Continuum Removal Research Articles

Improving SPAD spectral estimation accuracy of rice leaves by considering the effect of leaf water content

AbstractThe spectral range <1,000 nm is commonly used for estimating the chlorophyll content of plants. However, the influence of the response band of leaf water content in the Short‐Wave InfraRed on the chlorophyll estimate model is uncertain. We measured the Visible‐Near InfraRed‐Short Wave InfraRed spectral reflectance and SPAD values of rice (Oryza sativa L.) leaves at different growth stages. The vegetation indices (VIs), red edge positions (REPs), and spectral characteristic parameters (SCPs) were calculated. The correlation between the above parameters and SPAD was analyzed. The effects of chlorophyll and water content on the spectral reflectance absorption characteristics of rice leaves were analyzed by the PROSPECT model. SPAD spectral estimation models were established with different inputs using the Random Forest (RF) regression algorithm. Our results are described as follows: (a) The spectral reflectance after continuum removal of rice leaves was highly correlated with SPAD in 600–690, 720–760, 1,400–1,490, and 1,900–1,980 nm. (b) Compared with VIs and REPs, the SCPs after continuum removal including the Depth (D1) and Area (A1, A2) parameters had the highest correlation with SPAD. (c) By taking into account the influence of leaf water content (LWC), the accuracy of SPAD estimation may be enhanced. The RF SPAD estimate model with SCPs of all absorption valley as inputs performed the best. The findings reveal on the mechanism of the influence of LWC on the accuracy of chlorophyll spectra estimation. It provides technical support for estimating SPAD in rice leaves with high accuracy and for developing chlorophyll tachistoscopes.

An Indirect Inversion Scheme for Retrieving Toxic Metal Concentrations Using Ground-Based Spectral Data in a Reclamation Coal Mine, China

A reclamation coal mine in Baishui County of Shaanxi Province, China, was selected as the study area to develop a fast survey method for estimating soil heavy metal concentrations using spectral data. A portable object spectrometer manufactured by Analytical Spectral Devices (ASD) was used to measure soil spectral reflectance, and an X-ray fluorescence device was utilized to obtain the content of heavy metals. The Savitzky-Golay filter, first derivative reflectance (FDR), second derivative reflectance (SDR), continuum removal (CR), and continuous wavelet transform (CWT) were used to transform the original reflectance (OR) spectra for enhancing the spectral characteristics, respectively. Furthermore, correlation analysis was introduced to determine the characteristic bands and the correlations of heavy metals. Partial least squares regression (PLSR), extremely learning machine (ELM), random forest (RF), and support vector machine (SVM) were implemented for quantitatively determining relations between heavy metal contents and spectral reflectance. The outcomes demonstrated that the spectral transformation methods could effectively capture the characteristic bands and increase the relations between heavy metal contents and spectral reflectance. The relation between Fe and Ni was close with a relatively high correlation coefficient (r = 0.741). RF combined with CWT at the decomposition scales of 9 demonstrated the best performance with the highest Rv2 (0.71) and the lowest RMSEv (1019.1 mg/kg) for inferring Fe content. Ni content was inferred based on the close relationship between Fe and Ni. The result of RF was better than other methods with the highest Rv2 (0.69) and the lowest RMSEv (1.94 mg/kg) for estimating Ni concentration. Therefore, the RF model was chosen for mapping Fe and Ni contents in the study area. The present study revealed that the indirect inversion methods using spectral data can be effectively used to predict heavy metal concentrations. The outcomes supply a new perspective for retrieving heavy metal content based on hyperspectral remotely sensed technology.

Calibration of NOMAD on ExoMars Trace Gas Orbiter: Part 3 - LNO validation and instrument stability

The LNO channel is one of the 3 instruments of the NOMAD suite of spectrometers onboard the ExoMars Trace Gas Orbiter currently orbiting Mars. Designed to operate primarily at nadir at very high spectral resolution in the 2.3 ​ μ m–3.8 ​ μ m spectral region, the instrument observes the martian atmosphere and surface daily since March 2018. To perform an accurate calibration of the instrument, in-flight measurement needs to be integrated to account for potential change during the cruise phase and later during the mission. In a companion article, Thomas et al. this issue, PSS, 2021 proposed a method based on the use of 6 observation sequences of the sun by LNO to derive a self-consistent approach, assuming temporal stability. Here we report an alternative concept of calibration, model the instrument using basic principle, based on the comparison between each solar spectrum observed and a reference solar spectrum. The method has the advantages to allows testing of the temporal stability but also instrumental effects such as temperature. It encompasses the main transfer functions of the instrument related to the grating and the AOTF and the instrument line shape using 9 free parameters which, once inverted, allow the observations to be fitted with an acceptable Root Mean Square Error (RMSE) around 0.5%. We propose to perform a continuum removal step to reduce the spurious instrumental effect, allowing to directly analyze the atmospheric lines. This methodology allows quantifying the instrumental sensitivity and its dependence on temperature and time. Once the temperature dependence was estimated and corrected, we found no sign of aging of the detector. Finally, the parameters are used to propose an efficient calibration procedure to convert the LNO-NOMAD data from ADU to radiances with spectral calibration and the instrument line shape. A comparison with the method reported in Thomas et al. this issue, PSS, 2021 showed that both calibrations are in agreement mostly within 3%. • The LNO channel is 1 of 3 spectrometers in NOMAD on ExoMars Trace Gas Orbiter. • The LNO channel observes Mars in limb, nadir and solar occultation modes. • This article describes an alternative method of calibration for the LNO channel data. • The method is ​based on a comparison between observation of the sun by the LNO channel and a synthetic solar spectrum. • Temperature and time dependence of the instrumental responsivity were tested. Nadir observations can be calibrated radiometrically.

