J.M. Chesworth1, D.F McKillop1 and D.J. Spriggs2
1University of Zimbabwe, Department of Animal Science
P.O. Box 167, Mount Pleasant, Harare, Zimbabwe
2National Association of Dairy Farmers
P.O. Box 1241, Harare, Zimbabwe
Abstract
Introduction
Conclusion
References
The main limitations to the use of by-products in diets for farm animals are the uncertainty of the likely response in terms of animal production and their need for supplementation with other materials to provide a diet adequate for the needs of production. These problems are found at all levels of animal production from subsistence systems to commercial farming. In industrialised countries and some developing ones agro-industrial by-products are commonly incorporated into commercially compounded feeds on a basis of decisions made using computing methods. A computer package has been written to offer these same facilities at a relatively low-cost to the individual or group of farmers. The system has application in situations in which there exists some element of choice in the feeding of animals. There are three components to the system for calculating diets in this way.
This paper describes the development and application of one such system as a decision-making tool in the optimisation of utilisation of feeds, including agro-industrial by-products, on dairy farms in Zimbabwe. Copies of the computer package are available to organisations in developing countries for the cost of the media.
Application of feeding systems in small-scale agriculture
At the previous meeting of the African Research Network for Agricultural by-products held in Egypt, strong criticism was voiced of the application to small-scale farming operations of feeding standards, developed in industrialised countries (Preston, 1986). This criticism was based upon the fact that in some subsistence-type farming environments it has been shown that animal performance obtained after the feeding of combinations of poor quality materials was not accurately predicted from published laboratory assessments of these feeds.
If it is accepted that at very low levels of production it is inappropriate to use feeding systems but that it is entirely correct when dealing with animals on large-scale commercial enterprises then there must be a point of cross-over between the two. Most of the difficulties with the application of feeding standards stem from the fact that many of the forages and by-products that are used in subsistence agriculture are of relatively poor quality and that book values for their probable nitrate composition give a poor prediction of their usefulness. The reason for this is to be found in the so-called associative effects of feeds. These are often ignored in the formulation of rations in large-scale farming as, with high quality diets, they bring very limited changes. As the number of ingredients in a diet is reduced the chances of large interactive effects increase.
Given a very large number of dietary components there is a much greater chance that the range of materials will combine to i) support optimum rumen fermentation and ii) supply materials to the rumen hall and to the hind gut in approximately the right proportions. One of the objectives of feeding systems such as the combined ME and new protein systems of the ARC (1980, 1984) has been to give results which will provide good rumen fermentation leading to the uptake of appropriate materials from the gut. That it is not possible to transfer such rationing systems to many small-holder situations is due to the fact that the small number of materials of very unbalanced composition which are available do not combine satisfactorily to provide optimum rumen fermentation.
The dairy industry in Zimbabwe
Zimbabwe has a very well developed commercial dairy industry which is characterised by a relatively small number (approximately 540) of large-scale producers. The technology used on many of these farms is comparable to that found in industrialised countries; in addition cows are of European breeds and are usually fed on combinations of home-grown roughages together with factory-compounded concentrate feeds. Since Independence there has been a move towards the development of a small-scale dairy sector. The objectives here are twofold: (i) to provide a higher level of nutrition for the human population in these areas and (ii) to create some cash income for small-scale farmers. As a consequence of the large scale commercial industry the country already possesses a sophisticated distribution and marketing network serving a largely urban market. Much of the infrastructure in terms of bulk-handling facilities is already in place and in a number of areas points have been established for the collection of milk from small-scale producers.
