Many epidemics of root diseases involving soil fungi depend on the interplay between fungal growth and the spatial and temporal heterogeneity of the soil environment. Colonization or infection of a root occurs at fine scales with growth and movement of fungal mycelia through soil. However, epidemics are observed at coarser scales, and depend on a cascading spread through populations of roots. We briefly review conventional analyses of soil-borne epidemics and argue that these treat soil physical conditions at scales too coarse to be meaningful for interactions between soil, plants and fungi, and fail to consider the effect of soil physical conditions on the underlying epidemiological processes. Instead, we propose a conceptual epidemiological framework that integrates spatial scales and use this to review the effect of soil structure on the dynamics of soil-borne pathogenic fungi. Using the soil-borne fungal plant pathogen Rhizoctonia solani as an example, we demonstrate that invasion of fungi into host populations is critically affected by environmental conditions operating at each of two scales: (i) at the microscopic scale (μm − cm) the fungus preferentially explores certain pathways in soil, and small changes in soil physical conditions make the fungus switch from small, dense colonies to large, sparse and rapidly expanding ones; (ii) at the larger scale (cm − dm) a critical density of susceptible hosts is required, in excess of which fungi switch from non-invasive to invasive spread. Finally, we suggest that the approach will increase the applicability of research dealing with microscopic soil–plant–microbe interactions towards the solution of large-scale epidemiological problems.
Soil-borne fungal plant pathogens are important determinants in the dynamics of plant populations in natural environments and in epidemics in agricultural environments. Examples of economically important soil-borne fungal plant pathogens include Fusarium spp., Gaeumannomyces graminis, Verticilium spp., Phytophthora spp., Pythium spp. and Rhizoctonia solani. Despite low initial densities of inoculum in soil, these pathogens can cause complete destruction of plants and, occasionally, total loss of yield. Infections occur initially at microscopic scales (approximately 10−3 m) following growth and movement of fungal plant pathogens through soil in a region surrounding a root or other subterranean organ. The region surrounding susceptible hosts from which a pathogen can infect has been defined as the pathozone (Gilligan, 1990). Epidemics, however, occur at macroscopic scales of metres and larger with cascading spread through populations of roots or plants. To date, most knowledge of the role of soil physical conditions in epidemics is obtained from data collected at the field scale. Here correlations are sought between particular environmental conditions or agronomic practices and the outcome of epidemics. However, different soil physical conditions operate at macro- and microscopic scales, and at each scale small changes in soil physical conditions can influence the outcome of soil-borne epidemics.
We show that a thorough understanding of the role of soil physical conditions is achieved only through integration of these spatial scales. First, we illustrate some shortcomings of a macroscopic approach that ignores the range of spatial scales and associated heterogeneity. Second, we introduce a simple framework that is based on generic epidemiological processes across a range of spatial scales. We use this framework to identify three key processes at different scales that drive an epidemic, namely fungal growth, infection dynamics of a single host, and invasive spread through a population of hosts. Making use of the epidemiological framework, we review the role of soil physical conditions in each of those processes. In particular, we demonstrate that thresholds operate at different scales, and how these are fundamental to the patchy nature of epidemics as well as the apparent difficulty in predicting outbreaks.
Soil physics and epidemics: macroscopic correlations
The development of novel molecular and biochemical techniques in recent decades has enabled us to increase immensely our knowledge about the microscopic and physiological processes underlying plant disease. Examples include the mechanisms involved in infection (Keijer, 1996), morphological changes of plant cells following fungal attack (Schmelzer, 2002), interaction between pathogens and antagonistic organisms (Weller, 1988), or the induction of systemic resistance upon pathogen attack (Heil, 2001). Despite the increase in knowledge at this scale, the extent to which these processes are affected by soil conditions remains unresolved, as studies at this scale are typically performed under sterile conditions in the absence of soil. Studies into why pathogens invade some soil environments and fail to do so in others therefore remain dominated by a macroscopic approach. Here correlations are identified between the outcome of epidemics as observed in the field and specific soil properties (such as structure, organic matter, moisture, pH) as well as the soil microbial composition (Alabouvette et al., 1996; Sturz et al., 1997) or agronomic practices such as continuous cropping, crop rotations or tillage operations (Sturz et al., 1997). To illustrate this, we consider here the relationship between soil tillage and epidemics.
