<![CDATA[Plants need an ideal and healthy soil condition for their growth and a sufficient amount of soil macronutrients. To determine soil nutrients, several methods have been widely employed. Yet, most of them are based on solvent extraction, which is normally time-consuming, requires complicated sample preparation, and sometimes involves chemical materials. Thus, a novel, fast and simultaneous method is required as an alternative method used to predict soil macronutrients in a short period and without involving chemical materials. Near infrared spectroscopy (NIRS) can be considered for this need, since this method is fast, environmentally friendly, and non-destructive. Therefore, the main objective of this study is to apply an NIRS method to predict soil macronutrients (N, P, and K). The diffuse reflectance spectrum was acquired for soil samples in a wavelength range from 1000–2500 nm. Spectra data were corrected using a smoothing method whilst prediction models were developed using principal component regression (PCR) and partial least square regression (PLSR). Prediction accuracy and robustness were evaluated using these following statistical indicators: correlation coefficient (r), root mean square error (RMSEC) and residual predictive deviation (RPD). The results showed that NIRS was able to predict soil macronutrients simultaneously with a maximum correlation coefficient r = 0.97 for N prediction, r = 0.99 for P prediction, and r = 0.95 for K prediction. Thus, it may be concluded that an NIRS method is feasible to be applied as a novel, reliable and fast method to predict soil macronutrients (N, P, and K) simultaneously.]]>
Copyrights © 2019