VNIR estimation of heavy metals concentrations in suburban soil with multi-scale geographically weighted regression

Visible and near-infrared reflectance (VNIR) spectroscopy serves as a promising alternative to detect heavy metal contamination in suburban soils. However, the traditional regression method assumes a global rather than a non-stationary relationship between soil spectra and heavy metals. This study explored the spatially varying relationships between soil chromophores (e.g., SOM and TFe2O3) and heavy metals, and further investigated its implication for heavy metal estimation with VNIR. Specifically, we (i) explore the spatial patterns and covariation of soil chromophore and heavy metals, (ii) verify the feasibility of using multi-scale geographically weighted regression (MGWR) in the VNIR estimation of heavy metals, and (iii) investigate the potential mechanism underlying the models. Specifically, continuum removal and derivation were used to preprocess the reflectance spectra, and then a random frog algorithm was performed to select candidate feature wavelengths, followed by the removal of wavelengths with a correlation higher than 0.80 to avoid collinearity. Standard errors were calculated to further eliminate the variables with a high variance of coefficients. The remaining wavelengths were retained as inputs of the estimation models. The results demonstrated that MGWR outperformed its competitors, resulting in an increase in the coefficient of determination and a decrease in the residual sum of squares. Our study revealed the spatial heterogeneities of the relationships between heavy metals concentrations and soil spectra, and found that the spatial correlation between heavy metals and TFe2O3 (p < 0.01) was the main reason that Cr and Ni can be estimated by using VNIR. The MGWR method not only improves the capability of VNIR in the estimation of heavy metal concentrations, but promotes our understanding of the estimation mechanism.

Are topsoil spectra or soil-environmental factors better indicators for discrimination of soil classes?

Visible and near-infrared (Vis-NIR) spectroscopy (350–2500 nm) is often used to discriminate different soil classes. Discrimination variables included generally reflectance, topographic factors, soil texture, and climate. In the present study, more soil properties, soil organic matter, soil pH, bulk density and thickness, were considered to build soil classes model, and the most important discrimination variables were identified. Reasons for misclassification and challenges in elaborating the soil class map were analyzed. The topsoil spectra of 1432 soil samples of nine soil classes in northeast China were measured and converted into continuum removal (CR) curves, and principal component analysis (PCA) was applied to extract spectral information for reducing data redundancy. Other discrimination variables, soil-environmental factors, were acquired, including topographic factors, soil texture, climate, and four soil properties of each soil sample. The importance of soil–environmental factors and CR spectra was calculated for building models. Random forest (RF) models were trained to discriminate nine soil classes based on five kinds of variables, and they were the principal components (PCs) of reflectance, PCs of CR spectra, soil–environmental factors, PCs of CR spectra + soil–environmental factors, and optimum combinations between CR spectra and soil–environmental factors according to the importance. Optimum discrimination variables of each soil class were selected to build random forest models, and then the sum accuracy was calculated. Our study showed that: (1) it was challenging to acquire a high discrimination accuracy using reflectance and CR spectra for large-scale soil data; (2) soil–environmental factors were more suitable to discriminate between soil classes than reflectance and CR spectra in northeast China; (3) the order of importance of discrimination variables was soil texture > climate > pH > thickness > other soil–environmental factors; and (4) the RF model with optimum combinations between CR spectra and soil–environmental factors led to an accuracy of 81.5 % and a kappa coefficient of 0.77. We validated soil-environmental factors were better than topsoil spectra for discrimination of soil classes, and we found soil constituents in soil-environmental factors had high importance, and we successfully discriminated nine soil classes with 1432 soil samples by optimum combinations between topsoil spectra and soil–environmental factors at a large scale.

