Previous PageTable Of ContentsNext Page

Computer-based decision support tools for Australian farmers

Andrew D. Moore1, J.F. Angus1, Michael Bange2, Chris J Crispin1, John R. Donnelly1, Mike Freer1, Neville I. Herrmann1, Heidi E Ottey1, Dirk Richards2, Libby Salmon1, Maarten Stapper1 and Alexander Suladze1

1 CSIRO Plant Industry, GPO Box 1600, Canberra ACT 2601. www.csiro.au Email Andrew.Moore@csiro.au
2
CSIRO Plant Industry, Australian Cotton Cooperative Research Centre, Locked Bag 59, Narrabri NSW 2390

Abstract

A suite of computer-based decision support tools is available for use by farmers and advisers in the broadacre farming systems of southern Australia. The tools are widely used in the wheat, rice, cotton and grazing (sheep and beef cattle) industries and for mixed crop-livestock farms. They cover planning horizons from days to decades and provide information to maximise profit, such as tactics for insect control in cotton or N topdressing of wheat, or strategies such as stocking rate or lambing time of sheep. All rely on inputs from the user to start a simulation model and a database of weather records and parameters to drive the model. The crop-oriented tools address a single field at a time; the livestock-oriented tools can also deal with movement of herds around fields. Information provided in these tools is more specific to location, season and economics than alternatives provided as rules of thumb or blanket recommendations.

Media summary

A suite of decision support tools provides Australian farmers and their advisers with unique and timely information to maximise profit.

Key Words

Decision support, models, cotton, livestock, rice, wheat

Introduction

Australian farmers need to be more efficient than their counterparts elsewhere to be competitive on world markets. They must continuously improve the efficiency of their existing production systems and rapidly adopt new technology for more profitable farming systems. These changes must be implemented in a context of significant year-to-year weather and price variability. Tactical decisions such as timing of pesticide application or irrigation are always contingent on uncertain future weather, while strategic decisions are conditioned by climatic risk, especially risk of drought. Computer-based decision support tools (DS tools) are resources that farmers and their advisers can use to deal with these challenges. CSIRO Plant Industry has developed a range of DS tools for extensive and semi-intensive agriculture in southern Australia.

The decision support toolkit

The suite of agricultural DS tools developed within CSIRO Plant Industry covers extensive and semi-intensive enterprises and a range of time scales from days to decades (Table 1). All these tools (except MetAccess) contain mathematical models that run using site-specific values supplied by the user and weather records and parameters stored in a database, rather than relying on pre-computed values. This strategy has arisen from the diversity within Australian agricultural enterprises: each DS tool must accommodate a large space of environments and decisions. The weather database also provides structured access to the daily surface data collated by the Bureau of Meteorology for Australian weather stations via the MetAccess (Donnelly et al. 1997) decision tool. The entire data set for Australia is stored on a single CD. In maNage rice and CottonLOGIC, up-to-date weather data can be downloaded using the Internet. The distribution channels for the tools depend upon the structure of the relevant industry. The cotton and rice packages are distributed to growers and advisers through industry organisations. The other tools are marketed to the dispersed dryland cropping and grazing industries through agribusiness partners of CSIRO. This approach provides more flexibility than presenting pre-computed model results.

Table 1. Summary of the CSIRO Plant Industry decision support tools.

 

Purpose

Planning Horizon

Intended Users

Year of Release

GrazFeed

Nutritional responses of sheep and cattle at pasture

Days

Advisers, Producers

1990

CottonLOGIC

Irrigation planning and scheduling, pesticide timing and nitrogen fertilizer rate decision-making in cotton

Days-Months

Producers

1996

MetAccess

Analysis of weather and climate data

Days-decades

Producers-Scientists

1993

maNage Rice

Nitrogen fertilizer rate decision-making in rice

Months

Producers

1994

maNage Wheat

Nitrogen fertilizer rate decision-making in wheat

Months

Producers

2003

GrassGro

Analysis of sheep and cattle grazing enterprises

Months-Decades

Advisers

1997

Lime & Nutrient Balance

Budgets for multiple nutrients and protons in mixed farming systems

Years-Decades

Producers

2003

FarmWi$e

Systems analysis of mixed farming systems

Years-Decades

Advisory Service

2004

GrazFeed2

The GrazFeed DS tool assists in tactical nutritional management of sheep and cattle grazing at pasture and supplied with feed supplements. It uses a model of diet selection, feed intake and nutrition (Freer et al. 1997). Users provide estimates of the mass and digestibility of herbage, plus attributes of the grazing livestock; the model predicts the likely production of meat, wool or milk. Since 1990, GrazFeed has evolved from an application used mainly by advisers to one for independent use by graziers. The shift in the user base was largely brought about by a synergy with the PROGRAZE extension programme (Bell & Allen 2000). Over 1000 copies of GrazFeed have been distributed since its release.

