (VCPA), competitive adaptive reweighted sampling (CARS), and interval combination population analysis (ICPA) were compared, and eventually, the variable selection strategy of ICPA and CARS hybridization was firstly proposed for selecting the characteristic wavelengths from Vis-NIR spectra to build partial least squares (PLS) models. Results indicated that the ICPA-CARS-PLS models achieved satisfactory performance for the evaluation of matcha particle size (Rp=0.9376) and P/F ratio (Rp=0.9283). Hence the rapid, effectual, and nondestructive online monitoring, Vis-NIR reflectance spectroscopy in tandem with chemometric models is significant for the industrial production of matcha.

Mapping the Spectral and Mineralogical Variability of Lunar Breccia Meteorite NWA 13859 by VNIR Reflectance Spectroscopy

We provide the first visible near-infrared (VNIR) reflectance spectroscopy characterization of a sample from the lunar feldspathic breccia NWA 13859. The sample is a 2 mm thick slab with an approximate area of 35 ± 2 cm2. We characterized the spectroscopic variability by choosing and analyzing representative spots throughout the sample. Our results show a distinct spectral contribution from orthopyroxene, which, according to a preliminary mineralogical characterization, should be a minor phase in the sample. In a second step, we measured several spectra along a transect between a clast and the matrix. In order to oversample the signal, the points analyzed along the sample were partly superimposed to each other (80% areal superposition). The same approach was extended to a grid of points covering an area of about 8.6 cm2, with the goal of obtaining a spatial resolution of the spectral parameters higher than the instrument’s spot size. Using this strategy, we obtained 2D maps of spectral parameters, which allowed us to infer the major mineralogical composition of the sample.

Soil carbon content prediction using multi-source data feature fusion of deep learning based on spectral and hyperspectral images

Visible near-infrared reflectance spectroscopy (VNIR) and hyperspectral images (HSI) have their respective advantages in soil carbon content prediction, and the effective fusion of VNIR and HSI is of great significance for improving the prediction accuracy. But the contribution difference analysis of multiple features in the multi-source data is inadequate, and there is a lack of in-depth research on the contribution difference analysis of artificial feature and deep learning feature. In order to solve the problem, soil carbon content prediction methods based on VNIR and HSI multi-source data feature fusion are proposed. The multi-source data fusion network under the attention mechanism and the multi-source data fusion network with artificial feature are designed. For the multi-source data fusion network based on the attention mechanism, the information are fused through the attention mechanism according to the contribution difference of each feature. For the other network, artificial feature are introduced to fuse multi-source data. The results show that multi-source data fusion network based on the attention mechanism can improve the prediction accuracy of soil carbon content, and multi-source data fusion network combined with artificial feature has better prediction effect. Compared with two single-source data from the VNIR and HSI, the relative percent deviation of Neilu, Aoshan Bay and Jiaozhou Bay based on multi-source data fusion network combined with artificial feature are increased by 56.81% and 149.18%, 24.28% and 43.96%, 31.16% and 28.73% respectively. This study can effectively solve the problem of the deep fusion of multiple features in the soil carbon content prediction by VNIR and HSI, so as to improve the accuracy and stability of soil carbon content prediction, promote the application and development of soil carbon content prediction in spectral and hyperspectral image, and provide technical support for the study of carbon cycle and the carbon sink.

Reconstructing historical time-series of cyanobacteria in lake sediments: Integrating technological innovation to enhance cyanobacterial management

