CSIRO PUBLISHING
www.publish.csiro.au/journals/app
Australasian Plant Pathology, 2003, 32, 167–180
Integrated approaches to understanding and control of diseases and pests in field crops N. McRobertsA,D, G. HughesB and S. SavaryC A
Crop and Soil Management Research Group, SAC, Auchincruive, Ayr KA6 5HW, United Kingdom. B School of Biological Sciences, University of Edinburgh, Edinburgh EH9 3JG, United Kingdom. C Ecole Nationale Supérieure Agronomique de Rennes, CS 84215, 35042 Rennes cedex, France. D Corresponding author; email:
[email protected]
Presented as a Keynote Address at the 8th International Congress of Plant Pathology, Christchurch, New Zealand, 2–7 February 2003
Abstract. The idea that there is no such thing as an empty niche became well established in the ecological thinking in the second half of the 20th century. The implications of this ecological concept have been slow to be recognised by plant pathologists and other crop protection scientists despite the fact that they regularly attempt large-scale population management and the prevention of niche exploitation. Two questions that follow from taking an ecological view of crop disease management are what constitutes a niche, and to what extent can decision-makers choose the manner and extent of exploitation of the niche that they wish to protect? It is suggested that in developing IPM strategies, it is important to consider farmers and their wider socio-economic circumstances as part of the niche that is exploited by pests and diseases. This view arises from large-scale studies, in both temperate and tropical crop production systems, of concurrent epidemics of multiple pests and diseases, in which variation in farmers’ activities is as obvious as variation in the physical and biological environment. Further incentive for adopting this view comes from the fact that IPM strategies are implemented (or not) through farmers’ decisions and actions and so such strategies must be constructed with this filtering process in mind. Methodologies for developing robust IPM strategies are discussed and areas are noted in which further methodological development is needed, including modelling of competition among niche exploiters, formal analysis of adoption of IPM methodology by farmers, and use of information by decision-makers in crop protection. AP03206 eNIn.taeMglcr. tRaeodbraetsprocahesotdisease andpestcontrlo
Introduction Crop protection scientists are responsible for the development of methods that, when put into practice in agriculture, attempt population management over huge geographic areas. Frequently these attempts are successful, but the continuing loss of food and commodity crops to pests, weeds and diseases world-wide highlights that only partial success is achieved in many cases. From an ecological perspective, we should not find the failure of crop protection to deliver pest free crops surprising. Empty niches (if they exist at all) are the exception rather than the rule in nature and attempts at niche clearing are unlikely to be successful in the long-term. The aim of the paper is to re-invigorate discussion of the potential for integrated control, (particularly among plant pathologists who lag behind entomologists and weed ecologists in adoption of integrated thinking) to address the © Australasian Plant Pathology Society 2003
complex challenges facing global crop protection in the 21st century. We discuss disease and pest control as an ecological problem and attempt to link large-scale empirical studies of production constraints (McRoberts et al. 1995, 2000b; Savary et al. 2000a, 2000b) with ecological theory concerned with niche occupancy and resource use. Indeed, our discussion touches on a wide range of subjects from ecological theory on competition, through plant physiology, to socio-economics and decision theory. Comprehensive discussion of integrated control demands that such a range of subjects be included, and it may be that the breadth of topics encompassed by IPM is one of the factors which has led to recent reviews generally having a pessimistic tone (Jeger 2000; Way and van Emden 2000) regarding its future development. In discussing integrated control in its broadest sense, we give attention to each of the four corners of the disease 10.1071/AP03026
0815-3191/03/020167
168
Australasian Plant Pathology
N. McRoberts et al.
Table 1.
Alternative views of integrated control
‘Hey farmer farmer. Put away that DDT now. Give me spots on my apples. But leave me the birds and the bees. Please’ (Mitchell 1970) ‘some optimum combination and use of all known management strategies whereby farmers are offered a choice of chemical, biological and physical controls that may reinforce each other or work together to manage a particular pest.’ (Orr et al. 2001) ‘IPM is...based on environmental monitoring, biological monitoring, pest modeling, and a communication and delivery system [and] provides a continuous flow of weather and pest information, control tactics recommendations and evaluations, etc, from the field to IPM specialists and back to decision makers.’ (Whalon and Croft 1984)
tetrahedron suggested by Zadoks and Schein (1979) and stress that future development of integrated control must continue to include developments in areas of research related to all four corners and their interactions. In setting out this view of integrated control and how it can be better understood, the paper poses and discusses four questions: (1) what is integrated control? (2) why should anyone interested in it? (3) what is the ecological basis of integrated control and what can the discipline of epidemiology contribute to the subject? and (4) why is it sometimes difficult to enact? What is integrated control? Table 1 presents three quotations that illustrate concepts of integrated control, one drawn from popular culture and the other two from the scientific literature. As Jeger (2000) recently noted, many definitions of integrated control are in use, though most contain the basic principles first advocated by Stern et al. (1959). One of the central concepts in such descriptions of IPM is raised in the quotation from Joni Mitchell’s song ‘Big Yellow Taxi’ which highlights public concern with the potential negative effects of pesticide use on the wider environment. It makes it clear that utilities may differ significantly among different groups for different aspects of food production; Mitchell has a clearly expressed, though not formally stated, higher utility for biodiversity than for disease-free, unblemished apples. The subject of utilities will be raised again in examining growers’ responses to the provision of information on thresholds and disease management. A key point to be taken from Mitchell’s informal assessment of disease and pest control is that there is a balance to be struck between pest population regulation and other objectives. Whalon and Croft (1984) provide a concise account of the development of IPM in the North American apple industry that was, in part, prompted by the introduction and subsequent failure of DDT and other pesticides. For the past half century, crop protection methodology in industrialised countries, and to an increasing extent in developing countries, has been dominated by the use of pesticides; these are often deployed with the aim of emptying the niches which crops present to pests. Murdoch (1975) commented on the underlying strategy generally adopted in chemical pest control and noted that it had at least three characteristics of interest. First, it depends for success on exploiting the inability of a pest population to withstand the
instability caused by large-scale reductions in numbers of reproducing individuals. Secondly, it acts directly to inhibit population growth processes. Thirdly, the combined effect of the first and second points is that chemical control generally eschews mechanisms which regulate populations through density-dependent processes in favour of population control by induced instability. It would, of course, be a mistake to suggest that chemical control methods are the only type in which instability, rather than density-dependent regulation, is used as a means of control. For example, recent attempts to control citrus canker in South and North America (Gottwald et al. 2001; Gottwald 2003) provide ample evidence that physical control measures (in this case sanitation roguing) can be used to achieve disease eradication. The issue of balance is raised more formally in the quotation from recent work on the development of IPM programmes in the Blantyre Shire Highlands of Malawi by Orr et al. (2001). The authors place emphasis on the balance to be struck in developing IPM programmes in making all potential control methods for a pest available to growers. This definition of IPM might be called a traditional definition, and would likely be recognised by the majority of crop protection scientists as capturing the important concepts of IPM. Orr et al. (2001) discuss IPM strategies in a number of different ways and highlight two important issues. First, integrated control methods for an individual pest will nearly always have to take into account the fact that crops are subject to attack from pest complexes. Secondly, IPM is driven by knowledge. This second point links directly with the third definition presented in Table 1 that is taken from a discussion of the history of IPM in the North American apple industry by Whalon and Croft (1984). One view of IPM, then, is that it is a process by which unreflective control methods, which aim to control pest populations by induced instability, are replaced by knowledge- (information) based approaches that seek to exploit naturally occurring processes that regulate pest populations. Why should anyone be interested in integrated control? Pesticides and sustainability The motives for the development of IPM programmes have varied historically across continents and with crops (Teng and Savary 1992; Way and van Emden 2000). If we define
