Get Our e-AlertsSubmit Manuscript
Plant Phenomics / 2022 / Article

Research Article | Open Access

Volume 2022 |Article ID 9892728 | https://doi.org/10.34133/2022/9892728

Yanjie Li, Honggang Sun, Federico Tomasetto, Jingmin Jiang, Qifu Luan, "Spectrometric Prediction of Nitrogen Content in Different Tissues of Slash Pine Trees", Plant Phenomics, vol. 2022, Article ID 9892728, 10 pages, 2022. https://doi.org/10.34133/2022/9892728

Spectrometric Prediction of Nitrogen Content in Different Tissues of Slash Pine Trees

Received06 Jun 2021
Accepted21 Dec 2021
Published12 Jan 2022

Abstract

The internal cycling of nitrogen (N) storage and consumption in trees is an important physiological mechanism associated with tree growth. Here, we examined the capability of near-infrared spectroscopy (NIR) to quantify the concentration across tissue types (needle, trunk, branch, and root) without time and cost-consuming. The NIR spectral data of different tissues from slash pine trees were collected, and the concentration in each tissue was determined using standard analytical method in laboratory. Partial least squares regression (PLSR) models were performed on a set of training data randomly selected. The full-length spectra and the significant multivariate correlation (sMC) variable selected spectra were used for model calibration. Branch, needle, and trunk PLSR models performed well for the concentration using both full length and sMC selected NIR spectra. The generic model preformatted a reliable accuracy with R2C and R2CV of 0.62 and 0.66 using the full-length spectra, and 0.61 and 0.65 using sMC-selected spectra, respectively. Individual tissue models did not perform well when being used in other tissues. Five significantly important regions, i.e., 1480, 1650, 1744, 2170, and 2390 nm, were found highly related to the content in plant tissues. This study evaluates a rapid and efficient method for the estimation of content in different tissues that can help to serve as a tool for tree storage and recompilation study.

1. Introduction

Slash pine, which original from North America, has been widely planted in China for resin and timber production [1, 2]. Research on variation in the amount of allocated to reserves in plant tissues could provide new insights into various aspects of defense and contents with the changing seasons [3].

N, which relates to the protein synthesis, is one the most important macronutrients that support plant growth. Conversely, is also the most common limiting factor for plant and natural ecosystem growth and development. Plants collect and store from soil or atmospheric deposition as a reserve to maintain metabolism for growth [4]. Internal cycling allowed the plant store themselves for tree growth [5, 6]. However, the internal cycling amount of storage and consumption has been less studied, for example, when it starts to reserve [4], how it is related to the environmental factors such as season and soil fertility [7], how affects the ecophysiology process and contributes to different plant tissues for tree growth, [8], and how the variation of storage and consumption amounts in different plant tissues.

Therefore, it is necessary and helpful to understand the internal cycling process of in trees for plant ecology and management. However, the dearth of research on the variation of storage and consumption is mainly ascribed to the cost and time-consuming methods, e.g., a large number of samples need to be considered for multiple plant species and methods such as atomic absorption spectrometry [9, 10] and chromatography [11] need to be deployed.

Near-infrared spectroscopy (NIRS) has been wildly used for plant chemical component analysis [1216]. The C-H, N–H, and O–H bonds have a strong sensitivity in the range of NIR spectroscopy, which could vibrate, stretch, and bend when interacting with NIR spectra [17]. PLSR [18] has been highly used to predict the chemical content using NIR spectra. However, the application of NIR spectroscopy in analyzing the changes in content in different tree tissues is poorly understood. It is mainly concentrated on leaves under different environmental conditions [1921].

Here, we hypothesis that (1) the reflectance spectroscopy is feasible to characterize the content in different tree tissues using the PLSR model, (2) the predictability is decreased from one tissue models to another, (3) the most important regions of absorption features associated with content can be found using sMC variable selection, and more importantly, (4) generic (mixed tissue) calibration models can be used for predicting the variation of content across tree tissues.