The global rise of cyanobacterial blooms emphasizes the need to develop tools to manage water bodies prone to cyanobacterial dominance. Reconstructing cyanobacterial baselines and identifying environmental drivers that favour cyanobacterial dominance are important to guide management decisions. Conventional techniques for estimating cyanobacteria in lake sediments require considerable resources, creating a barrier to routine reconstructions of cyanobacterial time-series. Here, we compare a relatively simple technique based on spectral inferences of cyanobacteria using visible near-infrared reflectance spectroscopy (VNIRS) with a molecular technique based on real-time PCR quantification (qPCR) of the 16S rRNA gene conserved in cyanobacteria in 30 lakes across a broad geographic gradient. We examined the sedimentary record from two perspectives: 1) relationships throughout the entire core (without radiometric dating); 2) relationships post-1900s with the aid of radiometric dating (i.e., 210Pb). Our findings suggest that the VNIRS-based cyanobacteria technique is best suited for reconstructing cyanobacterial abundance in recent decades (i.e., circa 1990 onwards). The VNIRS-based cyanobacteria technique showed agreement with those generated using qPCR, with 23 (76%) lakes showing a strong or very strong positive relationship between the results of the two techniques. However, five (17%) lakes showed negligible relationships, suggesting cyanobacteria VNIRS requires further refinement to understand where VNIRS is unsuitable. This knowledge will help scientists and lake managers select alternative cyanobacterial diagnostics where appropriate. These findings demonstrate the utility of VNIRS, in most instances, as a valuable tool for reconstructing past cyanobacterial prevalence.

Vis-NIR Reflectance Spectroscopy and PLSR to Predict PCB Content in Severely Contaminated Soils: A Perspective Approach

Soil reclamation from polychlorinated biphenyls (PCBs) requires careful analysis in terms of their concentrations and spatial distribution. Conventional laboratory analysis, even if providing the careful evaluation of PCBs, is costly and time-consuming. Therefore, rapid and cost-effective techniques to replace traditional analytical approaches are required. The utility of visible-near infrared (vis-NIR) reflectance spectroscopy in conjunction with partial least square regression (PLSR) analysis was evaluated in this study. Spectral reflectance was measured in the laboratory on 28 soil samples collected in a highly contaminated area of southern Italy and chemically analysed to determine eighteen PCB congeners, their sum (PCBs18), and extractable organic halogen content (EOX). Spectroscopic data were pre-processed prior to data analysis by combining different methods. Using PLSR analysis, significant relationships were observed between the predicted and the measured content of PCBs18, EOX, and the percentage of several isomeric classes of PCBs. Although rigorous models could not be calibrated, due to the limited number of samples, the preliminary results of this study demonstrated that vis-NIR reflectance spectroscopy, coupled with PLSR, can be considered a promising method for a rapid and cost-effective prediction of PCBs.

Potential of GPR data fusion with hyperspectral data for precision agriculture of the future

Precision Agriculture (PA) requires accurate spatial and temporal information of soil properties at a very fine scale. Traditional soil characterization methods are time consuming, laborious and invasive and do not allow long-term repeatability of measurements. Ground Penetrating Radar (GPR) appears to be a particularly suitable methodology for characterizing soil and subsurface from a physical property point of view. Visible-near infrared-shortwave infrared (VIS-NIR-SWIR) reflectance spectroscopy has now become a widespread technique in soil analysis. Information on soil variability can be improved by the integration of data from multiple sensors. The overall objective of this paper was to examine the potential of fusing GPR data with hyperspectral data using multivariate geostatistics for delineating the management zones in the soil of an olive grove of centuries-old trees in Italy.A linear model of coregionalization (LMC) was individually fitted for the raw hyperspectral data and for GPR data including for each case a nugget effect and two spherical models at short scale and at longer scale. After that, one data set was obtained from the fusion of the two sensor data sets and a LMC was fitted for the combined data to be then used in factor cokriging. The application of this technique produced a delineation of the field into homogeneous zones, highlighting a wide southern-central zone, characterized by different granulometric and chemical properties. The proposed approach was then effective to discriminate areas with different properties by using multi-sensor data. It then has the potential to be used in PA.

