Individual variation in milk yield response to concentrate intake and
milking interval length
During the last century in the Netherlands milk production per cow has almost tripled.
Accordingly, the amount of concentrates yearly fed per cow strongly increased.
Furthermore, automation and robotisation has changed dairy management, especially by the
introduction of automatic concentrate feeders and milking systems. A new management
concept, emerging in the last decades, is Precision Livestock Farming (PLF). The objective
of PLF is to optimize livestock production, by on-line monitoring and control of the
production process, utilizing the technical possibilities of automation and robotisation.
Nowadays, individual settings for daily concentrate supply and milking frequency are based
on standards, ignoring individual variation in milk yield response on concentrate intake and
milking frequency. This leads to the main hypothesis for this thesis research that
profitability of dairy farming can be improved by utilizing information on individual
variation in response.
The first objective of this research was to quantify the individual variation in milk yield
response to concentrate intake and milking interval length, in order to assess the economic
prospects of applying individual optimal settings for concentrate supply and milking
In the first observational study (Ch. 2), data from 299 cows on four farms in the first 3
weeks of the lactation were collected. Individual response in daily milk yield to concentrate
intake was analysed by a random coefficient model. During the first three weeks of
lactation, considerable variation in individual milk yield response to concentrate intake was
found on all four farms. An economic simulation was carried out, based on the estimated
parameter values in the observational study. Individual economically optimized settings for
concentrate supply were compared with conventional strategies for concentrate supply
based on averaged population response parameters. Applying individual economic optimal
settings for concentrate supply during early lactation, potential economic gain ranges from
0.20 to 2.03 €/cow/day.
In a second observational study (Ch.3), data of normal uninterrupted milkings during one
week from 311 cows kept in 5 separate herds on one farm were collected. The data set
consisted of 4,915 records and random coefficient models were fitted to estimate the
individual effects of milking interval on daily milk yield and milking duration. Between
individuals, considerable variation in milk yield and milking duration was found in
response to milking interval. Based on the estimated individual response, a simulation was
carried out in order to optimize the utilization of an AMS for different herd sizes and
occupation rates. Applying optimal individual milking intervals for a herd of 60 cows and
an AMS operating at an occupation rate of 64%, the average milking interval reduced from
0.421 day to 0.400 day, the daily milk yield at the herd level increased from 1,883 to 1,909
kg/day, and milk revenues increased from 498 to 507 €/day. In the actual situation, the herd
consisted of 60 cows. A further increase of daily milk revenues per AMS was possible by
increasing the operation rate and/or herd size.
The conclusion is that between dairy cows there is a considerable variation in effects of
concentrate intake and milking interval length on milk yield and, consequently, milking
duration. A marked increase in economic profits of dairy production is possible by
improvement of the concentrate allocation and/or the utilisation of an AMS, applying
optimal individual settings based on the actual individual response in milk yield.
Development of adaptive models
The second objective was the development and testing of adaptive models for on-line
estimation of the actual individual response in milk yield to concentrate intake and milking
interval length. In Ch. 4 adaptive dynamic models for on-line estimation of the actual
individual milk yield response to concentrate intake and milking interval length were
evaluated. The parameters in these models may change over time and are updated through a
Bayesian approach for on-line analysis of time series. Time series data of daily milk yield
during the first 200 days of lactation from 17 cows were analysed with different adaptive
dynamic models. Three models were evaluated: a model with linear terms for concentrate
intake and length of milking interval, a model with linear and quadratic terms, and an
enhanced model in order to obtain more stable parameter estimates. The linear model was
only useful for forecasting milk production and the estimated parameters of the quadratic
model turned out to be unstable. The parsimony of the enhanced model lead to far more
stable parameter estimates.
In Ch. 5 an adaptive dynamic model was used for time series analysis of herd mean daily
milk yield, in order to quantify the impact of heat stress and to assess the potential for
monitoring and control of milk production. Time series data of daily milk yield from 2003
to 2006 were collected on six experimental research farms in The Netherlands. The impact
of heat stress was quantified in terms of critical temperature, duration and loss in milk
yield. The estimated critical temperature was 17.8 oC, the duration was 5.5 days, and loss in
milk yield 31.4 kg milk/cow/year, averaged over farms. Besides estimation of the impact of
heat stress, level and trend, including a weekly cyclical pattern were estimated to evaluate
the production process. The Bayesian approach for on-line analysis of time series
comprises also a procedure for the detection of potential outliers and other deteriorations
that might be promising for monitoring the production process. Outliers and other process
deteriorations are adequately detected by this monitoring procedure.
The conclusion is that on-line estimation of the actual individual response in milk yield and
milking duration is possible following a Bayesian approach for time series using an
adaptive dynamic model. Besides estimation of the actual response the Bayesian approach
adequately detects process deteriorations. Therefore, adaptive dynamic models provide a
useful tool for control and monitoring of the dairy production process.