The fact that the infrastructure exists for the farmers to get their produce to the market also means that they are able to buy foodstuffs from external suppliers and do not have to rely solely upon farm-produced materials. The largest schemes of this nature form part of planned resettlement schemes where the farmers would have access to land, water and amenities such as schooling. Many farming decisions are not available to the farmer; he is only allowed to crop a part of his land, the remainder must be used for animal production. As a part of the whole package, they are growing Napier grass, Pennisetum purpureum (approximately 2.23 ha per producer), and have a smaller area for cropping. Under Zimbabwean conditions Napier grass may be expected to yield between 4 and 10 tonnes of dry matter per hectare per year (Addison, 1952). Under most conditions this dry matter has a relatively low crude protein level. Cropping will undoubtedly produce residues which may also be incorporated into the animal production systems. Again these materials may be expected to be of low digestibility and low protein content. In addition to the home-produced feeds they are purchasing compounded feeds from stock-feed companies and thus are in the same market as their large-scale colleagues. Farmers on the resettlement schemes have costs to meet in producing their milk and in transporting it to the collection points and for this reason they have to work on similar, but smaller, budgetary considerations.
One of the factors which has underlined government policy towards the dairy industry over the past few years has been a notable squeezing of the margins between the permitted prices of compounded feeds and the producer price for milk. The same prices are paid to all producers irrespective of size of enterprise so the difficulties are common to all sizes of farms. These changes in financial structure have resulted, in the commercial sector, in a much greater interest in the home-mixing of feeds and the incorporation into them of cheaper materials, many of them agro-industrial byproducts. The aim is to balance what materials the farmer has with the minimum amount of purchased food. On the small-scale resettlement schemes the aim is just the same. These organised small-scale schemes do have one advantage over the more scattered communities which are typical of subsistence farmers: they are concentrated in one place and can use this opportunity to purchase larger quantities of feed or feed ingredients.
Availability of agro-industrial by-products in Zimbabwe
A large number of the ingredients available to the feed formulator are in fact agro-industrial by-products as these are generally cheaper than products which are available to the human food market. Within Zimbabwe many of the milling by products are not accessible in raw form to the farmer as they are used by the feedstuffs divisions of the milling companies which produce them. A list of some of the by-products currently available in Zimbabwe is presented as Table 1.
Availability of computing facilities
Computers are now increasingly available to farmers and extension workers for a whole variety of tasks. The cost of a machine of sufficient sophistication for the formulation of livestock diets has fallen dramatically in recent years. In the early years of the decade one would have cost about the same as an expensive limousine, now when freely available they are about the same price as a basic motor cycle.
In addition to the computer hardware the following components are needed:
1. a system for calculating the likely requirements of farm animals,2. a knowledge of the nutritional properties of each material which is likely to be fed expressed in the same terms as those in which animal requirements have been described, and
3. some way of choosing, from the information on feeds, the most advantageous mixture needed to satisfy the needs of animals.
Animal requirements
In designing a rationing system for farm animals the first consideration has to be the method used for the expression of animal requirements. There are a number of competing methods which have been developed in different countries of the world. In setting up the original computer program one consideration was to look for a system which would generate easily the values for the requirements of each nutrient. It was for this reason that we chose the British ARC method (ARC, 1980, 1984).
There are some disadvantages to the British scheme, the first one being the need to establish the quality of the diet before calculating the animal's needs and thus before the diet can be formulated. The technique has therefore to be an iterative one starting from a 'good guess' based on a knowledge of the materials likely to be available. The second is in the use of the protein rationing system. The theoretical basis of the ARC protein system (ARC, 1984) is not in serious question, it provides a useful conceptual framework which explains many of the apparent anomalies in previous schemes. Where it breaks down is in its need for an estimate of the likely fate in the rumen of any one protein source. Basically, the degradation of proteins will depend on the fermentative environment of the rumen and on the rate at which materials pass through. These will in turn depend upon the particular combination of feeds which enter the rumen, in other words it is necessary to know the result of the dietary formulation before starting to devise it.
