Cow key point detection in indoor housing conditions with a deep learning model.

Lameness in dairy cattle is a costly and highly prevalent problem that affects all aspects of sustainable dairy production, including animal welfare. Automation of gait assessment would allow monitoring of locomotion in which the cows' walking patterns can be evaluated frequently and with limited labor. With the right interpretation algorithms, this could result in more timely detection of locomotion problems. This in turn would facilitate timely intervention and early treatment, which is crucial to reduce the effect of abnormal behavior and pain on animal welfare. Gait features of dairy cows can potentially be derived from key points that locate crucial anatomical points on a cow's body. The aim of this study is 2-fold: (1) to demonstrate automation of the detection of dairy cows' key points in a practical indoor setting with natural occlusions from gates and races, and (2) to propose the necessary steps to postprocess these key points to make them suitable for subsequent gait feature calculations. Both the automated detection of key points as well as the postprocessing of them are crucial prerequisites for camera-based automated locomotion monitoring in a real farm environment. Side-view video footage of 34 Holstein-Friesian dairy cows, captured when exiting the milking parlor, were used for model development. From these videos, 758 samples of 2 successive frames were extracted. A previously developed deep learning model called T-LEAP was trained to detect 17 key points on cows in our indoor farm environment with natural occlusions. To this end, the dataset of 758 samples was randomly split into a train (n = 22 cows; no. of samples = 388), validation (n = 7 cows; no. of samples = 108), and test dataset (n = 15 cows; no. of samples = 262). The performance of T-LEAP to automatically assign key points in our indoor situation was assessed using the average percentage of correctly detected key points using a threshold of 0.2 of the head length (PCKh0.2). The model's performance on the test set achieved a good result with PCKh0.2: 89% on all 17 key points together. Detecting key points on the back (n = 3 key points) of the cow had the poorest performance PCKh0.2: 59%. In addition to the indoor performance of the model, a more detailed study of the detection performance was conducted to formulate postprocessing steps necessary to use these key points for gait feature calculations and subsequent automated locomotion monitoring. This detailed study included the evaluation of the detection performance in multiple directions. This study revealed that the performance of the key points on a cows' back were the poorest in the horizontal direction. Based on this more in-depth study, we recommend the implementation of the outlined postprocessing techniques to address the following issues: (1) correcting camera distortion, (2) rectifying erroneous key point detection, and (3) establishing the necessary procedures for translating hoof key points into gait features.

Productive life span and resilience rank can be predicted from on-farm first-parity sensor time series but not using a common equation across farms.

A dairy cow's lifetime resilience and her ability to recalve gain importance on dairy farms, as they affect all aspects of the sustainability of the dairy industry. Many modern farms today have milk meters and activity sensors that accurately measure yield and activity at a high frequency for monitoring purposes. We hypothesized that these same sensors can be used for precision phenotyping of complex traits such as lifetime resilience or productive life span. The objective of this study was to investigate whether lifetime resilience and productive life span of dairy cows can be predicted using sensor-derived proxies of first-parity sensor data. We used a data set from 27 Belgian and British dairy farms with an automated milking system containing at least 5 yr of successive measurements. All of these farms had milk meter data available, and 13 of these farms were also equipped with activity sensors. This subset was used to investigate the added value of activity meters to improve the model's prediction accuracy. To rank cows for lifetime resilience, a score was attributed to each cow based on her number of calvings, her 305-d milk yield, her age at first calving, her calving intervals, and the DIM at the moment of culling, taking her entire lifetime into account. Next, this lifetime resilience score was used to rank the cows within their herd, resulting in a lifetime resilience ranking. Based on this ranking, cows were classified in a low (last third), moderate (middle third), or high (first third) resilience category within farm. In total, 45 biologically sound sensor features were defined from the time series data, including measures of variability, lactation curve shape, milk yield perturbations, activity spikes indicating estrous events, and activity dynamics representing health events (e.g., drops in daily activity). These features, calculated on first-lactation data, were used to predict the lifetime resilience rank and, thus, to predict the classification within the herd (low, moderate, or high). Using a specific linear regression model progressively including features stepwise selected at farm level (cutoff P-value of 0.2), classification performances were between 35.9 and 70.0% (46.7 ± 8.0, mean ± SD) for milk yield features only, and between 46.7 and 84.0% (55.5 ± 12.1, mean ± SD) for lactation and activity features together. This is, respectively, 13.7 and 22.2% higher than what random classification would give. Moreover, using these individual farm models, only 3.5 and 2.3% of cows were classified high when they were actually low, or vice versa, whereas respectively 91.8 and 94.1% of wrongly classified animals were predicted in an adjacent category. The sensor features retained in the prediction equation of the individual farms differed across farms, which demonstrates the variability in culling and management strategies across farms and within farms over time. This lack of a common model structure across farms suggests the need to consider local (and evidence-based) culling management rules when developing decision support tools for dairy farms. With this study we showed the potential of precision phenotyping of complex traits based on biologically meaningful features derived from readily available sensor data. We conclude that first-lactation milk and activity sensor data have the potential to predict cows' lifetime resilience rankings within farms but that consistency between farms is currently lacking.

