pith. sign in

arxiv: 2607.00834 · v1 · pith:P33Y46Q6new · submitted 2026-07-01 · 💻 cs.LG

Spectroscopy Analysis with Machine Learning Regression for the Quantification of Carbon and Nitrogen Contents in Inceptisol and Oxisol Soil Types: Comparing Different Preprocessing and Validation methods as well as Feature Importance

Vinicius Herique Kieling , Guilherme Macedo Baggio , Felipe Augusto Bueno Rossi , Marco Antonio de Castro Barbosa , Dalcimar Casanova , Larissa Macedo dos Santos Tonial , Jefferson Tales Oliva This is my paper

Pith reviewed 2026-07-02 15:53 UTC · model grok-4.3

classification 💻 cs.LG
keywords NIR spectroscopymachine learningsoil carbonsoil nitrogenensemble learningSavitzky-Golay filterRPDportable spectrometer
0
0 comments X

The pith

Stacking ensemble models trained on filtered NIR spectra quantify carbon and nitrogen in two soil types with RPD above 2.0.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper evaluates machine learning models that use portable NIR spectral data to predict carbon and nitrogen content in Inceptisols and Oxisols. It tests multiple preprocessing steps, validation approaches, and ensemble methods to identify a reliable calibration pipeline. The stacking models reach RPD values greater than 2.0 with low overfitting, indicating that the spectra support non-destructive nutrient estimates. Performance differences between the two soil types are attributed to their distinct pedological properties.

Core claim

Stacking ensembles that combine PLS, SVR, and Ridge regressions as base learners with linear regression as the meta-model, applied to Savitzky-Golay filtered NIR spectra after NIPALS-Huber outlier removal, achieve RPD greater than 2.0 and low overfitting for C and N prediction in Oxisols and Inceptisols under 10-fold, leave-one-out, and Kennard-Stone validation.

What carries the argument

Stacking ensemble with PLS, SVR, and Ridge base models plus linear regression meta-model, applied after Savitzky-Golay filtering and NIPALS-Huber outlier removal on portable MyNIR spectra.

Load-bearing premise

The NIR spectra contain enough information to predict C and N content without major unmodeled interference from moisture, particle size, or other soil factors.

What would settle it

A new collection of soil samples from the same types, measured on the same device and processed identically, that yields RPD below 1.5 on the trained stacking models would falsify the quantification claim.

Figures

Figures reproduced from arXiv: 2607.00834 by Dalcimar Casanova, Felipe Augusto Bueno Rossi, Guilherme Macedo Baggio, Jefferson Tales Oliva, Larissa Macedo dos Santos Tonial, Marco Antonio de Castro Barbosa, Vinicius Herique Kieling.

Figure 1
Figure 1. Figure 1: Soil profiles analyzed in this study: (a) Inceptisol and (b) Oxisol. [7] [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Evolution of NIR analytical sensors [17]. [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Comparison between NIR spectra acquired by (A) portable equipment and (B) benchtop [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Simplified Fluxogram describing the methodology used in this study [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Illustrative representation of the NIR spectra acquisition using a portable instrument: (A) [PITH_FULL_IMAGE:figures/full_fig_p014_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Inceptisol Spectra Both plots suggest that there are some instrumental artifacts, such as distor￾15 [PITH_FULL_IMAGE:figures/full_fig_p015_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Oxisol Spectra tion, at the end of the spectra mentioned. Tables 2 and 3 provide numerical details regarding the spectra of both types of soil samples, such as the maximum, minimum, and mean values of NIR re￾flectance, as well as the wavelength interval of the portable device. Dimensions (shape) are also included, corresponding to the matrix in the programming envi￾ronment [PITH_FULL_IMAGE:figures/full_fi… view at source ↗
Figure 8
Figure 8. Figure 8: 20 Largest Coefficients in Absolute Value for Carbon prediction in Oxisol as a function [PITH_FULL_IMAGE:figures/full_fig_p053_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: 20 Smallest Coefficients in Absolute Value for C prediction in Oxisol as a function of [PITH_FULL_IMAGE:figures/full_fig_p054_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: 20 Largest Coefficients in Absolute Value for Nitrogen prediction in Oxisol as a func [PITH_FULL_IMAGE:figures/full_fig_p054_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: 20 Smallest Coefficients in Absolute Value for Nitrogen prediction in Oxisol as a [PITH_FULL_IMAGE:figures/full_fig_p055_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Spectra with Relevant (Green) and Non-Relevant (Red) Regions [PITH_FULL_IMAGE:figures/full_fig_p058_12.png] view at source ↗
read the original abstract