A survey amongst more than 70 published accounts shows that conservation tillage systems can variously increase (54%), decrease (42%) or occasionally (4%) have no effect on plant diseases. The effects of climate, crop specificity or type of disease are discussed in other reviews (Sturz et al., 1997; Sumner et al., 1981), but even after such analysis, contradictions remain as to whether conservation tillage is more likely to promote or inhibit plant diseases. Some generalizations can be made: soils with an unstable structure (such as silty soils) are more easily compacted, often less aerated and generally wetter. Under those conditions diseases induced by Phytophthora spp. or Verticillium prevail, while those induced by R. solani are suppressed (Alabouvette et al., 1996). Apparent contradictions, however, are more common. For example, compaction of soil can reduce foot rot of winter wheat caused by Microdochium nivale and Fusarium spp. (Colbach et al., 1996), and reduce sharp eyespot in winter wheat (Colbach et al., 1997). But compaction can enhance epidemics such as common root rot in peas (Fritz et al., 1995), crown and root rots in a crop rotation in spring barley and soybean (Sturz & Carter, 1995), and cereal root rot in humid climates (Sturz et al., 1997). A unified insight into the effects of tillage operations upon epidemics is therefore still lacking.
Shortcomings with a macroscopic approach
There are at least two potential arguments against trying to identify macroscopic correlations between the effect of environmental conditions (biotic or abiotic) and epidemics. First, conventional analyses focus on quantification of the effect of selected properties on the amount of diseased plants at the end of an epidemic. Such analyses fail to identify and quantify the underlying generic epidemiological processes. These include the survival of pathogens in the soil, the ability of pathogens to initiate infection of a single host (primary infections of individual roots) and the ability of pathogens to spread through an entire population (secondary infections throughout the root system). The outcome of an epidemic depends therefore upon the complicated interactions of host, disease and inoculum dynamics, each of which may be affected by environmental variables. Second, the physical properties of soils have been assessed in terms of macroscopic properties considered to be characteristic of a substantial volume of soil. Average measurements at these scales may be too coarse to be meaningful at the scale of microorganisms (Young, 1998; Young & Ritz, 2000). To advance our understanding of soil-borne fungal plant diseases we need:
1to identify the generic epidemiological processes that determine the dynamics of epidemics;
2to identify the biological interactions that drive the epidemiological processes and the various spatial scales at which these interact with the environment; and
3to design a framework that enables integration of these processes over a range of spatial scales.
Soil physics and epidemics: an epidemiological approach
We propose the use of parsimonious models that capture the essential dynamics of epidemics (Gilligan, 2002), and subsequently identify how the key processes are affected by soil physical conditions. Most epidemiological models are defined within a generic framework in which tissue passes through different states from susceptible (S) to infected (I) and removed (R), with an additional class for inoculum (X) for soil-borne epidemics (Figure 1). These models implicitly combine microscopic processes of infection of a single host with large-scale epidemics in a population of hosts. The models were originally developed for human and animal epidemics (Anderson & May, 1991) but have since been widely applied in botanical epidemiology (Jeger, 2000; Gilligan, 2002), including soil-borne epidemics (Gibson et al., 1999). However, most studies fail to account for soil physical properties, so our understanding of their effect on epidemiological processes is still weak (Otten et al., 2004a).
To discuss how soil physical conditions interact with epidemiological processes we consider a simplified model that includes only two classes for hosts, S and I. The main drivers of an epidemic are the microscopic transmission rates for primary (βp, representing infection of a host from soil-borne inoculum) and secondary infection (βs, representing infection of a host from an infected host), which both operate at the level of individual hosts, and the interactions between X, S and I within the population (Figure 1). Biological processes underlying these drivers include pathogen dispersal and the spatial distribution of the population of roots and inoculum in the soil environment. In this paper we focus on the important group of soil-borne fungal plant pathogens that are dispersed primarily by extension of fungal mycelium.
Identification of research questions within an epidemiological framework
The epidemiological approach helps to structure questions required to understand how soil physical conditions affect epidemics. We suggest the following, which we subsequently address in more detail.