Symposium review: Future of housing for dairy cattle.

The objective of this review was to describe recent changes and expected developments in housing systems for dairy cows. These new developments should create an appropriate production environment for modern high-producing dairy cows and stimulate dairy farming-related developments in management, agro-technology, and equipment. Increased labor efficiency has been an important driver of the change from tie-stall barns to cubicle barns (also known as freestall barns). In future housing systems, the natural behavior of cows, climate control, emissions of ammonia and greenhouse gases, reuse of waste, manure quality, the aesthetics of buildings in the landscape, and capital efficiency are becoming increasingly important elements. To address future requirements, new concepts beyond cubicle barns must be developed. Freewalk housing systems; that is, loose housing systems without cubicles, would meet some of these future demands. These systems operate with composting bedding material or artificial permeable floors as lying and walking areas. However, these barns are still in development. Combinations of cubicle and freewalk housing systems, together with other techniques being developed, might become a major future housing system. Other techniques and systems that are being explored according to sustainability criteria include the multi-climate shed, the CowToilet (Hanskamp AgroTech, Doetinchem, the Netherlands) to separate feces and urine, and multifunctional buildings. These buildings and techniques can be part of land-based or, less commonly, city-based farming systems, such as floating farms.

Spatial distribution of load induced soft-tissue strain in cattle claws.

Claw disorders in dairy cattle have negative effects on both animal welfare and farm profits. One possible cause of claw disorders is the high mechanical load that cattle encounter when walking and standing on hard concrete floors. It is currently unclear how high mechanical loading leads to claw disorders and lameness. It is hypothesized that mechanical loading leads to compression of the soft tissue in the claws, which may directly or indirectly lead to tissue damage. Roentgen stereophotogrammetry in combination with CT-reconstruction was used to detect deformations in the distal hind limbs of dissected specimens of dairy cows under a range of loading regimens. The load was recorded in 3D using a force plate. Even at moderate load levels, such as during standing, the soft tissue layer was considerably compressed (>10% of the initial thickness), especially where the sole rests on the floor. Compression increases with increased and/or prolonged load. Most importantly, the location of areas of highest compression coincides with the locations where sole ulcers are most often found. These findings provide insight into the etiology of bovine claw disorders, and may contribute to solutions to reduce them.

A new technique using roentgen stereophotogrammetry to measure changes in the spatial conformation of bovine hind claws in response to external loads.

Claw and locomotion problems are widespread in ungulates. Although it is presumed that mechanical overload is an important contributor to claw tissue damage and impaired locomotion, deformation and claw injury as a result of mechanical loading has been poorly quantified and, as a result, practical solutions to reduce such lesions have been established mostly through trial and error. In this study, an experimental technique was developed that allowed the measurement under controlled loading regimes of minute deformations in the lower limbs of dissected specimens from large ungulates. Roentgen stereophotogrammetric analysis (RSA) was applied to obtain 3D marker coordinates with an accuracy of up to 0.1 mm with optimal contrast and to determine changes in the spatial conformation. A force plate was used to record the applied forces in three dimensions. The results obtained for a test sample (cattle hind leg) under three loading conditions showed that small load-induced deformations and translations as well as small changes in centres of force application could be measured. Accuracy of the order of 0.2-0.3 mm was feasible under practical circumstances with suboptimal contrast. These quantifications of claw deformation during loading improve understanding of the spatial strain distribution as a result of external loading and the risks of tissue overload. The method promises to be useful in determining load-deformation relationships for a wide variety of specimens and circumstances.