Crystallinity effects on the vibrational spectral features of saponite: Implications for characterizing variable crystalline phyllosilicates on Mars

X-ray diffraction patterns of Martian mudstones acquired by Chemistry & Mineralogy X-Ray Diffraction (CheMin) aboard Mars Science Laboratory (MSL) suggest that the smectites detected in Gale crater have poor crystallinity. This finding poses an urgent question about the structural ordering of phyllosilicates found globally by Visible/Near-Infrared (VIS-NIR) spectroscopic orbital remote sensing on Mars, linked to their formation conditions. In this study, we synthesized saponite (Nz+x/z[M6][Si8-xAlx]O20(OH)4·nH2O, where M and N correspond to the divalent octahedral cations and the interlayer cations, respectively) with variable crystallinity, which bear structural similarities to the smectite discovered at Gale crater. Synthetic saponite was characterized using Field Emission Scanning Electron Microscopy (FE-SEM), powder X-ray diffraction (XRD), then studied the spectral features of these samples using nuclear magnetic resonance spectroscopy (NMR), Raman spectroscopy, and VIS-NIR reflectance spectroscopy. Our study revealed that the crystallinity of these saponite samples increased (as shown in the FWHM of XRD (060) peak), and is accompanied by (1) improved T-O-T layer stacking along the c axis (specified by the intensity and width of the XRD (001) peak); (2) increased uniformity of the SiO4 unit in tetrahedral sheets (based on the peak widths of Raman spectra and the well-resolved 29Si NMR peaks); (3) improved regularity in the distribution of Mg2+ in octahedral sites and thus the regularity of metal-OH bonds in octahedral sheet (based on the resolution of the NIR 2.2–2.4 μm band and the peak width of the XRD (060,330) peak); and (4) the increased Al in tetrahedral sites and decrease of Al in octahedral sites (the width of 27Al NMR peak), the crystallinity of saponite raised. Based on experimental observations, the first derivative spectra of metal-OH absorptions were proposed as a crystallinity index for smectite on Mars. Our results indicate that different vibration spectroscopy techniques can constrain the structural ordering of smectite on Mars and provide insight into their formation conditions.

Visible/near infrared reflectance (VNIR) spectral features of ion-exchangeable Rare earth elements hosted by clay minerals: Potential use for exploration of regolith-hosted REE deposits

Regolith-hosted ion-exchangeable rare earth element (REE) mineralization in South China dominate the global source of heavy REEs. The lack of studies on spectral features of ion-exchangeable REEs limits the application of visible/near infrared reflectance (VNIR, 350–2500 nm) spectroscopy for rapid identification of regolith-hosted REE deposits in prospective areas. This paper presents a systematic study on VNIR spectral features of ion-exchangeable REEs hosted by a series of synthetic REE-adsorbed clay minerals and natural regolith samples. The results show that ion-exchangeable Nd3+, Dy3+, Ho3+, Er3+ and Tm3+ display diagnostic absorption features at the bands of 730–870, 805, 641, 652, and 684 nm. Specifically, the Nd3+ related absorption bands of regolith samples at ~800 nm show a shoulder at ~802 nm that are distinguishable from unweathered primary REE(−bearing) minerals. The maxima of second derivative curves derived from the Nd3+ absorption at ~802 nm, M802_2nd, can be used to quantify variations of ion-exchangeable REEs in regoliths. Nevertheless, the absorption features of REE3+ ions could be subdued due to an overlapping problem caused by the broad absorption bands of Fe-oxides in the range of 700–900 nm. The obtained VNIR spectra of ion-exchangeable REEs provide critical supplement to spectral library for REE-bearing minerals. This study demonstrates that field-based VNIR spectroscopy can be deployed as a powerful tool for regional and deposit-scale exploration of regolith-hosted ion-exchangeable REE resources.