2. Materials and Methods

2.1. Sampling

A considerable variation among genotypes of plant and environment is needed to get an accurate model [22]. As such, during summer and winter 2016, we collected samples from the slash pine experimental plantation at Tianmu Mountains region in Hangzhou, China (30°42N, 120°30E). This experimental plantation was established in January 1993 at three sites (each site was planted with a distance of 2.0 m between trees). Each site contained 30 rows, in each row, 6 trees were planted. Needle, branch, trunk, and root samples were collected from three or four mature individuals per slash pine. Fresh and new needles and branches were collected from the upper crown using a 20-meter-high pruning shear. From each trunk, a stem phloem was extracted at the breast height. Root samples were collected from the coarse and fine roots with equal weight (20 g). All tissue samples were stored at -4°C and shipped to the laboratory to avoid respiration. Subsequently, samples were dried in oven at 60°C and then ground into powder using a Wiley mill fitted with a 2 mm screen, which make the powder have a consistent size smaller than 2 mm. Each sample was then stored into a tube container. A total of 1985 tree samples were collected, and four tissues were collected from each tree, which makes 1985 samples for each tissue. In total, 7940 samples were collected.

2.2. Spectral Collection

All of the tissue samples were placed into an air-dried room for a consistence of moisture content (~12%). Then, the powder of each sample from all tissues was used for NIR reflectance spectra collection using a NIR spectrometer (LF-2500, Spectral evolution, USA) with a 10 mm diameter contact fiber-optical probe. The probe was firstly placed on a standard whiteboard for calibration; for stability reasons, the probe was directly placed onto the powder surface using a holder while collecting spectra. The maximum depth of each powder was ~3 cm to make sure the light could not penetrate the sample. The length of spectra was ranged from 1100 to 2500 nm with a 8 nm resolution (totaling 175 wavelengths). The final absorbance spectrum (log 1/reflectance) was the average value of 32 scan times [23, 24].

2.3. N Content Analysis

All of the samples were analyzed for content using a total nitrogen analyser (TNM-1, Shimadzu, Columbia, MD) based on the Kjeldahl method [25] and calculated according to Gebauer et al. [26]. The content was reported here as the percentage of dry mass.

2.4. Model Calibration and Validation

PLSR models, which is one of the most common methods for model calibration based on spectral data [27, 28], were generated to predict content (% dry mass) of the needle, branch, trunk, root tissues, and a mixture of all tissues together. Spectra data were first processed using standard normal variation (SNV) and first derivative using Savitzky-Golay smoothing (window size:15 data points) [29] which increased result accuracy when comparing with other types of preprocessing methods taken into consideration in our study. For each plant tissue calibration, each model was run 200 times for performance evaluation. The dataset was split into 80% for training and 20% for testing. During the lace of no external validation on an independent test set that was collected in a different year or similar condition. To overcome this limitation, we performed a blockwise cross validation that was named test set validation, and the results were presented as calibration and cross-validation results. The training data were used for model train using leave-one-out cross-validation (LOO). The model performance was tracked by using the coefficient of determination and that root-mean-square error derived from calibration (R2C and RMSEC) and cross-validation (R2CV and RMSECV). We used sMC () [30] as a variable selection method to find out the best PLSR model performance with less spectral variables. This sMC-PLSR method could minimize the effect of irrelevant spectra variables and highlight the most relevant variables that respond to chemical components [3133]. Optimal latent variables were selected for each PLSR model to obtain the best model. The pls package [34] was used for PLSR and sMC-PLSR model calibration, and the plsVarSel [35] was used for sMC variables selection in R software version 3.1.2; [36].

3. Results

3.1. The Content Variation

Data of the mixture of all tissues have a reasonable large variation (2.7 to 18.2%). In both summer and winter, root has the lowest content (5.7 and 6.6%, respectively) compared with other tissues. Trunk has the largest content variation, followed by needle and branch tissues. Seasonal variation of content across different tissues is significant especially in branch and needle (Figure 1).

