Precision Nutrition Tools to Support Extensive Forage-Based Livestock Systems

Abstract

Precision nutrition is the future of the livestock nutrition industry and is expected to improve the health and performance of animals through individualized feeding and management. Several developments in precision nutrition exists for ration formulation, feedstuff analysis, and individualized feeding, particularly in intensive livestock systems. Although research in precision animal nutrition exists, there are few practical tools for grazing animals in pasture-based systems. The overarching goal of this work was to develop and evaluate tools to support precision nutrition for extensive, forage-based livestock systems. This goal was addressed through three complementary studies exploring relevant tools. In a preliminary project, the objective was to assess the accuracy of a handheld spectral sensing device for predicting the dry matter (DM), neutral detergent fiber (NDF), acid detergent fiber (ADF), and crude protein (CP) composition of hay. A follow-on experiment pilot tested the spectral sensing device for on-animal use to monitor the composition of hay consumed during normal feeding behavior. We explored these objectives through time-series observations of forage sampling of mixed-grass hay, which was scanned with a spectral sensor programmed to measure light reflectance at 18 wavelengths before bench chemistry analysis for each sample. The data collected were split into three parts and used in random forest regressions. We found that the resulting root mean square prediction errors (RMSPE) for each of the four models were promising, especially for the two fiber fractions, with the lowest error rates of 5.85% for NDF and 8.05% for ADF. We investigated the following objective by placing spectral sensors on halters worn by horses consuming hay and comparing the spectral readings to forage samples collected from where the animal was eating. Although a small dataset, the mounted sensor system showed promising results, with RMSPEs of 8.02% (DM), 5.07% (NDF), 4.52% (ADF), and 23.5% (CP). Further development on the halter system as well as extensive data collection on grazing animals and a variety of forage is necessary for confirming the practicality of this technology. In the second project, our objective was to perform a quantitative literature review to investigate dietary and feed factors affecting total-tract digestibility of dry matter (DMD), crude protein (CPD), neutral detergent fiber (NDFD), ether extract (EED), non-structural carbohydrates (NSCD), non-fiber carbohydrates (NFCD), and residual organic matter (rOMD). Additionally, we aimed to assess how our equations behaved for estimating digestible energy (DE) compared with existing modeling systems and to evaluate them against independently measured DE from the literature. We explored this objective through a literature review that yielded 54 studies, which were used to develop linear mixed-effect regressions, with five models derived for each nutrient using several explanatory variables. Models were selected based on their ability to explain dataset variation and stability when predicting DE in example rations. Two models were developed for DE estimation: one using measured data from the literature, and another using both measured data and calculated data from reference tables when values were not provided in the literature. We found that the models explained variation well. When evaluated against measured DE from 17 studies, our calculated system provided DE estimations similar to those of existing systems. Overall, this new approach offers an additional, practical tool for estimating energy supplies in equine diets. In the final study, our objective was to determine whether thermal imaging of body surface temperature could provide an objective means of body condition scoring (BCS) in mature horses of the Quarter Horse (QH) and Thoroughbred (TB) breed types, as well as in multiparous gestating beef cows. We explored this objective by capturing thermal images on one or both sides of each animal's body while five to eight trained scorers assigned BCS. Several covariates were monitored for their influence on assigned BCS, including cloud coverage, animal breed, individual scorer, and scorer's location. Random forest regressions were derived to evaluate the ability of the thermal camera to accurately BCS horses and cows with data split into three parts. After models were created for each body region, we found that the root mean square error (RMSE, % mean) ranged from 7.6% to 10.6% for horses and 6.81% to 13.4% for cows. We also assessed between-scorer variability by calculating the coefficient of variation (CV). The variability among the eight horse scorers ranged from 13% to 14% and 10.8% to 12.1% for the five cow scorers. Using surface temperature obtained through thermal imaging displays promise as an alternative method to objectively BCS horses and beef cows.