Near-Infrared (NIR) spectroscopy has emerged as a promising alternative to traditional soil analysis methods, offering advantages such as speed, low cost, and non-destructive testing. This work proposes a machine learning (ML) approach to calibrate predictive models for carbon (C) and nitrogen (N) content in Oxisols and Inceptisols, utilizing NIR spectral data acquired with a portable MyNIR device. Various preprocessing methods were evaluated, with the most effective being the Savitzky-Golay (SG) filter and a robust outlier removal method based on the Nonlinear Iterative Partial Least Squares (NIPALS) algorithm coupled with a Huber loss function. Multiple validation strategies were compared, including 10-fold cross-validation, leave-one-out, and holdout via the Kennard-Stone method, followed by standardization. Stacking ensemble learning models were employed, using Partial Least Squares (PLS), Support Vector Regression (SVR), and Ridge as base models, with linear regression as the meta-model. The models were evaluated using R2, Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Ratio of Performance Deviation (RPD) metrics. The performance gap between soil types suggests the influence of pedological characteristics. Furthermore, the models achieved an RPD > 2.0 with low overfitting, validating the potential of this approach for rapid C and N quantification. This study contributes to the optimization of sustainable agricultural practices, aligning with the demand for efficient and environmentally friendly analytical methods. The developed technique enables faster decision-making for producers and consultants based on organic matter content, fertility indicators, and nutrient availability.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper claims that stacking ensembles (PLS, SVR, Ridge bases with linear regression meta-learner) trained on Savitzky-Golay filtered and NIPALS-Huber outlier-cleaned NIR spectra acquired with a portable MyNIR device achieve RPD > 2.0 for predicting carbon and nitrogen contents in Oxisol and Inceptisol soils. It compares multiple preprocessing pipelines and validation strategies (10-fold CV, LOOCV, Kennard-Stone holdout), reports performance gaps between soil types as evidence of pedological influence, and concludes that the approach enables rapid, low-cost quantification.

Significance. If the reported RPD values prove robust to moisture/particle-size confounding and to truly independent test sets, the work would support portable NIR + stacking ensembles as a practical alternative to laboratory wet chemistry for soil C/N, with direct relevance to precision agriculture and sustainable nutrient management. The systematic comparison of preprocessing and validation choices adds incremental practical guidance.

major comments (3)
  1. [§2.2] §2.2 (Preprocessing pipeline): No drying protocol, moisture measurement, or explicit correction (e.g., water-band subtraction or external parameter orthogonalization) is described despite moisture and particle-size effects being dominant, well-documented interferents in NIR soil spectra. The central claim that SG + NIPALS-Huber spectra encode C/N content therefore rests on an untested assumption that these covariates are uncorrelated with the target labels within the collected sample.
  2. [§2.3] §2.3 (Validation strategies): The Kennard-Stone holdout and all cross-validation folds are performed on the same spectral matrix after a single standardization step; no geographically or temporally independent test set is reported. This setup leaves open the possibility that reported RPD > 2.0 values reflect within-dataset correlations rather than generalizable predictive power.
  3. [Results] Results section (performance tables): Sample sizes (n per soil type), exact hyperparameter values for PLS (latent variables), SVR (C, ε, γ), and Ridge (α), and full residual distributions or bootstrap confidence intervals on RPD are not provided, preventing assessment of whether the claimed performance gap between Oxisols and Inceptisols is statistically reliable or merely descriptive.
minor comments (2)
  1. [Abstract] The abstract states that 'the performance gap between soil types suggests the influence of pedological characteristics' without a supporting statistical test; this interpretive sentence should be moved to the discussion or qualified.
  2. [§2.4] Notation for the stacking meta-learner and the exact definition of RPD (which RPD formula is used?) should be stated explicitly in the methods rather than assumed from prior literature.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their thorough review and valuable suggestions. Below we provide point-by-point responses to the major comments. We have revised the manuscript to address the identified issues where feasible.

