SEA Working Paper 2000/03

Decision support for integrated weed management

David J. Pannell

Agricultural and Resource Economics, University of Western Australia, Nedlands WA 6907

Abstract

Integrated Weed Management (IWM) is generally taken to mean use of a diverse mixture of chemical and non-chemical weed management practices. It is a strategy that may or may not be helpful to farmers, given their particular individual circumstances and objectives. However, evaluating possible IWM strategies is complex and difficult for farmers. A decision support system, such as the Australian package RIM, eases some of these difficulties. RIM simulates ryegrass (Lolium rigidum) population dynamics, competition and economic costs and returns for any user-specified IWM strategy for a period of up to 20 years. It provides a number of insights about the economics and biology of IWM, including the following. Even if herbicides are available for use, the optimal combination of control practices may include non-chemical methods. If restrictions are placed on herbicide use (e.g. voluntary restrictions to delay resistance, legal restrictions to limit adverse side effects, or biological restrictions due to the onset of resistance) it is biologically feasible to replace the herbicides with a suite of non-chemical treatments. Individually these tend to be less effective than selective herbicides, so a greater number of treatments must be employed. Overall, the economic returns to farmers from a low-herbicide system are likely to be lower than a more herbicide-intensive system. If herbicide use is limited to low levels, the best available integrated strategy involves approximately the same average density of weeds as a herbicide-based system. Thus the economic losses due to reduced herbicide usage are not primarily due to differences in weed density, but to differences in total treatment costs. There appears to be no compelling case for reducing the reliance on herbicides in order to delay the time when they will be lost to resistance.

Key words: decision support, economics, herbicide resistance, modelling, Lolium rigidum

Introduction

Integrated weed management (IWM) is not easy to define precisely, but it clearly implies use of a diversity of weed control methods, including non-chemical methods. Many farmers already profitably combine a range of control methods, more so in some farming systems than others. In any farming system, with the onset of serious, multiple herbicide resistance, it becomes essential to broaden and diversify control methods, and so farmers are increasingly adopting strategies that would satisfy even extreme definitions of IWM.

However, farmers face a number of difficulties in their decision making about IWM strategies:

• the farmers are inexperienced and perhaps unfamiliar with a number of the control options;
• strategies must be evaluated over the longer term, not just for a single year;
• the long-term impacts of multiple control options are difficult to predict;
• the impacts of individual treatments within an integrated strategy are difficult to interpret from field observations;
• some strategies have indirect, as well as direct costs; and
• the number of possible combinations of treatments to be considered is vast.

Given these difficulties, IWM seems a topic for which computerised decision support systems (DDSs) could be especially valuable to farmers. However, DSSs for weed management have focussed primarily on herbicides and most have had a relatively short-term focus.

In this paper I will discuss the concept of IWM and its relevance to actual farm management decisions. Then I will briefly outline some difficulties in developing a DSS for IWM, and suggest that these difficulties explain the scarcity of such systems internationally. Finally, the Ryegrass Integrated Management (RIM) model is described and used to demonstrate some insights about IWM.

Should Integrated Weed Management Be An Objective?

A primary feature of IWM is diversity:

• diversity of treatment types,
• diversity of crop or pasture types grown in rotation, and
• diversity of stages in the weed’s life cycle when treatments should be applied.

Sometimes, IWM is discussed by scientists as if this diversity means that IWM strategies are intrinsically good and unambiguously worthy of adoption by farmers. This is confusing strategies with objectives. IWM is a strategy that may or may not be helpful to farmers in achieving their objectives. Farmers’ objectives for their farms are likely to include:

• profitability,
• avoidance of risk,
• family and personal objectives,
• resource conservation/sustainability.

IWM may, in some cases, be helpful to farmers in pursuing these objectives, but in cases where it is not, we should not be surprised if IWM is not embraced by farmers.

What the promotion of IWM to farmers does provide is encouragement to farmers to consider a broader range of treatment options than they might otherwise have considered. When assessing these treatments, farmers will consider them on their merits, in terms of how well they help farmers to meet their objectives of profitability, low risk, sustainability and so on. No single objective takes absolute precedence over the others, although evidence about farmer adoption of innovations indicates that profitability is a particularly important objective to farmers (e.g. Cary and Wilkinson, 1997; Lindner, 1987; Pannell, 1999; Sinden and King, 1990).