Table 1. Agro-industrial by-products available directly to farmers in Zimbabwe.
|
Grain products |
Straws and stovers |
|
Maize superdust |
Maize stover |
|
Cracked maize |
Sorghum stover |
|
Sorghum bran (red and white) |
Wheat straw |
|
(Note: wheat products are not generally available to the farmer directly) |
Barley straw |
|
Animal by-products |
|
|
Feather meal |
|
|
Poultry manure |
|
|
Meat and bone meal |
|
|
Blood meal |
|
|
Fish waste |
|
|
Oilseed and associated products |
Human food and beverage by-products |
|
Cottonseed meal (extracted) |
Bakery waste |
|
Cottonseed (whole) |
Peanut skins |
|
Cotton hulls |
Brewers grains |
|
Sunflower meal |
Navy (baked) bean waste |
|
Sunflower hulls |
Fruit pulp |
|
Soyabean meal |
Grape residues |
|
Groundnut hulls |
|
In evolving a practical scheme that could be used some ways around these had to be found. In relation to the dietary quality, a first approximation of a metabolisability (q) of 0.5 is taken. This represents an overall ME in the dry-matter of a little less than 10 and is probably fairly representative of the overall diets that may be offered to dairy animals in tropical environments. At the end of the formulation procedure the metabolisability of the diet is estimated by the program and if this value differs widely from the starting estimate the operator is warned to start the whole procedure again using the new value.
Estimates of protein requirements made using the British system depend upon two completely separate sets of figures. The first is an estimate of the likely needs of the animal for amino acids. These figures can be well predicted from physiological data. The second set of information concerns the ability of the feed to provide amino acids. Here the precision of estimates must be much lower. The main problem is that many of the byproducts which we might like to use have not yet been investigated.
When the ARC protein system was first published it was suggested that feeds should be divided up into four groups depending upon their likely degradabilities, each being assigned an assumed value. We have continued this procedure by ascribing new materials to the most similar group. To give some sort of latitude to the formulation in terms of protein quality we calculate independently the protein (UDP) and the rumen degradable protein (RDP) and sum them to produce the overall crude protein (CP) figure. A quite arbitrary 'safety margin' of 10% is added to the overall crude protein figure. In setting up the requirements for linear programming all three, UDP, RDP and CP are made 'greater-than' constraints. This means that an oversupply of protein is given but that this can be made up of either RDP or UDP .
There are two constraints which are set by the program package which are not derived from the ARC (1980, 1984) recommendations. These are for the minimum value of fibre in the diet and for the maximum value of fat. "either of these is likely to be a significant problem for the small-scale producer. Given his inputs of poor quality crop residues or of very fibrous grass then he is unlikely to be short of fibre. Similarly it is unlikely that he will have access to high fat materials such as unextracted soyabeans.
It was decided not to formulate in terms of trace elements; it is notoriously difficult to predict reliably the trace element composition of feeds particularly when buying feed ingredients from distant, and possibly unknown, sources. As a policy we recommend the incorporation of the recommended amount of a proprietary mineral mixture in "hat has been referred to as the "shot-gun" approach. This has the added advantage of reducing the number of constraints that may be applied and thus speeds solution of the linear programming matrix. The constraints applied to formulation are shown in Table 2.
Table 2. Constraints used in dietary formulation.
|
Component |
Unit | ||
|
Metabolizable energy |
MJ/kg |
dry |
matter |
|
Crude protein |
g/kg |
dry |
matter * |
|
Rumen-degradable protein |
" |
" |
" |
|
Undegradable protein |
" |
" |
" |
|
Calcium |
" |
" |
" |
|
Phosphorous |
" |
" |
" |
|
Magnesium |
" |
" |
" |
|
Fibre |
" |
" |
" |
|
Fat + |
" |
" |
" |
*Set to be 10% higher than the sum of rumen-degradable and undegradable proteins.
+Set as a maximum value.
Information of feed composition
To perform the calculation, figures are needed for the composition of each of the individual ingredients that may be considered for inclusion in the diet. In a large commercial organisation there may be several hundred ingredients that are available to the compounder. It is not possible for the company to analyse every batch of material as it enters the factory and indeed this may be a waste of time in terms of the known variability of sample composition between different parts of the same bulk delivery of a feed. Not only this, the feed manufacturer may not physically have some of the ingredients on his premises at the time at which the initial formulation is undertaken. Buying of a needed ingredient may actually follow the decision to incorporate it. The feed manufacturer thus makes his decisions on the incorporation of ingredients on a basis of information which he has gleaned from a wide variety of sources and which experience has shown him to be relatively reliable. As an example of the consistency of data relating to feed composition it is worthwhile looking at the detailed studies of the Rowett Research Institute's Feed Evaluation Unit (FEW, 1980). For example, 16 samples of maize from all over the world had ME values with a mean of 13.7 MJ/kg dry weight and coefficient of variation of only 6.6%; the average crude protein was 104.5 g/kg dry weight with a coefficient of variation of 5.0%. It is unlikely that changing the maize from the best one to the worst would have had any great effect on the cost of diets which incorporated them, nor upon the resulting animal production.