read point-by-point responses

  1. Referee: [§2.2] §2.2 (Preprocessing pipeline): No drying protocol, moisture measurement, or explicit correction (e.g., water-band subtraction or external parameter orthogonalization) is described despite moisture and particle-size effects being dominant, well-documented interferents in NIR soil spectra. The central claim that SG + NIPALS-Huber spectra encode C/N content therefore rests on an untested assumption that these covariates are uncorrelated with the target labels within the collected sample.

    Authors: We agree that moisture and particle-size effects are critical considerations in NIR soil analysis. The spectra in our study were acquired directly on field-collected samples using the portable MyNIR device without additional drying or correction steps, reflecting practical application conditions. We will revise section §2.2 to explicitly describe the sample collection and preparation procedures and add a discussion of potential confounding factors, including the assumption regarding uncorrelated covariates. This will also include recommendations for future work incorporating moisture correction techniques. revision: yes

  2. Referee: [§2.3] §2.3 (Validation strategies): The Kennard-Stone holdout and all cross-validation folds are performed on the same spectral matrix after a single standardization step; no geographically or temporally independent test set is reported. This setup leaves open the possibility that reported RPD > 2.0 values reflect within-dataset correlations rather than generalizable predictive power.

    Authors: The validation approaches used (10-fold CV, LOOCV, and Kennard-Stone holdout) follow standard practices for assessing model performance within the available dataset. We acknowledge that these do not constitute a fully independent external validation set from different geographic or temporal sources. We will revise the manuscript to explicitly note this limitation and emphasize that the reported performance applies to the studied soil types and conditions. Future extensions of this work will aim to include independent test sets. revision: yes

  3. Referee: [Results] Results section (performance tables): Sample sizes (n per soil type), exact hyperparameter values for PLS (latent variables), SVR (C, ε, γ), and Ridge (α), and full residual distributions or bootstrap confidence intervals on RPD are not provided, preventing assessment of whether the claimed performance gap between Oxisols and Inceptisols is statistically reliable or merely descriptive.

    Authors: We will update the Results section and associated tables to report the sample sizes for each soil type, the specific hyperparameter values selected for each base model, and include bootstrap-derived confidence intervals for the RPD metrics along with residual analysis to support the statistical reliability of the observed performance differences between soil types. revision: yes

Circularity Check

0 steps flagged

No circularity: standard empirical ML evaluation on held-out splits

full rationale

The paper describes collection of NIR spectra from soil samples, application of preprocessing (Savitzky-Golay, NIPALS-Huber), training of stacking ensembles (PLS, SVR, Ridge base learners with linear meta-learner), and reporting of R2/RMSE/MAE/RPD on 10-fold CV, LOOCV, and Kennard-Stone holdout splits. These steps constitute conventional supervised learning workflow; performance numbers are computed on samples excluded from each model's parameter estimation. No equations, uniqueness theorems, or self-citations are invoked that would make any reported quantity definitionally identical to the fitted inputs. The central claim (RPD > 2.0 validates the approach) rests on empirical results rather than reduction to prior definitions or self-referential fits.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The central claim rests on empirical fitting of multiple ML models whose hyperparameters are optimized on the target dataset, plus domain assumptions that NIR spectra linearly encode C and N after preprocessing. No new physical entities are introduced.

