4. Conclusion
This study concludes a decreasing trend of spectral reflectance with increasing levels of salinity for SAS. The best probable wavelengths were identified that showed potential absorption characteristics to predict soil properties. Overall, hyperspectral remote sensing (350-2500 nm) showed acceptable to excellent prediction accuracy for soil pH, EC, BD, N, Mg, Zn and B. For SOC, P, K, Na, Ca, S, Fe, Mn and Cu, the prediction results were non-reliable. The best prediction was obtained for soil salinity (electrical conductivity). Multivariate techniques like SVR, PLSR and PCR were found more robust compared to MARS and RF for prediction of properties in SAS.
Conflict of interest: There is no conflict of interest