3.2. Prediction Using Full Length Spectra

The content varied over 3-fold in branch and trunk, 2-fold in needle and root, respectively, which could enhanced the prediction accuracy on four types of slash pine tissues using full length NIR spectra, including individual tissues model and generic model (Figures 2(a), 2(c), and 3). Branch model produced the highest mean R2C of 0.71 (ranged from 0.67 to 0.74) and R2CV of 0.77 (ranged from 0.62 to 0.85), followed by needle model (mean R2C and R2CV were 0.67 (range: 0.64-0.72) and 0.73 (range: 0.54-0.82), respectively), trunk model (mean R2C and R2CV were 0.63 (range: 0.57-0.66) and 0.71 (range: 0.49-0.84), respectively), and generic model (mean R2C and R2CV were 0.62 (range: 0.60-0.64) and 0.65 (range: 0.56-0.73), respectively). The lowest R2C and R2CV were found in root model, with the mean R2C of 0.42 (ranged from 0.37 to 0.48) and R2CV of 0.42 (ranged from 0.1 to 0.65). However, the lowest RMSEC and RMSECV were found in needle model, with the mean RMSEC of 0.87% (range: 0.82-0.91) and RMSECV of 0.78% (range: 0.68-0.91), followed by root, branch, generic, and trunk model in sequence. The RMSEC and RMSECV of the five models were ranging from 0.82 to 1.67 and 0.68 to 1.85, respectively. Very small prediction error was obtained from the 100 simulated models for both individual tissues and generic. The number of latent variables in all of these models is lower than 10. Residual plots showed that all of the individual models and generic models tend to underestimate when values are low and overestimate when is high (Figure 4).

3.3. Prediction of Using the NIR Spectra Selected by sMC

PLSR models using sMC-selected NIR spectra variables (less than 15% percent of the full length spectra) yield a slightly better results in both individual tissues and generic (Figures 2(b), 2(d), and 5). The mean R2C and R2CV in models predicting the of both individual and generic tissues were varied from 0.42 to 0.71 and 0.43 to 0.77, respectively. Models predicting the N of branch, needle, trunk, and root had a mean R2C of 0.71 (range: 0.69-0.73), 0.67 (range: 0.61-0.71),0.63 (range: 0.591-0.65), and 0.42 (range: 0.37-0.48); mean R2CV of 0.77 (range: 0.63-0.86), 0.73 (range: 0.61-0.86),0.71 (range: 0.55-0.86), and 0.43 (range: 0.19-0.65); mean RMSEC of 1.25% (range: 1.20-1.29%), 0.86% (range: 0.79-0.93%), 1.61% (range: 1.54-1.69%), and 1.16% (range: 1.10-1.22%); and mean RMSECV of 1.23% (range: 0.95-1.43%), 0.78% (range: 0.62-0.91%),1.59% (range: 1.25-1.80%), and 1.15% (range: 0.89-1.47%), respectively. Generic model showed a promising prediction result with high mean R2C and R2CV of 0.61 (range: 0.59-0.63) and 0.65 (range: 0.56-0.73), and low mean RMSEC and RMSECV of 1.38 (range: 1.35-1.41) and 1.38 (range: 1.34-1.41), respectively. The sMC models vary larger in the validation set than the calibration set.

3.4. Performance Prediction of Individual Tissue Model Applying on Other Tissues

Individual tissue PLSR model produces a better prediction on their own tissues than on other tissues (Figures 6 and 7). However, generic model, based on both full length NIR spectra and sMC-selected NIR spectra, showed a promising result in predicting of other individual tissues with less bias (Figures 3, 4, and 5).

Figure 8 is the score plot of sMC_generic model for content prediction. It showed that different types of tissues could be clearly identified by the sMC_generic model, even though the sMC_generic model could efficiently predict the content in all tissues. Branch and root tissues may have similar construction, they are some overlap between them, and needle and most of the trunk are different from branch and root, which have been distributed as a single group, respectively.

3.5. Variable Selection of the NIR Spectra Applied in Prediction of Four Tissues

Five significantly important regions, i.e., 1650, 1744, 2170, and 2390 nm, were found highly influenced performance of model, both individual and generic. The region around 1480 nm has been recognized as very important among all five models, and the region around 1650 nm has a great significance in needle model than in other models. Regions like 1744 nm were shown to be less important in trunk model than other models. The region around 2390 nm showed a greater significance on the generic and needle model (Figure 9).