In setting up the linear programming system for use on farms a major difficulty has been in ascribing likely nutrient composition to many of the materials which have been found to be available in relatively small amounts. The collection of data has used a large number of sources and not all of these are consistent in the ways in which they calculate or even express their information. In a small country it is not possible to have analytical data on every single type of material, much less every batch that is released onto the farming market. Many of the materials which do become available are common to a number of countries and it ought to be possible to agree on one or two 'compromise' figures that describe the properties of each type of feed. Such effort would be more economically expended at an international level and as such would tend to reduce the problems associated with the various schemes of analysis employed in different countries.
The use of analytical figures which may not precisely reflect the nutrient value of the material under consideration is open to question. If, however, we look for instance at the metabolisable energy of a poor quality feed and consider it for incorporation into a diet, does it matter significantly if we assume its ME to be 8.5 rather than 7.5 MJ/Kg, the errors that we shall experience in arriving at first estimate of intake may be much greater than this. It must be emphasised that this is a management tool and only provides a likely first estimate that the farmer will have to modify in the light of experience.
Linking of the animal requirements and feed composition
The programs have been written to run on two different machines. The first release (Release 1.0) of the package was written for the small farm which might have access to a 'BBC Model B' computer manufactured in Britain by the Acorn Computers Limited (Cambridge, England).
With the rapid increase in the number of computers available in the agricultural sector in Zimbabwe it became obvious that the main type of machine that was and would continue to be available was the IBM personal computer (IBM/PC) and its derivatives and copies. With this in mind a second version of the program was written. The flexibility of the system was reduced by fixing the number of constraints to 9 and by limiting it to cattle diets. With the larger amounts of memory available on the IBM/PC machines it became possible to increase to a maximum of 30 the number of ingredients considered each time a diet is formulated.
Operation of the package
Animal requirements
The package consists of three elements. The first section calculates, according to the ARC (1980, 1984) recommendations, the likely daily nutrient requirements of cattle under different production systems for ME, crude protein, rumen-degradable protein, undegradable protein, calcium, phosphorous and magnesium. The user is then prompted to enter values which he considers appropriate for the maximum amount of fat and the minimum amount of fibre in the ration (there are default values of 50 and 150 g/kg dry matter respectively). The program also calculates a maximum value for the dry-matter intake of the animals using relationships that are described elsewhere in this paper. These relationships are ones which have been developed for the most part using cattle of European breeds and maintained under good nutritional conditions. The estimates may not be appropriate for all conditions so provision is made for the estimates to be reduced in the light of experience with local conditions. Once the nutritional conditions have been set, the display shows the likely nutrient densities (in MJ/Kg or g/kg dry-matter as appropriate). If any of these figures are apparently out of Line with practical diets the figure is 'starred' and the user asked to consider collecting the dietary requirements again with more realistic production targets. As an example, it is unlikely that diets with an ME content of more than about 13.5 MJ/kg could be formulated under practical condition. Once the operator is satisfied with the ration in terms of its composition the details are sorted as a data file until required for the third segment of the package.
Feed information
The second element of the package is a data handling system for the feedstuffs which works similarly to a standard spreadsheet. This program can create and maintain a file in which details are kept of all the ingredients which are likely to be encountered in the country in question. The maximum number of ingredients that may be so maintained at the moment is 98 but there is no reason why this number should not be increased if users find the current limit too restrictive. The filing system works rather like a commercial menu-driven spreadsheet. In addition to the nutritional cats on any material the price per kg of fresh weight must be entered. The individual user in his own country will have to build up this file for his own circumstances. There is on the diskette as released a trial version of the data file but this is included only for demonstration purposes and no guarantee is given as to the accuracy or applicability of the data.