Integrated approaches to diseases and pest control
sustainability simply as ‘the ability to persist over time’, the diverse motives noted by Way and van Emden (2000) can all be classed as attempts to improve the sustainability of agricultural production. The specific motives for the development of integrated control have focused on reducing the level of pesticide use in crop production (Whalon and Croft 1984; Webster 1985). Ultimately, the strength of the motivation provided by reduction in pesticide use rests on the dependence of pesticides on non-renewable fossil oil reserves. There is a fairly strong consensus that world-wide oil production is approaching the point where production will decline. The peak year for production in the USA occurred nearly 25 years ago (http://www.technocracy.org). Arguments about the use of integrated control to reduce pesticide use are, therefore, in a very direct way, arguments about sustainability. In response to the need to reduce the reliance of agriculture on fossil fuel and other non-renewable resources, Shiyomi (2001) recently called for the development of systems of production which replace such dependence with a reliance on ‘complex biological interactions’. In making this plea, Shiyomi (2001) pointed out that replacing fossil fuel use with biological interactions will be difficult because it requires a much greater understanding of how interactions within a system result in its observed behaviour. Shiyomi (2001) used an example (originally published by Levins and Vandermeer 1990) based on biological control of two grasshopper species by a partially cannibalistic mantis, to illustrate how even simple systems (in this case having just three components) can give rise to unforeseen behaviour when interactions change by relatively small amounts. In response to this feature of biological systems, Shiyomi (2001) suggested a prominent role for modelling in the development of sustainable methods of production. Other analysts of the development of integrated control are not so supportive of the strategic use of models (Way and van Emden 2000). We will return to the topics of reliance on interactions within systems and the role of models later, but Shiyomi’s (2001) point about increased understanding deserves further discussion here. Education and sustainability It has been noted previously (Jeger 2000) that thinking of integrated control only as a means of reducing pesticide use risks neglecting the many other benefits of adopting an holistic approach to pest management. One of the most important of these is the increase in level of understanding of production systems which using IPM fosters among growers. As already noted, Shiyomi (2001) expressed this increased understanding as a potential constraint, and the widespread failure of many IPM packages to reach high levels of adoption (Teng and Savary 1992) may well indicate that there is some basis for this concern. However, there are well-documented examples of cases where significant
Australasian Plant Pathology
169
increases in grower awareness of the complexity of the system have been achieved. For example, farmers who used EPIPRE system for a few years, when asked why they had given up using it, frequently mentioned that they had learned enough to make their own decisions (Zadoks 1985). The demonstrated success of farmer field schools in raising awareness of the activities of beneficial organisms in tropical rice production is similarly well documented (Kenmore 1996). There appears, then, to be a trade-off between the educational benefits which IPM brings with it, and the adverse effects that the requirement for learning has on the probability of IPM being adopted. Economics and sustainability Finally, an important justification for adopting IPM is reduction in production costs. Long-term systems experiments in the UK and elsewhere have demonstrated that integrated production systems for field crops tend to perform better financially than high-input systems when commodity prices are low (Jordan et al. 1985; Webster et al. 1999; McRoberts et al. 2000a). Within the European Union, integrated control methods (if they offer significant cost savings) may have an impact on the social sustainability of arable farming. A recent analysis on the viability of UK wheat production under plans proposed under the mid-term review of the EU Common Agricultural Policy (CAP) has indicated that, at a grain price of £60/t and the current cost of production of £71/t, 56% of wheat production will be unviable. With cost savings of 30%, the predicted percentage of viable wheat production rises to 90% (D. Oglethorpe, personal communication). Of course, financial motivations for adopting IPM are common to agriculture in developing and developed countries (Matteson et al. 1984; Heong and Escalada 1999; Orr et al. 2001). Discussion of some recent developments in decision-making is taken up in the final section of this paper, but we note here that difficulties in establishing whether particular courses of action are justified economically have been a major concern in the development of IPM programmes. However, failure to adopt the most appropriate (and this may not always be the cheapest) method may happen because of lack of familiarity with the biology of the pest (Orr et al. 2001), lack of access to capital to purchase pesticides or lack of familiarity with economic calculations (Franke et al. 2003), lack of awareness of the range of alternatives (Jeger 2000) or a failure to use available information or to link reduction in pesticide use with increased economic performance (McRoberts et al. 2000b). What is the ecological basis of integrated control? Components of the system Way and van Emden (2000) recently presented a pessimistic view of the role of strategic models in the
170
Australasian Plant Pathology
N. McRoberts et al.
development of IPM. These authors put forward a strong case for the superiority of empirical research in bringing real advances to the adoption of IPM. Here, we hope to show that a strategic view of integrated control has useful insights to offer and may serve to refresh interest in the subject of IPM among plant pathologists. The following discussion is focused on the exploitation of leaf area by airborne pests and diseases of field crops. Integrated control must take account of all four components of the disease tetrahedron: pathogen, host, environment and people. We will build up an overall picture of the typical cropping system by adding components in turn. Taking an ecological viewpoint, the host represents one or more resources, which will be exploited by various pests, depending on environmental conditions and the intervention of people. Resource utilisation by pests It has been pointed out many times (by Teng and Savary 1992; McRoberts et al. 1995; Savary et al. 1997, 2000a, 2000b; Jeger 2000 and by numerous other authors) that crops are rarely, if ever, attacked by a single pest or disease in isolation. Rather, they are subject to exploitation simultaneously by a varying number of different pests and diseases which, together with weeds, form pest complexes. Table 2 gives three examples of pest complexes drawn from temperate and tropical crops. The host and pests are intimately linked through their trophic relationships. Since van der Plank’s seminal epidemiological analysis (van der Plank 1963) plant pathologists have become used to viewing the temporal progress of resource exploitation by plant pathogens by plotting disease progress curves (DPC). Methods for comparing DPCs have shown steady elaboration over the 40 years since van der Plank’s book appeared (Campbell and Madden 1990; Hughes et al. 1997). However, since yield is derived predominantly from the healthy rather than the diseased portions of a crop, there is considerable justification for turning DPCs upside-down and focussing Table 2.