Thus, IWM, if it is practiced, can be seen as an artifact of the farmer’s decision process. It is the result of considering a wide range of treatment options and combining them in the "best" way. So, from the farmer’s perspective, the question is not, IWM or not. Rather, it is, treatment A or not, treatment B or not, treatment C or not, and so on. Of course, the treatments must be considered within the context of other treatments that will also be used.

If the result of this process is a diversified portfolio of treatments, it might reasonably be described as IWM. If not, there is no particular cause for alarm, provided that in making the decision, the farmer was reasonably well informed about the treatments and their impacts.

The exception to this perspective on IWM would be where non-IWM strategies involve substantial off-farm costs, such as risks to health or the environment from herbicide use. In this situation, these spillover effects may result in "market failure" (Pannell, 1994) providing potential grounds for government intervention. However, for such intervention to be effective, it almost certainly requires an approach stronger than merely informing farmers that they are causing adverse off-site impacts. Cary and Wilkinson (1997) and Sinden and King (1990) both identified the inadequacy of awareness raising as the sole vehicle for achieving behaviour change for issues with broader environmental impacts.

Decision Support Systems

As noted earlier, assessment of the multitude of weed treatment options available to farmers appears to be a problem that would benefit greatly from the availability of a computerised DSS. However, there has been a notable absence of such a DSS. The RIM model (Pannell et al. 1999), described below, appears to be the only example of a weed DSS that represents a comprehensive set of weed control options, including both chemical and non-chemical options.

Factors contributing to the apparent absence of similar DSSs outside Australia might include the following:

• Farmer interest: Southern Australia, and particularly the state of Western Australia where RIM was developed, has the most severe problem of herbicide resistance in the world. This has created unprecedented farmer interest in non-chemical weed control methods. In other regions without such severe resistance problems, the interest in and demand for a product like RIM would probably not be as great.
• Data: Developing such a DSS involves substantial requirements for data (Gerowitt 2000). Commonly, comprehensive data relating to key aspects of potential IWM systems is not available. Examples of hard-to-find information include the effectiveness of the full range of non-chemical control methods, and elements of weed population dynamics for the relevant species. The herbicide resistance problem has generated a major research effort in Australia that has provided information about most of the inputs needed for RIM.
• Expense: Development of RIM has occurred over nine years and has required inputs from many individuals (scientists, economists, consultants, farmers). It clearly has entailed a substantial commitment of resources. Fortunately, it has been well supported by the Grains Research and Development Corporation, a contestable research fund that distributes funds from farmer levies and matching government funds.
• Economic expertise: A key element of such a DSS, if it is to be of high value to farmers, is the representation of economic outcomes. In many places, expertise in the economic analysis of weed management is difficult to find, but Australia is fortunate in having a history of activity in this field (e.g. Pannell 1988, 1990; Pandey and Medd 1990; Abadi and Pannell 1991; Pandey et al. 1993; Gorddard et al. 1995, 1996; ).
• Need for multidisciplinary inputs: As noted above, development of RIM required extensive inputs from many disparate groups and individuals with different expertise. The difficulties of conducting research and analysis involving multiple disciplines is widely acknowledged (e.g. Pannell and MacAulay 1998). The team involved in development of RIM has had a history of successful cross-disciplinary collaboration in research (e.g. Pannell 1996).

This combination of reasons probably explains why RIM has been developed in Western Australia and not elsewhere.

The RIM Model

Overview of RIM

RIM is a decision support system – it is designed to provide information and insights to farmers to help them in their long-term decision making about management of ryegrass, the most important weed of crops in Australia. RIM allows the user to try out many different combinations of weed control treatments and observe their predicted impacts on ryegrass populations, crop yields and economic outcomes. Applications of early versions of RIM are presented by Stewart (1993) and Schmidt and Pannell (1996a, 1996b).

RIM represents a paddock. The user can specify whether or not the ryegrass population in the paddock is resistant to each herbicide group, or how many shots of each group are available before resistance will develop. A wide variety of non-chemical weed treatment options are included, so that as chemicals are lost, the next best substitute can be identified.

