Performance of near infrared spectroscopy of a solid cattle and poultry manure database depends on the sample preparation and regression method used

Abstract

Determining the chemical composition of animal manure rapidly is essential to manage fertilisation and decrease environmental pollution. Near infrared (NIR) spectroscopy is a non-destructive, inexpensive and rapid method to determine several components of manure simultaneously. This study investigated the ability of NIR spectroscopy to analyse the dry matter, total and ammonium nitrogen, phosphorus, calcium, potassium and magnesium contents in a database of heterogeneous cattle and poultry solid manures. The accuracy of calibration models obtained from different sample preparation methods (dried ground vs. fresh homogenized) and multivariate regression methods (partial least squares (PLS) vs. local regression) were compared. The results showed that using local regression with NIR spectra of fresh homogenized manure could predict dry matter (R2=0.99, RMSEV = 1.64%, RPD = 13.31), total (R2=0.98, RMSEV = 0.16%, RPD = 7.11) and ammonium nitrogen (R2=0.97, RMSEV = 0.042%, RPD = 5.57) and phosphorus (R2=0.95, RMSEV = 0.10%, RPD = 5.56) contents accurately.