Pannell and Zilberman (2001) presented a detailed review of past research on the adoption of agricultural innovations and discussed the issues in relation to herbicide resistance and integrated weed management. They identified 11 factors that have been found to be key constibutors to actual adoption decisions.

1. Land quality, soil conditions, soil type (Caswell and Zilberman 1986; Green, 1995; Wu and Babcock, 1998). Adoption of herbicides was complementary to the reduced-tillage practices and thus was particularly favoured in situations where soil erosion was a problem and topsoil was shallow.

2. Prices Increases in the price of fuel and labor tend to increase the adoption of herbicides (Miranowski and Carlson, 1993; Carlson and Wetzstein, 1993). Increases in commodity prices tend to increase the intensification of farming and thus increase adoption and use of herbicides (Helms et al., 1987; Miranowski et al., 1991). Higher product prices also tend to increase the use of land in locations of lower soil quality (de Gorter and Fisher, 1993) and this may further increase use of herbicides. These results highlight the importance of economic incentives as key drivers of adoption.

3. Government policies Since there has been an increased emphasis on soil erosion prevention, US governments have required certain environmental practices as conditions for participation in commodity programs that raise product prices. This has led to increases in the area of land with no or minimum tillage, and thus increased the reliance on herbicides (Uri, 1998; Wu and Babcock, 1998). These programs also include provisions to encourage reductions in the use of some herbicides, such as atrazine. Thus, their overall effect is to increase the use of herbicides but to modify them towards herbicides that are judged to be more environmentally benign.

4. Location For most technologies, the evidence points to higher adoption rates by producers who are closer to the nearest urban center than by producers who are further away (Rogers, 1995). This may be due to a lower transportation cost and/or closer contact with dealers and distributors of the technology or extension agents.

5. Uncertainty and risk Most farmers are risk averse (see Bar-shira et al., 1997) meaning that they are willing to give up some average profit in order to reduce fluctuations in, or uncertainty about, profit. In many cases, adoption of new technologies is gradual so that farmers can experiment with them and have more reliable assessments of their impacts. One of the main activities of marketing specialists and extension agents is to reduce farmers’ uncertainties about new technologies. This is achieved, for example, by the establishment of demonstration plots, demonstration activities, money back guarantees or warranties.

6. Observability and trialability These issues are closely related to the issue of uncertainty. If a technology is highly observable, farmer uptake through imitation of their neighbours is likely to proceed more rapidly. One can well imagine that the high observability of tractors enhanced their rate of spread through farming communities (although it clearly does not over-ride other considerations in a farmer’s final decision to adopt or reject an innovation). Perhaps even more important is the potential to trial the innovation on a small scale, so that information can be obtained, uncertainties reduced and skills developed without the risk of large financial costs if the technology turns out to be uneconomic or fails due to inexperience (Pannell 1999). A new herbicide is highly trialable; its impacts on weeds are readily observable, it can be readily compared to existing herbicides, and it can be used effectively on a very small scale.

7. Adjustment costs One of the key obstacles to the adoption of some new technologies is that can they require investment of time and effort from the farmers. One reason that the adoption of Integrated Pest Management (IPM) has been slow is the relatively high adjustment cost it entails (Wiebers, 1992). Another example is the adoption of tractors, which took about 40 years in the United States to complete (Cochrane, 1979). On the other hand, adoption of herbicides was less problematic, especially for farmers who already sprayed against insects and therefore required little up-front investment in machinery. The rapid and high rate of adoption of genetically modified herbicide-resistant crop varieties in North America and South America can partly be explained by the fact that these are technologies that require very little adjustment cost. The farmer simply replaces one seed variety with another and continues to use herbicides with which they are familiar (glyphosate in most cases).

8. Public acceptance Adoption of new technologies may be lowered if farmers perceive that they may endanger their capacity to sell their products at high prices, or that they may encounter legal problems or protests (e.g. genetically modified crops).

9. Education More educated farmers tend to adopt advanced technology earlier. Education has been shown to be a key factor explaining adoption of various forms of IPM (Wiebers, 1992) and soil testing (Wu and Babcock, 1998). However, farmers with less education may adopt more sophisticated technologies when they rely on private consultants or public extension. Huffman (1974) found that the impact of extension was higher for farmers who were less educated.