RIM is useful for a number of different types of users, including:

• farmers attempting to make decisions for their own weed management;
• private consultants, extension agents or agribusiness agronomists wishing to provide advice to their clients;
• facilitators running RIM with groups of farmers;
• scientists, students and others wishing to understand the management of herbicide-resistant ryegrass.

The enterprise options available for users to select are: wheat, barley, canola, lupins, volunteer pasture, sub-clover pasture, and cadiz pasture. These may be selected by the user in any agriculturally feasible sequence. There are inter-year impacts of one enterprise on another, depending on the sequence selected.

Biology

Weed population dynamics

The key factors driving the pattern of weed population change over time are as follows:

• Initial weed seed density.
• The timing of weed seed germination relative to the crop. Later germinating weeds produce fewer seeds per plant.
• Natural mortality of weeds or seeds.
• Seed production per plant.
• Impacts of weed and crop densities on seed production per plant.
• The timing and effectiveness of treatments to reduce weeds or seeds.

All of these factors are represented in RIM.

Competition between weeds and crops

The yield of a crop depends on the relative competitive abilities of that crop and of ryegrass, and the densities of each. The standard competition relationship for wheat yield as a function of ryegrass density is shown in Fig. 1.

Figure 1. Wheat yield as a function of weed density.

Crop-related variables

The following crop-related variables are represented.

• Standard weed-free yields for crops after a break of at least three years without a legume.
• Yield boosts for cereals after legumes or canola.
• Yield effects on crops of green manuring and swathing.
• Yield effects from disease in short rotations.
• Seeding rates.
• Savings in nitrogen fertiliser following lupins (a legume).
• Impacts of delayed sowing on yield.
• Parameters of the competition functions.
• Phytotoxic effects of herbicides on each crop.

Pasture-related variables

There are three types of pasture represented in RIM: a volunteer pasture, a phase pasture (assumed to be cadiz serradella) and a regenerating pasture (assumed to be sub-clover). For some purposes, cadiz and sub-clover are treated as being equivalent (e.g. treatment effectiveness, treatment costs).

RIM does not include detailed simulation of the population dynamics for each possible pasture species, so the biological impacts of a pasture phase on ryegrass populations are represented in a relatively simple (but effective) way. For each type of pasture, its impact on the ryegrass seed density under standard grazing conditions is specified by the user. The standard reduction in weed seeds is greater in a second or third consecutive year of pasture because the non-ryegrass components of the pasture stand are denser and more competitive by that stage.

Economics

Users of RIM quickly come to appreciate the importance of taking a long-term view on the economics of weed management. RIM highlights the potential for long-term benefits from short-term economic sacrifices. It allows you to assess trade-offs like this in a balanced way. It certainly does not presume that every preventative strategy is economic in the long term. This depends on a host of factors, including the cost of the strategy, its impact on weeds, prices of outputs and the initial weed seed density.

Taking a long-term view of economics poses some problems. In particular, how should you assess the overall economics of a strategy for which the gross margin changes dramatically from year to year? How can you validly compare costs and benefits that occur in different years, given the complexities of interest, tax, price trends and trends in yields?

The standard approach used by economists and financial analysts to assess long term investments involves a process called "discounting", which allows all costs and benefits to be expressed in the equivalent of their present day value. The costs and benefits of all strategies of interest would be discounted and added up, and the preferred strategy would be that with the highest "Net Present Value".

If the discount rate used is the bank interest rate (which it often is), then this process is equivalent to identifying the strategy that would result in the highest bank balance at the end of the period (assuming that all income is deposited in the bank account and accumulates interest, and all costs are withdrawn from the bank account and reduce the amount of interest earned). This "final bank balance" approach is the method used in RIM because we believe that most people find it easier to understand. The approach also makes it easier to include some realistic complexities that are often ignored in long-term financial analyses. RIM includes each of the following complexities in its calculations.

1. Tax is paid on interest earned. We represent the tax system very simply, because there is so much variability between farmers in their tax arrangements. RIM allows the user to specify a single marginal tax rate, which should be the rate of tax they would pay on any additional income earned above current income.
2. The inflation rate on sale prices in agriculture has historically been lower than the inflation rate on input purchase prices. This is commonly referred to as the "cost-price squeeze" and is the reason why farmers have had to improve their productivity in order to remain in business. We assume that this trend will continue for the time being so we set the inflation rate on crop and sheep product prices lower than the assumed inflation rate on input costs.
3. Yields increase over time. This is hard to predict, but over the long term is a very significant factor. In the standard RIM we have set the annual rate of yield increases to 1.0 percent for crops and 0.5 percent for sheep products. These relate to yield per hectare, not per sheep.