Application of a Fractional Order Differential to the Hyperspectral Inversion of Soil Iron Oxide

Iron oxide is the main form of iron present in soils, and its accumulation and migration activities reflect the leaching process and the degree of weathering development of the soil. Therefore, it is important to have information on the iron oxide content of soils. However, due to the overlapping characteristic spectra of iron oxide and organic matter in the visible-near infrared, appropriate spectral transformation methods are important. In this paper, we first used conventional spectral transformation (continuum removal, CR; standard normal variate, SNV; absorbance, log (1/R)), continuous wavelet transform (CWT), and fractional order differential (FOD) transform to process original spectra (OS). Secondly, competitive adaptive reweighted sampling (CARS) was used to extract characteristic wavelengths. Finally, two regression models (backpropagation neural network, BPNN; support vector regression (SVR) were used to predict the content of iron oxide. The results show that the FOD can significantly improve the correlation with iron oxide compared with the CR, SNV, log (1/R) and CWT; the baseline drift and overlapping peaks decrease with increasing the order of FOD; the CARS algorithm based on 50th averaging can select more stable characteristic wavelengths; the FOD achieves better results regardless of the modelling method, and the model based on 0.5-order differential has the best prediction performance (R2 = 0.851, RMSE = 5.497, RPIQ = 3.686).

Multi-temporal spectral reflectance of tropical savanna understorey species and implications for hyperspectral remote sensing

• Spectra of savanna understorey vegetation species over phenological seasons. • Separability of functional groups, and species, depended on season. • Informs species separation and timing for drone-based hyperspectral captures. • VNIR drone spectral range excludes foliar traits such as cellular water absorption. The relationship between species phenology and spectral separability is essential to determine the optimal remote sensing sampling period to maximise spectral separability of vegetation species. However, this relationship in many native grasses, introduced grasses and herbs, shrubs, and vine species in tropical savannas is unknown. We measured the in-situ hyperspectral response of monospecific vegetation stands of several understorey species representing different functional groups over phenological stages (throughout dry and wet seasons) using a non-imaging spectrometer. We present a spectral library of both native and introduced species of a tropical savanna environment. We analysed the data using continuum removal to highlight absorption features. Most understorey species displayed a photosynthetic spectral response with increased greenness at the end of the wet season that progressively declined as vegetation dried out. For some species, there were seasonally dependent differences in absorption features with spectral differences between the late wet and early dry season, and late dry and early wet seasons. We resampled the data to the spectral range typical of drone-mounted hyperspectral sensors (i.e., 150 bands between 400 and 1000 nm), which omits the water absorption features of the SWIR. These findings suggest an ideal sampling period for measuring outdoor canopy reflectance of understorey species, which will promote methodological improvements of hyperspectral data capture. The use of a VNIR only hyperspectral sensor will exclude the important regions of the spectrum for plant spectral identification including non-pigment bands for water, nitrogen and cellulose. The results have implications for determining the completeness of rehabilitation assessing reestablishment among indigenous species across mine sites under rehabilitation.

Hyperspectral Modeling of Soil Organic Matter Based on Characteristic Wavelength in East China

Soil organic matter (SOM) is a key index of soil fertility. Visible and near-infrared (VNIR, 350–2500 nm) reflectance spectroscopy is an effective method for modeling SOM content. Characteristic wavelength screening and spectral transformation may improve the performance of SOM prediction. This study aimed to explore the optimal combination of characteristic wavelength selection and spectral transformation for hyperspectral modeling of SOM. A total of 219 topsoil (0–20 cm) samples were collected from two soil types in the East China. VNIR reflectance spectra were measured in the laboratory. Firstly, after spectral transformation (inverse-log reflectance (LR), continuum removal (CR) and first-order derivative reflectance (FDR)) of VNIR spectra, characteristic wavelengths were selected by competitive adaptive reweighted sampling (CARS) and uninformative variables elimination (UVE) algorithms. Secondly, the SOM prediction models were constructed based on the partial least squares regression (PLSR), random forest (RF) and support vector regression (SVR) methods using the full spectra and selected wavelengths, respectively. Finally, optimal SOM prediction models were selected for two soil types. The results were as follows: (1) The CARS algorithm screened 40–125 characteristic wavelengths from the full spectra. The UVE algorithm screened 105–884 characteristic wavelengths. (2) For two soil types and full spectra, CARS and UVE improved the SOM modeling precision based on the PLSR and SVR methods. The coefficient of determination (R2) value in the validation of the CARS-PLSR (PLSR model combined with CARS) and CARS-SVR (SVR model combined CARS) models ranged from 0.69 to 0.95, and the relative percent deviation (RPD) value ranged from 1.74 to 4.31. Lin’s concordance correlation coefficient (LCCC) values ranged from 0.83 to 0.97. The UVE-PLSR and UVE-SVR models showed moderate precision. (3) The PLSR and SVR modeling accuracies of Paddy soil were better than those for Shajiang black soil. RF models performed worse for both soil types, with the R2 values of validation ranging from 0.22 to 0.68 and RPD values ranging from 1.01 to 1.60. (4) For Paddy soil, the optimal SOM prediction models (highest R2 and RPD, lowest root mean square error (RMSE)) were CR-CARS-PLSR (R2 and RMSE: 0.97 and 1.21 g/kg in calibration sets, 0.95 and 1.72 g/kg in validation sets, RPD: 4.31) and CR-CARS-SVR (R2 and RMSE: 0.98 and 1.04 g/kg in calibration sets, 0.91 and 2.24 g/kg in validation sets, RPD: 3.37). For Shajiang black soil, the optimal SOM prediction models were LR-CARS-PLSR (R2 and RMSE: 0.95 and 0.93 g/kg in calibration sets, 0.86 and 1.44 g/kg in validation sets, RPD: 2.62) and FDR-CARS-SVR (R2 and RMSE: 0.99 and 0.45 g/kg in calibration sets, 0.83 and 1.58 g/kg in validation sets, RPD: 2.38). The results suggested that the CARS algorithm combined CR and FDR can significantly improve the modeling accuracy of SOM content.