Genotype by environment interaction for livability of dairy calves from first parity cows.

Death of calves around parturition is a matter of concern for dairy farmers. Relatively high stillbirth rates and unfavourable trends have been reported for Holstein heifers in the Netherlands and several other countries. In our study, we investigated herd differences, genetic parameters and genotype by environment interaction for heifer calf livability. A large dataset with data from calvings between 1993 and 2012 of Dutch dairy farms was used. There were considerable differences between herds in livability of calves from heifers, with averages ranging from 74% to 95%. Both herds with relatively high and low averages showed the same negative trend between 1993 and 2012, with largest declines in herds with the lowest averages. We found that heritability and genetic variation of first parity livability were substantially larger in herd environments where the likelihood of stillbirth was high v. environments where stillbirth was at a low level. The genetic correlations between herd environment levels were all very close to unity, indicating that ranking of sires was similar for all environments. However, for herds with a relatively high stillbirth incidence selecting sires with favourable breeding values is expected to be twice as profitable as in herds with a relatively low stillbirth incidence.

Effects of flooring and restricted freestall access on behavior and claw health of dairy heifers.

Claw health, locomotion, feed intake, milk yield, body weight, activity, and lying and standing behavior of dairy heifers were monitored in a single dairy herd during the first 3 mo after calving. During the first 8 wk after calving, 2 treatments were applied: restricted freestall access by closing the stalls between 2300 h and 0500 h (yes or no) and alley flooring (concrete or rubber topped slatted floors). Apart from treatments, housing was identical. The animals were kept in small groups (n=4 to 6) in adjacent barn pens. Thereafter, the animals were kept in 1 group in a freestall section with concrete slatted floor and unrestricted access to the stalls for 5 wk. All animals were fed the same partial mixed ration. We hypothesized that (1) hard flooring causes high mechanical load of the claws and (2) restricted freestall access causes prolonged standing bouts and reinforced effects of hard flooring on claws. The heifers had only minor claw lesions before first calving, and the prevalence and severity of sole hemorrhages increased during the first 3 mo after calving (from 0.24 ± 0.08 to 1.18 ± 0.14 and from 0.04 ± 0.01 to 0.24 ± 0.02, respectively), particularly in the outer hind claws. Animals kept on rubber alley flooring had lower average hemorrhage scores in wk 9 (0.13 ± 0.03 vs. 0.21 ± 0.03) and wk 14 (0.20 ± 0.03 vs. 0.27 ± 0.03) after calving, had a slower feed intake (3.05 ± 0.14 vs. 3.46 ± 0.14 g/s) and spent more time feeding (7.3 ± 0.3 vs. 6.6 ± 0.3 min/h) than animals kept on hard concrete alley floors. Restricted freestall access resulted in fewer standing bouts per day (14.4 ± 1.0 vs. 17.9 ± 1.0) and more strides per hour (99.8 ± 5.4 vs. 87.2 ± 5.4) without changing overall standing time (15.0 ± 0.3 vs. 14.7 ± 0.3 h/d) and did not affect the occurrence of sole hemorrhages. The animals with no overnight freestall access spent more time standing (55.9 ± 0.9 vs. 35.8 ± 0.9 min/h) and feeding (7.8 ± 0.3 vs. 4.3 ± 0.3 min/h) between 2300 and 0500 h and less during the rest of the 24-h period (31.3 ± 0.8 vs. 37.0 ± 0.8 min/h and 6.8 ± 0.3 vs. 7.6 ± 0.3 min/h). Thus, the animals adapted to restricted freestall access, that caused increased overnight standing, by additional lying down during the day and used part of the extra standing time at night for feeding. The restrictions probably had only a minor effect on the mechanical load of their claws. Therefore, the first part of the hypothesis was confirmed and the second part was rejected.

