@article{fdi:010093566,
  title = {{A}ssessing soil texture classification accuracy based on {VNIR} lab spectroscopy},
  author = {{R}ajendra, {T}. {C}. and {G}omez, {C}{\'e}cile and {D}harumarajan, {S}. and {K}umar, {D}. {N}.},
  editor = {},
  language = {{ENG}},
  abstract = {{S}oil texture is an important soil parameter controlling various physical, chemical and biological soil properties. {V}isible {N}ear-{I}nfrared ({VNIR}) spectroscopy has garnered attention due to its simplicity, non-destructive nature, absence of hazards and rapidity. {D}ue to its popularity, numerous studies employ this technique without adhering to the unity constraint on predicted fractions. {T}his study aims to assess the accuracy of soil texture classification in the {USDA} textural triangle through laboratory {VNIR} spectra. {F}ive different approaches were evaluated in this study: i) four approaches ({A}1-{A}4), defined as regression-assisted classification techniques, were based on the {P}artial {L}east {S}quares {R}egression ({PLSR}) method to predict quantitative fractions followed by a texture classification based on the {USDA} texture triangle and ii) one approach ({A}5), defined as a direct classification method, was based on the {P}artial {L}east {S}quares {D}iscriminant {A}nalysis ({PLS}-{DA}) classifier to classify soil texture using spectra directly. {E}ach regression-assisted classification approach varies in predicting fractions and ensuring the unity constraint on the predicted fractions. {I}n approach {A}1, the clay, silt and sand fractions predicted by {PLSR} for each sample were normalized to ensure sum-to-unity. {I}n approach {A}2, the silt content was derived as residual from the clay and sand contents predicted by {PLSR} for each sample, ensuring unity. {I}n {A}pproach {A}3, the clay, silt and sand fractions were simultaneously predicted using a multi-output variant of {PLSR}. {A}pproach {A}4 employed {PLSR} on log-ratio transformed ({LRT}) fractions, enabling simultaneous prediction and inherently ensuring sum-tounity. {A}pproach {A}4 via {LRT} utilizes information about the relative fractions of soil texture instead of the absolute fractions. {F}or the regression-based fraction predictions, approaches ({A}1-{A}4) achieved similar performances, with mean coefficients of determination ({R}2) of 0.88-0.90 for clay ({RMSE}: 4.2-4.4 %), 0.82-0.84 for sand ({RMSE}: 6.1-6.5 %), but lower ({R}2 = 0.29-0.38) for silt ({RMSE}: 3.8-4.1 %). {A}pproach {A}2, which infers silt as a residual, yielded poorer silt predictions. {D}espite these quantitative differences, the resulting classification accuracies in the {USDA} texture triangle were high with overall accuracy of 71-71.8 %, average accuracy of 62.4-65.3 % and {C}ohen's {K}appa of 0.61-0.62 for {A}1-{A}4 while {A}5, attained only 56.4 % overall accuracy and {C}ohen's {K}appa of 0.42. {A}mong the regression-assisted methods, {A}pproach {A}4 using log-ratio transformations of clay, silt, and sand simultaneously enforced compositional constraints and matched the best classification performances ({OA} = 71.4 %, {AA} = 65.3 %, {K} = 0.62) while requiring fewer models. {T}his work highlighted that i) the four regressionassisted classification approaches provided comparable and correct performances of soil texture classification, ii) the direct classification approach provided modest performance, iii) regression-assisted classification approaches outperformed the direct classification approach, and iv) in any of the approaches, the misclassifications were typically into the neighbouring textural classes. {T}his study aided in the creation of accurate and effective approaches for classifying soil texture by evaluating their performance and suitability. {A}mong these, {A}pproach {A}4, involving {PLSR} with log-ratio transformation, displays promise and warrants further evaluation on broader datasets and potential application on airborne or spaceborne platforms.},
  keywords = {{PLSR} ; {L}og-ratio transformation ; {S}um-to-unity constraint ; {R}egression-assisted classification ; {D}irect classification},
  booktitle = {},
  journal = {{C}hemometrics and {I}ntelligent {L}aboratory {S}ystems},
  volume = {263},
  numero = {},
  pages = {105419 [14 p.]},
  ISSN = {0169-7439},
  year = {2025},
  DOI = {10.1016/j.chemolab.2025.105419},
  URL = {https://www.documentation.ird.fr/hor/fdi:010093566},
}