研究目的
To assess the predictive performance of midDRIFTS-PLSR models in quantifying various soil properties, including carbon fractions, in calcareous soils of Iran.
研究成果
The midDRIFTS-PLSR technique effectively predicted soil carbon fractions (TC, TIC, TOC) with high accuracy in calcareous soils, demonstrating its potential as a rapid and cost-effective method for soil monitoring. However, predictions for other properties like nitrogen fractions and texture were less accurate. Future research should involve larger sample sizes and address spectral interferences to improve model robustness.
研究不足
The study had a relatively small sample size (68 samples), which may affect model generalizability. Interference from carbonates in calcareous soils could reduce prediction accuracy for some properties. The heterogeneity of soil samples and spectral overlap of organic and inorganic compounds posed challenges for model calibration.
1：Experimental Design and Method Selection:
The study used mid-infrared diffuse reflectance Fourier transform spectroscopy (midDRIFTS) coupled with partial least squares regression (PLSR) to predict soil properties. The methodology involved collecting soil samples, performing laboratory analyses, and developing calibration and validation models.
2：Sample Selection and Data Sources:
A total of 68 soil samples were collected from three land uses (forest, range, and crop lands) in Lorestan Province, southwest Iran, from surface horizons (0–30 cm).
3：List of Experimental Equipment and Materials:
Equipment included a Tensor-27 mid-infrared spectrometer with a praying mantis diffuse reflectance chamber, a Vario-EL III elemental analyzer, a Multi N/C analyzer, mechanical sieves, ball mill, oven, and various chemicals for soil analysis.
4：Experimental Procedures and Operational Workflow:
Soil samples were oven-dried, crushed, sieved, and ball-milled. MidDRIFTS analysis was performed with spectral scanning in the range of 4000–600 cm?1. Calibration and validation subsets were created, and PLSR models were optimized using preprocessing methods and frequency region selection.
5：Data Analysis Methods:
Data were analyzed using SPSS and SigmaPlot software for descriptive statistics, correlation analysis, and regression. Model accuracy was evaluated using R2, RPD, and RMSE.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
