Hyperspectral-derived spectrotransfer functions for soil properties estimation in arid lands

Introduction Visible and near-infrared (Vis–NIR) spectroscopy (350–2,500 nm) offers a rapid, non-invasive alternative to traditional soil analysis. Methodology This study applied (Vis–NIR) spectroscopy to estimate physical and fertility-related attributes in 188 soil samples from an arid region in Egypt’s Qena Governorate. Reference methods determined sand, silt, clay, available nitrogen (AN), phosphorus (AP), and potassium (AK). Spectral reflectance was acquired under controlled laboratory conditions using an ASD spectroradiometer, and data were analyzed with Unscrambler X software (version 10.4). Partial least squares regression (PLSR) and stepwise multiple linear regression (SMLR) were used to calibrate and validate Spectro - transfer functions (STFs). PLSR regression coefficients (B values) identified statistically significant reflectance bands across the full spectral range, which were subsequently used to construct STFs. The predictive performance of both PLSR models and STFs was evaluated for estimating soil attributes. Results and discussion Findings indicated that PLSR generally achieved higher accuracy than STFs for most target properties. The AN content was determined with an acceptable degree of precision, as demonstrated by the calibration R 2 of 0.82 and the R 2 of 0.77. Furthermore, the R 2 for the validated and calibrated data sets was 0.63 and 0.74, respectively, for AP, indicating that the R 2 was fair in the validated data set and very good in the calibrated data set. In relation to AP, the findings showed that the R 2 was 0.75 and the calibration R 2 of estimating AP was 0.82. As a result, the AP content was calculated with a reasonable degree of precision. For the calibrated dataset, the created STFs predicted sand, silt, and clay with R 2 values of 0.89, 0.87, and 0.37, respectively. In contrast, the validated dataset’s R 2 values were 0.85, 0.84, and 0.64. PLSR models outperformed STFs in predicting soil properties. Conclusion This efficient approach supports sustainable land management, crop selection, and land-use planning in arid regions.

Перевод пока недоступен