Mapping leaf metal content over industrial brownfields using airborne hyperspectral imaging and optimized vegetation indices

Monitoring plant metal uptake is essential for assessing the ecological risks of contaminated sites. While traditional techniques used to achieve this are destructive, Visible Near-Infrared (VNIR) reflectance spectroscopy represents a good alternative to monitor pollution remotely. Based on previous work, this study proposes a methodology for mapping the content of several metals in leaves (Cr, Cu, Ni and Zn) under realistic field conditions and from airborne imaging. For this purpose, the reflectance of Rubus fruticosus L., a pioneer species of industrial brownfields, was linked to leaf metal contents using optimized normalized vegetation indices. High correlations were found between the vegetation indices exploiting pigment-related wavelengths and leaf metal contents (r ≤ − 0.76 for Cr, Cu and Ni, and r ≥ 0.87 for Zn). This allowed predicting the metal contents with good accuracy in the field and on the image, especially Cu and Zn (r ≥ 0.84 and RPD ≥ 2.06). The same indices were applied over the entire study site to map the metal contents at very high spatial resolution. This study demonstrates the potential of remote sensing for assessing metal uptake by plants, opening perspectives of application in risk assessment and phytoextraction monitoring in the context of trace metal pollution.

Visible Near-Infrared Reflectance and Laser-Induced Breakdown Spectroscopy for Estimating Soil Quality in Arid and Semiarid Agroecosystems

Visible near-infrared reflectance spectroscopy (VNIRS) and laser-induced breakdown spectroscopy (LIBS) are potential methods for the rapid and less expensive assessment of soil quality indicators (SQIs). The specific objective of this study was to compare VNIRS and LIBS for assessing SQIs. Data was collected from over 140 soil samples taken from multiple agricultural management systems in New Mexico, belonging to arid and semiarid agroecosystems. Sampled sites included New Mexico State University Agricultural Science Center research fields and several commercial farm fields in New Mexico. Partial least squares regression (PLSR) was used to establish predictive relationships between spectral data and SQIs. Fifteen soil measurements were modeled including the soil organic matter (SOM), permanganate oxidizable carbon (POXC), total microbial biomass (TMB), total bacteria biomass (TBB), total fungi biomass (TFB), mean weight diameter of dry aggregates (MWD), aggregates 2–4 mm (AGG > 2 mm), aggregates < 0.25 mm (AGG < 0.25 mm), wet aggregate stability (WAS), electrical conductivity (EC), calcium (Ca), magnesium (Mg), sodium (Na), and iron (Fe). Overall, calibrations based on measurements irrespective of locations performed better for LIBS and combined VNIRS-LIBS. Measurements separated according to locations highly improved the quality of prediction for VNIRS as compared to combined locations. For example, the prediction R2 values for regression of VNIRS were 0.19 for SOM, 0.30 for POXC, 0.24 for MWD, 0.15 for AGG > 2 mm, and 0.13 for EC in combined datasets irrespective of location. When separated according to locations, for one of the locations, the predictive R2 values for VNIRS were 0.48 for SOM, 0.70 for POXC, 0.67 for MWD, 0.60 for AGG > 2 mm, and 0.51 for EC. The prediction values varied with the sampling time for both LIBS and VNIRS. For example, the prediction values of some SQIs using VNIRS were higher in samples collected in winter for measurements, including SOM (0.90), MWD (0.96), WAS (0.66), and EC (0.94). Using the VNIRS, the corresponding predictive values for the same SQIs were lower for samples collected in the fall (SOM (0.61), MWD (0.45), WAS (0.46), and EC (0.65)). While this study illustrates the prospects of VNIRS and LIBS for estimating SQIs, a more comprehensive evaluation, using a larger regional dataset, is required to understand how the site and soil factors affect VNIRS and LIBS, in order to enhance the utility of these methods for soil quality assessment in arid and semiarid agroecosystems.