1Soil physics and fungal spread: at what scale and to what extent do soil physical conditions affect the spread of pathogenic fungi through soil?
2Linking mycelial spread to infection of a host: can we summarize the complicated interactions between mycelial spread, soil physical conditions and hosts in a simple epidemiological concept (pathozone) that relates fungal growth to infection of a single host?
3From infection of a single host to epidemics: can we scale-up from infection of a single host to spread through a population of hosts? What are the critical host densities required for fungal invasion, and how are these densities affected by soil physical conditions?
We show below how critical thresholds in soil physical conditions operate at the scale of fungal growth as well as the spread through a root system. The examples we discuss deal with an economically important soil-borne fungal plant pathogen, Rhizoctonia solani, for which molecular techniques for quantification in soil are available. The issues raised, however, have broader implications not only for other soil-borne diseases, but also for many other ecologically important processes involving fungi and bacteria in soil.
Soil physics and fungal spread
The soil provides the physical habitat in which biological interactions take place. The spatial heterogeneity of the structure of soil is manifest in various ways and at various scales: in tilled soil it can manifest itself in the form of beds of aggregates; in non-tilled soils it may appear as a system with planar pores or cracks between soil peds. Biopores, together with variation in soil physical properties associated with structural heterogeneity, such as the availability of water and oxygen, contribute further to the variability in the fungal environment. Despite the overwhelming evidence that disruptions of the soil structure have profound effects on microorganisms (Young & Ritz, 2000), we as yet know too little of the changes they bring about in the microenvironment to pinpoint controlling mechanisms. Microcosm studies in which some control can be exerted over the soil environment offer a way forward by enabling specific hypotheses to be tested. It is, however, essential that they include heterogeneity at scales that are relevant to fungal growth. In this section we introduce the spatial scales that we consider relevant for fungal colonies and discuss the effect of heterogeneity at various scales on fungal spread.
Soil structure and fungal length scales
It is essential to discuss structure and heterogeneity of the soil environment in relation to spatial scales. First, we define the mesoscopic and microscopic scales as, respectively, ‘of similar size to’ or ‘much smaller than’ the scale at which we observe fungal colonies (cm). Second, we introduce structure or heterogeneity in the context of these scales. We refer to a homogeneous environment as one that seems homogeneous at the mesoscopic scale. Typical measurements that can characterize the physical environment at this scale include bulk-density, particle-size or aggregate-size distributions, volumetric water content and matric potential. Closer examination of such a homogeneous sample at microscopic scales that are relevant to fungal hyphae with a typical diameter of 10 μm, however, reveals a tortuous network of pores through which hyphal spread occurs (Figure 2). Any mesoscopically homogeneous environment is therefore heterogeneous at smaller scales. Most techniques in soil physical research that can be used at the mesoscopic scale are not relevant at the microscopic scale, but advances in techniques such as thin sectioning (Tippkotter & Ritz, 1996) and computer-aided tomography (Gregory & Hinsinger, 1999) increasingly enable us to characterize the environment at these scales.
Based on these fungal length scales we discuss fungal spread in relation to three different environments: (i) homogeneous and single grain structure, which includes sand where the mesoscopic environment is formed by a spatial arrangement of solid primary particles; (ii) homogeneous and aggregated, which includes repacked sieved soils and beds of aggregates in tilled soils, with the mesoscopic environment formed by secondary particles (aggregates); and (iii) heterogeneous, which includes environments with either layered soils, large aggregates or structural features such as cracks, old root channels and biopores. The following examples are extracted from a series of experiments done with the ubiquitous soil-borne fungal plant pathogen and saprotroph R. solani that demonstrate the importance of heterogeneity at various scales for fungal spread.
Fungal spread through a homogeneous single grain structure
Perhaps the simplest physical environment we can construct for microbiological studies comprises sand with a narrow particle-size distribution, packed at a specific density. The geometry of the pore space through which fungal spread occurs is determined largely by the particle-size distribution of the sand, with pores containing either water or air. The simplicity of the structure enables us to address specific questions related to the effect of the physical environment upon fungal spread in replicated experimental systems. The following broad conclusions were drawn when we observed the spread of R. solani through simple homogeneous microcosms (Otten & Gilligan, 1998; Otten et al., 1999).
