Does host genotype diversity affect the distribution of insect and disease damage in willow cropping systems?


L. Peacock, IACR-Long Ashton Research Station, Department of Agricultural Science, University of Bristol, Long Ashton, Bristol BS41 9AF, UK (fax +44 1275 394007; e-mail


  • 1Planting of willow genotype mixtures for biomass production has been suggested as a non-chemical strategy for pest management. Basic information on spatial dynamics of important insects and pathogens is necessary for the effective deployment of host genotypes in a plantation. In 1998 and 1999 the degree and spatial distribution of damage by Melampsora spp. and Phratora vulgatissima were studied concurrently in a field trial containing monocultures of willow Salix genotypes with different willow rust and beetle susceptibilities, and two design mixtures (random or regular) of three or five genotypes.
  • 2For both years, there was more rust and beetle damage on plants in monocultures than in mixtures. There were significant differences in the vertical distribution of beetle damage along stems between plantation designs for the 2 years, yet only in 1999 for rust. Rust severity along stems was significantly correlated between leaves, indicating localized spread of disease. In contrast, beetle damage severity along stems was weakly correlated between alternate leaves and leaves farther apart, suggesting beetle movement from tree to tree.
  • 3In 1998, spatial distribution of rust was aggregated in 67% of plots studied and in 40% for beetles. There was no significant difference in the distribution of beetle damage between planting design, but rust was aggregated in 75% of mixtures and 33% of monocultures. While beetle damage distribution was similar between years, rust was aggregated in all monoculture plots but in only 8% of mixtures in 1999. The difference between years and design for rust was probably because of the later stage of rust development in 1999 and the delaying effect of mixtures on this development.
  • 4In 1998, there were significant negative correlations between the extent of rust and beetle damage on individual trees. However, only 20% of plots showed a significant spatial dissociation between these two types of damage. There were fewer discernible interactions in 1999.
  • 5Plantation design provides the basis for integrated control of rust and beetle damage within willow cropping systems by delaying the spread and development of both organisms. Both pest and disease can be assessed successfully simultaneously under field conditions, a major saving in human resources. Initial selection of willow mixture configuration should primarily take into account the effects of spatial design on rust development.


Diversity in vegetation composition and structure within cropping systems can influence pathogen and arthropod pest populations. While the overall effects may differ between systems and species, the published findings are of a general reduction in density and subsequent damage by the causal agent, compared with a system comprising only a single genotype (Chin & Wolfe 1984; Andow 1991). The concept that mixtures can be developed to produce a degree of functional diversity (diversity that limits pathogen and pest expansion), and hence a more sustainable form of cropping, has been discussed for a wide range of agricultural systems (Andow 1991; Finckh & Wolfe 1998). At present, mixtures have been deployed primarily in annual crops (Garrett & Mundt 1999) where, if pests or diseases develop rapidly, pesticides can be applied or the choice of genotypes used as components in the mixture can be changed the following season. In long-term perennial crops, such as willows grown as short-rotation coppice (SRC), it may not be possible to change constituent genotypes or feasible to use pesticides. In addition, the complex of pests and pathogens likely to attack such crops may not show the same effects in response to the use of mixtures compared with pest complexes of annual crops. Boudreau & Mundt (1997) argue that the most important mechanisms that contribute to changes in insect and disease levels in variety mixtures are ones that affect dispersal of these damaging agents. A passively dispersed agent, such as an air-borne fungal pathogen, will be dependent on wind or rain for its dispersal, whereas an actively dispersed agent, such as a highly mobile insect pest, can to some extent choose its direction and rate of movement. It is possible that the mode of dispersal, and subsequent damage, will be different for each agent within a common cropping system, and therefore that different pests may be affected in different ways by genotype mixtures.