Evaluation of Vis-NIR preprocessing combined with PLS regression for estimation soil organic carbon, cation exchange capacity and clay from eastern Croatia

This study aimed to evaluate (i) the potential of visible-near-infrared (Vis-NIR) preprocessing techniques combined with a partial least square (PLS) regression to estimate soil organic carbon (SOC), the cation exchange capacity (CEC), and clay content and (ii) the effects of the wavelengths selection on the predictive capability of the PLS models. The study was conducted on a total of 132 topsoil samples (0–25 cm) collected on arable land in Eastern Croatia. Raw reflectance data and six commonly used spectral preprocessing, namely, multiplicative scatter correction (MSC), standard normal variate (SNV), de-trending (DT), continuum removal (CR) and Savitzky–Golay first and second derivative (SG1.der and SG2.der) were compared to identify the best for predicting selected soil properties. The PLS models calibrated with the SG1.der preprocessing achieved the best predictive performance for all soil properties compared to other methods used. The R2, RPD and RMSE of the SG1.der -PLS model in the validation mode for SOC, CEC and clay were 0.81, 2.12, 2.57 g C kg−1; 0.78, 1.95, 1.80 cmol(+) kg−1 and 0.83, 2.20, 37.2 g kg−1, respectively. The lowest predictability had PLS models that used soil raw reflectance data. These results indicate that the appropriate choice of Vis-NIR preprocessing can optimize the PLS calibration performance for accurate predicting SOC, CEC and clay in arable soils of the study area. The selection of informative wavelengths (IW) obtained using the PLS regression β-coefficient approach resulted in a highly reduced number of wavelengths in the IW-PLS models. Informative wavelengths in the IW-PLS model for SOC lay in the full Vis-NIR spectrum but mainly in the near-infrared (NIR), whereas for the CEC and clay in the NIR spectrum. In addition, IW-PLS models have a lower number of principal components and almost identical prediction capability relative to models based on full spectra (FS–PLS). Our results suggest that the Vis-NIR spectroscopy is a powerful tool to estimate soil properties rapidly, non-destructively and inexpensively applicable in some real situations e.g., monitoring and updating information on soil properties for precise soil management and soil quality assessment.

Monitoring Leaf Nitrogen Accumulation With Optimized Spectral Index in Winter Wheat Under Different Irrigation Regimes.

Rapid and non-destructive estimation of leaf nitrogen accumulation (LNA) is essential to field nitrogen management. Currently, many vegetation indices have been used for indicating nitrogen status. Few studies systematically analyzed the performance of vegetation indices of winter wheat in estimating LNA under different irrigation regimes. This study aimed to develop a new spectral index for LNA estimation. In this study, 2 years of field experiments with different irrigation regimes were conducted from 2015 to 2017. The original reflectance (OR) and three transformed spectra [e.g., the first derivative reflectance (FDR), logarithm of the reciprocal of the spectra (Log(1/R)), and continuum removal (CR)] were used to calculate two- and three-band spectral indices. Correlation analyses and univariate linear and non-linear regression between transformed-based spectral indices and LNA were performed. The performance of the optimal spectral index was evaluated with classical vegetation index. The results showed that FDR was the most stable transformation method, which can effectively enhance the relationships to LNA and improve prediction performance. With a linear relationship with LNA, FDR-based three-band spectral index 1 (FDR-TBI1) (451, 706, 688) generated the best performance with coefficient of determination (R2) of 0.73 and 0.79, the root mean square error (RMSE) of 1.267 and 1.266 g/m2, and the ratio of performance to interquartile distance (RPIQ) of 2.84 and 2.71 in calibration and validation datasets, respectively. The optimized spectral index [FDR-TBI1 (451, 706, 688)] is more effective and might be recommended as an indicator for estimating winter wheat LNA under different irrigation regimes.

