Evaluation of Informative Bands Used in Different PLS Regressions for Estimating Leaf Biochemical Contents from Hyperspectral Reflectance

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.

Temporal instability of partial least squares regressions for estimating leaf photosynthetic traits from hyperspectral information

Partial least squares regression (PLSR) is applied increasingly often to predict plant photosynthesis from reflectance spectra. While its applicability across different areas has been examined in previous studies, its stability across time has yet to be evaluated. In this study, we assessed a series of PLSR models built upon three different band selection approaches (iterative stepwise, genetic algorithm, and uninformative variable elimination), in combination with different spectral transforms (original and first-order derivative spectra), for their stabilities in predicting the maximum carboxylation rate (Vcmax) and maximum electron transport rate (Jmax) from hyperspectral reflectance spectra at different temporal scales (seasonal and interannual). The results showed that both photosynthetic parameters can be estimated from leaf hyperspectral reflectance with moderate to good accuracy across different growing stages (R² = 0.45-0.84) and years (R² = 0.37-0.97). We further found that the iterative stepwise selection of informative bands when building PLSR models could greatly improve its predictive capacity compared with that of other PLSR models, especially those based on first-order derivative spectra. However, the selected bands of the models for both photosynthetic parameters were, unfortunately not consistent. Furthermore, we could not have identified any model with fixed spectra performed consistently across different seasonal stages and across different years. However, the blue spectral regions were popularly selected throughout the growing stages and in different years. The results demonstrate that leaf spectra-trait estimation using PLSR models varies with time and thus cast doubt over the use of a specific PLSR model to infer leaf traits across different temporal-spatial contexts. The development of a general applicable PLSR model is still in the works.

Quantitative Analysis of Routine Chemical Constituents of Tobacco Based on Thermogravimetric Analysis

As the most basic indexes to evaluate the quality of tobacco, the contents of routine chemical constituents in tobacco are mainly detected by continuous-flow analysis at present. However, this method suffers from complex operation, time consumption, and environmental pollution. Thus, it is necessary to establish a rapid accurate detection method. Herein, different from the ongoing research studies that mainly chose near-infrared spectroscopy as the information source for quantitative analysis of chemical components in tobacco, we proposed for the first time to use the thermogravimetric (TG) curve to characterize the chemical composition of tobacco. The quantitative analysis models of six routine chemical constituents in tobacco, including total sugar, reducing sugar, total nitrogen, total alkaloids, chlorine, and potassium, were established by the combination of TG curve and partial least squares algorithm. The accuracy of the model was confirmed by the value of root mean square error for prediction. The models can be used for the rapid accurate analysis of compound contents. Moreover, we performed an in-depth analysis of the chemical mechanism revealed by the result of the quantitative model, namely, the regression coefficient, which reflected the correlation degree between the six chemicals and different stages of the tobacco thermal decomposition process.

Selecting informative bands for partial least squares regressions improves their goodness-of-fits to estimate leaf photosynthetic parameters from hyperspectral data

The plant photosynthetic capacity determines the photosynthetic rates of the terrestrial biosphere. Timely approaches to obtain the spatiotemporal variations of the photosynthetic parameters are urgently needed to grasp the gas exchange rhythms of the terrestrial biosphere. While partial least squares regression (PLSR) is a promising way to predict the photosynthetic parameters maximum carboxylation rate (V_cmax) and maximum electron transport rate (J_max) rapidly and non-destructively from hyperspectral data, the approach, however, faces a high risk of overfitting and remains a high hurdle for applications. In this study, we propose to incorporate proper band selection techniques for PLSR analysis to refine the goodness-of-fit (GoF) in estimating V_cmax and J_max. Different band selection procedures coupled with different hyperspectral forms (reflectance, apparent absorption, as well as derivatives) were examined. Our results demonstrate that the GoFs of PLSR models could be greatly improved by combining proper band selection methods (especially the iterative stepwise elimination approach) rather than using full bands as commonly done with PLSR. The results also show that the 1st order derivative spectra had a balance between accuracy (R² = 0.80 for V_cmax, and 0.94 for J_max) and denoising (when a Gaussian noise was added to each leaf reflectance spectrum at each wavelength with a standard deviation of 1%) on retrieving photosynthetic parameters from hyperspectral data. Our results clearly illustrate the advantage of using the band selection approach for PLSR dimensionality reduction and model optimization, highlighting the superiority of using derivative spectra for V_cmax and J_max estimations, which should provide valuable insights for retrieving photosynthetic parameters from hyperspectral remotely sensed data.

