To test the applicability of faecal NIRS to real conditions, an experimental approach was undertaken across several representative dairy farms (N = 30) located ‘La Réunion’ island. From an ongoing survey, this approach consists to characterize the nutritional value of all feeds (grazed fresh forage, hay, silages and supplementary feeds) offered to the lactating herds, and to predict ingested diet from faecal NIRS models previously developed on a large experimental sheep faeces reference data-base.
The methodological objective was to evaluate if such a spectral database could be a useful reference to estimate dairy cow total dry matter intake and diet quality, and so predict the grazed grass intake with reasonable accuracy. According to preliminary results, the NIRS estimated total intake varied between 13.7 and 19 kg DM/day and in vivo organic matter digestibility ranged from 51.7 to 74.8 % with an averaged value of 66 %. The estimated grass intake varied between 0 to nearly 10 kg DM/d. On a spectral basis, dairy cows faeces were quite different from the sheep faeces reference database, with an averaged standardised distance (H) upper of 3.0 (H = 9.1; Hmin = 2.08 – Hmax = 19.22) but predicted intake appeared valid. Indeed, according to the feeding value of diets and lactating cow requirements, the NIRS predicted total intakes were well correlated to the level of milk production. Moreover, for four particular situations, the fresh grass was cut, distributed at the trough and total intake really measured. The correlation between predicted and measured values was high with R2 = 0.94 and standard error of regression = 0.469 kg DM/d. These initial results appear quite encouraging, although the methodology is still exploratory and needs to be validated across a larger set of data. As a low cost and rapid prediction technique, NIRS appears to be a potential methodology that could find many useful developments in the improvement of the knowledge of forage use in tropical conditions.