Dairy cow resilience: Data-driven detection and quantification of perturbations using daily milk weights.

Resilience is an animal's capacity to recover after a perturbation or maintain specific functions under stress. The increasing frequency of extreme weather events and labor shortages highlight the need to identify and select animals that can maintain production in unpredictable environments. The US dairy herds group cows into pens based on factors such as parity, lactation stage, reproductive status, and milk production. By coupling daily milk yield data with day-to-day pen location information, we can model management and environmental stressors affecting groups of cows and detect perturbations at the pen level. It also allows us to detect management and environmental perturbations that may occur each day at the pen level and subsequently measure the responses of individual cows to these stressors over a particular period. Our entire dataset included 62,580,945 daily milk weights and pen locations of 227,633 cows in parity 1, 2, or 3 from 204 herds representing 30 US states from 2018 to 2024. Individual lactation curves were fitted using polynomial quantile regression (0.5 quantile) to derive expected lactation curves. Perturbations were identified using a data-driven detection method and were based on residuals between mean expected and mean observed daily milk weights at the pen level. The initial dataset was stratified based on the severity and duration of the perturbation periods, considering 40 combinations that included severity levels from ≥3% to ≥7% milk yield loss and durations ranging from ≥3 to ≥10 d. Our resilience phenotype, delta milk yield (ΔMY) measured the change in a cow's mean daily milk production relative to her mean expected daily milk yield during an identified perturbation period. Variance components, heritabilities, and repeatabilities were estimated using a model with fixed effects for calving age, DIM, parity, and herd-year-season, and random effects for perturbation event, additive genetics, and permanent environment. Sire PTA correlations with TempVar (milk consistency across the full lactation) were calculated using the Calo's method to assess the relationship between resilience and consistency traits. Additionally, sire PTA Pearson correlations were estimated within comparable severity thresholds to determine the genetic correlations between sire PTA during perturbations with similar severities. Estimated h of ΔMY during perturbations ranged from 0.01 (0.00) to 0.20 (0.08), depending on the severity and duration of the perturbation, while sire PTA correlations between ΔMY and TempVar ranged from -0.51 (0.01) to -0.16 (0.03), indicating that more consistent cows have lower milk loss during perturbations. Our findings suggest that animals differ in their response to perturbations at the pen level in comparison to their contemporaries within the pen, and this measure of resilience using daily milk data is heritable. Identifying perturbations of varying severity and duration at the pen level can more effectively capture the management and environmental conditions affecting an individual cow, and resilience can be measured by comparing how her response differs from that of her contemporaries when exposed to stressful conditions. This enables the selection and management of more adaptable and sustainable cows capable of handling diverse challenges through a data-driven approach to detecting perturbations.