free parameters (2)
  • Hyperparameters of base learners (PLS latent variables, SVR C/epsilon/gamma, Ridge alpha)
    Chosen during model training to maximize reported metrics on the collected spectra.
  • Outlier threshold and NIPALS iteration settings
    Selected as part of the robust preprocessing pipeline to achieve the final RPD values.
axioms (2)
  • domain assumption NIR reflectance spectra contain sufficient quantitative information about soil C and N after standard preprocessing
    Invoked by the decision to use regression models on the spectral data.
  • domain assumption Cross-validation and Kennard-Stone holdout provide unbiased estimates of generalization error for these models
    Underlying the claim of low overfitting and RPD > 2.0.

pith-pipeline@v0.9.1-grok · 5881 in / 1643 out tokens · 33032 ms · 2026-07-02T15:53:12.497512+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

65 extracted references · 65 canonical work pages · 1 internal anchor

  1. [1]

    P. Nie, T. Dong, Y . He, F. Qu, Detection of soil nitrogen using near infrared sensors based on soil pretreatment and algorithms, Sensors 17 (2017) 1102

    work page 2017

  2. [2]

    D. W. Pribyl, A critical review of the conventional soc to som conversion factor, Geoderma 156 (2010) 75–83

    work page 2010

  3. [3]

    Van Raij, Fertilidade do solo e manejo de nutrientes; international plant nutrition institute (ipni): Piracicaba, s˜ao paulo, brazil, 2011, Google Scholar (2011) 420

    B. Van Raij, Fertilidade do solo e manejo de nutrientes; international plant nutrition institute (ipni): Piracicaba, s˜ao paulo, brazil, 2011, Google Scholar (2011) 420

    work page 2011

  4. [4]

    Sparks, Environmental soil chemistry academic press, California

    D. Sparks, Environmental soil chemistry academic press, California. USA (2003)

    work page 2003

  5. [5]

    Segnini, L

    A. Segnini, L. M. d. Santos, W. T. L. d. Silva, L. Martin-Neto, C. E. Bo- rato, W. J. d. Melo, D. Bolonhezi, Estudo comparativo de m ´etodos para a determinac ¸˜ao da concentrac ¸˜ao de carbono em solos com altos teores de fe (latossolos), Qu´ımica Nova 31 (2008) 94–97. 61

    work page 2008

  6. [6]

    H. G. d. Santos, P. K. T. Jacomine, L. H. C. d. Anjos, V . ´A. d. Oliveira, J. F. Lumbreras, M. R. Coelho, J. A. d. Almeida, J. C. d. Araujo Filho, H. N. Lima, F. A. Marques, J. B. d. Oliveira, T. J. F. Cunha, Sistema brasileiro de classificac ¸˜ao de solos, 6 ed., Embrapa, Bras ´ılia, DF, 2025. Dispon´ıvel em formato digital no portal da Embrapa

    work page 2025

  7. [7]

    edu/agricultural-life-sciences/soil-orders/ inceptisols, 2026

    University of Idaho, Inceptisols,https://www.uidaho. edu/agricultural-life-sciences/soil-orders/ inceptisols, 2026. Acessado em: 26 mai. 2026

    work page 2026

  8. [8]

    R. V . Rossel, T. Behrens, E. Ben-Dor, D. Brown, J. A. M. Dematt ˆe, K. D. Shepherd, Z. Shi, B. Stenberg, A. Stevens, V . Adamchuk, et al., A global spectral library to characterize the world’s soil, Earth-Science Reviews 155 (2016) 198–230

    work page 2016

  9. [9]

    R. B. A. Fernandes, I. A. d. Carvalho, E. S. Ribeiro, E. d. S. Mendonc ¸a, Comparison of different methods for the determination of total organic car- bon and humic substances in brazilian soils, Revista Ceres 62 (2015) 496– 501

    work page 2015

  10. [10]