attention on the change in the healthy proportion of the crop remaining over time. This view of the pests–crop interaction also provides a more explicit view of the utilisation of the host resource by the pest than does the traditional DPC. As an example, Fig. 1 shows results from surveys of pests present in commercial Scottish winter wheat crops during the mid-1990s. Visual assessments of disease severities in the canopy were carried out at regular intervals through the growing season. The data presented cover the period from (approximately) May to July in each of two growing seasons. Disease severities for the major foliar diseases were recorded using standard 0–9 scales. Healthy leaf area (HLA) was calculated as (9–Σi(xi)) where x is the severity score for each of the i diseases present and is plotted against the estimated crop growth stage (recorded on a standard decimal scale). The main foliar diseases present in the crops were powdery mildew and those caused by Septoria spp. There is a general tendency for HLA score to decline with time (from a mean of 7.6 to 5.2 in 1994/5 and from 8.1 to 5.8 in 1995/6 between the first and third samples shown) indicating increasing exploitation of the host by the pathogens present. However, note that in both seasons the variance in the HLA values increased with time (the vertical dispersion of the points increases) by a factor of 10 in 1994/5 and by a factor of 3 in 1995/6. Human interaction with the host–pest system The variation in HLA among crops (as depicted in Fig. 1) arises from the specific pest-host-environment interactions for each crop and is further affected by husbandry practices imposed by the grower. All of the crops used to generate Fig. 1 were treated with fungicides, with considerable variation in the choices of active ingredient, rate and timing of application. Such variation is to be expected in any set of real fields under management in a relatively intensive cropping system, but it begs the question of the extent to which the grower is responsible for altering the availability of resources to different species within the pest complex and
Examples of multiple pest complexes
Wheat leaf diseases: Western Australia
Winter oilseed rape pests and diseases: Scotland
Rice pest complex: South-east Asia
Powdery mildew Stem (black) rust Leaf (brown rust) Stripe (yellow) rust Yellow (tan) spot Septoria nodorum Septoria tritici
Light leaf spot Phoma leaf spot Phoma stem cancker Downy mildew Dark leaf spot Botrytis leaf spot Sclerotina stem rot Flea beetle Stem weevil Seed weevil Pollen beetle
Bacterial leaf blight Sheath rot Sheath blight Brown rot Leaf blast Neck blast Rice tungro Plant Hopper Whorl maggot Leaf folder
http://www.agric. wa.gov.au
McRoberts et al. (1995)
Savary et al. (2000a)
Integrated approaches to diseases and pest control
Australasian Plant Pathology
9
Healthy leaf area score
8 7 6 5 4 3 2 1
1994/5
0 30
40
50
60
70
80
90
Crop growth stage 9 Healthy leaf area score
8 7 6 5 4 3 2 1
1995/6
0 30
40
50
60
70
80
90
Crop growth stage Fig. 1. Healthy leaf area (HLA) recorded in surveys of winter wheat crops in north-east Scotland in two consecutive growing seasons, plotted against crop growth stage. HLA was calculated as the remainder from summed disease severity scores for powdery mildew, Septoria spp., yellow rust and brown rust. Severities for each disease were recorded on a 0–9 scale. The graphs show data from three sampling bouts in each season. Each point is the mean of ten separate samples per crop. The trajectories for HLA in specific crops are shown in each year, selected to show the crops with the highest and lowest HLA at the first and last samples and are indicated by the larger, darker symbols. Different sampling bouts are indicated by different symbol shapes.
171
thus, ultimately for selecting the type of problems which dominate in their crops. The close association between different cropping practices and sub-sets of the pest complex (injury profiles) has been clearly demonstrated in the tropical rice systems of Asia by Savary et al. (2000a, 2000b) who identified different classes of cropping practice and injury profiles from an extensive study of rice-cropping systems across south-east Asia. In all, six different classes of cropping practice and five different injury profiles were identified. The correlations between particular cropping practices, injury profiles and observed yields were very strong in certain cases. Injury profile 1 (dominated by stem rot and sheath blight) was found to be most closely associated with cropping practices 1, 2 and 4 that were all transplanted rice systems with good to moderate control of water supply. Injury profiles 3 (dominated by sheath rot, brown spot, leaf blast and neck blast) and 5 (overall, low damage) were closely associated with cropping practice 3 (rice–wheat systems with diverse crops preceding rice, generally low inputs and poor control of water supply). An analysis of the association between these injury profiles and yields suggested that cropping practices 1 and 4 are associated with high yields, while practice 3 is associated with low yields (Savary et al. 2000a, 2000b). To some extent, of course, both the injury profiles and the cropping practices used by growers are dependent on the climate and wider features of the cropping system and resource base. However, allowing for the limitations that climate and physical resources place on choice of cropping practice, and on the type of pests which will develop in a particular area, results such as those obtained by Savary et al. (2000a, 2000b) indicate the extent to which people are responsible for determining which pests gain access to the resources available to support their growth. One of the obvious ways in which the grower may influence the development of the pest complex is through the initial selection of the cultivar of crop that will be grown. In an ideal situation, cultivars would offer good resistance to all of the diseases that might be important. Use of resistant cultivars is almost universally proffered as the first step in implementing an integrated disease control programme. However, breeding for multiple disease resistance is a difficult process, particularly when it must be balanced against competing priorities in the breeding programme. The apparent difficulty in achieving this goal is illustrated in Fig. 2 which uses data from the UK Recommended List of winter wheat cultivars (http://www.hgca.org.uk) and the government of Western Australia web site on integrated management of wheat leaf diseases (http://www.agric. wa.gov.au). Data from both sources for cultivar resistance scores to several diseases was subject to principal components analysis. In both cases, the first axis contrasts cultivars with
172
Australasian Plant Pathology
N. McRoberts et al.
3 (a) PS
Second principal component (24% variance)
2 1 0
TY Fsp
BG -1
ST SN
-2 PR
-3
Competition among pests
-3 -2 -1 0 1 2 3 First principal component (33% variance) 3 (b)
Second principal component (25% variance)
2 1 0
BG ST
-1
PTr SN
PS -2 PR
a cultivar with a combined high level of resistance to both diseases from either set. An additional point of interest to note, however, is that within the Australian cultivars, there appears to be a close correlation of resistance scores for all three major rust diseases; this is indicated by the fact that the vectors for these diseases point in the same direction. In contrast, with the UK cultivars resistance to brown rust and to yellow rust appear to be uncorrelated, with the vectors for these two diseases pointing in opposite directions. The occurrence of pest complexes that simultaneously exploit a crop leads to the question of whether the individual members of the complex are in competition for resources and thus, to some extent, control each other? A second question which arises from the first is whether such competition can be exploited in integrated control?