Suitability of different multivariate analysis methods for monitoring leaf N accumulation in winter wheat using in situ hyperspectral data

Leaf nitrogen accumulation (LNA) is an important indicator of crop growth, and the real-time and nondestructive estimation of LNA by hyperspectral remote sensing is crucial for crop N management. In this study, five field experiments were conducted over four successive years at three ecological sites (Xinyang, Zhengzhou and Zhoukou) in Henan, China, incorporating different N amounts and wheat varieties. The canopy spectral data were obtained by a ground-based spectrometer, and the original spectral reflectance (R) was transformed using continuum removal (CR), first-derivative reflectance (FDR) and inverse-log reflectance (log (1/R)) methods. The effective bands of the above four transformed reflectance data were selected by the variable importance in projection (VIP) scores of the partial least square regression method (PLSR). Then, four multivariate regression methods (PLSR, stepwise multiple linear regression (SMLR), extreme learning machine (ELM) and support vector machine (SVM)) were used to develop an optimal forecasting model for LNA in winter wheat. The results show that FDR spectral data had the best comprehensive performance and were a better method for spectral data preprocessing. In the comparison of different nonparametric algorithms, the SVM model performed the best using the full wavelengths, in which R2cal/val and RMSEcal/val were 0.927–0.963/0.899–0.925 and 0.725–1.066/1.161–1.331 g m−2, respectively. The characteristic bands of FDR were mainly distributed in the visible light and red edge regions (445, 501, 546, 683, 747 and 771 nm, respectively), and these effective bands were consistent under different field conditions. The SVM regression method using the effective wavelengths of FDR exhibited excellent performance for LNA prediction with R2val = 0.875 and RMSEval = 1.294. In addition, the FDR-PLSR-SVM model produced satisfactory inversion results under different field conditions. The accuracy of the filling period was better than that of the reviving-anthesis growth period, which might be due to diminished N dilution and stable canopy structure during the reproductive stage. R2val and RMSEval of the FDR-PLSR-SVM model during the filling period were 0.914 and 1.156 g m−2 in Xinyang, 0.825 and 1.542 g m−2 in Zhengzhou and 0.796 and 1.902 g m−2 in Zhoukou, respectively. Considering the advantages of simple and efficient computer calculations, the FDR-PLSR-SVM algorithm is strongly recommended to predict the LNA in winter wheat. This study provides a reference basis for the selection of effective wavelengths for N nutrition diagnosis by unmanned aerial vehicles and satellite remote sensing in the future.

HYPERSPECTRAL INVERSION OF SOLUBLE SALT CONTENT IN MURAL PAINTING

Abstract. Mural painting is one of the carriers expressing history and culture. Due to the natural and anthropogenic factors, the salt in mural painting and environment is enriched in the surface layer with temperature change. It will induce irreversible diseases such as crispy alkali, which is not conducive to the survival of mural painting in the present. An efficient and non-destructive method to detect salt in murals is of great importance. Therefore, we proposed a method to predict the soluble salt content of mural paintings based on hyperspectral techniques. First, simulated samples with different salt concentrations were measured by a special spectroradiometer to acquire their spectra. Next, breakpoint correction and average smoothing preprocessing are performed and the data set is divided. Then, the spectra were enhanced by continuum removal (CR) and the logarithm of reciprocal (LR). The salt concentration was correlated with the spectra to extract 10 characteristic bands. Finally, the salt content prediction model was established by simple linear regression (SLR) and multiple linear regression (MLR). The accuracy of the model was evaluated with the coefficient of determination R2, root mean square error RMSE, and relative percent deviation RPD. The experimental results show that the best inversion fit is based on the combination of the CR-MLR model at the strong correlation bands of 420nm, 584nm, and 2379nm (Calibration Set R2 = 0.846, RMSE = 0.138, and RPD = 3.240). This paper provides a new technical means for the non-destructive detection of salt content in murals.

Combining different pre-processing and multivariate methods for prediction of soil organic matter by near infrared spectroscopy (NIRS) in Southern Brazil