Physicochemical indicators coupled with multivariate analysis for comprehensive evaluation of matcha sensory quality

The sensory quality of matcha is a pivotal factor in determining consumer acceptance. However, the human sensory panel test is difficult to popularize by virtue of professional requirements and inability to evaluate large samples. The analysis showed that physicochemical indicators of matcha were significantly related to sensory quality. Hence, principal component analysis (PCA) based on selected key physicochemical indicators was proposed to evaluate the sensory quality of matcha in this research. The eight key indicators were selected from twenty-four physicochemical indicators based on least absolute shrinkage and selection operator (LASSO) for the establishment of the PCA comprehensive evaluation model. The results demonstrated that the PCA comprehensive evaluation model achieved superior performance, with -0.895 r_c (correlation coefficient in calibration set) and -0.883 r_p (correlation coefficient in prediction set) for overall sensory quality. This work demonstrated that LASSO-PCA comprehensive evaluation as an objective protocol has great potential in predicting matcha sensory quality.

Rapid Detection of Pomelo Fruit Quality Using Near-Infrared Hyperspectral Imaging Combined With Chemometric Methods

Pomelo is an important agricultural product in southern China. Near-infrared hyperspectral imaging (NIRHI) technology is applied to the rapid detection of pomelo fruit quality. Advanced chemometric methods have been investigated for the optimization of the NIRHI spectral calibration model. The partial least squares (PLS) method is improved for non-linear regression by combining it with the kernel Gaussian radial basis function (RBF). In this study, the core parameters of the PLS latent variables and the RBF kernel width were designed for grid search selection to observe the minimum prediction error and a relatively high correlation coefficient. A deep learning architecture was proposed for the parametric scaling optimization of the RBF-PLS modeling process for NIRHI data in the spectral dimension. The RBF-PLS models were established for the quantitative prediction of the sugar (SU), vitamin C (VC), and organic acid (OA) contents in pomelo samples. Experimental results showed that the proposed RBF-PLS method performed well in the parameter deep search progress for the prediction of the target contents. The predictive errors for model training were 1.076% for SU, 41.381 mg/kg for VC, and 1.136 g/kg for OA, which were under 15% of their reference chemical measurements. The corresponding model testing results were acceptably good. Therefore, the NIRHI technology combined with the study of chemometric methods is applicable for the rapid quantitative detection of pomelo fruit quality, and the proposed algorithmic framework may be promoted for the detection of other agricultural products.

The Cooling and Humidifying Effects and the Thresholds of Plant Community Structure Parameters in Urban Aggregated Green Infrastructure

The cooling and humidifying effects of urban aggregated green infrastructure can provide essential services for city ecosystems, regulating microclimates or mitigating the urban heat island effect. However, the optimal thresholds of plant community structure parameters for maximizing the associated cooling and humidifying effects remain unclear. In this paper, we use the method of dummy variable regression to measure plant communities in an urban aggregated green infrastructure. By examining the relationships between the cooling and humidifying effects and plant community structure parameters (i.e., canopy density, porosity, and vegetation type), we introduce optimal thresholds for the parameters. We find that canopy density has a significantly positive correlation with both cooling and humidifying effects, while porosity has a positive correlation with cooling and a negative one with humidifying. Different vegetation types have distinct influences on cooling and humidifying effects. When the canopy density is between 0.81 and 0.85 and the porosity is between 0.31 and 0.35, the cooling and humidifying effects of the plant communities reach their peak. Additionally, the greening coverage rate and spatial types of urban aggregated green infrastructure have influences on cooling and humidifying effects. The findings can help us to better understand the relationships between plant community structure parameters and their temperature regulation functioning for urban aggregated green infrastructure. This study provides guidelines and theoretical references for the plant configuration of future urban green spaces.