Willows grown as SRC have been shown to be a suitable species as a source of renewable energy to provide an alternative to fossil fuels (Stott et al. 1985). However, SRC willows are attacked by a number of potentially debilitating diseases and pests (Hunter, Royle & Arnold 1996). Cropping systems that incorporate plantings of willow genotypes that differ in susceptibility to the most serious disease and pest have been proposed (Parker, Arnold & Royle 1996; Tabbush & Parfitt 1999). An understanding of the potential competition effects between neighbouring trees (Parker, Arnold & Royle 1996) should also be considered when choosing genotypes for use in mixtures.

Willow rust disease, caused by Melampsora spp., is a wind-borne and rain-splashed pathogen that is one of the main causes of yield loss in SRC willows. Melampsora exists as a number of species and pathotypes, each of which is capable of infecting a characteristic range of willow species and hybrid genotypes (Pei, Royle & Hunter 1996). First symptoms are often seen on susceptible genotypes in UK willow plantations from mid-May to early June (McCracken & Dawson 1990). Parker, Arnold & Royle (1996) have shown evidence for the highly aggregated nature of rust spread within a plantation consisting of plantings of a single willow genotype. Dispersal within a plantation is thus initially from a focal source, and subsequent development is dependent on microclimate conditions and suitability of host genotype for epidemic progress. It is known that the majority of rusts affecting willows are unable to complete their life cycle on the willow host but instead need to pass through an alternate host such as European larch Larix decidua (Mill.) (Pei, Royle & Hunter 1993), consequently infections of foliar rust arise each season from external sources. The use of genotype mixtures has been suggested in the control of this rust pathogen (Dawson & McCracken 1995; Parker, Arnold & Royle 1996; McCracken & Dawson 1998).

In coppiced willow, the most damaging invertebrate pests are chrysomelid beetles: the blue willow Phratora vulgatissima (L.) and the brassy willow beetle Phratora vitellinae (L.). Phratora vulgatissima prefers to feed on willows low in phenylglucosides and P. vitellinae prefers willows high in these compounds (Pasteels & Rowell-Rahier 1992). Differences in preference have been noted between favourable willow types under both laboratory and field conditions (Kendall et al. 1996) indicating that beetles actively select their preferred willow food source. A study of the responses of the willow beetle to mixed willow genotype plantations shows a reduction in the overall density, damage and oviposition in the mixtures compared with the monocultures (Peacock & Herrick 2000). Beetles overwinter mainly outside plantations (Kendall & Wiltshire 1998), enter the plantation during April, feed, mate and lay eggs. Phratora vulgatissima are known to form aggregations and are highly mobile (Peacock, Herrick & Brain 1999), the position of highest infestation moving throughout the planting during the season (Sage et al. 1999).

The coppice willow system is therefore subjected to serious diseases and pest infestations that enter from sources outside of the plantation early in the growing season. The success of a specific willow plantation design in reducing the damage caused by beetles and rust will depend to a large extent on how easily these organisms are able to disperse throughout the plantation. Because rust is sessile and passively dispersed, and beetles are highly mobile and actively disperse, requirements for effective plantation design for control of either may be very different. Basic information on the spatial dynamics of these differentially dispersed organisms is necessary for the effective deployment of willow genotypes in a plantation, as dispersal within the crop may have significant effects on the implementation of pest management techniques (Jeger 1999). This study assessed the damage incurred from beetles and rust within willow plantations with random and regular design configurations of three and five genotypes, by intensive, simultaneous, assessments of the same stems in mid-growing season for 2 consecutive years (1998–99). The patterns of damage were then compared and contrasted between the different willow plantation designs. The sampling technique used also permitted the assessment of possible interactions between rust and beetle damage.