Treatment options

There are a total of 35 different weed treatment options included in RIM (Table 1). They can be broken into four separate groups: selective herbicides (11), non-selective herbicides (5), non-chemical treatments (16) and user-defined treatments (3).

Table 1. Weed treatment options included in the RIM model.

* N = Non-selective herbicide, S = Selective herbicide, B = "Biological" treatment (non chemical)

In the case of herbicides, the user can specify the number of shots available for each herbicide group, prior to the onset of resistance. If ryegrass is fully resistant to a herbicide group, the limit for that group is set to zero.

Method of delivery

The primary vehicles for RIM to be delivered to farmers are:

1. workshops with farmers;
2. one-to-one sessions between farmers and their agronomic advisors, particularly in the private sector;
3. inclusion of general conclusions and messages from RIM in extension programs conducted by public sector or university extension personnel.

Based on our observation of other computerised DSSs in Australia and around the world, it is not our expectation that independent use of the package by farmers will be the primary means by which RIM has its impact. Nevertheless the package is available for purchase by farmers, and many copies have been sold. We have not yet undertaken any formal evaluation of its impacts via these different possible channels, but it is hoped that we are able to undertake such evaluations in future. Current research by Rick Llewellyn is investigating the impact of the RIM workshops with farmers on farmer knowledge and their farm management decisions.

Limitations

RIM will not automatically calculate which strategy is "best". Users evaluate strategies using experimentation and trial and error.

RIM does not represent year to year variation in weather, potential yield or herbicide performance. Yields in the model do vary from year to year due to the sequence of crops and pastures selected, and the level of weed competition. Climatic conditions do not rule out any of the treatment options. Users can self-impose constraints on the use of different treatments.

RIM represents only a single paddock. Some strategies may involve changes in machinery or livestock management that have impacts at the whole farm level. Similarly, RIM makes particular assumptions about the way that investments in machinery are financed. Farmers may need to further consider whole-farm cash flow implications of strategies outside of RIM before making adoption decisions.

Although considerable effort has been expended on data collection, there are still areas where the available information is relatively weak. This seems inevitable in such a comprehensive model. Sensitivity analysis (Pannell, 1997) is an important approach for evaluating the significance of data deficiencies. A related issue is the variation in biological and economic parameters between farms. The values included in the standard version of RIM are representative of a typical farm in a region of Western Australia, but need adjusting for other farm types and for other regions. Users can readily alter the parameter values to suit their particular situation. Further information on RIM is available at http://www.general.uwa.edu.au/u/dpannell/rim.htm

Insights on IWM

In this section, a number of insights into the performance of IWM strategies are outlined, all derived from results of the RIM model.

1. The selection of an integrated system of weed management practices is difficult. Even with the assistance of a DSS such as RIM, the number of potential combinations and sequences of practices is so vast as to be impossible to explore fully. If the task is difficult in a computer simulation, it must be dramatically more so in the real world.

2. A strategy to limit the usage of herbicides in order to preserve their useful life or reduce off-site impacts can involve substantial economic costs to farmers. The evidence for annual ryegrass in Australia is that such a strategy does not result in a greater number of uses of a particular herbicide before resistance is evident. Its advantage is purely in maintaining the potential to use the herbicide in subsequent years.

To illustrate the economic consequences, Table 2 shows the results of reducing reliance on a selective herbicide over a 10 year period. The scenario is simplified for illustrative purposes. It is based on the assumption that Group A herbicides (fops and dims) are the only selective herbicides available. No constraints are placed on the use of non-selective herbicides or non-chemical treatments, other than those that are required agriculturally. Results are shown for different intensities of use of the selective herbicides, ranging from 10 uses over the 10 years down to 2 uses. The lupin-wheat cropping rotation is used throughout.

Table 2. Consequence of restricting usage of selective herbicides.

* The number of years in which this treatment was applied is shown in brackets, if the usage is less than the maximum potential.

