Cow routines and behavioral responses are altered substantially following the installation of robot milking. The present study was designed to analyze the effect that switching from milking parlor to automatic milking system (AMS) had on the culling rate (due to various causes) of dairy cattle. For this purpose, culling records and causes for culling were tracked in 23 dairy farms in the Galicia region (NW Spain). The animals in these farms were monitored for 5 years. For the present study, that length of time was divided into three different stages, as follows: 2 years before switching from a milking parlor to AMS (stage 1), the 1st year following the implementation of AMS (stage 2) and the 2nd and 3rd years succeeding the implementation of AMS (stage 3). Cox models for survival analysis were used to estimate the time to culling due to different reasons during stage 1 in relation to stages 2 and 3. The data indicated that the risk of loss due to death or emergency slaughter decreased significantly following the installation of AMS. In contrast, the risk of culling due to low production, udder problems, infertility or lameness increased significantly. Low-production cows (such as cows in advanced lactation due to infertility) or sick cows (such as mastitic or lame cows) allegedly have a noticeable effect both on the performance and the amortization of the cost of AMS, which in turn would lead to a higher probability of elimination than in conventional systems.

The throughput of automatic milking systems (AMS) is likely affected by differential traffic behavior and subsequent effects on the milking frequency and milk production of cows. This study investigated the effect of increasing stocking rate and partial mixed ration (PMR) on the milk production, dry matter intake (DMI), feed conversion efficiency (FCE) and use of AMS by two genotypes of Holstein-Friesian cows in mid-lactation. The study lasted 8 weeks and consisted in a factorial arrangement of two genotypes of dairy cattle, United States Holstein (USH) or New Zealand Friesian (NZF), and two pasture-based feeding treatments, a low stocking rate system (2 cows/ha) fed temperate pasture and concentrate, or a high stocking rate system (HSR; 3 cows/ha) fed same pasture and concentrate plus PMR. A total of 28 cows, 14 USH and 14 NZF, were used for comparisons, with 12 cows, six USH and six NZF, also used for tracking of animal movements. Data were analyzed by repeated measure mixed models for a completely randomized design. No differences (P>0.05) in pre- or post-grazing herbage mass, DMI and FCE were detected in response to increases in stocking rate and PMR feeding in HSR. However, there was a significant (P<0.05) grazing treatment×genotype×week interaction on milk production, explained by differential responses of genotypes to changes in herbage mass over time (P<0.001). A reduction (P<0.01) in hours spent on pasture was detected in response to PMR supplementation in HSR; this reduction was greater (P=0.01) for USH than NZF cows (6 v. 2 h, respectively). Regardless of the grazing treatment, USH cows had greater (P=0.02) milking frequency (2.51 v. 2.26±0.08 milkings/day) and greater (P<0.01) milk yield (27.3 v. 16.0±1.2 kg/day), energy-corrected milk (24.8 v. 16.5±1.0 kg/day), DMI (22.1 v. 16.6±0.8 kg/day) and FCE (1.25 v. 1.01±0.06 kg/kg) than NZF cows. There was also a different distribution of milkings/h between genotypes (P<0.001), with patterns of milkings/h shifting (P<0.001) as a consequence of PMR feeding in HSR. Results confirmed the improved FCE of grazing dairy cows with greater milk production and suggested the potential use of PMR feeding as a tactical decision to managing HSR and milkings/day in AMS farms.

Two types of artificial neural networks, multilayer perceptron (MLP) and self-organizing feature map (SOM) were used to detect mastitis by automatic milking systems (AMS) using a new mastitis indicator that combined two previously reported indicators based on higher electrical conductivity (EC) and lower quarter yield (QY). Four MLPs with four combinations of inputs were developed to detect infected quarters. One input combination involved principal components (PC) adopted for addressing multi-collinearity in the data. The PC-based MLP model was superior to other non-PC-based models in terms of less complexity and higher predictive accuracy. The overall correct classification rate (CCR), sensitivity and specificity of this model were 90·74%, 86·90% and 91·36%, respectively. The SOM detected the stage of progression of mastitis in a quarter within the mastitis spectrum and revealed that quarters form three clusters: healthy, moderately ill and severely ill. The clusters were validated using k-means clustering, ANOVA and least significant difference. Clusters reflected the characteristics of healthy and subclinical and clinical mastitis, respectively. We conclude that the PC based model based on EC and QY can be used in AMS to detect mastitis with high accuracy and that the SOM model can be used to monitor the health status of the herd for early intervention and possible treatment.