Discriminating between true-positive and false-positive clinical mastitis alerts from automatic milking systems.

Automatic milking systems (AMS) generate alert lists reporting cows likely to have clinical mastitis (CM). Dutch farmers indicated that they use non-AMS cow information or the detailed alert information from the AMS to decide whether to check an alerted cow for CM. However, it is not yet known to what extent such information can be used to discriminate between true-positive and false-positive alerts. The overall objective was to investigate whether selection of the alerted cows that need further investigation for CM can be made. For this purpose, non-AMS cow information and detailed alert information were used. During a 2-yr study period, 11,156 alerts for CM, including 159 true-positive alerts, were collected at one farm in The Netherlands. Non-AMS cow information on parity, days in milk, season of the year, somatic cell count history, and CM history was added to each alert. In addition, 6 alert information variables were defined. These were the height of electrical conductivity, the alert origin (electrical conductivity, color, or both), whether or not a color alert for mastitic milk was given, whether or not a color alert for abnormal milk was given, deviation from the expected milk yield, and the number of alerts of the cow in the preceding 12 to 96 h. Subsequently, naive Bayesian networks (NBN) were constructed to compute the posterior probability of an alert being truly positive based only on non-AMS cow information, based on only alert information, or based on both types of information. The NBN including both types of information had the highest area under the receiver operating characteristic curve (AUC; 0.78), followed by the NBN including only alert information (AUC=0.75) and the NBN including only non-AMS cow information (AUC=0.62). By combining the 2 types of information and by setting a threshold on the computed probabilities, the number of false-positive alerts on a mastitis alert list was reduced by 35%, and 10% of the true-positive alerts would not be identified. To detect CM cases at a farm with an AMS, checking all alerts is still the best option but would result in a high workload. Checking alerts based on a single alert information variable would result in missing too many true-positive cases. Using a combination of alert information variables, however, is the best way to select cows that need further investigation. The effect of adding non-AMS cow information on making a distinction between true-positive and false-positive alerts would be minor.

Combining somatic cell count traits for optimal selection against mastitis.

Test-day records of somatic cell counts (SCC) can be used to define alternative traits to decrease genetic susceptibility to clinical mastitis (CM) and subclinical mastitis (SCM). This paper examines which combination of alternative SCC traits can be used best to reduce both CM and SCM and whether direct information on CM is useful in this respect. Genetic correlations between 10 SCC traits and CM and SCM were estimated from 3 independent data sets. The SCC traits with the strongest correlations with CM differed from those with the strongest correlations with SCM. Selection index calculations were made for a breeding goal of 50% CM and 50% SCM resistance using these correlations. They indicated that a combination of 5 SCC traits (SCC early and late in lactation, suspicion of infection based on increased SCC, extent of increased SCC, and presence of a peak pattern in SCC) gave a high accuracy, almost without loss, compared with the full set of 10 SCC traits. The estimated accuracy of this index was 0.91, assuming that the correlations had been estimated without error. To take errors in estimation into account, correlations were resampled from a normal distribution with mean and standard errors as originally estimated. The accuracy of the index calculated with the original correlations was then recalculated using the resampled correlations. The average accuracy based on 50,000 resamplings decreased to 0.81. Use of direct information on CM improved the accuracy (uncorrected for errors in correlations) only slightly, to 0.92.

Effects of two trimming methods of dairy cattle on concrete or rubber-covered slatted floors.