PG
-3 -3 -2 -1 0 1 2 3 4 First principal component (28% variance) Fig. 2. Biplots from principal components analyses of wheat cultivar resistance ratings to groups of fungal diseases. (a) Data from the 2003 UK recommended cultivar list (HGCA, 2003). (b) Data from the Government of Western Australia web site on integrated disease control (http://www.agric.wa.gov.au/). In both cases the first principal component separates cultivars with high resistance ratings for powdery mildew (caused by Blumeria graminis, BG) from those with high resistance ratings for Septoria nodorum (SN). Other diseases and pathogens included are indicated by the following abbreviations: Fsp, Fusarium species (head blight); PTr, Pyrenophora tritici-repentis (yellow (tan) spot); PG, Puccinia graminis (stem rust); PR, Puccinia recondita (brown rust); PS, Puccinia striiformis (yellow rust); ST, Septoria tritici; TY, Tapesia yallundae (eyespot).
high ratings for resistance to powdery mildew (caused by Blumeria graminis) with those with high resistance to Septoria nodorum. It does not appear to be possible to select
Evidence of control of pest populations by natural predators and enemies is abundant in the literature, and forms the basis of successful IPM programmes for many invertebrate pests (Whalon and Croft 1984; Jeger 2000; Way and van Emden 2000). Evidence that different pests (and particularly in the context of this paper, airborne diseases) might regulate each other’s populations through competition is less common. However, in attempting to draw together the evidence that such effects might be real and significant, and in trying to provide a strategic viewpoint from which to consider them, we might stimulate a renewed interest in integrated control that is badly needed, particularly in relation to foliar diseases of field crops. One type of evidence of competition among pests for which there is well-documented data is population increase after release from the competitive effects of a second population. One of the most well known examples of this phenomenon is the failure of the Chinese plan to remove the four pests during the Mao era in the 1950s. In this case, attempts to kill small birds (which were viewed as pests of crops) led to a reduction in crop yields when insect populations which the birds had also fed on were released from predation and caused greater damage to the crops than the combined effects of the birds and insects together. This example recalls Shyomi’s (2001) comments on the need for modelling to help in the understanding of even relatively simple biological systems. A more recent example of population increase after release from competition is provided by the interactions among pathogens giving rise to the stem base disease complex in temperate cereals. In this case it is known that control of the eyespot pathogen (Tapesia yallundae) early in the growing season can lead to subsequent colonisation by the sharp eyespot pathogen (Rhizoctonia cerealis) (Burnett 1999). Observations of population increase following release from competition are indirect evidence that competition between diseases exists. More direct attempts can be made to
Integrated approaches to diseases and pest control
Australasian Plant Pathology
measure competitive effects, and discussion of such attempts leads to an explicit statement of a strategic, ecological view of integrated control. Various authors have attempted to make direct assessments of competition between diseases, often describing the interactions through the use of Lotka-Volterra (L–V) equations. A generalised form of these equations (in terms of rates of population change) as presented by May (1973) is given here as Equation 1. m § · d N i = N i . ¨ k i .¦ α i, j .N i ¸ dt © j=1 ¹
(1)
Equation 1 states that the rate of change in an individual population Ni in a community of i such populations competing for a one-dimensional resource is proportional to the current population size, the population’s unique resource utilisation function (ki) and the sum of the competition effects of the other j species in the community, expressed through the competition coefficients αij. As Way and van Emden (2000) have pointed out, May (1973) described such models as ‘caricatures’ of reality. However, whether being a caricature is a bad thing depends on the purpose for which it is intended. As May (1973) also noted, models such as Equation 1 are useful as strategic tools ‘which sacrifice(s) precision in an effort to grasp at general principles... [and]... even though they do not correspond in detail to any single real community, aim to provide a conceptual framework for the discussion of broad classes of phenomena. Such [a] framework can serve a useful purpose in indicating key areas or relevant questions for the field and laboratory ecologist, or simply in sharpening discussion of contentious issues.’ Toward such strategic aims, analysis of the general properties of L–V equations has identified that, if they provide a useful description of real populations, then a stable environment permits the stable coexistence of many species. A corollary to this is that a randomly fluctuating environment (such as one in which control measures are used that disrupt intrinsic processes that regulate population size, as discussed above) limits the number of species which may coexist. Species numbers in a community are also determined by the sizes and shapes of the resource utilisation functions, ki of the available resource, K. Finally, for stable coexistence of many species, the competition coefficients, αij must be relatively small (May 1973). There have been few cases where L–V equations have been fitted to disease progress data. The earliest reported example was by Madden et al. (1987) who used the method to estimate parameters of disease progress curves for the incidence of tobacco vein mottling virus and tobacco etch virus over three seasons in a number of fields in Kentucky. Madden et al. (1987) concluded that competition effects
173
between the viruses were small, but found quite marked differences in the estimates of the upper limit of disease incidence for the two viruses. More recently, Weber (1996) used an L–V model to examine interactions between Blumeria graminis and Septoria nodorum on wheat leaves, while Ngugi et al. (2001) applied the method to the analysis of leaf blight and anthracnose of beans. Summarising the results, these analyses suggest that, generally, competition coefficients have been found to be small and somewhat variable within replicated and repeated experiments. These findings are in keeping with the idea of coexistence of multiple diseases on a single host. Towards a strategic view of integrated control From the above we have some tentative evidence from experimental studies that multiple disease complexes might be a specific example of a type of ecological community which can be described in general terms by Equation 1. Further justification for considering it in more detail must come from its value in raising testable hypotheses. In passing we note that the fraught issue of validation, in the strict quantitative sense in which it is often discussed, does not apply to the strategic use of models such as Equation 1. Following on from May’s (1973) comments on sacrificing precision for descriptive power, validation of models of this type comes in two forms. First, the value of such models comes from the extent to which, as formal systems, they are isomorphic (Hofstadter 1999) with the parts of the real world they represent. In a general discussion of the use of models in biology, Doucet and Sloep (1992) refer to this issue by raising the important point that the modeller should make a formal statement that the model represents reality. Without this statement, model and reality remain separate entities. Secondly, in common with tactical models, they are validated over time by common consent of their fitness for purpose. Again, in the case of strategic models this will be because they describe general patterns of behaviour in real systems to which they are isomorphic, and because they provide useful insights into the mechanisms that might be responsible for that behaviour. Having set out the conditions by which Equation 1 might be judged useful, an example of the sort of behaviour which this type of model might produce is shown in Fig. 3. In fact, the results in Fig. 3 were generated from a set of equations of the form shown in Equation 2, which is a simple extension of Equation 1 with a non-density-dependent control term, βi for each of the Ni. m § · d N i = N i . ¨ k i .¦ α i, j .N i ¸ − βi .N i dt © j=1 ¹
(2)
The details of the model output in Fig. 3 are not important in the current context. The particular formulation of the