4. Discussion

In our study, content was variable within different types of tissues. This variation highly affected the model accuracy. It has been reported that less variation of the response characteristics could result in lower model prediction quality [37]. The root content has a low range from 2.7 to 10.2%, and this could explain why the root model produced a lower accuracy of prediction than other models.

is one of the most important macronutrients in plants that strongly influence plant growth and quality. Root in charge with taking up the from the soil in the form of NH4+ and NO3- and mainly stored as proteins, such as ribulose-1, 5-bisphosphate carboxylase/oxygenase (rubisco) [38]. It has been reported that the shortwave infrared region (1300–2500 nm, SWIR) is highly sensitive to plant water, lignin, cellulose, and proteins, and the most frequent mechanism transition between spectra bands and proteins is the N-H stretch, including the 1st and 3rd overtone stretch [39]. There are numerous studies that performed the reflectance data from plant leaves to monitor the traits [40, 41]. However, less is conducted on the different types of tissues [39, 42]. With the high range mean of R2C (0.42-0.72) and R2CV (0.43-0.77) and low range mean of RMSEC (0.86-1.61%) and RMSECV (0.78-1.38%) obtained from the individual tissue sMC_(root, branch, needle, and trunk) models and sMC_generic model, our results clearly showed that the content in different types of tissues can accurately and with good level of confidence be predict using reflectance spectroscopy. Furthermore, a generic model can efficiently predict the content in different tissues. In fact, we found that the mean R2C of our results in needle model 0.67 (range: 0.61-0.71) was close to the result reported in turfgrass () leaves [43] and tea leaves () [44] but less than that in the olive leaf () [45] when using the VIS-NIR/NIR spectroscopy.

We found that different individual tissues did perform a reliable prediction PLSR model when we considered the same type of tissues. However, the prediction accuracy extremely decreased when the model was built on one tissue to predict the content in other types of tissues (Figures 6 and 7). Different tissues in trees have different chemical structures, which might result into an absorbance decrease of in the NIR spectra [46]. For example, in the needles, the presence of phenolics or tannins may interfere with the absorbance of content in the NIR spectra and lead to content prediction bias when other types of tissues model are applied. The PLS score plot showed that these four types of tissues can be clearly classified, meaning that besides the content, there are some more chemical or physical information that are different among different tissues, resulting in different absorbance in the specific regions of the NIR spectra.

Similarly, a generic model was made to predict NSC concentration in tree tissues such as root, stem, branch, and leaf of various tree species, and a high prediction accuracy was obtained with of 0.91 and RMSE of 1.34% [16]. De Bei et al. [47] also found that a robust universal generic model with trunk and leaf tissues of Chardonnay grapevine could be used to predict the total nonstructural carbohydrate (TNC) concentration in both leaves and trunks with a . However, in our study, the accuracy of generic model was slightly low compared to the branch, needle, and trunk model. Thus, it is necessary to build the individual tissue model for content prediction in each tissue themself. Our results reinforced the idea that content prediction in different types of tissues using NIR can be fast and accurate.

It is important to estimate the data distribution and model stability using multiple permutations method [48]. The permutations generated the prediction error (Figures 3, 4, 6, and 7 error bar) based on multiple times of calibration and could be used on the model uncertainty test.