10. Extension and consulting advice Empirical evidence that extension and consulting advice can have a measurable impact on farmer adoption of innovative practices is not as common as might be expected, given the large investment in these areas throughout the developed world. Nevertheless, a number of studies have identified a positive relationship between IPM adoption and participation in extension activities (e.g. Napit et al. 1988; Thomas et al. 1990; Harper et al. 1990; McNamara et al. 1991; Fernandez-Cornejo and Kackmeister 1996; Maumbe and Swinton 2000)

11. Marketing A key to the diffusion of a new technology is providing information about its existence and its features. A significant amount of the budget of herbicide manufacturers is dedicated to marketing their products through activities such as advertisements, demonstrations and trade shows, money-back guarantees and warranties, and product stewardship through salespeople (Kotler, 1997).

This by no means exhausts the list of factors that have been found or proposed to influence adoption, but it includes the main types of factors of influence mentioned in the literature (although not necessarily with the same language).

Considering the nature of herbicide resistance and its management, it is possible to make some observations about adoption of resistance prevention strategies based on lessons from the general adoption literature. Herbicide resistance, as a farm management issue is:

• complex;
• potentially expensive or difficult to prevent;
• a source of considerable uncertainty; and
• difficult to observe until herbicide resistance is advanced.

Complexity increases uncertainty, adjustment costs and, possibly, direct costs. Complexity in herbicide resistance management arises from several sources. Firstly, farmers faced with resistance can benefit from understanding the ecological and biochemical theories that explain the occurrence of resistance, but these theories may be difficult to fully understand for non-scientists. Secondly, the weed treatment options that can be used to substitute for herbicides and delay resistance are themselves complex to use relative to herbicides, and normally must be used in combinations in order to achieve adequate weed control. Thirdly, the number of potential combinations of treatments to be considered is vast, especially when the issue is considered as a problem spanning a number of years. In this case, the options include not just changes in weed kill methods but also, potentially, changes in land use; for example, pasture may be included in place of a crop to broaden the available range of weed control options.

The cost-effectiveness of herbicides is revealed by their ubiquitous use through cropping systems of agriculture in the developed world. If a farmer is considering changing to a lower intensity of herbicide use in order to delay the development of herbicide resistance, he or she will be very conscious that in most cases such a move will increase costs. The change will therefore be unattractive unless the farmer is convinced that there are sufficient offsetting benefits. There may actually not be sufficient offsetting benefits in some situations.

Even if there are sufficient offsetting benefits to justify early adoption, it may be very difficult for the farmer to determine this with sufficient certainty in time for preventative strategies to be put into place. For some examples of herbicide resistance development, full resistance is apparent within a small number of years (e.g. five years or less in many recorded cases for annual ryegrass, Lolium rigidum, in Australia). Even if the innovative technologies are trialable, few years of trialling are available before the potential for preventative action has past.

This problem is worsened by the difficulties of trialling some types of treatment recommended to delay resistance. The difficulties include the following:

(a) Where treatments are implemented in combinations (as they generally must be) it can be very difficult to determine from field observations alone the individual impacts of each element of the combination in order to assess their individual worthiness for inclusion in an integrated management system.

(b) Some treatments have impacts that are relatively difficult to observe even if implemented in isolation. For example, increasing the crop seeding rate affects the seed production of both crops and weeds (in opposite directions), and the latter is rather difficult to observe quantitatively in the field without tedious collection and counting of weed seeds in both standard and high seeding rate plots.

(c) The effectiveness of some alternative weed treatments is very sensitive to weather conditions or the quality of implementation, and so trials give highly variable results from time to time. Even if a treatment is beneficial in the long run, it may not appear so in a short-term trial, or it may take a long time before its value can be determined with adequate confidence.

All of these factors would tend to discourage the rapid adoption of integrated weed management systems, involving combinations of unfamiliar, complex, and expensive treatments that are difficult to trial. On the other hand, the nature of herbicide resistance is that, once it has developed, farmers have no choice but to alter their weed management systems. Thus, following the onset of resistance, most of the problems involved in encouraging farmers to change to some alternative system evaporate. The problem for farmers then becomes, which of the many possible alternative systems should best be adopted?

Extent of resistance Given the conclusion at the end of the last section, an important question influencing future efforts to promote IWM is the current extent and severity of herbicide resistance. We know that herbicides have been used intensively by most farmers as the primary means of weed control by most Australian crop producers for many years, and that herbicide resistance has been the obvious and inevitable consequence of that, but until the survey of Llewellyn and Powles (2001) we had no unbiased measure of extent and severity in Western Australia. The survey confirms that herbicide resistance is common and widespread in populations of annual ryegrass, but some of the findings may be considered surprising. Here are some results that surprised me.