This study monitored claw health, claw conformation, locomotion, activity, and step traits of cows from a single dairy herd that were trimmed according to the standard Dutch method or with an alternative "concave" trimming method. Half of the cows were kept in a stall section with concrete slatted floors in the alleys. The other cows were kept in a pen within the same housing with an identical concrete slatted floor in the alleys, but with a rubber top layer. All experimental cows were kept in the same environment for at least 3 mo before and after trimming. It was hypothesized that trimming for more-concave soles (i.e., with 3 to 5 mm of sole dug out under the claw bone) was preferred to the standard Dutch trimming with flat sole surfaces for cows kept in stalls with soft alley floors. None of the claw health or locomotion traits differed for the trimming methods. No interactions were found between flooring and trimming method. Floor effects were significant for several traits. Cows on the rubber-topped floors had significantly fewer sole hemorrhages (prevalence of 22 vs. 48% in mo 3) and larger claws (claw length 76.1 +/- 5.0 vs. 72.5 +/- 4.9 mm; heel height 49.3 +/- 6.3 vs. 46.0 +/- 6.4 mm; claw diagonal 129 +/- 6.4 vs. 125 +/- 6.9 mm), spent more time standing in the alleys (55.4 +/- 2.8 vs. 49.6 +/- 2.8%), and had higher activity (61.0 +/- 3.7 vs. 53.0 +/- 3.7 steps/h). This suggests greater claw comfort on rubber flooring compared with concrete flooring. Kinetic patterns during claw-floor contact while walking were similar for all treatments. During the double-support (stance) phase, claw-floor contact area increased to a maximum in the first 30% of double-support phase time, remained more or less stable until 80% of double-support phase time, and sharply decreased as the animal pushed off as shown by the change in center of pressure. A gradual change of center of pressure in the medial direction during double-support phase time was shown. The research hypothesis was rejected, but soft alley floors had subtle beneficial effects.

Characterization of distributions of somatic cell counts.

There is more useful information in distributions of somatic cell count (SCC) than is currently used in practice. Analysis of SCC of individual quarters (n = 450,834 quarter records of 133,102 cows) showed that the presence of pathogens did not change the peak of the SCC distribution. Instead, the percentages of observations in the tail changed. Probability density functions of specified sets of up to 5 standard distributions were then fitted on the number of records per class, using a maximum likelihood procedure. Analysis of cow SCC (2 data sets: n = 335,135 test-day records of 41,567 cows on 407 farms and n = 1,665,431 test-day records) showed that a mixture of a normal, a log-normal and an exponential density function (N+LN+E) best described the distribution of SCC. A mixture of 4 normal and an exponential distribution (4N+E) was also a good approximation. For this last mixture, each distribution could be associated with presence or absence of pathogens. The first 2 normal distributions appear to consist of uninfected cows and cows recovering from an infection, the third normal distribution may be associated with minor pathogens, and the fourth normal and the exponential distribution with major pathogens and persistent infections. Estimated percentages of records in each underlying distribution differed between parities, between stages of lactation, and between records with previous records being above or below 100,000 cells/mL. The categorical nature of cow-SCC can be utilized by deriving new traits such as the fraction of cow-SCC records in a lactation that are associated with an infection with a major pathogen.

The use of data from sampling for bacteriology for genetic selection against clinical mastitis.

One breeding objective of Dutch cattle breeders is to improve genetic resistance against clinical and subclinical mastitis. Because of a lack of direct mastitis information, udder health breeding values are based on indirect traits. Inclusion of direct information on clinical mastitis could improve reliability of breeding values. The aim of this study was to investigate whether data from milk samples sent in for bacteriology are potential sources of information for the occurrence of mastitis, which may be used in animal breeding, and if so how this data can be used. Although there are 2 separate flows of milk samples for bacteriology in the Netherlands, it was not considered necessary to account for the origin of the samples. In both flows, the majority of the samples are visually normal and flow-specific traits are highly correlated. Therefore, information from these flows is combined for genetic analysis. Nearly two-thirds of the bacteriology data could be linked to milk recording and pedigree records. Relatively few farmers (<3%) took 5 or more samples for bacteriology between January 1, 2003, and March 31, 2006. Their herds had, on average, greater milk production and lower cell counts than herds for which no samples were taken. However, the range and variation within both groups of herds for these variables was similar and there was a large overlap in sires used within both groups. Whether or not samples were taken for bacteriology turned out to be a potentially useful indicator for clinical mastitis at the cow level, because this trait had a strong positive genetic correlation with clinical mastitis registered by farmers (0.84 or 0.89, depending on the data set) and similar heritability (2%) and genetic variation. Also, genetic correlations of bacteriology with SCC traits were similar to those for farmer-registered clinical mastitis. An important advantage of these bacteriology data is that they are already collected routinely and stored in a central database in the Netherlands; this is not the case for registration of clinical cases. Thus, data from bacteriological culturing can be used for genetic improvement of udder health.