1Fungi spread predominantly, though not exclusively, through the air-filled pore volume, with further and faster spread with increased air-filled pore volume for otherwise identical media.
2Fungi spread preferentially along permeable soil surfaces when placed on such a surface, with faster expanding and up to four times larger fungal colonies, even though the same amount of biomass is formed (Figure 3).
3Changes in pore-size distribution at identical air-filled pore volume affect the rate and extent of colony growth, with more rapid and extensive spread in coarse media at identical air-filled pore volumes; and
4Radial spread of fungal colonies is more variable through soil than on surfaces.
These results clearly show a response of fungal growth to the physical environment. Perhaps more striking is that fungal spread can respond abruptly to some changes in soil physical conditions, with differential effects on the temporal (measured by biomass) and spatial dynamics (measured by extent and density of colony spread). For example, the relationship between air-filled pore volume and expansion of a fungal colony or biomass density of R. solani is highly non-linear. There is a very clear threshold value of air-filled pore volume less than which there is negligible fungal spread and more than which a colony expands rapidly (Figure 4). These results show that fungi spreading in a confined pore space produce small, dense fungal colonies, which switch rapidly to larger sparse colonies in a well-connected pore volume. Such switches in fungal growth had been observed previously in response to nutrients in the environment, where fungi can switch from an exploitative (in a rich environment) to an explorative mode, sending out a few sparse hyphae into the environment when nutrients become limited (Boddy, 1999; Boswell et al., 2003).
Fungal spread through homogeneous, aggregated soil
Narrow pore-size distributions are not common in natural soils. Pore spaces vary greatly both within and between aggregates, due mainly to varying clay content, organic matter content, and the degree of compaction caused by machinery and cultivation. Even within controlled environments one cannot prescribe soil structure precisely. One can, however, use bulk density and aggregate sizes as experimental variables to manipulate soil structure. Controlling these variables within specified ranges that are expected to occur in natural soils may subsequently identify the extent to which fungal spread can be affected. Using biological thin sections (Tippkotter & Ritz, 1996; Harris et al., 2003), which enable in situ visualization of fungal mycelium and soil structure, Harris et al. (2003) demonstrated the huge effect soil structure has on the spatial organisation of R. solani in soil (Figure 5). Colonization potential, defined as the volume of soil surrounding a source of inoculum within which a susceptible host has a finite probability of becoming colonized, was more than sixfold larger for soil packed at bulk densities > 1.4 g cm−3 compared with less dense soil (1.2 g cm−3) (Otten et al., 2001). These results accord with greater fungal densities in compacted field soil, suggesting that the findings have generality beyond the isolate used in this study (Glenn et al., 1987). Not surprisingly therefore, the spatial distribution of fungal hyphae in soil is patchy, in particular in the more porous soil (Harris et al., 2003). In such soil, larger voids within a soil are preferentially exploited by fungal mycelium; the denser volumes within the soil are partly avoided, even though these volumes are potentially accessible to fungal growth in denser soil containing no large voids (Figure 5).
Fungal spread through heterogeneous, structured soil
Microcosms for biological studies mostly comprise repacked sieved soils. Additional large-scale heterogeneity in the physical environment is rarely introduced. Yet cracks, old root channels and bio-pores are common features in most natural soils, in particular in reduced tillage systems where they can prevail for a long time. Filamentous fungi are well adapted to explore macro-pores via hyphal extension, and they can readily span air-gaps of several mm in vitro (Ritz, 1995) and in experimentally controlled arrays of aggregates (Toyota et al., 1996). Recently we reported on the effect of geometry of larger pores on fungal spread through soil by introducing them artificially into replicated microcosms (Otten et al., 2004c). Using biological thin sectioning techniques we showed that at the advancing edge of a fungal colony, mycelium occurs predominantly within or directly adjacent to cracks (Figure 6). Behind the advancing edge of the colony the fungal hyphae penetrated the surrounding bulk soil, but the results clearly showed that initial colonization occurs through rapid exploration of macro-porosity within soil. The preference for the crack was further enhanced if the conditions in the surrounding bulk soil were made less favourable. The precise mechanisms that make fungi follow these pathways preferentially in soil, however, remain a topic for further investigation. The preferential exploration of these soil features may prove decisive in the interactions between the soil, plants and fungi.
