A model for decision making under non-certainty on Canterbury light land farms

Abstract

Annual production on Canterbury light land sheep farms fluctuates markedly. A major reason for this is the extreme climatic variability both between years and between seasons. This has a major, effect on feed supply, and livestock farmers cannot plan their management programmes with any degree of certainty. Consequently, it is extremely difficult to maintain a system that balances feed supply and demand and maximises long-run profits. This problem is referred to in the text as the stocking rate problem. The study considers the significance of this problem and possible methods of allowing for non-certainty and thereby improving long-run profits through more accurate planning.

Non-certainty is most commonly accounted for by assessing expected values for input and output coefficients and using these values for planning. However in the case of Canterbury light land, the "expected" situation rarely eventuates and management plans are often sub-optimal and in some situations infeasible. Further, farmers are generally risk averse and therefore base long term policy decisions on worst outcomes so that feasibility is retained under all possible outcomes. Although management adjustments are made as conditions change, adjustment alternatives are limited by the predetermined long term policy. Therefore if conditions are better than expected, their full benefit cannot be realised.

An attempt is made to overcome these planning problems by developing a multi-period stochastic Linear Programming model using a case-study farm as a base. It allows management plans to be determined on the basis of both the current state of the farming system and likely future conditions (described by a range of discrete state of nature outcomes with associated probabilities). Thus the model output defines optimal management strategies together with the management action required if conditions change during the year.

It is shown that long-run profitability can be improved by more actively allowing for non-certainty when making management decisions. In fact expected or other single value estimates for input/output coefficients can lead to infeasible management plans. Management systems determined by the model are much more flexible than those commonly used on light land farms. In particular, stocking rates fluctuate according to both past and present environmental conditions. Finally with the addition of further management detail the model could be used to assist decision making on individual farms. Management strategies could be planned and updated as environmental and market expectations change.