Hamidreza Matinfar; H Mahmodzadeh; A Fariabi
Volume 10, Issue 2 , September 2018, , Pages 15-32
Abstract
Soil organic matter is one of the most important Physical and chemical properties of soil that it iscritical in determining the quality and management of soils. Quantify of soil organic carbon due to thehigh spatial variability and changes over time is difficult. Near-infrared-visible spectroscopy is ...
Read More
Soil organic matter is one of the most important Physical and chemical properties of soil that it iscritical in determining the quality and management of soils. Quantify of soil organic carbon due to thehigh spatial variability and changes over time is difficult. Near-infrared-visible spectroscopy is afeasible method to reduce the time and cost to check the organic carbon. The aim of this study was toevaluate soil organic carbon through near-infrared and visible spectroscopy with the statistical modelsPLSR and PCR. For this purpose, 40 soil samples from depths of 0 to 30 cm were collected bysystematic random method based on previous studies and determination of different classes of soils inthe region. Chemical analysis of soils was performed according to standard methods. Spectralreflectance of soil samples in the range of 350 to 2500 nm was measured then after applying thepreprocessing methods such as Savitzky and Golay filter, Soil organic carbon were calculated byprincipal component analysis (PCA), regression partial least squares (PLSR) and principal componentregression (PCR) models. The results of this study showed that the Savitzky and Golay filter was thestrongest preprocessing method for spectral data. Coefficients of determination (R2), root mean squareerror of Prediction (RMSE) and ratio of prediction to deviation (RPD) in the calibration andvalidation to predict organic matter, respectively, 0.97, 0.05, 5.09 and 0.85, 0.14, 2.78 respectively.Therefore, for dry and semi-arid soils of the PLSR model, it is more efficient to predict the organiccarbon of the soil. The results showed that the PLSR model has better performance than the PCRmodel in soil organic carbon estimation.