Samples within NIR spectra contain both useful and irrelevant chemical information, and the irrelevant information could reduce the model accuracy. Therefore, it is extremely valuable and useful to extrapolate the key spectra variables to reduce the model calibration error [49]. The sMC method in our study could efficiently select the spectra variables highly related with the content. The sMC_PLSR models for each tree tissues and generic tissues yielded a similar promising result to the full-length spectra with less spectra variables (44 to 115 variables among different tissue models; Figure 4). Additionally, the key wavelengths between 1400-1800 nm and 2100-2400 nm that are highly sensitive to the content were found by sMC method. Important peaks among these regions were found at 1480, 1650, 1744, 2170, and 2390 nm. In our study, the relevant bands that are related to or protein contents are not exactly what other studies have found. A reason for this could be that the different spectral sampling distances from spectral bands vary the number of bands in the spectral range. In addition, different preprocessing methods such as 1st and 2nd derivatives could reduce the band numbers. However, in the context of NIR spectroscopic analysis, the bands sensitive to found in our study can still be found in similar studies. Curran [50] and Kumar et al. [51] both reported that the distinct absorption peaks of proteins and/or are mainly sensitive to the SWIR region which support our study. Spectral bands sensitive to proteins or usually are also related to other components such as lignin (1690 nm and 1940 nm), cellulose (1940 nm), and starch (1690 nm and 1940 nm; [51]. The region around 1650 nm was mainly related to the first overtone of O-H strength which may be dominated by the cellulose in woody samples [52]. Wavelengths around 1480, 1744, 2170, and 2390 nm have been reported to be related to protein contents [47, 50, 53].

5. Conclusions

Our results have shown that we can successfully use NIR spectroscopy to characterize the content in tissues of different tree parts. The most important wavelength regions were found by the sMC variable selection method, which contributed to a consistent, promising, and robust calibration model across the four tissues. Our measurement of content in different tissues of trees using NIR spectroscopy is fast and accurate which could highly help for the internal cycling study in the future.

Data Availability

The data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest

The authors declare no conflicts of interest.

Authors’ Contributions

Yanjie Li designed the study, conducted the experiment, and wrote the manuscript. Qifu Luan and Jingmin Jiang supervised experiments at all stages and performed revisions of the manuscript, Honggang Sun supported the data collection and field experiment, and Federico Tomasetto performed revisions of the manuscript. All authors read and approved the final manuscript.

Acknowledgments

This work was funded by Fundamental Research Funds of CAF (CAFYBB2020SY008) and National Key R&D Program of China (2017YFD0600502-2).