The cost of restricting herbicide usage voluntarily increases as the restriction is tightened (see the bottom line of Table 2). In this example, if only two uses of selective herbicides are allowed over the 10 years, profit is reduced by 30 percent relative to one use per year. This is despite the inclusion of an array of non-chemical treatments to replace the herbicide.

Note that the cost indicated by comparing profit figures in different columns of Table 2 is an over-estimate of the cost of conserving herbicides. This is because it fails to account for the benefits of having herbicides available for continued use beyond the 10-year period. In other analyses (not reported here) we find that when these benefits are properly accounted for, the strategy of conserving herbicides is, in most cases, only slightly less profitable than the strategy of using them up more rapidly.

3. It is possible to maintain the continuous cropping rotation with reduced herbicide usage and a substantially altered weed management system. The altered system involves a greater diversity of treatment types. Each of these is individually less effective than herbicides, so a greater number of treatments must be employed. The fewer the number of herbicide applications, the greater the number of non-chemical treatments that are profitable to employ. The third row of Table 2 shows how total usage of non-chemical treatments falls as reliance on selective herbicides is increased.

4. Well-designed, economical strategies involving less reliance on selective herbicides result in almost the same average density of weeds as do herbicide-dominant strategies. This is consistent with survey results in Western Australia, which have found that weed densities in farmers’ paddocks with herbicide resistance are, on average, no greater than in non-resistant paddocks. Thus the economic difference between the scenarios is not primarily due to differences in weed densities, but to differences in total treatment costs.

5. There appears to be no compelling case for reducing the reliance on herbicides in order to delay the time when they will be lost to resistance. Even in cases where early adoption is in fact beneficial, it will be very difficult for farmers to determine that this is true. In the face of major uncertainties about such a change, and no compelling arguments in favour of it, most farmers are likely to maintain a more or less traditional, herbicide-based weed management system until forced to change by the full development of resistance.

6. It is often not possible to generalise about the desirability of a particular practice. Its attractiveness to farmers will depend on the context within which it will be used. That context includes the direct cost of the treatment, the weed density at the time of usage, the other treatments being employed, the sale price of outputs, and so on. This is illustrated in Table 2, where delaying seeding is often an economically attractive option in years when a selective herbicide is not used, but is not used at all in the strategies that involve 8 or 10 herbicide uses.

7. Some non-chemical practices may be economically viable even in the context of a herbicide-oriented system. In Table 2, the strategy of increasing crop seeding rates to increase competitive effects of crops against weeds is attractive in all scenarios, including the herbicide-rich scenarios.

8. When high levels of herbicide resistance develop, the farmer has no choice but to employ an IWM strategy involving diverse chemical and non-chemical practices. The column for 2 herbicide uses in Table 2 illustrates the kind of strategy that becomes economical in the context of low herbicide availability.

Conclusion

Use of a DSS such as RIM provides a number of valuable insights into the nature of the integrated weed management problem. As we have seen, comparisons and assessments of the many possible combinations of weed treatments are very difficult for farmers to undertake. RIM indicates that, at least for one farming system, large reductions in herbicide usage and their replacement by non-chemical control methods is biologically feasible, but economically unattractive to farmers. We expect that this result applies commonly.

The model also provides insights into why farmers behave as they do. The common observation that farmers do not fully embrace IWM until they are forced to by high levels of herbicide resistance is revealed to be a rational and reasonable response. It is probably not intransigent and irresponsible as sometimes portrayed by scientists or agronomists.

I have argued that IWM is not an objective in its own right, but one possible means of achieving the farmer’s objectives, such as profit, risk reduction, sustainability, etc. Furthermore, the decision problem for farmers is not whether to embrace IWM as a whole, but what use, if any, to make of the many possible treatment types that are available. Therefore, extension should focus on raising awareness of the treatment options and helping farmers to evaluate their suitability. A DSS like RIM is clearly a very valuable tool for achieving this.

Acknowledgements

I am grateful to my co-developers of the RIM model: Vanessa Stewart, Anne Bennett, Marta Monjardino, Carmel Schmidt and Steve Powles and to the many others who have contributed. I also acknowledge financial support from the Grains Research and Development Corporation, the University of Western Australia and the III IWSC.

References

Abadi Ghadim, A.K. and D.J. Pannell. 1991. The economic trade-off between pasture production and crop weed control, Agricultural Systems 36:1-15.

