Olivier Martin

Heatwaves increasingly threaten dairy production stability in the context of global warming. While the effects of heat stress on cows are well documented, the responses of herds to such hazards are poorly described. The main objectives of this study were to assess whether annual heat load has a quantifiable impact on the annual performance of dairy herds and to establish which herd characteristics influence the herd-scale response. Using a bottom-up approach linking hypotheses on animal performance to herd-scale outcomes, we quantified herd response and capacity to thrive during heatwaves (resilience) in 3,393 commercial dairy herds located in 4 contrasting climatic areas in France. Herd data were available from 12 to 24 years during the period 2000-2023. The impacts of heat load on annual herd performance variables (milk yield, fat and protein contents, duration of lactation, calving interval, recalving number, productive longevity of cows and herd size) were studied using the Cumulative Annual Temperature-Humidity Index (CATHI). This index, which assesses the annual heat load level at the farm location, was developed specifically for this study. A multivariate regression was used to study the marginal effects of CATHI on herd performance according to the herd's main calving season, geographical area, average milk production level, primiparous rate, and herd size. We found that, on an annual scale, heat load negatively impacted herds in terms of several key performance measures: average milk yield, lactation duration, recalving number, productive longevity and herd size. Herds located in areas with low exposure to heatwaves (lower CATHI values) showed declines in performance at least twice as large compared with herds located in areas more often exposed to heat stress, suggesting climatic acclimatization or climate adaptation. Calving seasonality also influenced the response of herds to annual heat load: herds with year-round calving showed 40% less performance loss than herds with grouped winter or fall calving. As expected, the effect of annual heat load on herd performance was greater for higher-producing herds. Breed composition, the proportion of primiparous cows in the herd, and herd size had little or no impact on herd response to annual heat load. These results demonstrate that both regional climatic conditions and calving distribution over the year modulated dairy herd response to annual heat load. Understanding how the coping abilities of cows and management flexibility interact in the herd could provide actionable levers for designing dairy herds that are more resilient to climate change.

We developed a clinical named entity recognition model to predict clinical relevance of pharmacist interventions (PIs) by identifying and labelling expressions from unstructured comments of PIs. Three labels, drug, kidney and dosage, had a great inter-annotator agreement (>60%) and could be used as reference labelization. These labels also showed a high precision (>70%) and a variable recall (50-90 %).

Automated monitoring of fertility in dairy cows using milk progesterone is based on the accurate and timely identification of luteolysis. In this way, well-adapted insemination advice can be provided to the farmer to further optimize fertility management. To properly evaluate and compare the performance of new and existing data-processing algorithms, a test data set of progesterone time-series that fully covers the desired variability in progesterone profiles is needed. Further, the data should be measured with a high frequency to allow rapid onset events, such as luteolysis, to be precisely determined. Collecting this type of data would require a lot of time, effort, and budget. In the absence of such data, an alternative was developed using simulated progesterone profiles for multiple cows and lactations, in which the different fertility statuses were represented. To these, relevant variability in terms of cycle characteristics and measurement error was added, resulting in a large cost-efficient data set of well-controlled but highly variable and farm-representative profiles. Besides the progesterone profiles, information on (the timing of) luteolysis was extracted from the modeling approach and used as a reference for the evaluation and comparison of the algorithms. In this study, 2 progesterone monitoring tools were compared: a multiprocess Kalman filter combined with a fixed threshold on the smoothed progesterone values to detect luteolysis, and a progesterone monitoring algorithm using synergistic control, PMASC, which uses a mathematical model based on the luteal dynamics and a statistical control chart to detect luteolysis. The timing of the alerts and the robustness against missing values of both algorithms were investigated using 2 different sampling schemes: one sample per cow every 8 h versus 1 sample per day. The alerts for luteolysis of the PMASC algorithm were on average 20 h earlier compared with the ones of the multiprocess Kalman filter, and their timing was less sensitive to missing values. This was shown by the fact that, when 1 sample per day was used, the Kalman filter gave its alerts on average 24 h later, and the variability in timing of the alerts compared with simulated luteolysis increased with 22%. Accordingly, we postulate that implementation of the PMASC system could improve the consistency of luteolysis detection on farm and lower the analysis costs compared with the current state of the art.