174
Australasian Plant Pathology
N. McRoberts et al.
Proportion leaf area occupied
0.5
(a)
Biotroph 1 Biotroph 2 Necrotroph Saprophyte
0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 0
10
20
30
40
50
60
Time (arbitrary units) 1 (b)
0.9 Proportion of leaf area
0.8 0.7 0.6 0.5 HLA Total diseased area
0.4 0.3 0.2 0.1 0 0
10
20
30
40
50
60
Time (arbitrary units) Fig. 3. Disease progress curves for individual pathogens (a) and summed green leaf area progress and disease curves (b) for a hypothetical Lotka-Volterra competition model for four fungal species competing for host plant resources.
model was prompted by current research in Scotland to understand the emergence of necrotic spotting in barley as a major new problem (see http://www.hgca.org.uk). Note, however, that the model allows for the inclusion of both competition between pathogens that reduce green leaf area and competitive members of the leaf microflora which are asymptomatic colonisers of available infection courts. Considering Equations 1 and 2 in more detail, we may formulate integrated control as an optimisation problem. Each of the Ni pathogens will have an associated damage
function that relates pathogen population size to yield loss (Hughes 1996 and see below). We can write these functions generally as Li = fi(Ni) where Li is yield loss resulting from occurrence of pathogen i and the fi are specific functions for each i. Additionally, each pathogen will have a cost of control Ci associated with it. A specific target for integrated control might then be to attempt to suppress the development of the pathogen for which dLi/dNi is greatest, subject to minimising ΣiCi . Alternatively, the approach might be to minimise ΣiLi subject to ΣiCi
Integrated approaches to diseases and pest control
selection of cultivars might affect disease development both directly, by resource limitation, and indirectly, via effects on the competitive interaction between diseases. Plant pathologists have not generally been used to considering either interactions among diseases, or the interactions of pathogens with their hosts in quite this way, but there appear to be a number of issues which might benefit from such an approach. As an added incentive to stimulate such research, it can be pointed out that in researching this issue with a practical imperative, plant pathologists have the opportunity also to contribute to an area of theoretical ecology (i.e. community dynamics) where lively competition between rival theories still exists. Manipulation of the pest complex through alteration of cropping practices to exploit competition among pests would depend for its success on the possibility that competitive exclusion would operate to a sufficient extent to allow control of more damaging pests through encouragement of less damaging ones. Such a view of disease control may be alien to many plant pathologists, but considering it as a possibility makes one important concept in integrated control fully explicit: i.e. there is no such thing as an empty niche, and control efforts aimed at niche clearing are not likely to be sustainable in the long term. Practitioners of biological control of invertebrate pests are well aware of the need to leave reservoir populations of prey species in order to prevent extinction of predators (Jeger 2000). While recognising their shortcomings in certain respects (see May 1973), L–V models provide a strategic explanation for this observation through the coupling of predator and prey population sizes. Placing the task of integrated control of diseases within the same framework, albeit through the use of competition rather than predator-prey equations, provides a commonality of approach which might help to increase exchange of concepts between disciplines. Ultimately, the success of any strategic or tactical effort to improve integrated control practices depends on the practical results of the improvements being adopted by growers so that, in terms of the analysis presented by Savary et al. (2000a, 2000b), cropping practices are altered. This issue introduces our discussion to the interface between research and extension. It is a key area in understanding integrated control in practice, and leads to the final question raised at the start of this paper. Why is it sometimes difficult to enact? According to Rölling (1988), the two questions most frequently asked by extension workers about growers are (1) How can I get them where I want them?, followed by (2) Why don't they do what I want? In the context of this paper, the first question is analogous to how can I get them to adopt integrated control?, and the second is analogous to why won't they adopt integrated control when I suggest it?
Australasian Plant Pathology
175
Here we consider the situation in which integrated control might be introduced as an innovation to replace, for example, prophylactic use of pesticides. Reasons why farmers fail to adopt, or to retain, the use of innovations have been studied extensively (Rölling 1988; Fujisaka 1994; Rogers 1995) and are often discussed in relation to integrated control (Zadoks 1985; Jeger 2000; McRoberts and Hughes 2001; McRoberts et al. 2000b). The FAO suggests, as a rule of thumb, that adoption of an innovation is unlikely unless it offers at least a 2 : 1 financial advantage over existing practices (http://www.fao.org). This position may be a little pessimistic, but it certainly suggests that much of the effort exerted to demonstrate financial parity between integrated production systems and intensive cropping in Europe has been aiming too low if financial incentives are to be a primary factor in stimulating adoption. There is, however, a specific issue relating to the nature of integrated control, or at least the way in which it has often been presented, which is directly inhibitory to its adoption. This relates to the central role that threshold-based decision-making often takes in integrated control programmes. Recalling the quotation from Whalon and Croft (1984) given in Table 1 highlights the importance of information to successful integrated control. Much of the information which has been demanded by IPM programmes over the years has concerned estimates of pest population densities from samples, or weather data as inputs to disease prediction tools so that these can be linked to an action threshold for pest control. The importance of the development of economic threshold theory to crop protection should not be underestimated. However, as a general tactical means to determine action, it is not without problems. From a strong initial conceptual basis (Stern et al. 1959), threshold theory has developed, particularly in applied entomology, into a sophisticated branch of crop protection. Recent developments of the theory have addressed issues such as: spatial heterogeneity in pest populations (Hughes 1996), methods for testing decision tools (Twengstrom et al. 1998; Hughes et al. 1999) and the incorporation of prior information and subjective assessments of risk into decision-making (McRoberts and Hughes 2001; Yuen and Hughes 2002). It is the last of these points, and its interaction with the nature of the information which is used in decision-making in integrated control, which we wish to discuss here. Although information is undoubtedly important to implementation of successful integrated control, growers’ attitudes to information and use of it are by no means straightforward. Fig. 4 presents a summary of results from a survey of farmers’ attitudes and behaviour carried out in Scotland during the mid 1990s (McRoberts et al. 2000b). The correlations among the group of four attitudinal variables (the rectangular boxes) identified in Fig. 4 suggest that farmers who had an external locus of control (i.e. a belief
176
Australasian Plant Pathology
N. McRoberts et al.
Uses prophylactic sprays 0.30
0.40
Maximises yield 0.44
Recent increase in debt
0.36 0.37 External locus of control
0.31
Information seeking 0.45
0.43
Traditional attitude
0.40
0.40
0.34
Poor coping with policy change
Fig. 4. Correlations between attitude domains (in rectangles) and self-reported behaviour (elipses) for a sample of Scottish arable farmers. The values on the arrows are standard correlation coefficients. All values are significant at P ≤ 0.05.
