Modelling lactation in pasture-based dairy cows varying in production potential

The main objective of the thesis was to quantify the genetic and phenotypic determinants of variation in milk yield and composition, model the lactation pattern of pasture-based dairy cows varying in genetic potential for milk production and the comparative evaluation of empirical, mechanistic and random regression models as tools for management decisions. More than half a million lactation and pedigree data from Tasmanian dairy farms were sourced mainly from the TasHerd Milk Recording Organisation and the Elliott Research and Demonstration Station. The data were analysed using non-linear, generalised linear, mixed linear, multi-trait and random regression procedures in SAS and ASReml. Initial and the incline to peak but not peak and total milk yield were significantly influenced by sire EBV choice. Early lactation milk yield potential was highly correlated with peak and total milk yield and could be used as an early indicator of a cow's genetic merit. Genetic (sire estimated breeding value (EBV) and cow production level in early lactation), physiological (age, parity and body weight), environmental (season and year of calving, lactation stage, nutrition and herd), factors influenced production traits. In addition days to first test-day post-partum, lactation length, number of test-days and their interactions affected curve shapes. Heritability of 305d milk, fat, protein and somatic cell counts were 0.41, 0.37, 0.32 and 0.28 respectively. Phenotypic correlations between milk and component yields ranged from -0.03 to 0.92, while genetic correlations ranged between 0.034 and 0.85. Fourteen lactation functions including 8 empirical, 4 mechanistic and 2 semi-parametric types were fitted to test-day milk and milk composition yields. Empirical models adequately modeled the lactation of homogeneous group of cows but had varying error biases in fitting individual cow's profiles. Random regression, including cubic spline, models attained acceptable goodness of fit and permitted simultaneous evaluation of factors affecting curve shapes. Significant contributions of the thesis to lactation modeling are the identification of suitable functions and the introduction of a new empirical model for pasture-based systems. High positive correlation between parameter c of this model with peak milk yield and lactation persistency suggests that it has the potential for future dairy genetic improvement. The knowledge of factors affecting curve shapes in pasture-based systems will be relevant in developing appropriate management strategies to mitigate early lactation production stress and maintain persistency. Desirable as it is none of the tested mechanistic functions performed well. Suggestions for future work are; further research into the potential of existing mechanistic models to fit data across production systems and establishing a basis for understanding the physiological basis of empirical models. Lack of herd level management input inconsistent data recording pattern and incomplete test-date records were major obstacles of the study. Similarly, lack of economic indices made profitability modelling and overall farm economic analysis difficult. These constitute gaps in the current lactation data collation systems.