TY  - JOUR
AU  - Yang, Xing
AU  - Jiang, Bixia
AU  - He, Yanjie
AU  - Yue, Rui
AU  - Bao, Ruiqian
AU  - Lv, Gaoqiang
AU  - Ma, Fei
AU  - Du, Changwen
PY  - 2026
DA  - 2026/09/24
TI  - ATR-FTIR Spectral Characteristics and Their Application to the Classification of Cultivated Karst Soils in Shilin County
JO  - Digital Intelligence in Agriculture
T2  - Digital Intelligence in Agriculture
JF  - Digital Intelligence in Agriculture
VL  - 2
IS  - 3
SP  - 179
EP  - 190
DO  - 10.62762/DIA.2026.518539
UR  - https://www.icck.org/article/abs/DIA.2026.518539
KW  - Karst calcareous soil
KW  - spectral characteristics
KW  - soil classification
KW  - ATR-FTIR
KW  - PLS-DA
AB  - To enable rapid, non-destructive analysis and functional classification of karst calcareous soils in Shilin County, Kunming, this study acquired mid-infrared attenuated total reflectance Fourier-transform infrared (ATR-FTIR) spectra in the 4000-900\cmi{} range from 409 soil samples. After standardized preprocessing, the samples were preliminarily partitioned into four spectral groups by K-means clustering, and a partial least squares discriminant analysis (PLS-DA) model was constructed. Variable importance in projection (VIP) scores were further used to identify key spectral features distinguishing the classes. The spectra were dominated by peaks at 3620, 1650, and 1020\cmi{}, and no obvious calcite absorption at $\sim$1430\cmi{} was observed, suggesting strong carbonate leaching (decalcification weathering). K-means clustering divided the samples into four groups (G1--G4), interpreted from their spectral features as organic matter-mineral impoverished, mineral-dominated, organic matter-mineral enriched, and organic matter impoverished types, respectively, possibly reflecting spectral differentiation along a rocky desertification gradient. The PLS-DA model achieved an overall accuracy of 93.9% under nested cross-validation, with precision and recall exceeding 92% for G1--G3. VIP analysis identified the silicate skeleton region, the organic matter/water region, and the hydroxyl region as key bands, suggesting that gradients in clay mineral, organic matter, and bound water contents contribute most to spectral differentiation. These findings indicate that ATR-FTIR spectroscopy has potential for rapid, non-destructive classification of karst calcareous soils and can provide a technical pathway for soil survey and rocky desertification monitoring.
SN  - 3069-3187
PB  - Institute of Central Computation and Knowledge
LA  - English
ER  - 

@article{Yang2026ATRFTIR,
  author = {Xing Yang and Bixia Jiang and Yanjie He and Rui Yue and Ruiqian Bao and Gaoqiang Lv and Fei Ma and Changwen Du},
  title = {ATR-FTIR Spectral Characteristics and Their Application to the Classification of Cultivated Karst Soils in Shilin County},
  journal = {Digital Intelligence in Agriculture},
  year = {2026},
  volume = {2},
  number = {3},
  pages = {179-190},
  doi = {10.62762/DIA.2026.518539},
  url = {https://www.icck.org/article/abs/DIA.2026.518539},
  abstract = {To enable rapid, non-destructive analysis and functional classification of karst calcareous soils in Shilin County, Kunming, this study acquired mid-infrared attenuated total reflectance Fourier-transform infrared (ATR-FTIR) spectra in the 4000-900\cmi{} range from 409 soil samples. After standardized preprocessing, the samples were preliminarily partitioned into four spectral groups by K-means clustering, and a partial least squares discriminant analysis (PLS-DA) model was constructed. Variable importance in projection (VIP) scores were further used to identify key spectral features distinguishing the classes. The spectra were dominated by peaks at 3620, 1650, and 1020\cmi{}, and no obvious calcite absorption at \$\sim\$1430\cmi{} was observed, suggesting strong carbonate leaching (decalcification weathering). K-means clustering divided the samples into four groups (G1--G4), interpreted from their spectral features as organic matter-mineral impoverished, mineral-dominated, organic matter-mineral enriched, and organic matter impoverished types, respectively, possibly reflecting spectral differentiation along a rocky desertification gradient. The PLS-DA model achieved an overall accuracy of 93.9\% under nested cross-validation, with precision and recall exceeding 92\% for G1--G3. VIP analysis identified the silicate skeleton region, the organic matter/water region, and the hydroxyl region as key bands, suggesting that gradients in clay mineral, organic matter, and bound water contents contribute most to spectral differentiation. These findings indicate that ATR-FTIR spectroscopy has potential for rapid, non-destructive classification of karst calcareous soils and can provide a technical pathway for soil survey and rocky desertification monitoring.},
  keywords = {Karst calcareous soil, spectral characteristics, soil classification, ATR-FTIR, PLS-DA},
  issn = {3069-3187},
  publisher = {Institute of Central Computation and Knowledge}
}