Not all of the ingredients which are stored in the main matrix file will be available to the user at any one time. For this reason when making practical diets the user chooses a subset of ingredients that he already has or to which he might have access. The maximum number that can be chosen for consideration on entry into the diet is 30. On a farm, a cooperative or in a small feed manufacturing company it is unlikely that this figure will need to be exceeded.
Restrictions on inclusion
At this stage the user may choose to set limits on the inclusion of materials into the ration. For instance, in the case of a roughage of low quality it is often prudent to set a maximum amount to the proportion of the overall diet that this can represent due to the fact that it is likely that the animal will have a very restricted voluntary intake of this material. On the other hand, if the farmer has a very large stock of some material, it may be that he has to include at least some of it in his diets to ensure that it does not go to waste. The inclusion of this ingredient will have to be limited in the opposite direction, in other words the user sets a minimum amount that must be incorporated in the diet. It has to be admitted that limits of this nature can often prove to be obstacles to the successful formulation of diets and the tool must be used with great care. One other way of promoting the inclusion of an ingredient into a diet is to reduce its price to a ridiculous value and see if the least cost formulation includes or excludes it. If the ingredient is excluded even at a very low price it may be that the prudent farmer will consider throwing it away as its use is likely to cost rather than save money.
Dry-matter intakes of roughages
It is a fairly clear limitation to the use of roughages of low digestibility that it is difficult to assess the likely intake of such materials by ruminants. Intakes of some of the poorer materials may be as low as 0.5% of body mass whereas intakes as high as 1.8% of body mass have been achieved by animals given improved roughages supplemented with sources of high quality concentrates. Obviously in designing diets for livestock the likely intake has to be a major consideration.
In assessing dry-matter intakes of dairy animals we have used models that have been developed and used with great success in industrialised countries. A selection of such regressions is presented in Table 3.
Table 3. Equations used for prediction of total dry-matter in take of cows.
|
A. |
TDMI1 = 0.10MY + 0.015LW |
|
B. |
TDMI2 = 0.2MY + 0.22LW |
|
C. |
TDMI3 = 0.076 + 0.404CDM + 0.013LW - 0.129WL + 4.120log WL + 0.14MY |
|
D. |
TDMI = 3.476 + 0.404CDM + 0.013LW - 0.129WL + 4.120log WL + 0.14MY |
|
E. |
TDMI4 = 27.8 + 106.5q x LW0.75/1000 |
|
F. |
TDMI5 = 116.8 - 46.6x x LW0.75/1000 |
|
Where TDMI is total daily dry-matter intake, CDM is concentrate dry-matter intake/day, MY is daily milk yield in kg/day, WL is week of lactation, LW is animal liveweight in kg, q = diets of 0.5, 0.6 and 0.7 kg/d of DMI. | |
Notes :
1,2. For cows on low and medium quality diets equations A and B are used, the latter for high yielding cows (defined as a daily milk yield of over 15 kg/day)3. For lactation cows receiving high quality diets (defined as those with an overall metabolisability of 0.65 and above) equation D is used, the quantity of concentrates being calculated on a sliding scale assuming a metabolisability of 0.55 for the roughage component and 0.7 for the concentrates.
4,5. The program selects equations E and F for non-lactating cows receiving coarse and fine roughages respectively.
For the origins of equations A-D see Caird and Holmes (1986). Equations E and F are from ARC (1980).
Preliminary results obtained recently in Zimbabwe in a large scale trial in which a variety of treated roughages were fed to growing heifers on dairy farms are shown in Table 4. In each of these trials the animals were fed a high quality concentrate mixture of maize, cottonseed and soyabean meals and minerals at the rate of approximately 1% of live body mass. It can be seen from these figures that dry-matter intakes ranged from about 1.4 to 1.8% of live body mass. Most of these materials were of fairly high quality prior to chemical treatment and thus effects of treatment are not as great as might be expected. On the basis of these figures and those to be found in the literature we have made the assumption that roughage comes in 4 categories: poor, mediocre, good or excellent. The likely intakes of these materials as dry-matter have been set at 0.8, 1.0, 1.4 and 1.8% of live body mass. When setting the restrictions upon the inclusion of materials in the rations the appropriate figure is entered as a maximum value for the chosen roughage. This is likely to be subject to enormous errors but it has to be remembered that the whole purpose of the package is not to provide a precise prophesy of the amount of each dietary constituent to be consumed, it is a management tool to assist in decision making.