Alternative somatic cell count traits as mastitis indicators for genetic selection.

The aim of this study was to define alternative traits of somatic cell count (SCC) that can be used to decrease genetic susceptibility to clinical and subclinical mastitis (CM and SCM, respectively). Three kinds of SCC traits were evaluated: 1) lactation-averages of SCC, 2) traits derived from the proportion of test-day SCC above 150,000 cells/mL, and 3) patterns of peaks in SCC. Genetic parameters for these SCC traits and their genetic correlation with CM and SCM were estimated; CM and SCM were scored as binary traits. Two data sets (A and B) depending on CM recording were available. After editing, subset A contained 28,688 lactations from 21,673 cows in 394 herds. Subset B contained 56,726 lactations of 30,145 cows in 272 herds. Variance components for sire and permanent animal effects were estimated. Estimated heritabilities for all mastitis traits were around 0.03. Heritabilities for SCC traits ranged from 0.01 for patterns of peaks in SCC to 0.13 for lactation-average SCC. Genetic correlations between SCC traits and CM or SCM ranged from 0.55 to 0.93 for CM and from 0.55 to 0.98 for SCM. High genetic correlations were estimated between CM and SCC averaged over 250 d in milk (0.87), and between SCM and presence of test-day SCC >150,000 cells/mL (0.98) in subset A. In subset B, a high genetic correlation was estimated between CM and an SCC peak with a quick recovery (0.93) and between SCM and SCC averaged between 151 and 400 d (0.95). Partial genetic correlations were calculated to investigate the additional information of the alternative SCC traits, compared with lactation-average SCC. They showed that some traits remain informative for CM and others for SCM. Therefore, use of information from a combination of different SCC traits may be more successful in improving overall udder health than the traditional single SCC measure.

[Measurement of the status of trace elements in cattle using liver biopsy samples]. / Het bepalen van de status van de spoorelementenvoorziening bij rundvee met behulp van leverbiopten.

Serum, plasma, or urine samples are usually used for the measurement of the trace elements copper; zinc, iron, selenium, because these samples are easy to obtain; however; these samples are not always appropriate. For example, it is not possible to measure molybdenum, the major antagonist of copper; in blood or urine. Therefore measurement of trace elements in liver tissue is considered the gold standard. For the assessment of selenium the method of choice remains determination of glutathion peroxidase in erythrocytes and for the assessment of magnesium determination of magnesium in urine. We determined the accuracy and repeatability of measuring trace elements in liver biopsies and whole liver homogenates. The levels of trace elements measured were similar in both preparations (92% agreement). Liver biopsy in live animals is a relatively simple procedure but not common in The Netherlands. Reference levels of trace elements, classified as too low, low, adequate, high, and too high, were established on the basis of our research and information in the literature. In a second study we investigated the practical aspects of obtaining liver tissue samples and their use. Samples were collected from cattle on a commercial dairy farm. Liver biopsy provided additional information to that obtained from serum and urine samples. We prepared a biopsy protocol and a test package, which we tested on 14 farms where an imbalance of trace minerals was suspected. Biopsy samples taken from 4 to 6 animals revealed extreme levels of trace elements.

Effects of genotype by environment interactions on milk yield, energy balance, and protein balance.