Ecological and epidemiological consequences
Traditionally, fungal biomass in soil is expressed in terms of colony forming units or hyphal lengths per gram soil. Densities of 10–1000 m g−1 soil have been reported (Alexander, 1977), which gives the impression that fungi are ‘everywhere’ and that mean densities are an appropriate representation. The examination of soil thin sections shows that the spatial distribution of fungal biomass in soil is sparse, and that there is a degree of non-randomness in the way fungi explore the soil. There are regions in an aggregated soil that are preferentially followed by invading fungi, and fungal invasion is strongly mediated by soil structure. The behaviour is of great ecological and epidemiological importance, with beneficial as well as inhibitory effects for fungal plant pathogens. For example, higher fungal densities as induced by changes in bulk density or wetness may enhance the contact between inoculum and hosts, and therefore give rise to an increase in primary infection if the inoculum is dense. Secondary infection, however, benefits from rapid fungal expansion, which would occur for more porous soils and drier conditions. Preferential spread through old root channels and biopores further enhances the probability of infection, as they are the very same pathways that are preferentially explored by roots. The effect of soil physical conditions upon an epidemic therefore depends critically on the balance between primary and secondary infection that is driving the epidemic. This balance may be subject to spatial heterogeneity (due to clustering of resident soil inoculum carried over from a previous crop) and to temporal heterogeneity (typified by seasonal variations or by shorter fluctuations following rainfall). These examples also show that soil structure is an important factor in epidemics and might be a missing link in our attempts to extrapolate experimentation from the laboratory to the field, as in most laboratory studies the spatial structure that plays such a key role in fungal exploration and biotic interactions is often removed.
Pathozone: linking mycelial growth with infection
Soil close to hypocotyls or roots generally has physical, chemical and biological properties that differ from the bulk soil. This zone has been referred to as rhizosheath (for soil directly adhered to roots), or rhizosphere (for soil under the influence of roots). Its epidemiological equivalent is referred to as the pathozone. The pathozone is a region of soil surrounding a root, seed or hypocotyl within which a fungal propagule must occur if it is to have some chance of infecting the host (Gilligan, 1990). Growth of fungal mycelium within this region leads to infection. The pathozone is a dynamic region that can be characterized by the furthest extent from the host, the probability of infection at the host surface and the change in probability of infection with distance (Gilligan & Bailey, 1997) (Figure 5).
Within the pathozone, mycelial growth can lead to infection of a susceptible host. Gilligan & Bailey (1997) showed how the spatial distribution of fungal mycelium in a colony spreading over a surface could be used to predict the probability of infection of radish seedlings. The extent and clustering of mycelium within a fungal colony, resulting from microscopic processes for growth, death and branching of hyphae, crucially determine the infection process. Soil physical conditions alter these physiological processes and affect the morphology of a fungal colony. Yet, spatial distribution of fungal mycelium in soil such as shown in Figure 5 remains difficult to quantify.
The characteristics of the pathozone can also be quantified directly for fungal spread through soil as the number of successful colonizations of a target placed at predetermined distances from the site of inoculation in replicated soil samples. The effect of soil physical conditions is then captured in the dynamics of colonization and can be described by simple empirical models (Otten et al., 2001) (Figure 5). Changes in bulk density from 1.3 to 1.5 g cm−3 resulted in smaller, slower expanding pathozones (Otten et al., 2001) (Figure 5). The spread of fungi through soil occurs in three dimensions (not in just the one dimension shown in Figure 5), and so the difference in fungal colonization is substantial: the volume of soil that is colonized is more than twice as large for the more porous soil. Otten et al. (2001) observed a similar effect for differences in aggregate-size distribution of the soil, with faster, further but more variable expanding fungal colonies with increasing aggregate sizes. The dynamics of colonization within the pathozone are further mediated by additional heterogeneity in the form of cracks or larger pores, with cracks acting as either ducts or barriers for fungal spread introducing more variation and clustering of mycelium in the fungal colony (Otten et al., 2004c).
