Twenty-five years or more since the arrival of the first electronic models to assist in feeding the pig, new examples of the application of modelling continue to appear. One of the latest involves the commercial launch this year of a swine simulation software package that is being used across Canada by the company that reckons it is the country's largest animal feed manufacturer.
Maple Leaf Animal Nutrition (MLAN), an independent operating company of Maple Leaf Foods, has 18 feed mills to manufacture premixes and complete feeds, including nursery diets. It supplies these products to its own pork production chain handling about 2.2 million pigs per year in addition to selling to other producers nationwide.
MLAN has organised three regional teams to cover the east, centre, and west of the country for sales of its Shur-Gain and Landmark Feeds brands. The swine sales representatives began training on a new suite of software in February in order to help pork producers make better decisions on nutrition, production, and economics for their operations. By June the company decided it was ready for a full roll-out of the pig model called Watson, as part of a suite of applicationsalso including dairy and beef modelsrunning off a web-based platform.
Science of modelling
Although the name Watson is new, the science behind the model dates back at least to the end of the 1980s.
"I published the basic framework of this model in 1994 and I have spent more than 14 years working on the various biological components of the theory behind it," recalls Dr Neil Ferguson, formerly of the University of Natal in South Africa until moving three years ago to the Agresearch centre of Maple Leaf Foods in Canada. "So the biology of the model has existed for some time, although of course it has been updated continually over the years on the basis of new information. The main focus of development over the last two to three years has been to take that theoretical framework and put it into a commercial form, in other words working with the programmers to translate it into the form that could be used on MLAN's existing software platform."
Dr Ferguson refers to Watson as the ‘next generation' of pig models. Where others in the past have focused on pig growth and predicting the animal's requirements for nutrition as influenced by the breeding background, environment, and nutritional inputs, he explains, Watson goes the next step to incorporate the influences of the health and well-being of the pig in its considerations. Watson includes a least-cost formulator to allow optimal dietary solutions to be determined. In addition, a diagnostic tool helps suggest the source of possible limiting factors in the farm's production system.
Pig unit-based data
The necessary information to run the model is collected from participating producers by means of a questionnaire supplemented where necessary by a walk-through tour of the particular unit's facilities. A health scorecard is compiled from the unit's history for pig mortality, incidence of respiratory and enteric disease, and the degree of general stress experienced by the pigs in each phase of production.
The process of running data through simulations is generally not done during a farm visit, but rather at a later time after their collection. Dr Ferguson says this allows the sales representative time to set up for a particular customer and to run simulations until they are satisfied that the performance predictions accurately reflect the present production results. Once this process had been concluded, he adds, the representative will work in co-operation with the customer to address agreed business objectives and concerns, as well as to identify realistic options for improvement.
Dr Ferguson comments that the members of the sales team are trained to analyse information away from the unit and to return when they have the results for discussion. Watson enables understanding and benchmarking how the unit performs currently and then identifies the limiting factors that have the greatest influence on that unit. Ultimately, continues Maple Leaf's pig modeller, Watson shows the producer where the unit's performance can go and which strategies should be helpful in moving in that direction.
Maple Leaf's training regimen for its representatives begins with a two-day course on the basics of pig nutrition and physiology before their introduction to the theory and navigation of models over a subsequent period also of two days. This primer is followed by coaching and mentoring, done mainly by team managers over a period of at least three months, to familiarise the representatives in practical aspects of data collection, input and assessment. Finally each of MLAN's three regions has held follow-up meetings once or twice per month to share experiences and resolve any ongoing difficulties.
The teams learn, for example, how the built-in Least Cost Formulator of the Watson package can be used to extrapolate the analysis of a customer's own grain resources into diet designs for that farm using appropriate premixes from the MLAN product range. More deeply, they discover that the model has applications in devising the optimal programme for feeding the pig so its body composition at slaughter attracts the highest payment from the abattoir.
Mechanistic versus empirical modelling
At its simplest, the MLAN swine sales representatives were told at an introductory briefing session, a model is a series of mathematical equations that attempts to represent a system or process. The simplest forms are empirical in the sense that they assume unchanging influences on pig's growth and therefore seek to predict weight gain on little more than a knowledge of daily feed intake. Watson illustrates the alternative of a so-called ‘mechanistic model', one which considers the genetic potential of the particular pig for different types of tissue growth and qualifies that according to both feed intake and environmental limitations. From the data put into a mechanistic model it becomes possible to predict how the various tissues of the body will grow and hence the proportions of lean and fat in the eventual carcase.
Most of the other nutrition models used commonly in the feed industry are empirical, according to training material from Maple Leaf. Mechanistic models have a greater opportunity for more applications and are potentially more accurate because they rely on a penetrative understanding of the biology of the pig in addition to the physical and chemical reactions involved in the utilisation of nutrients for body maintenance and growth.
Watson is different again by being the result of over 10 years of research and validated in some 20 trials in real-life situations across Canada with variations in animal genotype, health, housing/environment and feeding, Dr Ferguson remarks. "Our validation trials have demonstrated already that the model is accurate. Typically, the difference between actual barn performance and predictions made by Watson was less than 4%. Ever since the launch this year we have seen that making it available to our sales teams has grown the volume of our pig feed business as well as helping our producers to improve the performance of their pigs.
Specific per pig genotype
"Here in Canada we have been able to define 27 generically named genotypes as representing the breeding material used by Canadian swine producers," Dr Ferguson comments. "It means that a producer could use Watson to select the most appropriate genotype for his farm."
In fact this model can simulate any type of production system for growing pigs that you would expect to find in Canada, says Maple Leaf. It allows up to 26 environmental phases as well as 26 dietary phases to be included in one simulation, for pigs weighing from 5 kilograms to 250 kg. Small parts of the production cycle can be simulated, or the complete process from entry to the nursery at weaning until the animal leaves for slaughter.
"The whole point about Watson is that it is a tool for diagnosing how the performance of the pigs in the nursery or the finishing house might be improved," Dr Ferguson declares. "Sometimes the diagnosis might suggest something quite simple, such as increasing the frequency of feeding from two times to three times daily. A nutritional deficit might be identified, such as including too little soybean meal in the diet. But there is also the opportunity to undertake quite complicated analyses.
"The model first compares the daily performance of the simulated pigs with their maximum lean growth potential. Next it examines the key factors impacting lean growth, including the stocking density in pens and the health status as well as nutrient composition and form and the pig's physical environment. The factor that most constrains growth on any given day in the particular unit is identified and further simulations show how performance would change with the constraining factor improved.
"Over the next two years we will concentrate on allowing users to become familiar with the model and on making updates as appropriate. Because it is web-based, we can change the computer programme without having to amend multiple copies of it."