Improving characteristic band selection in leaf biochemical property estimation considering interrelations among biochemical parameters based on the PROSPECT-D model

At present, many studies have mainly focused on analyzing the sensitivity and correlation to select characteristic bands. However, the interrelations between biochemical parameters were ignored, which may significantly influence the accuracy of biochemical concentration retrieval. The study aims to propose a new band selection method and to focus on the improving magnitude of characteristic band combination in leaf trait estimation when taking interrelations among different traits into consideration. Thus, in this study, firstly a ranking- and searching-based method considering the sensitivity and correlation between different wavelengths, which can enhance the reliability of spectral band selection, was proposed to select a subset of characteristic bands for leaf structure index and five leaf biochemical parameters (including chlorophyll (Chl), carotenoid (Car), leaf dry matter per area (LMA), equivalent water thickness (EWT), and anthocyanin (Anth)) based on the PROSPECT-D model. These characteristic bands were then validated based on a physical model for retrieving five biochemical properties using one synthetic dataset and six experimental datasets on leaf-level spectra. Secondly, and more innovatively, to explore interrelations among different biochemical parameters, trait-trait band combinations were adopted to retrieve and analyze how the five biochemical participants above affected each other. The results demonstrated that the combination of LMA (809 and 2278 nm), EWT (1386, 1414, and 1894 nm) is more beneficial in LMA and EWT estimation than respective retrieval: LMA-EWT band combination retrieval improves R² by 0.5782 and 0.1824 in two datasets, respectively, compared with solely LMA characteristic bands retrieval. What's more, the accuracy of Chl, EWT, Car, and Anth estimation can be also improved when considering interrelations between biochemical parameters. The experimental results show that the ranking- and searching-based method is an effective and efficient way to select a set of spectral bands related to the foliar information about plant traits, and trait-trait combinations, which focus on exploring latent interrelations between leaf traits, are useful in furthering improve retrieval accuracy. This research will provide notably advanced insight into identifying the spectral responses of biochemical traits in foliage and canopies.

Data-Driven Methods for the Estimation of Leaf Water and Dry Matter Content: Performances, Potential and Limitations

Leaf equivalent water thickness (EWT) and dry matter content (expressed as leaf mass per area (LMA)) are two critical traits for vegetation function monitoring, crop yield estimation, and precise agriculture management. Data-driven methods are widely used for remote sensing of leaf EWT and LMA because of their simplicity, satisfactory accuracy, and computation efficiency, such as the vegetation indices (VI)-based and machine learning (ML)-based methods. However, most of the data-driven methods are utilized at the canopy level, comparison of the performances of the data-driven methods at the leaf level has not been well documented. Moreover, the ML-based data-driven methods generally adopt leaf optical properties directly as their inputs, which may subsequently decrease their ability in remote sensing of leaf biochemical constituents. Performances of the ML-based methods cooperating with VI are rarely evaluated. Using the independent LOPEX and ANGERS datasets, we compared the performances of three data-driven methods: VI-based, ML-reflectance-based, and ML-VI-based methods, for the estimation of leaf EWT and LMA. Three sampling strategies were also utilized for evaluation of the generalization of these data-driven methods. Our results evidenced that ML-VI-based methods were the most accurate among these data-driven methods. Compared to the ML-reflectance-based and VI-based methods, the ML-VI-based model with support vector regression overall reduced errors by 5.7% (41.5%) and 1.8% (12.4%) for the estimation of leaf EWT (LMA), respectively. The ML-VI-based model inherits advantages of vegetation indices and ML techniques, which made it sensitive to changes of leaf biochemical constituents and capable of solving nonlinear tasks. It is thus recommended for the estimation of EWT and LMA at the leaf level. Moreover, its performance can further be enhanced by improving its generalization ability, such as adopting techniques on the selection of better wavelengths and definition of new vegetation indices. These results thus provided a prior knowledge of the data-driven methods and can be helpful for future studies on the remote sensing of leaf biochemical constituents.

Estimation of Soil Arsenic Content with Hyperspectral Remote Sensing

With the continuous application of arsenic-containing chemicals, arsenic pollution in soil has become a serious problem worldwide. The detection of arsenic pollution in soil is of great significance to the protection and restoration of soil. Hyperspectral remote sensing is able to effectively monitor heavy metal pollution in soil. However, due to the possible complex nonlinear relationship between soil arsenic (As) content and the spectrum and data redundancy, an estimation model with high efficiency and accuracy is urgently needed. In response to this situation, 62 samples and 27 samples were collected in Daye and Honghu, Hubei Province, respectively. Spectral measurement and physical and chemical analysis were performed in the laboratory to obtain the As content and spectral reflectance. After the continuum removal (CR) was performed, the stable competitive adaptive reweighting sampling algorithm coupled the successive projections algorithm (sCARS-SPA) was used for characteristic band selection, which effectively solves the problem of data redundancy and collinearity. Partial least squares regression (PLSR), radial basis function neural network (RBFNN), and shuffled frog leaping algorithm optimization of the RBFNN (SFLA-RBFNN) were established in the characteristic wavelengths to predict soil As content. These results show that the sCARS-SPA-SFLA-RBFNN model has the best universality and high prediction accuracy in different land-use types, which is a scientific and effective method for estimating the soil As content.