that their fate was not in their own hands) also tended to have traditional attitudes (e.g. a belief that passing the farm on to a family member was an important goal), to cope poorly with changes in policy related to farming, but they were likely to actively seek information. These attitudes were linked to a set of self-reported behaviours which indicated that such farmers were likely to have experienced an increase in debt during the 5 year period prior to the survey, they were likely to attempt to maximise yields and to use prophylactic pesticide applications. Across the sample as a whole, 30% of respondents reported that they always used pesticides prophylactically, while 42% reported that they never or rarely used thresholds to help in making decisions about pesticide use. Given that the state advisory service was active in promoting the use of reduced pesticide doses and in making decisions on the basis of crop inspections during the period of the research, what this suggests is that collecting information does not necessarily lead to changes in behaviour on the part of farmers. An important underlying factor in observations such as those presented in Fig. 4 may be the way in which individuals
use information to update beliefs. A general framework for analysing this process is provided by Bayes’ theorem, which sets out the algebra required to combine prior beliefs with new information in order to arrive at new (posterior) beliefs about uncertain events (Howson and Urbach 1989; Adams 1998). Applications of Bayes’ theorem in this context generally assume that separate pieces of information acquired by a decision-maker are independent. However, Chaterjee and Eliashberg (1990) note that decision-makers considering adoption of a new innovation do not treat successive pieces of information from the same source as if they equally informative; later pieces of information are perceived as having already been partly included in earlier pieces of information. An analogous situation occurs in the process of hypothesis testing in science. Generally, we are less convinced by the evidence of many repetitions of one type of experiment to falsify a hypothesis than we are by concurring evidence from several different types of experiment. The implications for extension effort to increase rates of adoption are clear — variation in the presentation of relevant information will increase the rate of adoption.
Integrated approaches to diseases and pest control
Australasian Plant Pathology
Yuen and Hughes (2002) discuss the use of Bayes’ theorem for updating prior beliefs in the specific context of information required to make a decision about whether to take action to control a particular pest. Yuen and Hughes (2002) describe the way in which case-control methodology can be adapted from medical diagnostic evaluation to the context of crop disease management. The potential of this methodology to improve the practical use of sampling schemes and other disease prediction systems is potentially very great indeed. However, in spite of these methodological advances, when sampling plans and decision tools are linked to the use of action thresholds, they include a self-limiting mechanism which may serve to restrict their adoption. Before an exposition of the problem, we briefly introduce some notation. First, note that the odds of an event are described formally as the probability that the event will occur divided by the probability that it will not. This can be written as odds(E) = P(E)/(1–P(E)). Next we assume that we have some tool which is intended to aid in decision making by providing evidence as to whether the event has occurred. The tool can be a sampling scheme, or simple statistical prediction rule, or a complex simulation model. The key point of interest is that the tool will not be capable of perfect performance and we can characterise its ability to make correct predictions by its likelihood ratio (LR) with respect to the event in question. For example, suppose that the event in question is the occurrence of disease above a threshold at which action will be taken to control it and that the decision tool is a sampling scheme with known long-term performance that is used to obtain an estimate of the disease intensity. Writing D+ for the actual occurrence of disease intensities above the threshold, T+ for the prediction of disease at or above the threshold, D– for the actual occurrence of disease below the threshold, and T– for a prediction of disease below the threshold, the LR of a positive prediction from the sampling scheme can be expressed as shown in Equation 3: LR(D+,T+) = P(T+|D+)/(1–P(T–|D–))
(3)
The conditional probabilities on the right hand side of Equation 3 are, in the numerator, the probability of predicting that disease is above the threshold, given that disease intensity actually is at or above the threshold, and, in the denominator, the probability of predicting that disease is below the threshold, given that disease actually is below the threshold. The numerator is known as the sensitivity of the decision tool and the denominator is 1 minus the specificity. Yuen and Hughes (2002) and Hughes et al. (1999) give details on how to calculate these quantities. Finally we note that Bayes’ theorem provides a means to use the LR(D+,T+) combined with a prior estimate of P(D+) to arrive at a posterior estimate of the odds(D+) based on experience and given that data gathered from sampling
177
indicates that disease intensity is at, or above, the threshold. This is shown in Equation 4 odds(D+|T+) = LR(D+,T+)•odds(D+)
(4)
Now consider two farmers with contrasting assessments of the prior odds of disease being at, or above, the threshold. Farmer A considers disease at damaging levels to be quite unlikely and has a subjective evaluation of the prior odds, odds(D+)A = 0.25 (equivalent to a value of P(D+) = 0.2), while farmer B is less optimistic and considers the odds(D+)B = 4 (equivalent to a value of P(D+) = 0.8. Working with various methods for predicting Sclerotinia stem rot in oilseed rape, Yuen and Hughes (2002) found that the best predictor for confirming the presence of disease had a value of LR(D+,T+) = 7.00. Placing our two hypothetical farmers into the position where they have used the predictive system for Sclerotinia stem rot and obtained a positive prediction, their posterior odds will be A = 0.25 × 7.00 = 1.75 (equivalent to a posterior probability P(D+ | T+) of 0.64), and B = 4 × 7.00 = 28 (equivalent to a posterior probability P(D+ | T+) of 0.96). Thus, the positive prediction confirms the expectation of the pessimistic farmer (B) but works counter to the expectation of the optimist (A). Intuitively, we might expect that the level of uncertainty concerned with knowing whether disease actually is above the threshold has decreased for farmer B, but increased for farmer A. These changes in the level of uncertainty related to the binary event (disease at, or above, the threshold, or not) can be quantified using Shannon’s (1948) information entropy equation. A recent account of this work is given in Aleksander (2002). Shannon’s equation states that the information required to gain knowledge of an uncertain event is related to the probability of the event by the formula Ii = Σi –pi(log2pi) where, the pi are the probabilities of i mutually exclusive outcomes for an event. The relationship between the probability of the event and the value of I for the case of i = 2 (the situation faced by our farmers) is shown in Fig. 5. Note that I measures information in bits. Inspection of Fig. 5 shows that entropy is maximal when p = 0.5, which makes intuitive sense. Points representing the prior and posterior positions of the two farmers are also shown on Fig. 5. It can be seen that farmer B has moved down the right hand side of the curve and requires less information to be certain that disease is above the threshold after using the sampling scheme than before using it. Farmer A, however, has moved up the left hand side of the curve and is in a position of higher entropy after using the sampling scheme than before using it. This might be described as a tension between the farmer’s prior expectation of low disease intensity and the contradictory evidence of its being above the threshold. A further point to note, in relation to the use of sampling schemes linked to economic thresholds, is the way in which the certainty of the results of sampling behaves in relation to
178
Australasian Plant Pathology
N. McRoberts et al.
threshold-based IMP programmes. However, from an extension perspective, if the mechanics of a decision-making procedure are substituted by the black box ‘Trust us, you don't need to know how this works, just follow the instructions’ one can see that extra effort will be required to convince generally risk-averse growers of the value of adopting such methods. As we have already noted above, the job is made even harder by the diminishing ability of information from the same source to corroborate a proposition.