Using visible near-infrared reflectance spectroscopy (VNIRS) of lake sediments to estimate historical changes in cyanobacterial production: potential and challenges

Cyanobacterial blooms are increasing worldwide and have negative impacts on aquatic ecosystems and the services they provide to human societies. A lack of long-term environmental monitoring data, however, has prevented the development of a baseline perspective against which drivers of the increasing frequency and severity of cyanobacterial blooms can be identified. In this study, we evaluate application of spectroscopy-based models to infer historical trends in cyanobacterial abundance from lake sediment cores. Using an amendment series (n = 15) of a sediment matrix spiked with increasing amounts of mixed cyanobacterial culture from 0 to 50 parts per thousand (‰), taxonomically diagnostic carotenoids were measured using visible near-infrared reflectance spectroscopy (VNIRS) and conventional but more costly and time-consuming high-performance liquid chromatography (HPLC). A partial least squares regression model was developed to correlate amendment series VNIR spectra to ‰ of added cyanobacteria. Despite challenges in differentiating carotenoid pigments because of overlapping absorption peaks, applications of the resulting 2-component model (r2 = 0.93, RMSEP = 0.23‰) to sediment cores from four Ontario lakes yielded temporal trends that were significantly correlated with downcore HPLC measures of cyanobacterial pigments in three out of four cases. Although our method is simplistic and may be improved in the future with more complex algorithms employing derivative analysis, we present our results as a possible stepping-stone towards spectral reconstruction of cyanobacterial production. Our study provides proof-of-concept that refinement of a method applying VNIRS to detect cyanobacterial carotenoids in lake sediments has the potential to be an important, rapid and non-destructive assessment tool for research and management of cyanobacterial blooms.

At-line Prediction of Gelatinized Starch and Fiber Fractions in Extruded Dry Dog Food Using Different Near-Infrared Spectroscopy Technologies.

Simple SummaryStarch is a non-fibrous carbohydrate that represents an important percentage of pet food composition. The degree of its gelatinization, due to the cooking process, can be a useful indicator of starch digestibility in the diet. Moreover, fiber fractions are important for animals’ health and nutritional status, so pet food industry is interested in the development of an easy and cost-effective method to measure these parameters. Results of this study revealed the applicability of visible/near-infrared spectroscopy to predict total and gelatinized starch, neutral detergent fiber, acid detergent fiber, and acid detergent lignin in pet food. On the other hand, near-infrared transmittance technology showed a scarce accuracy. The developed prediction models for total and gelatinized starch and fiber fractions using visible/near-infrared spectroscopy could be applied during the manufacturing process to perform quality controls.This study aimed to assess the feasibility of visible/near-infrared reflectance (Vis-NIR) and near-infrared transmittance (NIT) spectroscopy to predict total and gelatinized starch and fiber fractions in extruded dry dog food. Reference laboratory analyses were performed on 81 samples, and the spectrum of each ground sample was obtained through Vis-NIR and NIT spectrometers. Prediction equations for each instrument were developed by modified partial least squares regressions and validated by cross- (CrV) and external validation (ExV) procedures. All studied traits were better predicted by Vis-NIR than NIT spectroscopy. With Vis-NIR, excellent prediction models were obtained for total starch (residual predictive deviation; RPDCrV = 6.33; RPDExV = 4.43), gelatinized starch (RPDCrV = 4.62; RPDExV = 4.36), neutral detergent fiber (NDF; RPDCrV = 3.93; RPDExV = 4.31), and acid detergent fiber (ADF; RPDCrV = 5.80; RPDExV = 5.67). With NIT, RPDCrV ranged from 1.75 (ADF) to 2.61 (acid detergent lignin, ADL) and RPDExV from 1.71 (ADL) to 2.16 (total starch). In conclusion, results of the present study demonstrated the feasibility of at-line Vis-NIR spectroscopy in predicting total and gelatinized starch, NDF, and ADF, with lower accuracy for ADL, whereas results do not support the applicability of NIT spectroscopy to predict those traits.