    Shamrikova, E

    E. Shamrikova, E. Vanchikova, B. Kondratenok, E. Lapteva, S. Kostrova, Problems and limitations of the dichromatometric method for measuring soil organic matter content: a review, Eurasian Soil Science 55 (2022) 861–867

    work page 2022

  11. [11]

    D. W. Nelson, L. E. Sommers, Total carbon, organic carbon, and organic matter, in: D. L. Sparks, A. L. Page, P. A. Helmke, R. H. Loeppert, P. N. Soltanpour, M. A. Tabatabai, C. T. Johnston, M. E. Sumner (Eds.), Methods of Soil Analysis: Part 3 Chemical Methods, number 5 in SSSA Book Se- 62 ries, Soil Science Society of America and American Society of Agro...

    work page 1996

  12. [12]

    Jones Jr, Kjeldahl method for nitrogen determination, Micro-Macro Pub- lishing, Inc., 1991

    J. Jones Jr, Kjeldahl method for nitrogen determination, Micro-Macro Pub- lishing, Inc., 1991

    work page 1991

  13. [13]

    J. M. Bremner, Determination of nitrogen in soil by the Kjeldahl method, The Journal of Agricultural Science 55 (1960) 11–33

    work page 1960

  14. [14]

    S ´aez-Plaza, T

    P. S ´aez-Plaza, T. Michałowski, M. J. Navas, A. G. Asuero, S. Wybraniec, An overview of the Kjeldahl method of nitrogen determination. Part I. Early history, chemistry of the procedure, and titrimetric finish, Critical Reviews in Analytical Chemistry 43 (2013) 178–223

    work page 2013

  15. [15]

    Blanco, I

    M. Blanco, I. Villarroya, Nir spectroscopy: a rapid-response analytical tool, TrAC Trends in Analytical Chemistry 21 (2002) 240–250

    work page 2002

  16. [16]

    Gopal, M

    J. Gopal, M. Muthu, Handheld portable analytics for food fraud detection, the evolution of next-generation smartphone-based food sensors: The jour- ney, the milestones, the challenges debarring the destination, TrAC Trends in Analytical Chemistry 171 (2024) 117504

    work page 2024

  17. [17]

    Zareef, M

    M. Zareef, M. Arslan, W. Ahmad, M. M. Hassan, M. Shoaib, Q. Ouyang, H. Li, M. M. Hashim, S. Javaria, Q. Chen, Applications of benchtop and portable spectroscopy techniques for food quality monitoring, Spectrochim- ica Acta Part A: Molecular and Biomolecular Spectroscopy 344 (2026) 126713

    work page 2026

  18. [18]

    J. T. Oliva, V . Kieling, F. Rossi, E. O. Rodrigues, G. Guarneri, L. dos San- tos Tonial, Comparison between portable and bench-top near-infrared spec- 63 troscopy for corn silage characterization using partial least square and sup- port vector regression methods, Journal of Chemometrics 39 (2025) e70073. E70073 CEM-25-0139

    work page 2025

  19. [19]

    R. A. Viscarra Rossel, R. N. McGlynn, A. B. McBratney, Determination of reference values for nir spectra of soils, Geoderma 137 (2006) 70–82

    work page 2006

  20. [20]

    J. M. Soriano-Disla, L. J. Janik, D. J. Allen, M. J. McLaughlin, Evaluation of the performance of portable visible-infrared instruments for the prediction of soil properties, Biosystems Engineering 161 (2017) 24–36

    work page 2017

  21. [21]

    Nocita, A

    M. Nocita, A. Stevens, B. van Wesemael, M. Aitkenhead, M. Bachmann, B. Barth`es, E. Ben Dor, D. J. Brown, M. Clairotte, A. Csorba, et al., Soil spectroscopy: An alternative to wet chemistry for soil monitoring, Advances in Agronomy 132 (2015) 139–159

    work page 2015

  22. [22]

