The use of near-infrared spectroscopy (NIRS) to determine the energy value of individual feedstuffs and mixed diets for pigs

Abstract

To have quick and accurate feed quality information of ingredients and diets is of outmost importance to ensure that optimal nutritional requirements are met without providing an overabundance of nutrients, which increases costs and environmental pollution in wastes. Calibration models were developed and validated from near-infrared spectroscopy (NIRS) scans of 762 feedstuffs that have measured chemical composition and were assessed in balance experiments with pigs for measurements of the digestibility of dry matter, organic matter, nutrients, energy (DDE), and metabolizable energy (ME). For model development, the samples were grouped into ‘like’ sample types consisting of cereals, supplemental ingredients and feed mixtures (diets) as well as one group (total) to determine if it was possible to have one model for individual feedstuffs and mixed diets. The total models for most chemical constituents, digestibility of nutrients, DDE, and ME had good fits between predicted and measured values with R2 in the order of 0.90–0.99 for most chemical constituents, 0.86 for the DDE, and 0.84 for ME. The accuracy of the total models (SEP) varied between 8.7 and 48.6 g/kg DM for the chemical constituents, between 2.2% and 12.6% for digestibility of different nutrients, and between 0.33 and 0.68 MJ/kg DM for gross energy and ME, respectively. A standard equation was used to estimate ME (eME) from digestible fat, crude protein, crude fiber, and nitrogen-free extract whereby the R2 of eME was improved from 0.89 when eME was calculated from table value components and table digestibility estimates to 0.94 when estimated from NIRS-estimated components and NIRS digestibility nutrient estimates or estimated directly from NIRS calibration. For comparison, eME calculated from measured components and measured digestibility was 0.98. Models developed on the total dataset, including individual ingredients and diets, performed well and offers a greater robustness by representing a larger range of samples; however, models developed on just the cereals had greater precision. Near-infrared spectroscopy has proven to be a useful tool to estimate biological values associated with pig feedstuffs such as the DDE and ME where NIRS estimates were more accurate than the currently used estimation methods and can be used to quickly and inexpensively assess pig feedstuffs in the future.