CottonLOGIC

Australian cotton growers have a long history of using computer-based decision support (Hearn and Bange 2002). The CottonLOGIC package is a suite of three software tools for cotton production. NutriLOGIC assists cotton growers to determine the economically optimum nitrogen fertilizer rate for their crops based on soil or petiole nitrate tests. The model within NutriLOGIC estimates the optimum rate from these measures of N status, adjusted for soil type and sampling time.

HydroLOGIC is an irrigation management tool for furrow-irrigated cotton crops. Using HydroLOGIC irrigation managers can evaluate the consequences of different irrigation strategies on yield and water use using a range of simple plant and soil moisture measurements. Specifically, HydroLOGIC can be used to estimate the optimum cotton cropping area given a certain water allocation; schedule the next irrigation; conduct scenario analysis exploring the impact of timing of irrigations with different allocations of water; and benchmark the performance of previous cotton crops. HydroLOGIC employs the OZCOT model of cotton growth, development and yield (Hearn 1994). In large-scale commercial evaluations, HydroLOGIC recommendations compared favourably with standard irrigation practices for yield and fibre quality with full water allocation, and the tool was able to improve crop water use efficiency when water was limited.

Chemical pesticides are a large part of the cost of cotton production. Successful adoption of the principles of Integrated Pest Management (IPM) is vital to maintain economic viability, to prevent or delay the emergence of pesticide-resistant insect populations and to assist in meeting community concerns about environmental effects of pesticide use (Wilson et al. 2004). EntomoLOGIC is the pest management component of CottonLOGIC that supports the cotton industry’s efforts in adopting IPM. Users select sample areas in cotton fields and collect information on the types of beneficial and 'pest' insects present, their stage of development and quantity. The EntomoLOGIC software is then used to predict future pest numbers and indicates when pest numbers exceed defined economic thresholds. Cotton pest managers can then use this information to make their own decisions on when and how to control pests.

CottonLOGIC also contains options for record keeping and reporting of crop operations and an insect identification library. The number of registered CottonLOGIC users has increased steadily from 200 in 1995 to over 1100 in 2004. A recent survey (Bange et al. 2004) showed that CottonLOGIC was used across 51% of the 404 000 ha of cotton grown in Australia in 2002. Of this usage, 93% was for keeping records of insect and operational data, while 69% was to assist with management decisions using models embedded in CottonLOGIC.

maNage Rice

This started as a tool to advise on topdressing N fertilizer on irrigated rice in the Australian Riverina, based on variety, sowing date, plant N status, risk of cold damage, grain price and N-fertilizer cost. It presents options to growers as expected yields, gross margins and probability of achieving a target return, based on the variability due to cold damage. It has broadened to include real-time calculations of water requirement and harvest scheduling. A recent addition is a library of images of physiological disorders, pests and diseases, similar to the library in CottonLOGIC. A start has been made to enable users to designate zones within a field, guided, but not controlled by, paddock mapping information. This feature initially directs sampling of plant N status to different zones and then runs the underlying model on the zones.

maNage Wheat

maNage Wheat provides decision support about the timing and rate of N fertilizer to wheat crops for a particular paddock and season. It simulates growth, development, water balance and N balance of a crop after the user provides initial conditions of N status, sowing date, variety and seasonal weather to date. The system can be used strategically for pre-sowing N fertilizer, or tactically for topdressing when there is more information available about water supply. Output is expressed as expected yield and protein content for the range of seasons simulated, the gross margin for these mean values and the probability of the gross margin exceeding a user-supplied target. A related package called ShowDevel presents model-based predictions of the phasic development of wheat crops (Stapper & Lilley 2003); it has as yet only had a restricted release.

GrassGro

The GrassGro DS tool (Moore et al. 1997) allows strategic and tactical management options for sheep and cattle enterprises to be tested, taking into account variability in pasture supply. Users provide descriptions of soil physical parameters, the mass and phenological state of each pasture species, the breed and initial weights of livestock and the management system; the complete grazing system is then simulated using the Freer et al. (1997) ruminant biology model coupled with a soil moisture budget, a multi-species pasture growth model and a management model (Moore et al. 1997). Because of its relatively high degree of complexity, GrassGro is targeted to industry advisors. The DS package includes a training component and follow-up support for its 200 users. GrassGro is used in six Australian universities to teach pasture agronomy from a systems perspective (University of New England 2003; H. Daily, pers. comm.).

Lime & Nutrient Balance

This tool was commissioned by GRDC as a computer implementation of a slide-rule based Lime and Nutrient Calculator developed by Agriculture Western Australia. It provides a 10-year summary of inputs and outputs of all mineral nutrients. It calculates losses from removal as product and leaching, and gains as fertilizer (input by the user), in rainfall and N-fixation. The package also contains simple supply-demand budgets for N and P fertilizer and lime, based on soil tests, yield targets provided by the user, and parameters for the major soil types. It does not consider weather-risk, timing of application or economics, and is intended as a simple introduction to computer systems.