Linear program
The third segment of the program consists of the linear programming itself. The program calls for the name of the file on which details of the ingredients have been stored and the file on which the diet is specified. The user can then look at the matrix of data which is to be used to formulate his diet. It is also possible for data to be edited at this stage, for instance the user can change the specification of the feed to see what effect this has on the composition of the diet or even on its feasibility. After the user has finalised his choice of values the program continues to calculate the least cost ration. The results are expressed in terms of dry-matter and fresh matter.
Experience in the use of the programs
The suite of programs was originally written with the small farm in mind. It has been tested over a period of two years on a group of small- to medium-sized dairy farms in the Chequtu-Kadoma area of Zimbabwe through the co-operation of the National Association of Dairy Farming in Zimbabwe. Much of this area has a mean annual rainfall of approximately 650-800 mm and some is in natural region III. Some of the farms have limited areas of irrigation but many are involved in dry-land farming. It was found that most of the dairy farmers in the area were purchasing their feeds from one of four very large commercial compounders. The initial part of the study involved taking the manufacturers' published specification of feeds and formulating similar products. The saving to the farmer of home-mixing of feeds as against the cost of purchasing a ready made product varied from 15 to 45% depending on the feed. In general, savings were greatest with those feeds which had lower nutrient densities.
Table 4. Intakes of treated roughages by heifers (Holstein and Friesian breeds) on dairy farms in Zimbabwe.
|
Site |
Roughage source |
Treatment |
Intake* |
|
Ruwa |
Maize stover |
Chopped |
1.6 |
|
Ruwa |
Maize stover |
Chopped, urea added at feeding |
1.6 |
|
Ruwa |
Maize stover |
Urea incubated |
1.8 |
|
Ruwa |
Maize stover |
NaOH (4%) |
1.8 |
|
Nyamandhlovu** |
Maize stover |
Chopped |
1.8 |
|
Nyamandhlovu** |
Maize stover |
Chopped, urea added at feeding |
1.8 |
|
Nyamandhlovu** |
Maize stover |
Urea incubated |
1.8 |
|
Nyamandhlovu** |
Maize stover |
NaOH (4%) |
1.8 |
|
Norton |
Wheat straw |
None |
1.3 |
|
Norton |
Wheat straw |
Urea added at feeding |
1.4 |
|
Norton |
Wheat straw |
Urea incubated |
1.4 |
|
Norton |
Wheat straw |
NaOH (4%) |
1.5 |
|
Chegutu |
Veld hay |
Chopped |
1.6 |
|
Chegutu |
Veld hay |
Chopped, urea added at feeding |
1.5 |
|
Chegutu |
Veld hay |
Urea incubated |
1.8 |
|
Chegutu |
Veld hay |
NaOH (4%) |
1.8 |
* Dry-matter intakes are expressed as percentage of livemass, mean values for pen of ten animals per treatment (figures rounded to one decimal point). Preliminary results of Chesworth, Smith and Spriggs.** The stover used in this trial was from green maize production and was exceedingly 'sweet'.
The next stage was to use the programs as a tool in the determination of the overall policy of the farm. Farmers were encouraged to list the quantities of materials that they expected to have available on their farms. These materials, principally low digestibility products such as veld grasses and crop residues such as maize stover and soyabean hay, were then forced into formulation both passively by setting their prices down to exceptionally low levels (typically the direct material costs of harvesting them) and by setting their levels as constraints in the formulation. Overall diets formulated in this way included roughage materials which were not normally mixed and included in a complete diet. Animals were allowed access ad libitum to the roughage and the higher quality materials were fed as a mixed meal in line with the normal feeding policy on that farm. For this reason the formulation program has the facility to give the composition of the required diet after the subtraction of one (release 1.0) or more (release 2.0) ingredients. The use of the program led to much more complete utilization of the materials available on the farm.
