Chemometric and multivariate statistical analysis of FTIR spectra for predicting electrical, physicochemical, and structural properties of saline soils

Soil monitoring is crucial for sustainable agriculture and environmental protection because changes in physical and chemical properties directly influence soil fertility, nutrient availability, water retention, and overall ecosystem stability. Salinity accumulates soluble salts in soil and disrupts soil structure, impairs plant water and nutrient uptake, and poses a serious threat to soil health and agricultural productivity. Soil properties like electrical, physical, and chemical were analyzed using Fourier Transform Infrared (FTIR) spectroscopy, an automated PC-based X-band microwave technique, and standard laboratory methods. Knowing these properties enables early detection of soil degradation and provides support in land management. To enhance the prediction of soil properties, Chemometric and multivariate statistical methods, including Principal Component Analysis (PCA), Hierarchical Cluster Analysis (HCA), and Partial Least Squares Regression (PLSR), were applied to datasets derived from normalized first-order derivative FTIR spectra. The PCA findings show 96.13% of the overall variation in soil minerals and organic matter. Further, the Partial Least Squares (PLSR) regression approach was used to correlate the FTIR spectral data with the electrical and physicochemical properties of the soil. The results show high to very high predictive performance (R²) with acceptable Root-Mean-Square Errors (RMSE), further supported by RPIQ values indicating excellent prediction for dielectric constant (RPIQ = 4.29) and A.C. conductivity (RPIQ = 4.03), and comparatively lower performance for dielectric loss (RPIQ = 1.41) and tangent loss (RPIQ = 1.91). For the electrical properties of soil, dielectric constant (R² = 0.954) and A.C. conductivity (R² = 0.953) exhibit higher predictive power compared to dielectric loss (R² = 0.801) and tangent loss (R² = 0.741). In the physicochemical property models, calcium (R² = 0.988) shows the highest predictive power, while phosphorus (R² = 0.751) shows the lowest. The FTIR spectroscopy results coupled with chemometric predictive models provides a rapid, cost-effective, and reliable approach for assessing soil health, making it a valuable tool for large-scale monitoring and sustainable agriculture land management.