    Stenberg, R

    B. Stenberg, R. A. Viscarra Rossel, A. M. Mouazen, J. Wetterlind, Visible and near infrared spectroscopy in soil science, Advances in Agronomy 107 (2010) 163–215

    work page 2010

  23. [23]

    Recena, V

    R. Recena, V . M. Fern´andez-Caban´as, A. Delgado, Soil fertility assessment by vis-nir spectroscopy: Predicting soil functioning rather than availability indices, Geoderma 337 (2019) 368–374

    work page 2019

  24. [24]

    Workman Jr, L

    J. Workman Jr, L. Weyer, Practical guide to interpretive near-infrared spec- troscopy, CRC press, 2007

    work page 2007

  25. [25]

    Z. Wang, W. Huang, J. Li, S. Liu, S. Fan, Assessment of protein content and insect infestation of maize seeds based on on-line near-infrared spectroscopy 64 and machine learning, Computers and Electronics in Agriculture 211 (2023) 107969

    work page 2023

  26. [26]

    J. Liu, X. Luo, D. Zhang, C. Wang, Z. Chen, X. Zhao, Rapid determination of rice protein content using near-infrared spectroscopy coupled with feature wavelength selection, Infrared Physics & Technology 135 (2023) 104969

    work page 2023

  27. [27]

    Sisouane, M

    M. Sisouane, M. Cascant, S. Tahiri, S. Garrigues, M. El Krati, G. E. K. Boutchich, M. Cervera, M. de La Guardia, Prediction of organic carbon and total nitrogen contents in organic wastes and their composts by infrared spectroscopy and partial least square regression, Talanta 167 (2017) 352– 358

    work page 2017

  28. [28]

    Bedin, M

    F. Bedin, M. Faust, G. Guarneri, T. S. Assmann, C. Lafay, L. Soares, P. de Oliveira, L. dos Santos Tonial, NIR associated to PLS and SVM for fast and non-destructive determination of C, N, P, and K contents in poultry litter, Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 245 (2021) 118834

    work page 2021

  29. [29]

    Sharififar, K

    A. Sharififar, K. Singh, E. Jones, F. I. Ginting, B. Minasny, Evaluating a low-cost portable nir spectrometer for the prediction of soil organic and total carbon using different calibration models, Soil Use and Management 35 (2019) 607–616

    work page 2019

  30. [30]

    Benedet, W

    L. Benedet, W. M. Faria, S. H. G. Silva, M. Mancini, J. A. M. Dematt ˆe, L. R. G. Guilherme, N. Curi, Soil texture prediction using portable x-ray fluorescence spectrometry and visible near-infrared diffuse reflectance spec- troscopy, Geoderma 376 (2020) 114553. 65

    work page 2020

  31. [31]

    Y . Tang, E. Jones, B. Minasny, Evaluating low-cost portable near infrared sensors for rapid analysis of soils from south eastern australia, Geoderma Regional 20 (2020) e00240

    work page 2020

  32. [32]

    Priori, N

    S. Priori, N. Mzid, S. Pascucci, S. Pignatti, R. Casa, Performance of a portable ft-nir mems spectrometer to predict soil features, Soil Systems 6 (2022)

    work page 2022

  33. [33]

    N. M. Dhawale, V . Adamchuk, S. Prasher, R. V . Rossel, A. Ismail, Evalu- ation of two portable hyperspectral-sensor-based instruments to predict key soil properties in canadian soils, Sensors (Basel, Switzerland) 22 (2022)

    work page 2022

  34. [34]

    J. K. Carvalho, J. M. Moura-Bueno, R. Ramon, T. F. Almeida, G. Naibo, A. P. Martins, L. S. Santos, C. Gianello, T. Tiecher, Combining different pre- processing and multivariate methods for prediction of soil organic matter by near infrared spectroscopy (nirs) in southern brazil, Geoderma Regional 29 (2022) e00530

    work page 2022

  35. [35]