Cary, J.W. and R.L. Wilkinson. 1997. Perceived profitability and farmers’ conservation behaviour, Journal of Agricultural Economics 48:13-21.

Gerowitt, B. 2000. The impact of decision models on integrated weed management, Proceedings, III International Weed Science Congress, Foz do Iguacu, Brazil, 6-11 June 2000.

Gorddard, R.J., D.J. Pannell, and G.L. Hertzler. 1995. An optimal control model of integrated weed management under herbicide resistance. Australian Journal of Agricultural Economics 39:71-87.

Gorddard, R.J., D.J. Pannell, and G.L. Hertzler. 1996. Economic evaluation of strategies for management of herbicide resistance. Agricultural Systems 51:281-298.

Lindner, R.K. 1987. Adoption and diffusion of technology: an overview, In: Technological Change in Postharvest Handling and Transportation of Grains in the Humid Tropics, B.R. Champ, E. Highley and J.V. Remenyi (eds.), ACIAR Proceedings No. 19, pp. 144-151.

Pandey, S. and R.W. Medd. 1990. Integration of seed and plant kill tactics for control of wild oats: An economic evaluation, Agricultural Systems 34:65-76.

Pandey, S., R.K. Lindner, and R.W. Medd. 1993. Towards an economic framework for evaluating potential benefits from research into weed control, Journal of Agricultural Economics 44:322-334.

Pannell, D.J. 1988. Weed management: a review of applied economics research in Australia, Review of Marketing and Agricultural Economics 56:255-269.

Pannell, D. J. 1990. An economic response model of herbicide application for weed control, Australian Journal of Agricultural Economics, 34:223-241.

Pannell, D.J. 1994. Economic justifications for government involvement in weed management: a catalogue of market failures, Plant Protection Quarterly 9:131-137.

Pannell, D.J. 1996. Lessons from a decade of whole-farm modelling in Western Australia, Review of Agricultural Economics 18:373-383. http://www.general.uwa.edu.au/u/dpannell/dpap961f.htm (61K)

Pannell, D.J. 1997. Sensitivity analysis of normative economic models: Theoretical framework and practical strategies, Agricultural Economics 16:139-152. http://www.general.uwa.edu.au/u/dpannell/dpap971f.htm (100K)

Pannell, D.J. 1999. Social and economic challenges in the development of complex farming systems, Agroforestry Systems 45:393-409. http://www.general.uwa.edu.au/u/dpannell/spap972f.htm (65K)

Pannell, D.J. and T.G. MacAulay. 1998. Practical and economic issues in multidisciplinary research. Proceedings of the Bioeconomics Workshop, Post-Australian Agricultural and Resource Economics Society, 22 January, Armidale, New South Wales, ABARE, Canberra, pp. 1-12.

Pannell, D.J., V.A.M. Stewart, A.L. Bennett, M. Monjardino, C.P. Schmidt, and S.B. Powles. 1999. RIM 99 User’s Manual, A Decision Tool for Integrated Management of Herbicide-Resistant Annual Ryegrass, University of Western Australia, Nedlands, Australia.

Schmidt, C. and D.J. Pannell. 1996a. Economic issues in management of herbicide resistant weeds, Review of Marketing and Agricultural Economics 64:301-308.

Schmidt, C. and D.J. Pannell. 1996b. The role and value of herbicide-resistant lupins in Western Australian agriculture, Crop Protection 15:539-548.

Sinden, J.A. and D.A. King. 1990. Adoption of soil conservation measures in Manilla Shire, New South Wales, Review of Marketing and Agricultural Economics 58:179-192.

Stewart, V.A.M. 1993. Economics of integrated strategies for managing herbicide resistant ryegrass under continuous cropping, Undergraduate dissertation, Faculty of Agriculture, University of Western Australia.

Citation: Pannell, D.J. (2000). Decision Support for Integrated Weed Management. Proceedings, III International Weed Science Congress, Foz do Iguacu, Brazil, 6-11 June 2000. http://www.general.uwa.edu.au/u/dpannell/dss4iwm.htm.

RIM web page (for further information and order form)

SEA News issue #7

The SEA News index is at http://welcome.to/seanews

Copyright © 2000 David. J. Pannell
Last revised: May 21, 2003.