New method for rapid identification and quantification of fungal biomass using ergosterol autofluorescence

This research reports on the development of a method to identify and quantify fungal biomass based on ergosterol autofluorescence using excitation-emission matrix (EEM) measurements. In the first stage of this work, several ergosterol extraction methods were evaluated by APCI-MS, where the ultrasound-assisted procedure showed the best results. Following an experimental design, various quantities of the dried mycelium of the fungus Schizophyllum commune were mixed with the starchy solid residue (BBR) from the babassu (Orbignya sp.) oil industry, and these samples were subjected to several ergosterol extraction methods. The EEM spectral data of the samples were subjected to Principal Component Analysis (PCA), which showed the possibility to qualitatively evaluate the presence of ergosterol in the samples by ergosterol autofluorescence without the addition of any reagent. In order to assess the feasibility of quantifying fungal biomass using ergosterol autofluorescence, the EEM spectral data and known amounts of fungal biomass were modeled using partial least squares (PLS) regression and a procedure of backward selection of predictors (AutoPLS) was applied to select the Excitation-Emission wavelength pairs that provide the lowest prediction error. The results revealed that the amount of fungal biomass in samples containing interfering substances (BBR) can be accurately predicted with R²CV = 0.939, R²P = 0.936, RPDcv = 4.07, RPDp = 4.06, RMSECV = 0.0731 and RMSEP = 0.0797. In order to obtain an easy-to-understand equation that expresses the relationship between fungal biomass and fluorescence intensity, multiple linear regression (MLR) was applied to the VIP variables selected by the AutoPLS method. The MLR model selected only 2 variables and showed a very good performance, with R²CV = 0.862, R²P = 0.809, RPDcv = 2.18, RPDp = 2.35, RMSECV = 0.137 and RMSEP = 0.138. This study demonstrated that ergosterol autofluorescence can be successfully used to quantify fungal biomass even when mixed with agroindustrial residues, in this case BBR.

Machine Learning Algorithms to Predict Forage Nutritive Value of In Situ Perennial Ryegrass Plants Using Hyperspectral Canopy Reflectance Data

Nutritive value (NV) of forage is too time consuming and expensive to measure routinely in targeted breeding programs. Non-destructive spectroscopy has the potential to quickly and cheaply measure NV but requires an intermediate modelling step to interpret the spectral data. A novel machine learning technique for forage analysis, Cubist, was used to analyse canopy spectra to predict seven NV parameters, including dry matter (DM), acid detergent fibre (ADF), ash, neutral detergent fibre (NDF), in vivo dry matter digestibility (IVDMD), water soluble carbohydrates (WSC), and crude protein (CP). Perennial ryegrass (Lolium perenne) was used as the test crop. Independent validation of the developed models revealed prediction capabilities with R2 values and Lin’s concordance values reported between 0.49 and 0.82, and 0.68 and 0.89, respectively. Informative wavelengths for the creation of predictive models were identified for the seven NV parameters. These wavelengths included regions of the electromagnetic spectrum that are usually excluded due to high background variation, however, they contain important information and utilising them to obtain meaningful signals within the background variation is an advantage for accurate models. Non-destructive field spectroscopy along with the predictive models was deployed infield to measure NV of individual ryegrass plants. A significant reduction in labour was observed. The associated increase in speed and reduction of cost makes targeting NV in commercial breeding programs now feasible.

Effect of different regression algorithms on the estimating leaf parameters based on selected characteristic wavelengths by using the PROSPECT model

In this study, the characteristic wavelengths of leaf biochemical parameters (including carotenoid content, chlorophyll ${a} + { b}$a+b content, dry matter content, equivalent water thickness, and leaf structure parameter) were obtained through a sensitivity analysis based on a physical model. Then, performance of the selected characteristic wavelengths for monitoring leaf biochemical contents (LBC) was analyzed by using the following six popular regression algorithms: random forest, backpropagation neural network, support vector regression, radial basic function neural network, partial least-squares regression, and Gaussian process regression of different parameter values/kernel functions/training functions. In addition, the optimal parameters of each regression algorithm for estimating LBC were determined. Lastly, the effect of different regression algorithms on the accuracy of LBC estimation using four different data sets was also discussed. The results demonstrated that the selected 10 characteristic wavelengths combined with the Gaussian process regression model can be efficiently applied in estimating LBC.