Future developments
Linear programming as it is used in most of the feed industry is a method which has only one clear-cut objective: it chooses the unique combination of ingredient materials which are the cheapest way of satisfying a series of fixed criteria. The feed industry does not have any duty to look further than that. The objectives of the farmer may be much more complex than those of the manufacturer. He has a number of criteria which are poorly defined and about which he has to make decisions. The decisions are made on the basis of imperfect data and are not all equal in their importance to him. As a simple example of this, compare the meeting of criteria for metabolisable energy with those for calcium. To increase the ME of a feed from 9 to 13 MJ/kg dry-matter would result in a major expense. The latter diet would probably cost several times the former. On the other hand an increase in calcium incorporation from 9 to 13 8/kg dry-matter would probably, on its own, add little to the cost of a ration. It is one of the weaknesses of linear programming that it will work just as hard to meet each of these criteria. Decisions in the real world of farming are not as clear-cut as those in the manufacturing industries and really need appropriate techniques to reflect this. There is a development of classical linear programming methods which is called linear goal programming. In this, the overall criteria are similar to those set up in linear programs but in goal programming the problem is expressed as a series of desirable objectives which can be set in order of priority. For instance the most important priority might be that a diet is produced with an ME value of 11 MJ/kg and with crude protein level which differs little from 160 g/kg. This is in fact the most important part of the specification of the diet; for the given level of production the animal needs its food and whatever price is necessary must be paid. Reducing the price of the diet then is of less importance. The third objective therefore becomes to set the overall price of the diet as near as possible to the cost of the cheapest ingredient. This is much closer to the decision-making process which is adopted empirically by farmers.
There are differences in the way in which two systems arrive at feasible solutions; the older method uses the technique of adding and subtracting 'slack' values of zero cost to allow for the over- or under-supply of nutrients. In goal programming a series of deviations from the goals or objectives (the right hand side of classical LP) are set. The solution uses an iterative procedure to reduce the overall deviation from the desired values in order of priority.
Experiences with the preliminary versions of goal programming have not shown any great deviations from the solutions that would be reached using more conventional linear programs. Advantages are likely to be more apparent when the croup of ingredients which is available to the procedure is deficient in one or more of one of the minor components of the diet.
Linear programming has proved itself to be a useful tool in the formulation of diets in situations where some choice is available in respect of the ingredients that may be incorporated into diets for productive livestock. The technology is available and the facilities are becoming more common even in developing countries. In order to be able to use the system there will be an increasing need for the compilation of feed information, not only on the feeds themselves but also on their properties after modification.
Future developments will include the adoption of more appropriate optimisation methods that more closely parallel the empirical thinking needed to make rational decisions in animal production.
Addison, K.B. 1952. The effects of various cultural and manurial treatments on Napier fodder. Rhod. Agric. J. 53(4): 491.
ARC (Agricultural Research Council). 1980. The nutrient requirements of ruminant livestock. Agricultural Research Council, London, England.
ARC (Agricultural Research Council). 1984. The protein requirements of ruminant livestock. Agricultural Research Council, London, England.
Caird, L, and Holmes, W. 1986. The prediction of voluntary intake of grazing cows. J. Agric. Sci., Camb. 107:43.
FEU (Feed Evaluation Unit). 1980. Third Annual Report of the Feed Evaluation Unit. Rowett Research Institute, Aberdeen, Scotland.
LP88. 1984. Linear Programming for the IBM Personal Computer. Eastern Software Products, Inc., Virginia, USA.
Preston, T.R. 1986. Strategies for optimizing the utilization of crop residues and agro-industrial by-products for livestock feeding in the tropics. In: T.R. Preston and M.Y. Nuwanyakpa, (eds), Towards optimal feeding of agricultural by-products to livestock in Africa. Proceedings of a workshop held at the University of Alexandria, Egypt, October 1985. International Livestock Centre for Africa, Addis Ababa, Ethiopia.
