Diurnal observations of feeding choices in grazing horses correctly predict their daily diet composition

Although it is time consuming, performing direct observations of the feeding choices of grazing herbivores remains the reference method because it does not require any destructive vegetation sampling and limits bias associated with the use of indirect markers. It is also the only method for characterizing the use of vegetation patches. These observations are usually performed during the diurnal period only; however, equids graze 16 h per day on average, with 20–50% of their grazing occurring during the nocturnal period. The objective of this paper was to test the representativeness of diurnal observations of horses’ feeding choices to characterize their daily diet. We used an experimental dataset of 54 paired observations of the diurnal and daily (24-h period) diet composition in horses grazing a pasture with differing sward characteristics according to the stocking rate and season. Four classes of bites were considered based on the sward height and vegetation stage, which are key factors affecting selection by horses. The H0 indicated that the 24-h observations of horses’ feeding choices among the four bite types could be correctly predicted by diurnal observations. Using a chi-square test of adequacy to a multinomial distribution, we expected the set of variables to follow a chi-square distribution, with k-1 (k = four bite types) degrees of freedom, under H0. We first used a Kolmogorov-Smirnov test, which showed adequate observed values relative to the expected distribution (p = 0.67). Then, a Bayesian approach was used to estimate the chi-square distribution with parameter υ that best fit our data, and we found that the probability that H0 was true, which was defined as p (υ ≤ 3), was equal to p = 0.92. An analysis of the chi-square residuals did not reveal any bias in the prediction of the different bite types. Finally, the Bray-Curtis similarity indexes used to characterize prediction quality were very high (i.e., > 90%) regardless of the stocking rate and season. All these results indicate a high level of confidence in the hypothesis that the effects of sward structure on the daily diet composition of grazing horses can be predicted by diurnal observations. This methodological advance will help increase the number of observation days and animals in future studies and accurately define the number of factors that can be tested simultaneously.