Soil total nitrogen estimation of alpine grassland using visible/near-infrared spectra: a comparison of multivariate techniques with different spectral transformations

Rapid, precise, and quantitative assessment of soil total nitrogen (TN) of alpine grassland is crucial for monitoring the environmental conditions of grasslands. The right choice of preprocessing and calibration methods may improve the model performance of soil property estimation in visible/near infrared reflectance spectroscopy (VNIRS). VNIRS (350 to 2500 nm) measurements were made of 80 soil samples from the northeast part of the Tibetan Plateau. To reduce the signal noise, the Savitzky–Golay technique was implemented and six spectral transformations were evaluated with the aim of identifying the best transformation for predicting soil TN. The performance of three linear calibration models, namely, stepwise multiple linear regression, principal component regression, partial least squares regression (PLSR), and a combination of these models with a nonlinear back propagation neural network (BPNN) was tested for accuracy in the measurement of soil TN. Our results indicate the following: (1) models using the derivative transformations of spectra as inputs perform better than those using nonderivative transformations, and the first derivative of the reciprocal of reflectance [(1 / R) ′ ] is the best spectral transformation for soil TN estimation; (2) models based on the PLSR and companion BPNN-latent variables (LVs) yield better predictions and smallest errors of soil TN. We demonstrate that (1 / R) ′ is the best spectral transformation and consistently improves the estimation models of soil TN. Among multivariate techniques, the BPNN-LVs is recommended for estimation of soil TN with great accuracy.

Simultaneous prediction of soil properties from VNIR-SWIR spectra using a localized multi-channel 1-D convolutional neural network

The use of visible near-infrared and shortwave-infrared (VNIR-SWIR) diffuse reflectance spectroscopy for the estimation of soil properties is increasingly maturing with large-scale soil spectral libraries (SSLs) of laboratory spectra developed across the globe. Such an SSL is the publicly available LUCAS topsoil database with approximately 20,000 soil samples encompassing 23 countries of the European Union. A wide variety of machine learning tools have been applied to the LUCAS SSL to predict some of the soil samples’ physicochemical properties with different degrees of accuracy. In this paper, we developed and examined the use of a novel one-dimensional convolutional neural network (CNN) to simultaneously predict ten physicochemical properties of the LUCAS SSL. Leveraging on the use of multiple-input channels it uses as model inputs the absorbance spectra along with some pre-processed spectra developed using standard techniques. Moreover, it exploits the use of local spectral neighborhoods to perform an adaptive error-correction mechanism. This novel localized multi-channel 1-D CNN was applied to all the available physicochemical properties of the LUCAS SSL and was statistically compared with the current state-of-the-art where it was shown to statistically outperform its counterparts, as well as with other CNNs where it exhibited the best performance. In particular, for the mineral soil samples, the RMSE for the Clay content was 4.80% (R2 0.86), for soil organic carbon the RMSE was 10.96 g kg−1 (R2 0.86), while for total nitrogen the RMSE was 0.66 g kg−1 (R2 0.83).

Rapid estimation of soil cation exchange capacity through sensor data fusion of portable XRF spectrometry and Vis-NIR spectroscopy

Soil cation exchange capacity (CEC) is a critical property of soil fertility. Conventionally, it is measured using laboratory chemical methods, which involve complex sample preparation and are time-consuming and expensive. Previous studies have investigated nondestructive and rapid methods for determining soil CEC using proximal soil sensors individually, including portable X-ray fluorescence (PXRF) spectrometry and visible near-infrared reflectance (Vis-NIR) spectroscopy. In this study, we examined the potential of the fusing data from PXRF and Vis-NIR to predict soil CEC for 572 soil samples from Yunnan Province, China. The CEC of the samples ranged from 5.42 to 50.25 cmol kg−1. Both partial least-squares regression (PLSR) and support vector machine regression (SVMR) were applied to predict soil CEC with individual sensor datasets and a fused sensor dataset for comparison. The root mean squared error (RMSE), coefficients of determination (R2), and ratios of performance to interquartile range (RPIQ) were used to evaluate the performance of the models. Results showed that: (1) SVMR performed better than PLSR on single sensor datasets and the fused sensor dataset, in terms of RMSE, R2, and RPIQ; and (2) both PLSR and SVMR based on the fused sensor dataset had better predictive power (RMSE = 4.02, R2 = 0.72, and RPIQ = 2.23 in PLSR model; RMSE = 3.02, R2 = 0.82, and RPIQ = 2.31 in SVMR model) than those based on any single sensor dataset. In summary, the fused sensor data and SVMR showed great potential for estimating soil CEC efficiently.

