Modelling alternative dryland sheep systems

Abstract

On the east coast of New Zealand, sheep and beef cattle are increasingly confined to dry hills and un-irrigated flat land as land suitable for irrigation is converted to other uses. Dryland farming is subject to significant variability in temperature and particularly rainfall between and within years with the most important climate risk being the point at which soils dry out and pasture production ceases in late spring/summer. Improving pasture and animal performance and, in particular, the consistency of productivity and profitability in the face of a highly variable climate is complicated. The challenge is to utilise the 3-5 month window of opportunity for production between August and the end of the year to best advantage and without compromising the ability to feed ewes well in late summer/autumn prior to mating. Key variables in this context are high lamb growth rates in order to finish as many lambs as possible before the risk of dry conditions becomes too high, and flexibility to respond to the growing conditions as they unfold. The objectives of this research were to investigate, and demonstrate, opportunities for improving dryland sheep systems through increased lamb output, high pasture quality and utilisation, and flexibility to respond to climate and feed conditions. The research included a farm scale trial, adaptation of an existing sheep farm simulation model LincFarm (Cacho et al., 1995; Finlayson et al., 1995) to replicate the trial situation, development of an algorithm for optimising de-stocking interventions in response to dry conditions, and evaluation of the long term implications on productivity, profitability and risk of a range of policy options using the model. This report provides a summary of the field trials, and describes extensions to and evaluation of the LincFarm sheep systems model and its use to analyse a combination of stock and pasture options at different stocking rates based on the field trial. Development of the de-stocking algorithm and evaluation of a range of risk responses are presented elsewhere (Gicheha et al., 2013 a, b).