    Nawar, A

    S. Nawar, A. M. Mouazen, Combining mid infrared spectroscopy with stacked generalisation machine learning for prediction of key soil proper- ties, European Journal of Soil Science 73 (2022) e13323

    work page 2022

  36. [36]

    Saberioon, A

    M. Saberioon, A. Gholizadeh, A. Ghaznavi, S. Chabrillat, V . Khosravi, En- hancing soil organic carbon prediction of lucas soil database using deep learning and deep feature selection, Computers and Electronics in Agri- culture 227 (2024) 109494

    work page 2024

  37. [37]

    Z. Xu, X. Zhao, X. Guo, J. Guo, Deep learning application for predicting soil 66 organic matter content by vis-nir spectroscopy, Computational Intelligence and Neuroscience 2019 (2019) 3563761

    work page 2019

  38. [38]

    Gozukara, A

    G. Gozukara, A. E. Hartemink, J. Huang, J. A. M. Dematt ˆe, Prediction accuracy of pxrf, mir, and vis-nir spectra for soil properties—a review, Soil Science Society of America Journal 89 (2025) e70028

    work page 2025

  39. [39]

    Walkley, I

    A. Walkley, I. A. Black, An examination of the degtjareff method for deter- mining soil organic matter, and a proposed modification of the chromic acid titration method, Soil science 37 (1934) 29–38

    work page 1934

  40. [40]

    F. C. da Silva, Manual de an ´alises qu´ımicas de solos, plantas e fertilizantes. (2009)

    work page 2009

  41. [41]

    Tedesco, C

    M. Tedesco, C. Gianello, C. Bissani, H. Bohnen, S. V olkweiss, An ´alise de solo, plantas e outros materiais (in Portuguese), Boletim t ´ecnico 5 (1995)

    work page 1995

  42. [42]

    Barra, S

    I. Barra, S. M. Haefele, R. Sakrabani, F. Kebede, Soil spectroscopy with the use of chemometrics, machine learning and pre-processing techniques in soil diagnosis: Recent advances–a review, TrAC trends in analytical chemistry 135 (2021) 116166

    work page 2021

  43. [43]

    Sharma, A study on data scaling methods for machine learning, Interna- tional Journal for Global Academic & Scientific Research 1 (2022) 31–42

    V . Sharma, A study on data scaling methods for machine learning, Interna- tional Journal for Global Academic & Scientific Research 1 (2022) 31–42

    work page 2022

  44. [44]

    Isaksson, T

    T. Isaksson, T. Næs, The effect of multiplicative scatter correction (msc) and linearity improvement in nir spectroscopy, Applied Spectroscopy 42 (1988) 1273–1284. 67

    work page 1988

  45. [45]

    Barnes, M

    R. Barnes, M. S. Dhanoa, S. J. Lister, Standard normal variate transforma- tion and de-trending of near-infrared diffuse reflectance spectra, Applied spectroscopy 43 (1989) 772–777

    work page 1989

  46. [46]

    A. F. Goetz, G. Vane, J. E. Solomon, B. N. Rock, Imaging spectrometry for earth remote sensing, science 228 (1985) 1147–1153

    work page 1985

  47. [47]

    D. A. Skoog, F. J. Holler, S. R. Crouch, Principles of Instrumental Analysis, 7th ed., Cengage Learning, Boston, MA, 2017

    work page 2017

  48. [48]

    Wenjun, S

    J. Wenjun, S. Zhou, H. Jingyi, L. Shuo, In situ measurement of some soil properties in paddy soil using visible and near-infrared spectroscopy, PLOS ONE 9 (2014) e105708

    work page 2014

  49. [49]

