Development and validation of an optimized spiking strategy for cross‐layer transfer of near‐infrared soil organic matter models
Soil Science Society of America Journal · 2026-11 · 90 (6)
DOI: 10.1002/saj2.70388 ↗Kurz gefasst
Fachartikel in der Zeitschrift Soil Science Society of America Journal, online erschienen am 06.11.2026. Verfasst von Yinuo Kong und 7 weiteren Autor:innen. Die Arbeit nennt 49 Literaturangaben. Gefördert von: National Natural Science Foundation of China.
Abstract (Originalsprache)
Soil organic matter (SOM) is crucial for soil fertility and is strongly linked to soil productivity. The heterogeneity of soil at different depths leads to variations in spectral characteristics. This study addresses the challenge of using near‐infrared (NIR) spectroscopy to estimate SOM across heterogeneous soil profiles. Analyzing spectral data from soil cores (≈150 cm), samples were categorized into surface (0–30 cm), subsurface (>30 cm), and complete profile datasets. Optimal models for each layer were achieved using partial least squares regression with Savitzky–Golay smoothing and variable selection via the uninformative variable elimination‐successive projections algorithm. These layer‐specific models accurately predicted SOM within their own layer but performed poorly when directly applied to the contrasting layer. The spiking method significantly improved this cross‐layer prediction. To balance cost and accuracy, we recommend incorporating 20 (surface) and 30 (subsurface) spiked samples into the respective calibration sets. This increased the prediction coefficient of determination () by 11.0% and 1.6% and reduced the root mean square error of prediction by 29.7% and 6.7%, compared to unspiked models. The effectiveness of spiking was inversely related to calibration set size, and an extra‐weighting technique offered no additional benefit for small sets. Furthermore, calibration models constructed from the combined profile dataset generally provided more robust predictions across various depths than models based on individual layers. This work provides a practical framework for accurate SOM estimation throughout the soil profile and validates spiking as an effective strategy to enhance the transferability of NIR spectroscopy models between distinct soil layers.
Förderung
- National Natural Science Foundation of China
Zitieren
APA
Kong, Y., Chen, X., Zheng, W., Li, R., Jin, C., Yin, X., Hao, S., & Zhou, P. (2026). Development and validation of an optimized spiking strategy for cross‐layer transfer of near‐infrared soil organic matter models. Soil Science Society of America Journal, 90(6). https://doi.org/10.1002/saj2.70388
Vancouver
Kong Y, Chen X, Zheng W, Li R, Jin C, Yin X, et al. Development and validation of an optimized spiking strategy for cross‐layer transfer of near‐infrared soil organic matter models. Soil Science Society of America Journal. 2026;90(6). doi:10.1002/saj2.70388
BibTeX
@article{Kong2026,
author = {Kong, Yinuo and Chen, Xiyuan and Zheng, Weijian and Li, Ronghao and Jin, Chengqian and Yin, Xiang and Hao, Shanshan and Zhou, Peng},
title = {Development and validation of an optimized spiking strategy for cross‐layer transfer of near‐infrared soil organic matter models},
journal = {Soil Science Society of America Journal},
year = {2026},
volume = {90},
number = {6},
doi = {10.1002/saj2.70388}
}
EUVIMED