Materials and methods

Field study

This study utilized an existing 6-year-old experimental plantation at Long Ashton Research Station, Bristol, UK (51°25′28″ N, 2°39′49″ W), the detailed design of which has been published previously by Parker, Arnold & Royle (1996) and Peacock, Herrick & Brain (1999). The planting was composed of five willow genotypes [Salix burjatica (Nassarov) ‘Korso’, S. × dasyclados (Wimm.), S. stipularis (Sm.), S. viminalis (L.) ‘Bowles Hybrid’ and S. viminalis‘Mullatin’] planted as monocultures, and four mixture combinations, two with three willow genotypes (S. burjatica‘Korso’, S. × dasyclados and S. viminalis‘Bowles Hybrid’) (three-way) and two with all five willow genotypes (five-way). All plots were 20 m long (1 m between plants) by 20 rows wide (1 m between rows), comprising 400 plants per plot. One each of the three-way and five-way mixtures was planted randomly (random), in the other each genotype was planted in complete rows (regular). There were thus nine different plantings, and these were laid out in three complete randomized blocks. The five willow genotypes were originally chosen for differences in rust susceptibility, S. burjatica‘Korso’ and S. stipularis being the most susceptible to foliar rust and S. × dasyclados the least susceptible (Hunter, Royle & Arnold 1996). There are differences in beetle-feeding preferences between these genotypes, with S. × dasyclados and S. viminalis‘Mullatin’ the most, and S. burjatica‘Korso’ the least, preferred by P. vulgatissima (Kendall et al. 1996), the major willow beetle identified within this willow plantation (Peacock & Herrick 2000).

At the time of the 1998 assessment willows had 2-year-old regrowth, and for the 1999 assessments 1-year-old growth, having been coppiced during the winter of 1996/97 and 1998/99, respectively. The current year's growth in both years commenced in early April.

Sampling techniques

In 1998, S. viminalis‘Bowles Hybrid’ was selected for study as representative of a willow attacked by both rust and beetle. This genotype was sampled within all plantation designs containing it. During the first week in June, assessments were made on every S. viminalis‘Bowles Hybrid’ plant in the random mixes (three-way n = 130, five-way n = 78) and in the five-way regular mixture (n = 80), on every second plant in each row in the three-way regular mix (n = 60), and on alternate plants in every other row for the monoculture (n = 100). All three replicate blocks were assessed. The current year's growth from one main stem per tree was assessed, the number of leaves recorded, and the percentage leaf area loss due to beetle-feeding damage on each fully expanded leaf estimated visually. The furled leaves at the top of each assessed stem were assigned a relative feeding damage rating of zero (0), slight (< 5%) (1), moderate (5–20%) (2) or high (> 20%) (3) damage. Five leaves were sampled at equal intervals down each of these shoots and assessed visually for severity of foliar rust as percentage leaf area covered by active rust pustules. This was done using a generic leaf grid made up of 1% divisions (Parker, Perry & Royle 1993). Current year's stem growth was green in colour, whereas the previous year's growth was brown and suberized, so they were easily distinguished.

Field observations showed a lower incidence of rust in 1999 than 1998, and therefore assessments were delayed until the first week in July. Beetle-feeding damage and rust severity were assessed on S. viminalis‘Bowles Hybrid’ and S. × dasyclados. For each genotype, 24 plants [every third plant of a row (six plants) for every four rows] were assessed. A total of five leaves was randomly sampled at intervals down the stem and assessed visually for rust severity and beetle-feeding damage as in the previous year.

Dividing the stem into sections gave a vertical, third-dimension, to the spatial distribution of beetle and rust damage. The fourth quarter (base) or the fifth leaf corresponded to leaves that emerged in April, and the first quarter or first leaf represented leaves that emerged in June 1998 or July 1999.

Statistical methods

All analyses were done in genstat 5 (genstat Committee 1993), except for those that used spatial analysis by distribution indices (SADIE) (Perry 1998).

Beetle-feeding damage and rust severity

The different plantings formed a control (monoculture) plus factorial structure [mixture (three-way or five-way) × structural composition (random or regular)]. The position along the stems was analysed as a split-plot treatment. The mean damage rating of shoot tips per plot and the log percentage beetle-feeding damage and log rust severity were then subject to analysis of variance (the transformations were used to stabilize the variance). To investigate the relationships at different positions along the shoot for rust severity or beetle-feeding damage, correlations were calculated for each plot and tested for significance. Data were combined when no significant differences were found between specific designs. For instance, if no differences were found between regular and random configurations, data are presented for overall three-way and five-way mixtures, the spatial configuration for each density treatment being combined. Differences in the variability of beetle and rust damage between trees were determined for each plantation design by calculating variances and assessing them for significance.

