Comparing Regression Models based on Soil Moisture States using NIR Spectroscopy

Soil affects the quality and productivity of crops. Understanding the properties of soil is important for producing healthy crops. Conventional soil analysis methods are known for their precision, yet they tend to be inefficient due to their time-consuming and costly nature, primarily conducted within laboratory settings. Hence, research is underway to predict soil properties rapidly and non-destructively through the interaction between electromagnetic radiation and the soil surface, based on techniques like Diffuse Reflectance Spectroscopy (DRS). Yet, the accuracy of soil property analysis can be affected by various factors like soil moisture content and texture. The soil samples collected for this study were obtained from salty paddy fields in Hwaseong-si, Gyeonggi-do, South Korea. The measured parameters included pH, Electrical Conductivity (EC), Mg²⁺, Ca²⁺, Soil Organic Matter (SOM), Total Nitrogen (TN), Total Organic Carbon (TOC), Silt, Clay and Moisture Content (MC). Spectral data from both moist and dry soil were collected using the ASD Field Spec PRO4 spectrometer. The collected spectral data underwent preprocessing using the Standard Normal Variate (SNV) technique, followed by Partial Least Squares Regression (PLSR) analysis. The soil samples showed improved RPD values after drying, indicating that appropriate adjustments for soil moisture content could enhance the PLSR analysis model for soil property prediction based on DRS in salty paddy fields.