    Rinnan, F

    ˚A. Rinnan, F. van den Berg, S. B. Engelsen, Review of the most common pre-processing techniques for near-infrared spectra, Trends in Analytical Chemistry 28 (2009) 1201–1222

    work page 2009

  50. [50]

    S. Wold, K. Esbensen, P. Geladi, Principal component analysis, Chemomet- rics and intelligent laboratory systems 2 (1987) 37–52

    work page 1987

  51. [51]

    Savitzky, M

    A. Savitzky, M. Golay, Smoothing and differentiation of data by simplified least squares procedures, Analytical chemistry 36 (1964) 1627–1639

    work page 1964

  52. [52]

    E. J. Bjerrum, M. Glahder, T. Skov, Data augmentation of spectral data for convolutional neural network (cnn) based deep chemometrics, arXiv preprint arXiv:1710.01927 (2017)

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  53. [53]

    Geladi, B

    P. Geladi, B. R. Kowalski, Partial least-squares regression: a tutorial, Ana- lytica Chimica Acta 185 (1986) 1–17. 68

    work page 1986

  54. [54]

    Raschka, V

    S. Raschka, V . Mirjalili, Python machine learning: Machine learning and deep learning with Python, scikit-learn, and TensorFlow 2, Packt publishing ltd, 2019

    work page 2019

  55. [55]

    R. d. A. Ferreira, G. Teixeira, L. A. Peternelli, Kennard-stone method out- performs the random sampling in the selection of calibration samples in snps and nir data, Ci ˆencia Rural 52 (2021) e20201072

    work page 2021

  56. [56]

    R. W. Kennard, L. A. Stone, Computer aided design of experiments, Tech- nometrics 11 (1969) 137–148

    work page 1969

  57. [57]

    T. F. Boucher, M. V . Ozanne, M. L. Carmosino, M. D. Dyar, S. Mahadevan, E. A. Breves, K. H. Lepore, S. M. Clegg, A study of machine learning regression methods for major elemental analysis of rocks using laser-induced breakdown spectroscopy, Spectrochimica Acta Part B: Atomic Spectroscopy 107 (2015) 1–10

    work page 2015

  58. [58]

    Stone, Cross-validatory choice and assessment of statistical predictions, Journal of the royal statistical society: Series B (Methodological) 36 (1974) 111–133

    M. Stone, Cross-validatory choice and assessment of statistical predictions, Journal of the royal statistical society: Series B (Methodological) 36 (1974) 111–133

    work page 1974

  59. [59]

    D. C. Montgomery, G. C. Runger, Applied statistics and probability for en- gineers, John wiley & sons, 2010

    work page 2010

  60. [60]

    M. Khan, S. Noor, Performance analysis of regression-machine learning algorithms for predication of runoff time, Agrotechnology 8 (2019) 1–12

    work page 2019

  61. [61]

    E. D. Louw, K. I. Theron, Robust prediction models for quality parameters in japanese plums (prunus salicina l.) using nir spectroscopy, Postharvest Biology and Technology 58 (2010) 176–184. 69

    work page 2010

  62. [62]

    C. J. Willmott, K. Matsuura, Advantages of the mean absolute error (MAE) over the root mean square error (RMSE) in assessing average model perfor- mance, volume 30, Inter-Research, 2005

    work page 2005

  63. [63]

    T. Chai, R. R. Draxler, Root mean square error (rmse) or mean absolute error (mae)?, Geoscientific Model Development 7 (2014) 1247–1250

    work page 2014

  64. [64]

    Saeys, A

    W. Saeys, A. M. Mouazen, H. Ramon, Potential for onsite and online anal- ysis of pig manure using visible and near infrared reflectance spectroscopy, Biosystems Engineering 91 (2005) 393–402

    work page 2005

  65. [65]

    K. H. Esbensen, P. Geladi, A. Larsen, The rpd myth. . . , NIR news 25 (2014) 24–28. 70

    work page 2014