Background: Understanding the effects of environment on livestock provides valuable information on how farm animals express their production potential, and on their welfare. Ruminants often face perturbations that affect their performance. Evaluating the effect of these perturbations on animal performance could provide metrics to quantify how animals cope with their environment and therefore, better manage them. In dairy systems, milk production records can be used to evaluate perturbations because (1) they are easily accessible, (2) the overall dynamics throughout the lactation process have been widely described, and (3) perturbations often occur and cause milk loss. In this study, a lactation curve model with explicit representation of perturbations was developed. Methods: The perturbed lactation model is made of two components. The first one describes a theoretical unperturbed lactation curve (unperturbed lactation model), and the second describes deviations from the unperturbed lactation model. The model was fitted on 319 complete lactation data from 181 individual dairy goats allowing for the characterization of individual perturbations in terms of their starting date, intensity, and shape. Results: The fitting procedure detected a total of 2,354 perturbations with an average of 7.40 perturbations per lactation. Loss of production due to perturbations varied between 2% and 19%. Results show that the number of perturbations is not the major factor explaining the loss in milk yield over the lactation, suggesting that there are different types of animal response to challenging factors. Conclusions: By incorporating explicit representation of perturbations, the model allowed the characterization of potential milk production, deviations induced by perturbations (loss of milk), and thereby comparison between animals. These indicators are likely to be useful to move from raw data to decision support tools in dairy production.

Automated monitoring of fertility in dairy cows using milk progesterone is based on the accurate and timely identification of luteolysis. In this way, well-adapted insemination advice can be provided to the farmer to further optimize the fertility management. To properly evaluate and compare the performance of new and existing data-processing algorithms, a test dataset of progesterone time-series that fully covers the desired variability in progesterone profiles is needed. Further, the data should be measured with a high frequency to allow rapid onset events, such as luteolysis, to be precisely determined. Collecting this type of data would require a lot of time, effort and budget. In the absence of such data, an alternative was developed using simulated progesterone profiles for multiple cows and lactations, in which the different fertility statuses were represented. To these, relevant variability in terms of cycle characteristics and measurement error was added, resulting in a large cost-efficient dataset of well-controlled but highly variable and farm-representative profiles. Besides the progesterone profiles, information on (the timing of) luteolysis was extracted from the modelling approach and used as a reference for the evaluation and comparison of the algorithms. In this study, two progesterone monitoring tools were compared: a multiprocess Kalman filter combined with a fixed threshold on the smoothed progesterone values to detect luteolysis, and a progesterone monitoring algorithm using synergistic control ‘PMASC’, which uses a mathematical model based on the luteal dynamics and a statistical control chart to detect luteolysis. The timing of the alerts and the robustness against missing values of both algorithms were investigated using two different sampling schemes: one sample per cow every eight hours versus one sample per day. The alerts for luteolysis of the PMASC algorithm were on average 20 hours earlier compared to the ones of the multiprocess Kalman filter, and their timing was less sensitive to missing values. This was shown by the fact that, when one sample per day was used, the Kalman filter gave its alerts on average 24 hours later, and the variability in timing of the alerts compared to simulated luteolysis increased with 22%. Accordingly, we postulate that implementation of the PMASC system could improve the consistency of luteolysis detection on farm and lower the analysis costs compared to the current state of the art. Interpretative Summary: Validation of luteolysis monitoring tool for dairy cows. Adriaens: In this study, the performance of two monitoring algorithms to detect luteolysis using milk progesterone measurements was validated on a simulated dataset of realistic milk progesterone profiles. The synergistic control-based algorithm, PMASC, was able to identify luteolysis almost simultaneously with its occurrence. It was found to be more robust against missing samples and less dependent on the absolute milk progesterone values compared to a multiprocess Kalman filter combined with a fixed threshold. This research showed that implementation of PMASC could improve progesterone-based fertility monitoring on farm.