In Brazil, most guidelines for soil nitrogen (N) fertilization are based on the soil organic matter (SOM) content, which is one of the most laborious and expensive parameters in routine soil analysis. Recently, several studies have been made to generate and evaluate alternative methods to estimate SOM content, such as those based on near infrared spectroscopy (NIRS). However, the accuracy of NIRS is highly affected by spectral pre-processing techniques and multivariate methods. Moreover, the models are highly site-specific, indicating the need to develop regional calibrations. The objective of this study was to evaluate the effect of pre-processing and multivariate methods in SOM prediction using NIRS. A total of 2388 soil samples from Southern Brazil were analyzed for SOM content by carbon oxidation with a sulfochromic solution, followed by spectrophotometric colorimeter determination (modified Walkley-Black method) and by a NIR-spectrometer (1200–2400 nm). Seven pre-processing techniques were tested, including: Savitzki-Golay derivative (SGD), continuum removal (CR), detrend (DET), binning (BIN), smoothing (SMO), and the standard normal variate (SNV). The pre-processing techniques were combined with four multivariate models including multiple linear regression (MLR), partial least squares regression (PLSR), principal component regression (PCR), support vector machine (SVM), random forest (RF) and gaussian process regression (GPR). The performance of the predictions was evaluated by the coefficient of determination (R2) and the root mean squared error (RMSE). Moreover, SOM predictions obtained from the models were compared with the wet digestion method in order to evaluate the capability to correctly classify the SOM content for N fertilization. Overall, comparing the multivariate models, the best predictions according to R2 were found for SVM > GPR > PLSR > MLR > PCR > RF, in decreasing order. The best prediction was obtained with the combination of SVM model and SNV spectral pre-processing (R2 = 0.70, RMSE = 0.44, RPIQ = 2.26). For SVM and GPR, only DET, SNV, BIN and CR improved predictions when compared to raw spectra. When using the linear model MLR, the prediction of SOM was improved only with SNV, SMO and DET, compared to raw spectra. For the PLSR model, the prediction of SOM was improved with DET and SNV. Results estimated by NIRS reached 85% accuracy when compared to values of SOM obtained by the modified Walkley-Black method.

Combining Different Transformations of Ground Hyperspectral Data with Unmanned Aerial Vehicle (UAV) Images for Anthocyanin Estimation in Tree Peony Leaves

To explore rapid anthocyanin (Anth) detection technology based on remote sensing (RS) in tree peony leaves, we considered 30 species of tree peonies located in Shaanxi Province, China. We used an SVC HR~1024i portable ground object spectrometer and mini-unmanned aerial vehicle (UAV)-borne RS systems to obtain hyperspectral (HS) reflectance and images of canopy leaves. First, we performed principal component analysis (PCA), first-order differential (FD), and continuum removal (CR) transformations on the original ground-based spectra; commonly used spectral parameters were implemented to estimate Anth content using multiple stepwise regression (MSR), partial least squares (PLS), back-propagation neural network (BPNN), and random forest (RF) models. The spectral transformation highlighted the characteristics of spectral curves and improved the relationship between spectral reflectance and Anth, and the RF model based on the FD spectrum portrayed the best estimation accuracy (R2c = 0.91; R2v = 0.51). Then, the RGB (red-green-blue) gray vegetation index (VI) and the texture parameters were constructed using UAV images, and an Anth estimation model was constructed using UAV parameters. Finally, the UAV image was fused with the ground spectral data, and a multisource RS model of Anth estimation was constructed, based on PCA + UAV, FD + UAV, and CR + UAV, using MSR, PLS, BPNN, and RF methods. The RF model based on FD+UAV portrayed the best modeling and verification effect (R2c = 0.93; R2v = 0.76); compared with the FD-RF model, R2c increased only slightly, but R2v increased greatly from 0.51 to 0.76, indicating improved modeling and testing accuracy. The optimal spectral transformation for the Anth estimation of tree peony leaves was obtained, and a high-precision Anth multisource RS model was constructed. Our results can be used for the selection of ground-based HS transformation in future plant Anth estimation, and as a theoretical basis for plant growth monitoring based on ground and UAV multisource RS.

Visible-NIR spectral characteristics and grade inversion model of skarn-type iron ore

Iron ore grade determination is a key step in mining production, and its efficiency has a significant impact on mining efficiency and profitability. At present, the most commonly used method for ore grade determination is chemical analysis, which not only is expensive but also has a long assay cycle and lag effect relative to the ore allocation process and cannot effectively reduce the loss depletion rate of ore mining. Therefore, in situ iron ore grade data determination technology based on visible and near-infrared hyperspectral data analysis is an effective way to solve the above problems, but the poor quality of raw hyperspectral data and low accuracy of the ore grade inversion model are still one of the bottlenecks in the application of this technology. In this paper, 225 ore samples were collected from Hongling skarn-type iron ore in Chifeng, Inner Mongolia Autonomous Region, China. The visible and near-infrared hyperspectral data of the sample set and the grade data obtained from chemical analysis were used as data sources. The raw spectral data were first smoothed, and then the smoothed spectral data were processed using two algorithms, continuum removal (CR) and standard normal variate transform (SNV). Then, two dimensionality reduction algorithms, principal component analysis (PCA) and genetic algorithm (GA), were used to reduce the dimensionality of the spectral data before and after preprocessing, and the data sources were obtained after the four different preprocessing combination algorithms were processed. Finally, several quantitative inversion models were established based on the random forest algorithm and the backpropagation neural network (BPNN) algorithm. The stability, accuracy and credibility of the models were evaluated by four metrics: the coefficient of determination (R2), root mean square error (RMSE), residual prediction deviation(RPD) and mean relative error (MRE). The results showed that the quantitative inversion model (SNV-PCA-BPNN) established based on the BP neural network algorithm using the data processed by SNV and PCA for dimensionality reduction was the best, with an R2 of 0.99, an RMSE of 0.0048, an MRE of 1.10% and a RPD of 37.70, providing an effective method for the high-precision inversion of iron ore grades.