References

  1. M. Yi, T. Jia, L. Dong et al., “Resin yield in Pinus elliottii Engelm. is related to the resin flow rate, resin components and resin duct characteristics at three locations in southern China,” Industrial Crops and Products, vol. 160, p. 113141, 2021. View at: Google Scholar
  2. S. Zhang, J. Jiang, and Q. Luan, “Index selection for growth and construction wood properties in Pinus elliottii open-pollinated families in southern China,” Southern Forests: a Journal of Forest Science, vol. 80, pp. 209–216, 2018. View at: Google Scholar
  3. C. Wang, Z. Chen, S. Unteregelsbacher et al., “Climate change amplifies gross nitrogen turnover in montane grasslands of Central Europe in both summer and winter seasons,” Global change biology, vol. 22, pp. 2963–2978, 2016. View at: Google Scholar
  4. P. Millard, “Ecophysiology of the internal cycling of nitrogen for tree growth,” Zeitschrift für Pflanzenernährung und Bodenkunde, vol. 159, no. 1, pp. 1–10, 1996. View at: Google Scholar
  5. P. Millard and M. F. Proe, “Nitrogen uptake, partitioning and internal cycling in Picea sitchensis (bong.) Carr. as influenced by nitrogen supply,” New Phytologist, vol. 125, pp. 113–119, 1993. View at: Google Scholar
  6. E. K. S. Nambiar and D. N. Fife, “Nutrient retranslocation in temperate conifers,” Tree Physiology, vol. 9, pp. 185–207, 1991. View at: Google Scholar
  7. P. Millard and M. Proe, “Leaf demography and the seasonal internal cycling of nitrogen in sycamore (Acer pseudoplatanus L.) seedlings in relation to nitrogen supply,” New Phytologist, vol. 117, pp. 587–596, 1991. View at: Google Scholar
  8. J. D. Aber and C. T. Driscoll, “Effects of land use, climate variation, and N deposition on N cycling and C storage in northern hardwood forests,” Global Biogeochemical Cycles, vol. 11, no. 4, pp. 639–648, 1997. View at: Publisher Site | Google Scholar
  9. D. L. G. Borges and J. A. Holcombe, “Graphite furnace atomic absorption spectrometry,” in Encyclopedia of Analytical Chemistry, pp. 1–20, Chichester, UK: John Wiley & Sons, Ltd, 2017. View at: Publisher Site | Google Scholar
  10. X.-F. Hu, A.-Q. Wu, F.-C. Wang, and F.-S. Chen, “The effects of simulated acid rain on internal nutrient cycling and the ratios of Mg, Al, Ca, N, and P in tea plants of a subtropical plantation,” Environmental Monitoring and Assessment, vol. 191, no. 2, p. 99, 2019. View at: Publisher Site | Google Scholar
  11. A. Rivero-Villar, P. H. Templer, V. Parra-Tabla, and J. Campo, “Differences in nitrogen cycling between tropical dry forests with contrasting precipitation revealed by stable isotopes of nitrogen in plants and soils,” Biotropica, vol. 50, pp. 859–867, 2018. View at: Google Scholar
  12. M. Forina, P. Oliveri, L. Bagnasco et al., “Artificial nose, NIR and UV–visible spectroscopy for the characterisation of the PDO Chianti Classico olive oil,” Talanta, vol. 144, pp. 1070–1078, 2015. View at: Google Scholar
  13. A. Guillemain, K. Dégardin, and Y. Roggo, “Performance of NIR handheld spectrometers for the detection of counterfeit tablets,” Talanta, vol. 165, pp. 632–640, 2017. View at: Google Scholar
  14. Y. Li, L. A. Apiolaza, and C. Altaner, “Genetic variation in heartwood properties and growth traits of Eucalyptus bosistoana,” European Journal of Forest Research, vol. 137, pp. 565–572, 2018. View at: Google Scholar
  15. C. Malegori, E. J. N. Marques, S. T. de Freitas, M. F. Pimentel, C. Pasquini, and E. Casiraghi, “Comparing the analytical performances of micro-NIR and FT-NIR spectrometers in the evaluation of acerola fruit quality, using PLS and SVM regression algorithms,” Talanta, vol. 165, pp. 112–116, 2017. View at: Google Scholar
  16. J. A. Ramirez, J. M. Posada, I. T. Handa et al., “Near-infrared spectroscopy (NIRS) predicts non-structural carbohydrate concentrations in different tissue types of a broad range of tree species,” Methods in Ecology and Evolution, vol. 6, pp. 1018–1025, 2015. View at: Google Scholar
  17. L. Bokobza, Origin of Near-Infrared Absorption Bands, Near Infrared Spectroscopy: Principles, Instruments, Applications. Wiley-VCH, Berlin, 2002.
  18. P. Geladi and B. R. Kowalski, “Partial least-squares regression: a tutorial,” Analytica Chimica Acta, vol. 185, pp. 1–17, 1986. View at: Google Scholar
