Unveiling the potential of proximal hyperspectral sensing for measuring herbage nutritive value in a pasture-based dairy farm system : a thesis presented in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Agriculture and Horticulture at Massey University, Manawatū, New Zealand
The aim of this thesis was to unveil the potential of proximal hyperspectral sensing for measuring herbage nutritive value in a pasture based-dairy farm system. Hyperspectral canopy reflectance and herbage cuts as well as data on herbage and supplement allocation, and milk production were collected regularly from Dairy 1 farm at Massey University during the 2016-17 and 2017-18 production seasons. Milk, fat and protein yields and body condition score of cows were measured at monthly herd tests while live weights were recorded daily. Calibration equations determining herbage the nutritive value traits digestible organic matter in dry matter, metabolisable energy (ME), crude protein, neutral detergent fibre and acid detergent fibre from hyperspectral canopy reflectance data were developed and validated using partial least squares regression. Canopy reflectance calibration models were able to determine the various herbage nutritive value traits with R2 values ranging from 0.57 to 0.78. Variation of herbage nutritive value traits were mostly explained by month within production season (42.7% of variance among traits) followed by random error (33.4%), production season (13.1%) and paddock (10.7%). The relative importance of herbage nutritive value and other herbage quantity and climate-related variables in driving performance per cow in the herd was determined using multiple linear regression. Herbage metabolizable energy explained 20% to 30% of milk, fat and protein production per cow while herbage quantity and climate- related factors were relatively less important (below 15%). Random regression models were used to model lactation curves of milk, fat, protein and live weight to estimate daily ME requirements of individual cows. The daily ME estimated requirements was nearly a fifth above or below the daily mean ME supplied. The deviation of the daily ME estimated requirements of a cow from the actual ME supplied per cow in the herd was mostly explained by the observations made within a cow rather than between cows or breeds. Variation in herbage nutritive value in addition to the within and between cow variation of ME estimated requirements were high enough to justify the use of proximal hyperspectral sensing as measurement tool to assist with feed allocation decision-making. However, the potential of this technology could be further enhanced using more precise technologies to allocate herbage to individual cows or groups of cows. The potential benefits of more precise feed allocation will result in more efficient grazing management and thus improved utilisation of herbage and hence milk production.