Spatial distribution in plots

To analyse the spatial distribution of rust and beetle within the various plantation designs for both years, the recently developed SADIE was used (Perry 1998). Briefly, this SADIE algorithm gives an index of aggregation (Ia) and the probability of aggregation (Pa). Values of Ia > 1 indicate spatial aggregation, of < 1 indicate regularity, and those approximating 1 indicate randomness. SADIE tests are two-tailed, so Pa should be treated accordingly. SADIE has previously been used for studies on the spatial distribution of both insects and slugs (Bohan et al. 2000) and also for fungal pathogens (Turechek & Madden 2000).

As the SADIE tests are designed for count data, the incidence of damaged leaves with greater than 0·2% severity for rust was used in both years, as well as the number of leaves with beetle-feeding damage from a sample of leaves (20 in 1998, five in 1999). This severity level was chosen to include the lower overall rust damage in mixtures and thus give sufficient values for the SADIE analysis. Due to the nature of field variability and interactions under natural conditions that can influence spatial systems of pests and diseases, probabilities of a spatial pattern with P < 0·10 were considered to be an indication of spatial correlation and association.

Rust–beetle interactions

The sampling technique allowed for an investigation of any correlation between the levels of rust and beetle damage. The relationship in damage between these two agents was tested directly, first by calculating correlations between the average percentage damage on each plot, then in more detail by determining the correlation of damage by position along stems. Scatter plots of log(average rust severity + 0·1) (y) vs. log(average beetle damage + 0·05) (x) were made. Estimating the relationship between y and x is complex as each variable has an associated variance. In the ‘errors in both variables’ case, estimation of the regression line is dependent on having the ratio of the variances, which was unknown; thus the two extreme regression lines corresponding to the ratio of variances being zero or infinite are presented. The true estimated line would lie between these two. If these correlations were significant, a SADIE test was used to compare the spatial association of rust and beetle (Perry 1998). This algorithm gives an index of association (It) and a probability of association (Pt). Values of It > 1 indicate spatial association, and of < 1 dissociation.


Beetle-feeding damage and rust severity

Beetle-feeding damage

In 1998, there were significant differences in the vertical distribution of damage along stems between plantation designs. Damage on shoot tips due to beetle feeding (as measured by damage rating) was significantly greater (F = 4·90, d.f. = 2, 8, P < 0·05) for S. viminalis‘Bowles Hybrid’ plants in the monoculture (mean 0·66) than for plants in the mixtures (0·41, three-way regular; 0·45, five-way regular; 0·37, three-way random; 0·38, five-way random; SED = 0·114 on 8 d.f.). When the stem below furled leaves was divided into quarters, there was a significant difference in the pattern of feeding damage level along the length between random and regular spatial configurations (F = 4·14, d.f. = 6, 4005, P < 0·001). This difference was reflected in the progressive decrease in beetle damage in the mixture as a percentage of the monoculture, with the exception of the three-way regular design, which was similar to the monoculture (Fig. 1). This figure also showed that the mean percentage beetle-feeding damage for the monoculture increased up to the third quarter, with a decrease in the fourth. The mean percentage damage for the mixtures could be calculated by multiplying the monoculture value by the relative percentage of the monoculture for the mixture and position of interest.

Figure 1.

Beetle-feeding damage, as a percentage of the monoculture, in each quarter of Salix viminalis‘Bowles Hybrid’ stems sampled in 1998 as a component of a mixture of three or five genotypes planted in regular rows of genotypes or randomly. The mean damage (% damage + 0·05) for the monoculture is presented as a histogram (right-hand y-axis).