Increases in genetic merit for milk yield are associated with increases in mobilization of body reserves. This study assessed the effects of genotype by environment (GxE) interactions on milk yield and energy and protein balances. Heifers (n = 100) with high or low genetic merit for milk yield were milked 2 or 3 times a day and received rations of low or high caloric density. The management factors were selected to induce substantial differences in milk production levels and model different management strategies. The 2 x 2 x 2 factorial arrangement enables the assessment of the effects of genotype, environment, and GxE interactions. Mean daily energy-corrected milk production in the first 100 d in milk varied between 21.8 and 35.2 kg among the groups. The experimental factors affected milk production in the presumed direction. Ration was the most determinant factor on milk production. Effects of milking frequency and genetic merit were significant only in the groups that were fed rations with high caloric density. Signs for severe negative energy balances, protein balances, and low body condition scores, all of which may be indicative of health risks, were not concentrated in the highest producing cows. Feed caloric density and milking frequency had stronger effects on energy balances and protein balances, with unfavorable effects of low caloric density feed and an extra milking. This emphasizes the possible effect of mismanagement on animal health risks. High genetic merit cows had significantly lower postpartum body condition scores. Genotype x environment interactions existed, but more information is needed to determine if cows of different genetic merit for milk yield are differently at risk for disease under specific conditions. High milk production levels per se will increase allostatic load, but need not compromise the health status of relatively young cows. Ongoing one-sided selection for high yield may be combined with good animal health, but because high genetic merit for milk yield seems intrinsically connected to the allocation of resources from maintenance toward milk, this puts increasing demands on farmers' time and management skills.

Effects of management and genetics on udder health and milk composition in dairy cows.

Milk production per cow has increased significantly as a result of breeding, feeding, and other management factors. This study aims to address concerns about udder health risks for low- and high-producing dairy cows. In a 2 x 2 x 2 factorial design, Holstein-Friesian heifers (n = 100) of low or high genetic merit for milk production, milked 2 or 3 times a day, and fed a mixed ration with low or high energy content, were compared during the first 14 wk of lactation. Milk composition and cell counts were determined weekly; quarter milk samples for bacteriology were taken in wk 2, 8, and 14; and teat condition was scored in wk 2, 6, 10, and 14 during the experiment. The experimental factors resulted in substantial differences in milk production between treatment groups (24.1 for low vs. 25.6 kg/d for high genetic merit; 23.3 for 2 times vs. 26.5 kg/d for 3 times daily milking; and 20.9 for low-energy ration vs. 29.0 kg/d for high-energy ration). Ration composition was the most important determining factor for milk production, but did not affect cell counts or intramammary bacterial infections, although cows that received low-energy rations had rougher teat ends than cows receiving high-energy rations. This indicates that high production itself is not a major risk factor for udder health in the first lactation. A higher milking frequency impaired teat condition and improved cell counts in general, but did not clearly influence bacteriological status. High genetic merit was related to higher cell count, more Staphylococcus, and less Bacillus and other environmental pathogens in cultures and did not affect teat condition. The effects of milking frequency and feeding on udder health were similar for cows with high and low genetic merit. Genetic selection on milk production, without taking udder health into account, reduces udder health. As a result, maintaining udder health will require increasing the skills and time of dairy farmers who have to divide their attention to more cows when farm sizes increase, or selection should put more emphasis on udder health traits.

Sensors and management support in high-technology milking.

Two directions can be distinguished in the development of high-tech milking equipment: 1) high-capacity milking parlors with a high throughput of cows per person per hour and 2) automatic milking systems in which manual labor is replaced by a milking robot. High-capacity milking parlors are developed in such a way that one operator is able to milk many cows, partly by automation and partly by optimization of available labor. In such parlors, one operator can milk up to 125 cows per hour. This means that there are only a few seconds available for udder preparation. In an automatic milking system, a robot takes over all manual labor during milking. Currently available systems have one robot arm working with one milking stall (one-stall system) or one robot arm working with more milking stalls (multiple-stall systems). Cows have to go to the automatic milking system voluntarily. Therefore, there is a large variation in milking intervals. Moreover, a large variation between milkings and between cows was observed in milk flow rate, machine-on time and udder preparation time. Both developments in high-tech milking have effects on the milk ejection. The small amount of time dedicated to udder preparation in high-capacity milking parlors has negative effects on the milk ejection, among others leading to more bimodal milk flow curves and longer machine-on time. In automatic milking systems, the variation in time between udder preparation and cluster attachment and in milking frequency might have an effect on milk ejection. Lactation physiology can play a role in solving the questions around milk ejection in high-tech milking systems. The introduction of high-tech milking systems makes decision support systems using sensors necessary. These systems should assist in detection of abnormal milk and mastitis. To a lesser extent, diseased cows need to be brought to the attention of the dairy farmer. Some sensors are currently available for this purpose, but they do not fulfill all demands. In the near future other sensors might be developed. It is important that this development is demand driven and not technology driven. Lactation physiology can play an important role in the determination of milk components useful for automatic detection.