Continuum removal and self-absorption correction incalibration-free laser-induced breakdownspectroscopy for efficient elemental analysis

Even though laser-induced breakdown spectroscopy (LIBS) has emerged as a powerful analytical technique, the broad continuum emission and self-absorption effects in laser-produced plasmas (LPP) limit the accuracy of the LIBS technique in multi-elemental compositional analysis. In this work, we developed an algorithm to detect and remove the broad continuum emission, which usually originates from free-free and free-bound transitions. To eliminate the continuum, the segment-wise background correction method (using identified continuum parts of varied range) was used. The spectral interference of lines is more likely to be found in LIBS spectra, especially with low-resolution spectrometers. A Lorentzian curve fitting method was used to resolve closely spaced emission lines. The ‘internal reference self-absorption correction (IRSAC)’ method was introduced to correct the reabsorption effects in LPPs. When these methods are applied to LIBS data of bronze alloy, more accurate quantitative findings are obtained, with a major component accuracy error of less than 10% when compared to its reference abundance.

Using soil library hyperspectral reflectance and machine learning to predict soil organic carbon: Assessing potential of airborne and spaceborne optical soil sensing

Soil organic carbon (SOC) is a key variable to determine soil functioning, ecosystem services, and global carbon cycles. Spectroscopy, particularly optical hyperspectral reflectance coupled with machine learning, can provide rapid, efficient, and cost-effective quantification of SOC. However, how to exploit soil hyperspectral reflectance to predict SOC concentration, and the potential performance of airborne and satellite data for predicting surface SOC at large scales remain relatively underknown. This study utilized a continental-scale soil laboratory spectral library (37,540 full-pedon 350–2500 nm reflectance spectra with SOC concentration of 0–780 g·kg−1 across the US) to thoroughly evaluate seven machine learning algorithms including Partial-Least Squares Regression (PLSR), Random Forest (RF), K-Nearest Neighbors (KNN), Ridge, Artificial Neural Networks (ANN), Convolutional Neural Networks (CNN), and Long Short-Term Memory (LSTM) along with four preprocessed spectra, i.e. original, vector normalization, continuum removal, and first-order derivative, to quantify SOC concentration. Furthermore, by using the coupled soil-vegetation-atmosphere radiative transfer model, we simulated twelve airborne and spaceborne hyper/multi-spectral remote sensing data from surface bare soil laboratory spectra to evaluate their potential for estimating SOC concentration of surface bare soils. Results show that LSTM achieved best predictive performance of quantifying SOC concentration for the whole data sets (R2 = 0.96, RMSE = 30.81 g·kg−1), mineral soils (SOC ≤ 120 g·kg−1, R2 = 0.71, RMSE = 10.60 g·kg−1), and organic soils (SOC > 120 g·kg−1, R2 = 0.78, RMSE = 62.31 g·kg−1). Spectral data preprocessing, particularly the first-order derivative, improved the performance of PLSR, RF, Ridge, KNN, and ANN, but not LSTM or CNN. We found that the SOC models of mineral and organic soils should be distinguished given their distinct spectral signatures. Finally, we identified that the shortwave infrared is vital for airborne and spaceborne hyperspectral sensors to monitor surface SOC. This study highlights the high accuracy of LSTM with hyperspectral/multispectral data to mitigate a certain level of noise (soil moisture <0.4 m3·m−3, green leaf area < 0.3 m2·m−2, plant residue <0.4 m2·m−2) for quantifying surface SOC concentration. Forthcoming satellite hyperspectral missions like Surface Biology and Geology (SBG) have a high potential for future global soil carbon monitoring, while high-resolution satellite multispectral fusion data can be an alternative.