  19. J. A. Hardin, M. W. Smith, P. R. Weckler, and B. S. Cheary, “In situ measurement of pecan leaf nitrogen concentration using a chlorophyll meter and vis-near infrared multispectral camera,” HortScience, vol. 47, pp. 955–960, 2012. View at: Google Scholar
  20. A. H. James, W. S. Michael, R. W. Paul, and S. C. Becky, “In situ measurement of pecan leaf nitrogen concentration using a chlorophyll meter and vis-near infrared multispectral camera,” HortScience horts, vol. 47, pp. 955–960, 2012. View at: Google Scholar
  21. X. Yang, J. Tang, and J. F. Mustard, “Beyond leaf color: comparing camera-based phenological metrics with leaf biochemical, biophysical, and spectral properties throughout the growing season of a temperate deciduous forest,” Journal of Geophysical Research: Biogeosciences, vol. 119, pp. 181–191, 2014. View at: Google Scholar
  22. R. L. Lindroth and S. B. S. Clair, “Adaptations of quaking aspen (Populus tremuloides Michx.) for defense against herbivores,” Forest Ecology and Management, vol. 299, pp. 14–21, 2013. View at: Publisher Site | Google Scholar
  23. D. Gillon, C. Houssard, and R. Joffre, “Using near-infrared reflectance spectroscopy to predict carbon, nitrogen and phosphorus content in heterogeneous plant material,” Oecologia, vol. 118, pp. 173–182, 1999. View at: Google Scholar
  24. M. Min, W. S. Lee, Y. H. Kim, and R. A. Bucklin, “Nondestructive detection of nitrogen in Chinese cabbage leaves using VIS–NIR spectroscopy,” HortScience, vol. 41, pp. 162–166, 2006. View at: Google Scholar
  25. R. B. Bradstreet, “Kjeldahl method for organic nitrogen,” Analytical Chemistry, vol. 26, pp. 185–187, 1954. View at: Google Scholar
  26. G. Gebauer, M. Schulumacher, B. Krstić, H. Rehder, and H. Ziegler, “Biomass production and nitrate metabolism of Atriplex hortensis L. (C 3 plant) and Amaranthus retroflexus L. (C 4 plant) in cultures at different levels of nitrogen supply,” Oecologia, vol. 72, pp. 303–314, 1987. View at: Google Scholar
  27. G. P. Asner, R. E. Martin, D. E. Knapp et al., “Spectroscopy of canopy chemicals in humid tropical forests,” Remote Sensing of Environment, vol. 115, pp. 3587–3598, 2011. View at: Google Scholar
  28. K. L. Bolster, M. E. Martin, and J. D. Aber, “Determination of carbon fraction and nitrogen concentration in tree foliage by near infrared reflectances: a comparison of statistical methods,” Canadian Journal of Forest Research, vol. 26, pp. 590–600, 1996. View at: Google Scholar
  29. W. H. Press and S. A. Teukolsky, “Savitzky-Golay smoothing filters,” Computers in Physics, vol. 4, pp. 669–672, 1990. View at: Google Scholar
  30. T. N. Tran, N. L. Afanador, L. M. Buydens, and L. Blanchet, “Interpretation of variable importance in partial least squares with significance multivariate correlation (sMC),” Chemometrics and Intelligent Laboratory Systems, vol. 138, pp. 153–160, 2014. View at: Google Scholar
  31. G. Aliakbarzadeh, H. Parastar, and H. Sereshti, “Classification of gas chromatographic fingerprints of saffron using partial least squares discriminant analysis together with different variable selection methods,” Chemometrics and Intelligent Laboratory Systems, vol. 158, pp. 165–173, 2016. View at: Google Scholar
  32. L. M. Kandpal, S. Lohumi, M. S. Kim, J.-S. Kang, and B.-K. Cho, “Near-infrared hyperspectral imaging system coupled with multivariate methods to predict viability and vigor in muskmelon seeds,” Sensors and Actuators B: Chemical, vol. 229, pp. 534–544, 2016. View at: Google Scholar
  33. Y. Li and C. Altaner, “Predicting extractives content of Eucalyptus bosistoana F. Muell. Heartwood from stem cores by near infrared spectroscopy,” Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, vol. 198, pp. 78–87, 2018. View at: Google Scholar
  34. B. Mevik, R. Wehrens, and L. Hovde, “Partial least squares and principal component regression,” R package version, vol. 2, no. 3, 2015. View at: Google Scholar
  35. T. Mehmood, K. H. Liland, L. Snipen, and S. Sæbø, “A review of variable selection methods in partial least squares regression,” Chemometrics and Intelligent Laboratory Systems, vol. 118, pp. 62–69, 2012. View at: Google Scholar
