EUVIMEDEUVIMED entdecken
Crossref

Rapid assessment of soil nutrients under continuous Gastrodia elata cropping using near- and mid-infrared spectroscopy combined with multi-strategy machine learning

Wentao Zhang, Baiheng Xie, Jinfang Ma, Xiangyu Zhang, Furong Huang, Caiyun Wang

Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy · 2026-12 · 363 · 128453

DOI: 10.1016/j.saa.2026.128453 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 2026-12. Verfasst von Wentao Zhang und 5 weiteren Autor:innen. Beteiligte Einrichtung: Department of Optoelectronic Engineering, Jinan University, Guangzhou, Guangdong 510632, China. Hauptthemen laut MEDLINE: Soil, Gastrodia, Machine Learning, Nutrients. Die Arbeit nennt 40 Literaturangaben. Gefördert von: Science and Technology Program of Guizhou Province.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

To enable rapid and non-destructive quantitative assessment of soil nutrients in Gastrodia elata continuous-cropping fields, soils with different cultivation histories were collected from a G. elata experimental site in Bijie, Guizhou, China, including never planted (WZZ), continuous cropping for one year (LZ1), continuous cropping for two years (LZ2), and G. elata-konjac rotation (TMLuZ). Reference concentrations of soil organic matter (SOM), total nitrogen (TN), alkali-hydrolyzable nitrogen (HN), total phosphorus (TP), available phosphorus (AP), total potassium (TK), and available potassium (AK) were determined, and near-infrared (NIR) and attenuated total reflectance mid-infrared (ATR-MIR) spectra were acquired. Spectral preprocessing was performed, followed by variable selection using uninformative variable elimination (UVE) and competitive adaptive reweighted sampling (CARS). Partial least squares regression (PLSR) and random forest (RF) models were developed for single-modality spectra, and three fusion schemes-data-level, feature-level, and decision-level-were further designed. Model performance was evaluated on an independent validation set using RV2, RMSEV, and RPIQV, with RPIQV=1.4 as the usability threshold. The results demonstrated substantial differences in predictability among nutrient indicators. SOM, TN, and TK achieved stable performance on the independent validation set. The best SOM model was obtained from the NIR single-modality approach (PLSR, RV2=0.939, RPIQV=4.87), while data-level fusion yielded comparable accuracy (PLSR, RV2=0.936, RPIQV=4.75). The best results for TN and TK were both achieved using the ATR-MIR single-modality strategy (TN: RF, RV2=0.866, RPIQV=4.10; TK: PLSR, RV2=0.849, RPIQV=4.08). HN and AK reached a moderate prediction level (HN: RPIQV=2.42; AK: RPIQV=2.33). In contrast, TP and AP did not meet the usability threshold under any strategy (best RPIQV for TP = 1.16; for AP = 0.81). Overall, single-modality NIR and ATR-MIR models can provide a rapid quantitative screening route for SOM, TN, and TK in continuous-cropping fields, whereas the benefits of fusion were indicator-dependent. Phosphorus-related indicators still require further optimization.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

SoilGastrodiaMachine LearningNutrientsSpectroscopy, Near-InfraredLeast-Squares AnalysisNitrogenPhosphorusPotassiumRandom ForestSpectrophotometry, Infrared

Stichwörter der Autor:innen

Attenuated total reflectance mid-infraredIndependent validation setNear-infraredRPIQSoil nutrientsSpectral fusion

Substanzen

SoilNitrogenPhosphorusPotassium

Autor:innen und Institutionen

  • Department of Optoelectronic Engineering, Jinan University, Guangzhou, Guangdong 510632, China.
  • Guangdong Jiangmen Chinese Medicine College, Jiangmen, Guangdong, China.
  • Bijie Institute of Traditional Chinese Medicine, Bijie 551700, China; Guizhou Key Laboratory for Germplasm Innovation and Resource-Efficient Utilization of Dao-di Herbs, Bijie 551700, China.
  • Department of Optoelectronic Engineering, Jinan University, Guangzhou, Guangdong 510632, China; International Cooperation Joint Laboratory for Optoelectronic Hybrid Integrated Circuits, Jinan University, Guangzhou, Guan…
  • Bijie Institute of Traditional Chinese Medicine, Bijie 551700, China; Guizhou Key Laboratory for Germplasm Innovation and Resource-Efficient Utilization of Dao-di Herbs, Bijie 551700, China. Electronic address: wangcaiy…

Förderung

  • Science and Technology Program of Guizhou Province

Zitieren

APA

Zhang, W., Xie, B., Ma, J., Zhang, X., Huang, F., & Wang, C. (2026). Rapid assessment of soil nutrients under continuous Gastrodia elata cropping using near- and mid-infrared spectroscopy combined with multi-strategy machine learning. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 363, 128453. https://doi.org/10.1016/j.saa.2026.128453

Vancouver

Zhang W, Xie B, Ma J, Zhang X, Huang F, Wang C. Rapid assessment of soil nutrients under continuous Gastrodia elata cropping using near- and mid-infrared spectroscopy combined with multi-strategy machine learning. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy. 2026;363:128453. doi:10.1016/j.saa.2026.128453

BibTeX
@article{Zhang2026,
  author = {Zhang, Wentao and Xie, Baiheng and Ma, Jinfang and Zhang, Xiangyu and Huang, Furong and Wang, Caiyun},
  title = {Rapid assessment of soil nutrients under continuous Gastrodia elata cropping using near- and mid-infrared spectroscopy combined with multi-strategy machine learning},
  journal = {Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy},
  year = {2026},
  volume = {363},
  pages = {128453},
  doi = {10.1016/j.saa.2026.128453}
}
Kostenlos starten →