Diffuse reflectance spectroscopy for rapid estimation of soil Atterberg limits

Conventional methods for the determination of Atterberg limits based on laboratory tests are time-consuming. Reflectance spectroscopy which is fast and cost-effective has been reported as an alternative method for estimating Atterberg limits. To assess the efficiency of the method, 60 forest soil samples were measured by visible-near infrared reflectance spectroscopy (VNIR) within the range of 350–2500 nm. The Savitzky-Golay (SG) smoothing method for spectral data was used to reduce signal noise and the first derivative reflectance (DR) was applied to accentuate the features and assemble the data for use in the modeling process. In this study, raw reflectance (R) and first derivative reflectance (DR) data were used to estimate Atterberg limits. All two-band combinations of three types of spectra (ratio index, normalized difference index and difference index) as well as PLS and PLS-BPNN were calculated to extract optimum waveband combination and calibrations between Atterberg limits and reflectance spectra. The correlation coefficients between Atterberg limits and reflectance spectra indicated both positive and negative correlation at various wavelengths across the spectrum. The highest correlation for liquid limit (r = −0.86), plastic limit (r = −0.76) and plasticity index (r = −0.7) was found at 1921, 1938 and 1913 nm, respectively. In addition, the results showed the spectral indices technique as a good method for estimating soil plasticity and improving the prediction accuracy of Atterberg limits. In the current study, PLS and PLS–BPNN were also used to develop the calibration models for estimating Atterberg limits using R and DR. The results demonstrated that the prediction accuracy of PLS-BPNN for both R and DR was higher than that of PLS. Correlation coefficients between shrinkage limit (SL) and reflectance spectra showed weak correlation, and overall reflectance had the same correlation. The results represented that this method could be used to estimate soil plasticity properties quickly, non-destructively, and with minimal sample preparation and high accuracy. These advantages may be more important, especially when a huge number of samples should be tested and analyzed within a short period of time.

Estimation of soil pH using PXRF spectrometry and Vis-NIR spectroscopy for rapid environmental risk assessment of soil heavy metals

Environmental risk of heavy metals (HMs) in soil is commonly assessed by the different risk screening values of HMs under different pH based on soil environmental quality standards. To explore and establish a reliable, rapid and cost-effective method for detailed soil environmental quality survey with high-density sampling in the large-scale area is of significance for theoretical and practical research. In present study, using data from Yunnan Province, China, rapid analysis of soil HMs were conducted via portable X-ray fluorescence (PXRF) spectrometry, and that of soil pH was estimated by applying PXRF and visible near-infrared reflectance (Vis-NIR) spectroscopy data to partial least-squares regression (PLSR) and support vector machine regression (SVMR). Then we compared soil HM contamination grades calculated by conventional laboratory analysis data with those of rapid analysis data. It was found that soil HMs (i.e., As, Pb, Cu, and Zn) were successfully estimated by PXRF with high coefficient of determination (R2) above 0.97 (P < 0.001). SVMR with fused sensor dataset (here PXRF and Vis-NIR) provided the best predictive model for soil pH estimation (R2 = 0.86; the ratio of performance to deviation (RPD) = 2.21; the ratio of performance to interquartile distance (RPIQ) = 3.09). The Kappa coefficient of the classification was 0.91, a very high-level consistency between the assessment of soil HM contamination grades calculated by rapid analysis data and that of conventional laboratory analysis data. Therefore, our study suggested a promising method to rapidly detect soil HM contamination under different pH intervals, which would considerably reduce the financial burden of detailed soil HM survey great sampling number in large-scale areas.