Information required for certainty (bits)
1
0.8
0.6
0.4
Conclusion 0.2
0 0
0.2
0.4
0.6
0.8
1
Probability of disease at or above threshold Fig. 5. The relationship between Shannon’s information measure, I, and the probability of an event with two possible outcomes. Points are marked showing the information required by two farmers to decide whether disease is at or above or below a pre-specified threshold prior to (solid symbols) and after (open symbols) using a risk algorithm discussed by Yuen and Hughes (2002). The squares represent a farmer with a low prior probability of disease, the diamonds represent a farmer with a high prior probability of disease.
the true intensity of disease and the position of the threshold. Formally, the probability from using a sampling scheme of concluding that disease intensity is below the pre-specified threshold varies as a function of the true disease intensity and is known as the operating characteristic (OC) of the scheme. When the OC is plotted as a function of true disease intensity it typically takes the form of a backward sigmoid curve which has its point of inflection at the threshold on the ordinate and at p = 0.5 on the abscissa. As we have seen, for a binary event (i.e. true disease intensity at, or above, the threshold, or below the threshold) when p = 0.5, I = 1 and the uncertainty with respect to the event is maximal. Practically, this has the unwelcome result that the decision-maker becomes more uncertain of whether the disease is above or below the threshold the closer the true disease intensity gets to the threshold. At the danger of stretching an analogy to physics too far, one might conclude from the above comments that thresholds are subject to a form of uncertainty principle. The closer the true disease is to the threshold, the less useful the threshold is as a definite decision point for concluding that action is needed. Conversely, when true disease intensity is either much greater or much less than the threshold, the need for threshold-based sampling is likely to be lower anyway. Of course, the decision-maker can, and probably should, be shielded from the technical details of threshold theory and there are many examples of successful
Integrated control continues to attract theoretical and practical interest among crop protection scientists and policy makers. From a strategic perspective the subject area has great value in integrating a diverse range of disciplines. There has been less interest historically among plant pathologists than entomologists (or indeed weed ecologists) in developing and promoting integrated control. However, taking the concept of the disease tetrahedron (Zadoks and Schein 1979) as a starting point, we hope to have demonstrated that integrated control offers rich areas of both practical and theoretical research for plant pathologists to pursue in collaboration with colleagues from many other disciplines. The opportunity to carry out research of both theoretical depth and practical value in both developed and developing countries will hopefully be a strong stimulus to young plant pathologists to contribute to the future of methods for the integrated understanding and control of diseases and pests of crops. Acknowledgements The authors thank the organising committee of the 8th International Congress of Plant Pathology for the opportunity to present a version of this paper at the ICPP2003 in Christchurch, New Zealand, and Professor Larry Madden for useful discussions. SAC receives financial support from the Scottish Executive Environment and Rural Affairs Department (SEERAD). References Adams EW (1998) ‘A primer of probability logic.’ (CLSI Publications: Stanford, CA) Aleksander I (2002) Understanding information bit by bit. In ‘It must be beautiful: great equations of modern science’. (Ed. G Farmelo) (Granta: London) Amorim L, Bassanezi RB, Berger RD, Hau B, Bergamin, A (2003) Considering physiological effects of leaf pathogens for better disease management. In ‘Proceedings of the 8th International Congress of Plant Pathology. Abstract C21.2.’. Vol. 1. p. 91. Bassanezi RB, Amorim L, Bergamin A, Hau B, Berger RD (2001) Accounting for photosynthetic efficiency of bean leaves with rust, angular leaf spot and anthracnose to assess crop damage. Plant Pathology 50, 443–452.
Integrated approaches to diseases and pest control
Bastiaans L (1991) Ratio between virtual and visual lesion size as a measure to describe reduction in leaf photosynthesis of rice due to leaf blast. Phytopathology 81, 611–615. Bastiaans L, Roumen EC (1993) Effect on leaf photosynthetic rate by leaf blast for rice cultivars with different types of and levels of resistance. Euphytica 66, 81–87. Boote KJ, Jones JW, Mishoe JW, Berger RD (1983) Coupling pests to crop growth simulators to predict yield. Phytopathology 73, 1581–1587. Burnett F (1999) ‘The use of fungicide sequences to maximise the control of eyespot in cereals and minimise the risk of sharp eyespot.’ Home Grown Cereals Authority Report No. 200, London, UK. Campbell CL, Madden LV (1990) ‘An introduction to plant disease epidemiology.’ (John Wiley and Sons: New York, NY) Chatrerjee R, Eliashberg J (1990) The innovation diffusion process in a heterogeneous population: a micromodelling apporach. Management Science 36, 1057–1079. Doucet P, Sloep P (1992) ‘Mathematical modelling in the life sciences.’ (Ellis Horwood: New York, NY) Franke AC, McRoberts N, Marshall G, Singh S, Malik RK, Nehra A (2003) A survey of Phalaris minor in the Indian rice–wheat system. Experimental Agriculture (in press). Fujisaka S (1994) Learning from six reasons why farmers do not adopt innovations intended to improve sustainability of upland agriculture. Agricultural Systems 45, 409–425. Gary G, Jeuffory MH, Ney B, Tivoli B (1998) Effects of Ascochyta blight (Mycosphaerella pinodes) on the photosynthesizing leaf area and the photosynthetic efficiency of the green leaf area of dried-pea (Pisum sativum). Plant Pathology 47, 473–479. Gottwald TR (2003) Epidemiological analysis applied to disease eradication: Citrus canker in Florida and Brazil. In ‘Proceedings of the 8th International Congress of Plant Pathology. Abstract C21.3’. Vol. 1. p. 92. Gottwald TR, Hughes G, Graham JH, Sun X, Riley T (2001) The citrus canker epidemic in Florida: the scientific basis of regulatory eradication policy for an invasive species. Phytopathology 91, 30–34. Heong KL, Escalada MM (1999) Quantifying rice farmers’ pest management decisions: beliefs and subjective norms in stem borer control. Crop Protection 18, 315–322. Hofstadter D (1999) ‘Godel, Escher Bach an eternal golden braid.’ 