Hyperspectral Prediction of Salt in Soil by PLS Regression Method: A Research

Abstract

Soil salinization becomes universal problems mostly in arid and semiarid irrigated agriculture areas. Soil salinity is a process which affects the quality of soil and reduces crop yields and agriculture production. The soil salt content adversely affects the soil physical property including soil water content. The Visible and Near-Infrared Reflectance Spectroscopy provides improved estimation of soil salinity as fast approach to the characterization of soil salt content with spectral resolution of 350-2500 nm. The Partial Least Square Regression Method (PLSR) is frequently used to determine Soil Salt Content (SSC) obtains from the spectral data. The Result shows that the 550nm, 850 nm, 1430nm, 1918nm, 2052nm wavelength which is highly sensitive to salt content and model based on Partial Least Square Regression PLSR can only make approximate predictions for First Derivative RMSE (Root Mean Square Error) = 0.0282-0.0365, R2 (Coefficient of Determination) = 0.9313-0.9051 and for Continuum Remove RMSE (Root Mean Square Error) = 0.0280-0.0386, R2 (Coefficient of Determination) = 0.9313-0.8939.