Robotic milking and its effect on fertility and cell counts.

The aim of this study was to analyze the effect of robotic milking (RM) on fertility and somatic cell counts (SCC) among dairy herds participating in the national Dutch milk recording system. It was hypothesized that RM, and a higher milking frequency in general, would have negative effects on fertility, due to expected and supposed deeper negative energy balance (NEB). Herds increasing milking frequency from two to three times daily consistently had increased production. Milk production during RM was intermediate between the amounts obtained by milking twice versus three times a day. Milking three times a day and the associated higher production had no significant effect on reproductive measures such as nonreturn rate at 56 d post insemination (NR56) or days to first service. Although RM did not affect NR56, use of the robot was associated with an increase in days to first service. An increase in milking frequency from two to three times daily did not affect SCC, but SCC were significantly increased after milking with the robot. Robotic milking has a significant positive effect on production and no negative effect on fertility as measured by NR56. The effect of RM in increasing days to first service appears due to reasons other than increased production and a more NEB. Increased SCC during RM is potentially of concern. From the data available, the relationship of RM to clinical mastitis could not be determined but this aspect needs further attention.

Detection model for mastitis in cows milked in an automatic milking system.

Automated detection of diseases (such as mastitis) in dairy cows might be an alternative for detection by observation during milking - especially when using an automatic milking system (AMS). An outline of a detection model is given. This detection model includes time-series models for two variables (milk yield and electrical conductivity of milk), with interpolation on previous values. The model is flexible in the number of variables actually used. Parameter values and the residual variances are updated by linear regression after each milking. Alerts for mastitis are given when the residuals fall outside given confidence intervals. A data set with 111 cows for 16 months (on average, 58 lactating cows per day) was used to test the model. Depending on the chosen confidence interval, 42-44 out of 48 cases of clinical mastitis were detected; the remaining cases were not detected because not all data needed were available. These results were better than the results obtained with the model usually used on the farm. The number of false-positive alerts depended on the chosen confidence interval and was higher than the number found with the model usually used.

Relationship among estradiol, cortisol and intensity of estrous behavior in dairy cattle.

Economic profitability of a dairy farm is based, in part, on the calving interval of the cows. The optimal interval is 365 d. To achieve this, the cow needs to be pregnant within 85 d post partum. The first and most problematic step in this process is the determination of the optimal time for insemination, which is based on estrous behavior. The expression of estrous behavior, however, is at a low level in modem dairy herds, resulting in low detection rates and longer calving intervals. In the present study, a point scale was used to monitor postpartum, nonpregnant cows for estrous symptoms. Frequent blood samples were taken around estrus, and the cows were fit with pedometers to measure their activity. Correlations between the occurrence of symptoms of estrus and levels of estradiol and cortisol were then analyzed. Standing heat, the standard symptom of estrus, was observed in only 53% of the cows. A high correlation of 0.7 was found between estradiol concentration and estrous behavior. This was empasized by the fact that the estradiol level reached its highest level of 7.76 + 2.39 (SD) pg/mL at the same time as the highest behavior score. The highest pedometer readout lagged 8 h behind this moment. Cortisol levels did not exceed the physiological levels in rest situations but showed an increase at the time estrous behavior was at its maximum. The present study showed that standing heat is not the primary symptom for detecting estrus in cows. Pedometers are a useful aid but they have to be read several times a day. The high correlation between the visual symptoms of estrus and estradiol concentrations indicates that visual estrus detection is an efficient, reliable way to determine the right time for insemination.