Visible-near-infrared absorbance spectroscopy for rapid estimation of leaf nitrogen contents of Philippine rice cultivars

Abstract

Accurately estimating the leaf nitrogen contents of rice leads to more economical use of fertilizers, and reduces negative impacts of high fertilization rates on the environment. This study was aimed at developing visible-near-infrared absorbance spectroscopy prediction models for leaf nitrogen contents of rice. A field experiment was established comprising of four Philippines lowland rice cultivars receiving four levels of urea nitrogen fertilizer from 0 to 240 kg-N ha−1. Leaf samples were taken over a 5-week period around the panicle initiation stage, immediately dried and pulverized. Some 100 samples were scanned using a Foss NIRSystem 6500 from 400 to 2,498 nm, and analyzed for Kjeldahl-nitrogen in a laboratory. The full wavelength model with no spectral pretreatment gave the best model with a standard error of performance (SEP) and R2 of 0.142% and 0.958, respectively. Re-scanning of the samples suggested a better model from 700 to 2,498 nm with standard normal variate spectral pretreatment with a standard error of cross-validation (SECV) and R2 of 0.117% and 0.973, respectively.