  36. R Core Team, R: A Language and Environment for Statistical Computing, R Foundation for Statistical Computing, Vienna, Austria, 2017.
  37. S. Wold, M. Sjöström, and L. Eriksson, “PLS-regression: a basic tool of chemometrics,” Chemometrics and Intelligent Laboratory Systems, vol. 58, pp. 109–130, 2001. View at: Google Scholar
  38. R. J. Ellis, “The most abundant protein in the world,” Trends in biochemical sciences, vol. 4, pp. 241–244, 1979. View at: Google Scholar
  39. K. Berger, J. Verrelst, J.-B. Féret et al., “Crop nitrogen monitoring: recent progress and principal developments in the context of imaging spectroscopy missions,” Remote Sensing of Environment, vol. 242, p. 111758, 2020. View at: Google Scholar
  40. J. Goffart, M. Olivier, and M. Frankinet, “Potato crop nitrogen status assessment to improve N fertilization management and efficiency: past–present–future,” Potato Research, vol. 51, pp. 355–383, 2008. View at: Google Scholar
  41. R. F. Muñoz-Huerta, R. G. Guevara-Gonzalez, L. M. Contreras-Medina, I. Torres-Pacheco, J. Prado-Olivarez, and R. V. Ocampo-Velazquez, “A review of methods for sensing the nitrogen status in plants: advantages, disadvantages and recent advances,” Sensors, vol. 13, pp. 10823–10843, 2013. View at: Google Scholar
  42. P. Hansen and J. Schjoerring, “Reflectance measurement of canopy biomass and nitrogen status in wheat crops using normalized difference vegetation indices and partial least squares regression,” Remote sensing of environment, vol. 86, pp. 542–553, 2003. View at: Google Scholar
  43. G. E. Bell, B. M. Howell, G. V. Johnson, W. R. Raun, J. B. Solie, and M. L. Stone, “Optical sensing of turfgrass chlorophyll content and tissue nitrogen,” HortScience, vol. 39, pp. 1130–1132, 2004. View at: Google Scholar
  44. X. Li and Y. He, “Discriminating varieties of tea plant based on vis/NIR spectral characteristics and using artificial neural networks,” Biosystems Engineering, vol. 99, pp. 313–321, 2008. View at: Google Scholar
  45. N. Rotbart, Z. Schmilovitch, Y. Cohen et al., “Estimating olive leaf nitrogen concentration using visible and near-infrared spectral reflectance,” Biosystems Engineering, vol. 114, pp. 426–434, 2013. View at: Google Scholar
  46. M. Blanco and I. Villarroya, “NIR spectroscopy: a rapid-response analytical tool,” TrAC Trends in Analytical Chemistry, vol. 21, pp. 240–250, 2002. View at: Google Scholar
  47. R. De Bei, S. Fuentes, W. Sullivan, E. J. Edwards, S. Tyerman, and D. Cozzolino, “Rapid measurement of total non-structural carbohydrate concentration in grapevine trunk and leaf tissues using near infrared spectroscopy,” Computers and Electronics in Agriculture, vol. 136, pp. 176–183, 2017. View at: Google Scholar
  48. J. J. Couture, A. Singh, K. F. Rubert-Nason, S. P. Serbin, R. L. Lindroth, and P. A. Townsend, “Spectroscopic determination of ecologically relevant plant secondary metabolites,” Methods in Ecology and Evolution, vol. 7, pp. 1402–1412, 2016. View at: Google Scholar
  49. J. Workman Jr. and L. Weyer, Practical Guide and Spectral Atlas for Interpretive Near-Infrared Spectroscopy, CRC press, Boca Raton Florida, USA, 1st edition, 2012.
  50. P. J. Curran, “Remote sensing of foliar chemistry,” Remote Sensing of Environment, vol. 30, pp. 271–278, 1989. View at: Google Scholar
  51. L. Kumar, K. Schmidt, S. Dury, and A. Skidmore, “Imaging spectrometry and vegetation science,” in Imaging spectrometry, pp. 111–155, Springer, 2002. View at: Google Scholar
  52. M. Schwanninger, J. C. Rodrigues, and K. Fackler, “A review of band assignments in near infrared spectra of wood and wood components,” Journal of Near Infrared Spectroscopy, vol. 19, pp. 287–308, 2011. View at: Google Scholar
  53. R. F. Kokaly, “Investigating a physical basis for spectroscopic estimates of leaf nitrogen concentration,” Remote Sensing of Environment, vol. 75, pp. 153–161, 2001. View at: Google Scholar

Copyright © 2022 Yanjie Li et al. Exclusive Licensee Nanjing Agricultural University. Distributed under a Creative Commons Attribution License (CC BY 4.0).

 PDF Download Citation Citation
Views35
Downloads44
Altmetric Score
Citations