Many difficulties remain to be overcome before soil physical properties can be fully included in this concept. Close to the surface of most subterranean organs, soil is physically, chemically and biologically different from the bulk soil. It is likely that these microscopic heterogeneities affect the extent and the shape of the pathozone. For example, the probability of infection of R. solani on bean is enhanced by leakage of exudates from the seed after germination. Compared with the probability of colonizing an inert host from the same distance, the probability of infecting a bean seedling is increased and is closely related to the radial distance from the seed (Figure 7). The effect of soil physical conditions upon fungal spread is here not the only environmental factor affecting the pathozone, as the effect of soil physical conditions and microbial interactions upon diffusion of exudates from root and seed and subsequent stimulation of pathogen spread become also important. Note that close to the surface, inoculum placed in proximity to the hypocotyls was not likely to cause infection, possibly as the emerging bean disrupted the fungal mycelium.
The pathozone is to some extent a black box that hides the detailed dynamics of hyphal growth, and local interactions with soil structure, microbial competitors and root exudates that lead to colony behaviour, as well as the plant–pathogen interactions that lead to infection. Nevertheless, it is a convenient way to summarize the complicated growth dynamics through soil and soil–plant–pathogen interactions, including the variation that is characteristically associated with these processes.
Percolation: from infection of a single host to invasion of a population
Epidemics caused by soil-borne fungal plant pathogens occur when pathogens spread invasively through a susceptible population. The spatial distribution and availability of susceptible hosts therefore critically determines the ability to invade a population. The spatial distribution of roots in soil reflects the morphology of roots as well as soil structure. Lateral distributions of roots can seem to be random, clustered, or clustered at small scales and apparently random at larger scales (De Willigen & van Noordwijk, 1987; Tardieu, 1988). There are few data on spatial distributions of roots, but reported values of distances between main root axes range from 0.95 to 1.32 cm for winter wheat (de Willigen & Noordwijk, 1987), to <2 cm in the top 40 cm and >6 cm at a depth of 80 cm for maize (Tardieu, 1988), with distances two to four times greater in layers between rows. Whereas the precise values vary widely amongst crops and soil types, these results demonstrate that: (i) for lateral spread of disease in a field the pathogen has to spread up to several centimetres through soil before finding a new root, and (ii) there are layers in soil and patches within layers where root density is substantially less. Depending on the ability of the pathogen to spread through soil, these distributions may restrict pathogen invasion and prevent the outbreak of an epidemic. We therefore need to link the morphology of fungal colonies as mediated by soil physical properties with fungal invasion into structurally heterogeneous populations. The need to scale up from fungal colonies to fungal spread through populations is particularly important as small changes in fungal growth and probabilities of infection of single hosts can lead to sudden and very large differences in invasion into a population. This is because the system is not linear.
The characteristics of the pathozone enable us to scale up from individual fungal colonies to spread through a population. For example, Kleczkowski et al. (1997) included pathozone dynamics in a probabilistic spatial contact process (probabilistic cellular automaton) and successfully predicted both the mean and variance of replicated damping-off epidemics. More recently, Bailey et al. (2000) and Otten et al. (2004b) used percolation theory together with experiments to demonstrate that thresholds for invasion can be directly predicted from fungal morphology. They first summarized local spread by the probability of making contact with a susceptible host at specified distances using the pathozone concept. Progressive invasion by colony expansion then depends on the spatial distribution of susceptibles within a population (Figure 8). The pathogen continues to spread as long as it makes contact with these susceptible sites, creating an expanding patch. However, the fungus will stop spreading if it fails to make contact with new susceptibles, and this results in a finite patch. The outcome of such a stochastic process can be described by the theory of percolation. Bailey et al. (2000) were the first to show experimentally that a critical probability of spread between neighbouring susceptible sites is required in excess of which invasion of a population can occur. Importantly, the ability of a fungus to invade was not determined by the furthest extent of fungal growth, but instead by the distribution and density of fungal biomass as captured by the pathozone. In a second paper, Otten et al. (2004b) showed that fungal invasion into a population can be stopped by rendering a threshold proportion of sites within a population of susceptibles resistant to infection (Figure 9). They also showed that increasing the proportion of protected sites in a susceptible population decreases the connectivity between infected and susceptible sites and hence retards invasion. The significance of these findings is that the extent to which control strategies are applied or are spatially successful can be a critical component of disease management. Both examples clearly demonstrate that (i) invasiveness of fungal spread can be predicted from fungal colony morphology together with the geometry of the root or host population, and (ii) small changes in fungal morphology or distribution of susceptible hosts (which are both mediated by soil physical conditions) can make a fungus switch between invasive and non-invasive spread.