20th Anniversary Edition. (Penguin Books: London, UK) Howson C, Urbach P (1989) ‘Understanding scientific reasoning: the Bayesian approach.’ (Harcourt: New York, NY) Hughes G (1996) Incorporating spatial pattern of harmful organisms into crop loss models. Crop Protection 15, 407–421. Hughes G, McRoberts N, Burnett FJ (1999) Decision-making and diagnosis in disease management. Plant Pathology 48, 147–151. Hughes G, McRoberts N, Madden LV, Nelson SC (1997) Validating mathematical models of plant disease in space and time. IMA Journal of Mathematics, Applied Medicine and Biology 14, 85–112. Jeger MJ (2000) Bottlenecks in IPM. Crop Protection 19, 787–792. Jordon VWL, Hurcheon JA, Donaldson GV, Farmer DP (1985) Research into and development of integrated farming systems for less intensive arable crop production. In ‘Integrated crop protection: towards sustainability?’ (Eds D Atkinson, RG McKinlay) BCPC Symposium 63, 287–296. Kenmore PE (1996) Integrated pest management in rice. In ‘Biotechnology and integrated pest management’. (Ed. GJ Persley) (CAB International: Wallingford, UK) Levins R, Vandermeer JH (1990) The agroecosystem embedded in a complex ecological community. In ‘Agroecology’. (Eds CR Carroll,
Australasian Plant Pathology
179
JH Vandermeer, PM Rosset) (McGraw Hill Publishing Company: New York, NY) Madden LV, Nutter FW Jnr (1994) Modelling crop losses at the field scale. Canadian Journal of Plant Pathology 17, 124–135. Madden LV, Pirone TP, Raccah B (1987) Temporal analysis of two virus diseases increasing in the same tobacco fields. Phytopathology 77, 974–980. Matteson PC, Altieri MA, Gangé WC (1984) Modification of small farmer practices for better pest management. Annual Review of Entomology 29, 383–402. May RM (1973) ‘Stability and complexity in model ecosystems (2nd edn).’ (Princeton University Press: Princeton, NJ) McRoberts N, Davies DHK, Evans KA, Foster GN, McKinlay RG, Wale SJ (1995) The role of synoptic models in the development of crop protection systems for sustainable crop production. In ‘Integrated crop protection: towards sustainability?’ (Eds D Atkinson, RG McKinlay) BCPC Symposium 63, 439–446. McRoberts N, Foster GN, McCracken D, Sutherland K, Davies DHK (2000a) An evaluation of returns from crop management schemes based on farmer practice. Aspects of Applied Biology 62, 277–284. McRoberts N, Foster GN, Sutherland A, Willock J, Edwards-Jones G (2000b) Do hunter-gatherers make good farmers? Proceedings, BCPC Conference 2000, 1193–1198. McRoberts N, Hughes G (2001) Killing or culling? Is it possible to manage weeds as a resource? Proceedings, BCPC Conference 2001, 383–390. Mitchell J (1970) Big Yellow Taxi. Ladies of the Canyon. Reprise Records 7599274501. Murdoch WW (1975) Diversity, complexity, stability and pest control. Journal of Applied Ecology 112, 795–802. Ngugi HK, King SB, Holt J, Julian AM (2001) Simultaneous temporal progress of sorghum anthracnose and leaf blight in crop mixtures with disparate patterns. Phytopathology 91, 720–729. Orr A, Mwale B, Saiti D (2001) What is an integrated pest management ‘strategy’? Explorations in southern Malawi. Experimental Agriculture 37, 473–494. Rogers EM (1995) ‘Diffusion of innovations (4th edn)’. (The Free Press: New York, US) Rölling N (1988) ‘Extension science, information systems in agricultural development.’ (Cambridge University Press: Cambridge, UK) Savary S, Srivastava RK, Singh HM, Elazegui FA (1997) A characterisation of rice pests and quantification of yield losses in the rice-wheat system of India. Crop Protection 16, 387–398. Savary S, Wilocquet L, Elazegui FA, Teng PS, Du PV, Zhu D, Tang Q, Huang S, Lin X, Singh H, Srivastava RK (2000a) Rice pest constraints in tropical Asia: characterization of injury profiles in relation to production situations. Plant Disease 84, 341–356. Savary S, Wilocquet L, Elazegui FA, Castilla N, Teng PS (2000b) Rice pest constraints in tropical Asia: quantification of yield losses due to rice pests in a range of production situations. Plant Disease 84, 357–369. Shannon CE (1948) A mathematical theory of communication. Bell Systems Technical Journal 27, 379–423. (See http://www.math. psu.edu/gunesch/Entropy/infcode.html) Shiyomi M (2001) Utilization of biological interactions and matter cycling in agriculture. In ‘Structure and function in agroecosystem design and management’. (Eds M Shiyomi, H Koizumi) (CRC Press: Boca Raton, FL) Stern VM, Smith RF, van den Bosch R, Haygen KS (1959) The integrated control concept. Hilgardia 29, 81–101. Teng PS, Savary S (1992) Implementing the systems approach in pest management. Agricultural Systems 40, 237–264.
180
Australasian Plant Pathology
N. McRoberts et al.
Twengstrom E, Sigvald R, Svensson C, Yuen JE (1998) Forecasting Sclerotinia stem rot in spring sown oilseed rape. Crop Protection 17, 405–411. Van der Plank JE (1963) ‘Plant diseases: epidemics and control.’ (Academic Press: New York, NY) Way MJ, van Emden HF (2000) Integrated pest management in practice — pathways towards successful application. Crop Protection 19, 81–103. Whalon ME, Croft BA (1984) Apple IPM implementation in North America. Annual Review of Entomology 29, 435–470. Weber GE (1996) Modelling interactions between epidemics of Erysiphe graminis and Septoria nodorum on wheat. Zeitschrift für Pflanzenkrankheiten und Pflanzenshutz 103, 364–376. Webster JPG (1985) Analysing the economic consequences of sustainability in farming systems. In ‘Integrated crop protection: towards sustainability?’ (Eds D Atkinson, RG McKinlay) BCPC Symposium 63, 3–12.
Webster JPG, Bowles RG, Williams NT (1999) Estimating the economic benefits of alternative pesticide usage scenarios: wheat production in the United Kingdom. Crop Protection 18, 83–89. Yuen JE, Hughes G (2002) Bayesian analysis of plant disease prediction. Plant Pathology 51, 407–412. Zadoks JC (1985) EPIPRE, a computer-based decision support system for pest and disease control in wheat: its development and implementation in Europe. Plant Disease Epidemiology 2, 3–39. Zadoks JC, Schein RD (1979) ‘Epidemiology and plant disease management.’ (Oxford: New York, NY)
http://www.publish.csiro.au/journals/app