The necessary unequivocal and generally accepted definitions of normal and abnormal milk are not available. A precise definition is needed in order for companies to develop sensors to detect and sort abnormal milk at the time of milking. Experts at a workshop defined abnormal milk to be that from cows whose foremilk had changed in homogeneity or was coloured by blood. The objectives of this paper were: firstly, to explore how different groups of people scored the appearance of foremilk; and secondly, to develop a method suitable as an objective reference for testing of manual and automatic detection systems. Consumers, farmers and advisors did not agree on the visual appearance of normal, watery, clotty milk, or milk with blood, and experience is needed to score the visual appearance of foremilk correctly. It seems reasonable to expect a sensitivity of at least 70% for detection of abnormal milk during foremilking. Filter sizes 0·05, 0·07, 0·1, 0·2, 0·5, 1·0, and 2·0 mm were used to filter milk from cows with visually abnormal foremilk. If clots appeared in the foremilk, clots appeared on all size filters, but the filter with pore size 0·1 mm was the easiest to read and work with. The filter method is not reliable in identifying quarters with watery, yellowish, or bloody milk, whereas the method seems consistent, and at least as good as scoring of visual appearance in finding clots in the milk. Clots should show clearly on the filter to be counted as abnormal milk. All clinical cases with clots in the foremilk can be found on the filter and such cases have high somatic cell count (SCC). Both trained and untrained persons using the filter method can score normal and abnormal foremilk with a high specificity (>90%) and a high sensitivity (>80%). The filter method is recommended as a reference for scoring the homogeneity of foremilk.

Sorting of normal and abnormal milk at time of milking is done visually for conventional milking systems, but more concrete standards are needed when milking is done in automatic milking systems (AMS). Several panel tests were carried out to find out how different consumer groups, milkers and advisors look at and respond to the visual appearance of milk mixed with blood, in order to set a limit for what they think is acceptable. It is concluded from the test panel results that milk samples with 0·4% or more of blood all will be scored as pink and samples with 0·1% blood (about 6 μM-haemoglobin or 100 mg/l) can be visually detected if they are compared with milk samples without blood. The consumer group scored fewer of the samples with 0·1% blood as normal than did the professional groups. The test panel scored 65% of the samples with 1% blood as normal when milk was presented in a black strip cup, which is the reference method when foremilking takes place in a conventional parlour. Only 2% of the milk samples with 2% blood (about 120 μM-haemoglobin or 2000 mg/l) were scored as normal in a black strip cup and should consequently be detected by conventional as well as automatic systems. One model of AMS was tested for its ability to detect and separate milk coloured by blood. The model separated milk with [ges ]6 μM-haemoglobin. Milk mixed with blood injected into the milk stream for a short time at the beginning of milking was not separated. We lack data on how blood is naturally expelled into milk and in what amount. We propose that cow composite milk with >6 μM-haemoglobin should be separated because at this level milk will have a red tinge.

Food can be used in an automatic milking system (AMS) to encourage frequent attendance to, and modify behaviour in, an AMS. An AMS was designed such that a cow had to pass from a bedded area through the AMS to enter the feeding area. A visit resulted in milking if the interval since the last milking was greater than 4 h.

Fourteen Holstein-Friesian dairy cows were allowed to attend voluntarily an AMS for 15 h/day for three 8-day periods. In period 1 the cows were given forage in the feeding area and concentrates in the bedded area but in period 2 this was reversed. Period 3 repeated period 1. The cows were also divided into two groups of seven, only one of which was given 1 kg of concentrates when milked in each period in a cross-over design.

Visits per cow per day increased when the cows were given forage in the feeding area (6·0 v. 4·1 (s.e.d. 0·08) P < 0·05) but milkings per cow per day were only slightly different (2·6 v. 2·4 (s.e.d. = 0·02) P = 0·052). Feeding forage in the feeding area modified forage-eating behaviour by reducing the number of bouts (4·9 v. 7·6 bouts per cow per day, P < 0·05) and total forage eating time (209 v. 289 min per cow per day (s.e.d. = 33·6), P < 0·05). The cows also tended to spend less time lying (529 v. 620 min per cow per day (s.e.d. = 43·2) P = 0·051).

There was no effect of feeding/not feeding concentrates in the milking stall on visits per cow per day or milkings per cow per day. However, feeding tended to create more shuffling behaviour during the automatic teat cup attachment process (3·4 v. 6·7 shuffling bouts per milking (s.e.d. 2·07) P = 0·053) but there were no effects on the levels of kicking, successful teat cup attachment rate, vocalization, elimination, hesitations in mounting a step or poor rear leg positions. The cows tended to adjust to a more accessible stance during both the automatic attachment process (from 17% of milkings exhibiting poor leg positions at the start of attachment, to 10% at the end, P < 0·01) and over the course of the experiment (from 26% to 10%, P < 0·02). There was a trend for cows who were not fed to milk out quicker than those who were fed for all milkings per day combined (733 s v. 811 (s.e.d. 43·3) s P = 0·074) but there were no other effects on yield, milk flow rate from the udder or time taken to milk out for all milkings through the day combined or just the first milking of the day.

Feeding concentrates in the feeding area can be used as an alternative to feeding forage in the feeding area, but there is no need to feed concentrates in the milking stall. Feeding cows while they were being milked did not improve their behaviour or milking characteristics.