FarmWi$e

FarmWi$e is a new generic simulation tool for agricultural enterprises that gives the user full control over the structure of the modelled system. It is designed to facilitate the analysis of complex agricultural management questions, for example in situations where management activities are strongly dependent on seasonal conditions (e.g. livestock trading policies) and in exploratory studies of novel livestock production systems as a precursor to real-world trials. Any process model can be employed within FarmWi$e, because simulations are implemented within a "common modelling protocol" that handles information exchange between sub-models. Management activities are conceptualized as a set of simple "events" that alter the state of a sub-model. The series of events that takes place in a simulation is governed by a set of rules that describe conditions under which management events will take place.

Discussion & Conclusions

Economic impact of the tools

The impact of three of CSIRO Plant Industry’s DS tools has been assessed recently by independent economic analysts (the Centre for International Economics, Sydney). The resulting benefit:cost ratios were high: 11.2 for maNage Rice, 18.5 for EntomoLOGIC and 79 for GrazFeed.

Implementation issues

All our DS tools are Microsoft Windows applications, with the exception of a version of CottonLOGIC for Palm OS handheld computers (Bange et al. 2004) which was requested by clients. At present we use the Internet to provide updates and support, and the original versions are supplied on CD. Some require a signed licence, but there is a growing confidence that clicking on a licence during installation provides sufficient legal protection. Our software development has become more efficient and robust over time through the adoption of formal software design processes and the re-use of data, models and interface elements.

Some lessons learnt

We learned much from SIRATAC, the pioneering DS tool for cotton (Hearn and Bange 2002). One lesson is that advisers as well as producers should be involved in the process of delivery of DS information. Another is to accept that DS tools may have a fixed lifespan, because users learn to anticipate the outputs. Many of our DS tools are delivered as a series of versions with new features that keep the interest of users (Hearn and Bange 2002).

Some of our tools (CottonLOGIC, maNage-Rice) were developed in response to customer demand; others (GrazFeed, GrassGro) were the result of scientists identifying an opportunity. Our experience is that “science push” and “customer pull” can both work: what matters is that the resulting tool makes an existing decision-making process more reliable or helps to generate a new, useful decision-making process (as was the case with GrazFeed and the PROGRAZE programme).

Our tools provide assistance both directly to producers or indirectly, via advisers, depending on the characteristics of the industry and the complexity of the problems addressed. For all but the simplest DS tools, we have found that an appropriate level of training and user support must be provided if the software is to be adopted. If conducted well, the training and support processes generate a cycle in which user feedback guides further software development – and identifies questions requiring research that, when completed, can be immediately implemented.

Acknowledgements

The development of these DS tools has been supported by the Research and Development Corporations for Grain, Cotton, Wool and Rural Industries, and Landmark Ltd.

References

Bange MP, Deutscher SA, Larsen D, Linsley D, and Whiteside S (2004). Handheld decision support system facilitates improved insect pest management in Australian cotton systems. Computers and Electronics in Agriculture. 43, 131-147.

Deutscher SA, Bange MP and Rochester IJ (2001). Testing NutriLOGIC, a decision aid for nitrogen fertilizer management in cotton. In Proceedings of the 10th Australian Agronomy Conference, Hobart, TAS, Australia. http://www.regional.org.au/au/asa/2001/3/c/deutscher.htm.

Donnelly JR, Moore AD, and Freer M (1997). GRAZPLAN: decision support systems for Australian grazing enterprises. I. Overview of the GRAZPLAN Project and a description of the MetAccess and LambAlive DSS. Agricultural Systems 54, 57-76.

Freer M, Moore AD and Donnelly JR (1997). GRAZPLAN: decision support systems for Australian grazing enterprises. II. The animal biology model for feed intake, production and reproduction and the GrazFeed DSS. Agricultural Systems 54, 77-126.

Hearn AB (1994). OZCOT: a simulation model for cotton crop management. Agricultural Systems 44, 257-299.

Hearn AB and Bange MP (2002) SIRATAC and CottonLOGIC: persevering with DSSs in the Australian cotton industry. Agricultural Systems 74, 27-56.

Moore AD, Donnelly JR and Freer M (1997). GRAZPLAN: decision support systems for Australian grazing enterprises. III. Pasture growth and soil moisture submodels, and the GrassGro DSS. Agricultural Systems 55, 535-582.

Stapper M and Lilley JM (2003). Evaluation of SIMTAG and NWHEAT in simulating wheat phenology in southeastern Australia. Proceedings of the 11th Australian Agronomy Conference, Geelong, Victoria, (Australian Society of Agronomy). http://www.regional.org.au/au/asa/2001/1/d/stapper.htm.

University of New England (2003). eDSServe. http://edsserve.une.edu.au.

Wilson LJ, Mensah RK and Fitt GP (2004). Implementing integrated pest management in Australian cotton. pp 97-118 in ‘Insect Pest Management: Field and Protected Crops’ (eds AR. Horowitz and I Ishaaya). (Springer-Verlag: Berlin).

Previous PageTop Of PageNext Page