@article{fdi:010083200,
  title = {{U}sing carbonate absorbance peak to select the most suitable regression model before predicting soil inorganic carbon concentration by mid-infrared reflectance spectroscopy},
  author = {{G}omez, {C}{\'e}cile and {C}hevallier, {T}iphaine and {M}oulin {E}smard, {P}atricia and {A}rrouays, {D}. and {B}arth{\`e}s, {B}ernard},
  editor = {},
  language = {{ENG}},
  abstract = {{M}id-{I}nfrared reflectance spectroscopy ({MIRS}, 4000-400 cm(-1)) is being considered to provide accurate estimations of soil inorganic carbon ({SIC}) contents, based on prediction models when the test dataset is well represented by the calibration set, with similar {SIC} range and distribution and pedological context. {T}his work addresses the case where the test dataset, here originating from {F}rance, is poorly represented by the calibration set, here originating from {T}unisia, with different {SIC} distributions and pedological contexts. {I}t aimed to demonstrate the usefulness of 1) classifying test samples according to {SIC} level based on the height of the carbonate absorbance peak at 2510 cm(-1), and then 2) selecting a suitable prediction model according to {SIC} level. {T}wo regression methods were tested: {L}inear {R}egression using the height of the carbonate peak at 2510 cm(-)(1), called {P}eak-{LR} model; and {P}artial {L}east {S}quares {R}egression using the entire {MIR} spectrum, called {F}ull-{PLSR} model. {F}irst, our results showed that {F}ull-{PLSR} was 1) more accurate than {P}eak-{LR} on the {T}unisian validation set ({R}-val(2) = 0.99 vs. 0.86 and {RMSE}val = 3.0 vs. 9.7 g kg(-1) , respectively), but 2) less accurate than {P}eak-{LR} when applied on the {F}rench dataset ({R}-test(2) = 0.70 vs. 0.91 and {RMSE}test = 13.7 vs. 4.9 g kg(-1), respectively). {S}econdly, on the {F}rench dataset, predictions on {SIC}-poor samples tended to be more accurate using {P}eak-{LR}, while predictions on {SIC}-rich samples tended to be more accurate using {F}ull-{PLSR}. {T}hirdly, the height of the carbonate absorbance peak at 2510 cm(-1) might be used to discriminate {SIC}-poor and {SIC}-rich test samples (<5 vs. > 5 g kg(-1)): when this height was > 0, {F}ull-{PLSR} was applied; otherwise {P}eak-{LR} was applied. {C}oupling {P}eak-{LR} and {F}ull-{PLSR} models depending on the carbonate peak yielded the best predictions on the {F}rench dataset ({R}-test(2) = 0.95 and {RMSE}test = 3.7 g kg(-1)). {T}his study underlined the interest of using a carbonate peak to select suitable regression approach for predicting {SIC} content in a database with different distribution than the calibration database.},
  keywords = {{S}oil inorganic carbon ; {M}id-infrared reflectance spectroscopy ; {P}artial least squares regression ; {L}inear regression ; {N}ational dataset ; {TUNISIE} ; {FRANCE}},
  booktitle = {},
  journal = {{G}eoderma},
  volume = {405},
  numero = {},
  pages = {115403 [12 ]},
  ISSN = {0016-7061},
  year = {2022},
  DOI = {10.1016/j.geoderma.2021.115403},
  URL = {https://www.documentation.ird.fr/hor/fdi:010083200},
}