It is apparent that the already substantial changes in morphology of fungi as mediated by soil physical conditions can be even further amplified if we consider the effect of those conditions on the spread through a susceptible population. These non-linear responses might contribute to patchiness in the field, even if the physical conditions are fairly homogeneous. They may also account for sudden outbreaks of epidemics caused by small temporal changes in conditions, such as an increase in wetness following rain. Extensions of this work are now required towards spread in three-dimensional space, through dynamically changing population structures of susceptible hosts (notably for inclusion of root growth and death) within a spatially and temporally changing soil environment. Percolation theory is still a subject of active research in statistical physics, and is increasingly applied in biological systems, with applications ranging from bacterial movement (Li et al., 1996) to large-scale spread of epidemics (Newman, 2002; Meyers et al., 2003). It continues to contribute significantly towards integration of our knowledge on small-scale infection processes and pathogen movement into the outbreak of large-scale epidemics.
The spatial and temporal dynamics of microorganisms in a complex heterogeneous and competitive environment dictate the outcome of many ecologically and economically important processes in soil, such as organic matter dynamics, successful remediation of contaminated soil, invasion of beneficial organisms or the spread of an epidemic. The pore space provides the environment in which biological, physical and chemical interactions occur, and through which microorganisms and roots spread to interact and exploit resources in soil. We know that soil structure affects microbiological processes, but this knowledge has not yet led to a level of understanding that would enable successful and predictable management of these processes in soil. Research should now focus more on how small-scale interactions relate to large-scale phenomena observed in fields, and how large-scale manipulation of soil structure is experienced by microorganisms in the rhizosphere. With recent progress in precision agriculture, together with advances in molecular diagnostics that focus on small-scale behaviour, understanding how intervention at smaller scales affects ecosystems is growing.
We have reviewed primarily the effect of soil physical properties, and soil structure in particular, on the dynamics of soil-borne pathogenic fungi and the outcome of epidemics. However, other properties can be considered in a similar way. For example, pathozone dynamics have been used to analyse the effects of antagonists (Bailey & Gilligan, 1997), host exudates (Figure 7) and genetic differences within pathogen species (Golldack et al., 2004). Here we used the epidemiological approach to identify two processes operating at different spatial scales that drive soil-borne epidemics, namely the spread of fungal colonies leading to infection of a single host, and invasive spread through a population of hosts. Soil physical conditions affect infection at both scales. There is great potential for soil physical conditions to mediate fungal growth, with small changes in the soil physical environment provoking large and abrupt changes in fungal morphology. At the coarser scale, soil conditions can mediate the spatial distribution of roots, and we showed how small changes to a population structure could make a fungus switch from non-invasive to invasive spread. This leads to the important conclusion that whilst small-scale heterogeneity in soil physical conditions regulates fungal growth dynamics and infection of a single host, the outcome of an epidemic depends on the non-linear relationship between distributions of hosts, such as root geometry, and the probability for a fungus to spread from an infected to a susceptible host. Whereas it is important that research remains focused on the physical habitat in which soils, plants and microbes interact, we must not isolate such microscopic interactions from the ecological processes under investigation. We suggest that the epidemiological framework introduced here will increase the applicability of research dealing with microscopic interactions to the solution of large-scale epidemiological problems.
The work described in this paper was made possible by two grants from the Biotechnology and Biological Research Council, linked with Silsoe Research Institute and the Scottish Crop Research Institute, and a grant from the Natural Environmental Research Council. We thank the colleagues of these institutes as well as members and former members of the Epidemiology and Modelling Group in Cambridge for their contributions to this work. Notable amongst these are Professor Iain Young, Professor Karl Ritz, Mrs Kirsty Binnie, Mr Christopher Watts, Dr Douglas Bailey, Dr Adam Kleczkowski, Dr John Ludlam, Dr Darroch Hall and Mrs Anne Bates.