Prediction of soil organic matter using different soil classification hierarchical level stratification strategies and spectral characteristic parameters

Whether a finer soil classification hierarchical stratification strategy and the spectral characteristic parameters (SCPs) that describe the shape of the spectral curve can be used to improve the prediction accuracy of soil organic matter should be clarified. We measured the visible, near-infrared and shortwave infrared (VIS-NIR-SWIR, 400 – 2500 nm) spectral reflectance of 322 topsoil samples. The spectral reflectance was converted to continuum removal curves, and then, SCPs were extracted based on the curves. According to the results of the Second National Soil Survey of China, the samples were divided into 4 great groups or 8 genus, and a variety of stratification strategies were constructed based on great group (GR-S), genus (GE-S), spectral similarity (SS-S) and decision tree model (DT-S). A local random forest model was established to evaluate the performance of different stratification strategies and input variables. Our results are described as follows: (1) In different stratification strategies, the SOM prediction model based on DT-S exhibits the highest accuracy, followed by the SOM prediction models based on GE-S and SS-S; the SOM prediction model based on GR-S exhibits the lowest accuracy; (2) among the different input variables, the root mean squared error (RMSE) and coefficient of determination (R2) of the best SOM model predicted by SCPs are 5.18 g kg−1 and 0.89, respectively. Compared with the original reflectance based on the nonstratified strategy, the RMSE decreases by 4.88 g kg−1 and R2 increases by 0.32. The study results highlight the advantages of refining the soil hierarchy, which is helpful for identifying the differences in soils at the regional scale and analysing the relationship between stratification results and the characteristics of the soil environment to obtain a highly accurate prediction model.

Reflectance spectroscopy for assessing heavy metal pollution indices in mangrove sediments using XGBoost method and physicochemical properties

Single and composite pollution indices (SPI and CPI) of heavy metals are calculated with background heavy metal concentrations. They are efficient evaluation indicators of sediment pollution levels and could be directly compared among different regions and habitat types, however, they are rarely investigated in mangrove sediment with visible and near-infrared spectroscopy (VNIRS). With mangrove sediment samples collected from five regions and four habitat types, this study aimed to develop estimation models of heavy metal (Cr, Ni, Cu, Zn and Pb) pollution indices using XGBoost (extreme gradient boosting) method, and further to explore the effect of physicochemical properties (Fe, organic carbon (OC), clay and salinity) on the estimation accuracy. Kruskal-Walls test showed that regions and habitat types were significant factors affecting the six pollution indices. Compared with first derivative, second derivative and continuum removal reflectance, standard normal variate (SNV) reflectance was more accurate for estimating SPIs and CPI using XGBoost model. The XGBoost models with sensitive wavelengths of SNV reflectance outperformed that with whole spectrum (410 bands) in estimating the six pollution indices, and the ranking of estimation accuracy was CPI > SPICu > SPIZn > SPIPb > SPINi > SPICr. Compared with the XGBoost model with sensitive wavelengths alone, the model with additional physicochemical properties could increase R2 and RPD (the ratio of standard deviation to RMSE of validation in the test subset) values by 0.795–7.282% and 1.199–8.780%, and decrease RMSE values by 4.101–26.866% in estimating SPICr, SPINi and SPICu. Moreover, Fe and OC had greater effect on the estimation accuracy than clay and salinity. We conclude the integration of XGBoost and physicochemical properties holds potentials for accurately assessing heavy metal pollution indices in mangrove sediment with VNIRS, which can provide methodological basis for evaluating environment quality and ecological risk in mangrove ecosystem at the plot or landscape scale.

Estimating the Forage Neutral Detergent Fiber Content of Alpine Grassland in the Tibetan Plateau Using Hyperspectral Data and Machine Learning Algorithms

The neutral detergent fiber (NDF) content is a key factor in the forage palatability and livestock digestibility of alpine grassland. Traditional methods of measuring NDF are time-consuming and laborious, while multispectral remote sensing often has difficulty capturing subtle changes due to its broad spectral channels. In this study, we evaluate NDF using hyperspectral data and environmental factors (EFs) during 2016-2019. The artificial neural network (ANN), support vector machine (SVM), and random forest (RF) are used to construct models. The results show that: 1) compared with the correlation analysis, the least absolute shrinkage and selection operator (LASSO) regression can more effectively select the important variables for estimating NDF and achieve significant dimension reduction; 2) estimation accuracy of the models (coefficient of determination (R²) between 0.44 and 0.51, relative root mean square error (RMSE) between 4.31% and 4.61%) based on first derivative (FD) spectra is relatively better than the models based on original spectra (OR), log transformation (Log), and continuum removal (CR) spectra (R² between 0.35 and 0.46, RMSE between 4.39% and 4.92%). The RF model based on FD-Log-vegetation indices (VIs)-EFs is the best model (R² = 0.62, RMSE = 3.98%); and 3) compared with the EFs, hyperspectral feature plays a pivotal role in estimating NDF, with a total contribution of 81.66%, and red-edge and shortwave infrared (SWIR) regions are significant for estimating NDF. In general, the combination of hyperspectral data with the EFs significantly improves the accuracy of estimating NDF. This study provides important guidance for future efforts on estimating NDF of natural grassland using remote sensing and machine learning algorithms.