The extent to which selective herbicides are still effective against ryegrass in many population is surprisingly high. For example, less than half of the randomly selected ryegrass populations exhibited any resistance to diclofop methyl, and only 23 percent of paddocks had resistance in more than 20 per cent of the population. For chlorsulfuron, 64 percent of populations exhibited any resistance and only 38 per cent had reasonably severe resistance. The herbicide clethodim remains effective as a selective ryegrass control option in most populations across all surveyed areas.

Llewellyn and Powles (2001) concluded that the opportunity to conserve the effectiveness of at least one ryegrass selective herbicide is available to the vast majority of growers across all surveyed areas. From the perspective of adoption, their findings indicate that even in regions of high cropping intensity, relatively few farmers face a herbicide resistance problem that is so severe as to absolutely require adoption of additional non-herbicide control methods. Furthermore, there are still some areas where the most resistance-prone herbicides are still effective across most of the cropping land.

Another result that would surprise some is that farmers who have weed populations with the common forms of herbicide resistance maintain ryegrass densities that are no higher than in situations where herbicides are still fully effective. This implies that the main cost of herbicide resistance is due to higher control costs rather than higher weed numbers. Interestingly, this is consistent with results from economic (Pannell 2001).

Farmers perceptions and intentions Llewellyn et al. (2002) surveyed farmers in Western Australia to ascertain their knowledge and perceptions of herbicide resistance. They found that the majority of growers involved in the study have some experience with resistance development on their property and that most farmers have a reasonably good understanding of resistance and its management. Furthermore, most of their perceptions about quantitative aspects of resistance are reasonably accurate. This was true for issues including:

• the cause of resistance
• the current extent and severity of resistance,
• the number of applications of selective herbicides that is possible before resistance is likely to develop.

Scientists who wish to promote the preemptive adoption of IWM in order to delay onset of resistance may be discouraged to learn that farmers with such good knowledge of resistance and personal experience with it, generally choose to maintain their usage of herbicides as their primary weed control method for as long as this is possible. It is particularly interesting that they expect it to be possible for the foreseeable future. Indeed, only 13 per cent of these farmers predict that they will be less reliant on herbicides in ten years time than they are currently.

It is possible that this is partly explained by apparent misperceptions about two issues that Llewellyn et al. (2002) identified:

• The likelihood of new herbicides with novel modes of action becoming available for resistant weeds. The vast majority of growers (86 per cent) expected a new ryegrass-effective, wheat-selective herbicide to be available within 10 years, with 52 per cent of growers expecting such a product to be available in less than six years.
• The likelihood of resistant populations of weeds reverting to a state of herbicide susceptibility if use of herbicide is temporarily suspended. Surprisingly, 46 per cent of growers in one region perceive reversion to be likely, and almost 14 per cent of all surveyed growers perceive reversion to be very likely.

Both of these views would tend to discourage early adoption of costly preventative strategies (i.e. IWM). Perhaps extension activities to counter these apparent misperceptions could result in more early adoption of IWM. On the other hand, recent economic modelling of the issue, reported in the next section, suggests that changing these perceptions may not be sufficient.

RIM is a multi-period simulation model representing the biology, technology and farm-level financial aspects of integrated weed management in dryland farming systems of southern Australia. Its general features include:

• a 20 year time frame;
• representation of up to seven different crop or pasture based enterprises in a user-specified sequence;
• 35 different weed control options that can be selected in any technically feasible combination and in any technically feasible sequence;
• detailed representation of relevant biological relationships;
• representation of relevant financial details, including input costs, output prices, machinery costs, interest rates, yields, and environmental costs for certain treatments;
• ability to specify the herbicide resistance status of the weeds with respect to each herbicide mode-of-action group.

The biological processes, variables and relationships represented include:

• the density of living, ungerminated seeds in or on the soil through each year;
• a germination pattern of weed seeds over time within each year;
• mortality or removal of seeds, seedlings, or adult plants or prevention of seed production through various processes, including both natural and human-applied mechanisms;
• inter and intra-species competition impacts on crop yields, and weed seed production;
• damage to crop production from weed control treatments, including phytotoxic damage from herbicides, and mechanical losses due to treatments such as green manuring;

The original version of RIM (Pannell et al. 1999) included ryegrass as the only weed, but more recently, Monjardino et al. (2002) have added wild radish. Here I will present a selection of results from RIM that relate to the adoption of IWM by farmers.

Table 2. RIM results with and without early adoption of Integrated Weed Management (wheat:wheat:lupin rotation; 20 year time frame; herbicide applications available: 4 "fops" and "dims", 2 sulfonyl ureas, 4 triazines).