The relationship between different positions (quarters) along the stem for beetle-feeding damage gave significant (r = 0·22–0·62, P < 0·05, n = 90) positive correlations between 80% of the neighbouring positions. Furthermore, there was evidence of correlation between alternate sections, although less strong, with increasing distance between positions. As there were no differences between planting design for these correlations, it was not thought necessary to present them.

In 1999, S. viminalis‘Bowles Hybrid’ plants grown in a monoculture (mean 0·35%, n = 72) had significantly (F = 17·25, d.f. = 1, 4, P < 0·05) greater beetle-feeding damage than those grown in the five-way mixtures (0·20%, n = 142). As for the previous year, there were significant differences in the pattern of feeding damage along shoot lengths between the monoculture and the mixtures for S. viminalis‘Bowles Hybrid’ (F = 2·89, d.f. = 4, 1396, P < 0·05), with differences between the three-way and five-way mixtures as well (F = 2·87, d.f. = 4, 1396, P < 0·05). This was again reflected in the progressive decrease in damage as a percentage of the monoculture with position up the stem (Fig. 2). The differences in overall damage on plants of S. × dasyclados grown in various designs were not significant (0·38%, n = 72 monoculture; 0·29%, n = 282 mixtures). There were, however, highly significant differences in the pattern of damage along stems between the monoculture and overall mixtures (F = 3·75, d.f. = 4, 1408, P < 0·01). Relative to the monoculture, damage was more severe on stem position two and three in three of the mixtures (Fig. 3).

Figure 2.

Beetle-feeding damage, as a percentage of the monoculture, in 1999 on five leaves per stem (sampled equidistant down stems) of Salix viminalis‘Bowles Hybrid’ plants in mixtures of three or five willow genotypes planted in regular rows of genotypes or randomly. The mean damage (% damage +0·05) for the monoculture is presented as a histogram (right-hand y-axis).

Figure 3.

Beetle-feeding damage, as a percentage of the monoculture, in 1999 on five leaves per stem (sampled equidistant down stems) of Salix × dasyclados plants in mixtures of three or five willow genotypes planted in regular rows of genotypes or randomly, as a percentage of the monoculture. The mean damage (% damage +0·05) for the monoculture is presented as a histogram (right-hand y-axis).

Rust severity

In the first year of assessment (1998), rust damage was significantly greater (F = 95·73, d.f. = 1, 8, P < 0·001) on S. viminalis‘Bowles Hybrid’ plants in monoculture (1·50%, n = 300) compared with those in the three-way (0·39%, n = 574) or five-way (0·26%, n = 478) designs. Overall, plants in five-way mixtures were less severely diseased (although not significantly) (F = 4·62, d.f. = 1, 8, P < 0·06) than those in three-way plantings. There was no significant difference in the levels of disease recorded in random or regular configurations within three-way or five-way designs. The overall mean damage on the uppermost leaves (0·52%, n = 1352) within the plantation was significantly greater (F = 9·86, d.f. = 4, 5333, P < 0·001) than on leaves from all the other positions (0·45%, n = 5408). However, no differences were detected between these lower leaves. Correlations of rust severity between leaves along the stem for each plantation design were nearly all significantly positive for adjacent leaves, alternate leaves and those further apart (P < 0·05 for 143 out of 150 correlations). There was no significant difference between mixture designs in the pattern of rust severity along stems, even though levels of severity were significantly less than the monoculture (Fig. 4).

Figure 4.

Rust severity, as a percentage of the monoculture, in 1998 on five leaves per stem (sampled equidistant down stems) of Salix viminalis‘Bowles Hybrid’ plants in mixtures of three or five willow genotypes planted in regular rows of genotypes or randomly. The mean damage (% damage +0·1) for the monoculture is presented as a histogram (right-hand y-axis).

As in the previous year, there was an overall significant difference (F = 16·23, d.f. = 1, 8, P < 0·01) in rust damage between plantation designs but not between spatial configuration of S. viminalis‘Bowles Hybrid’ in 1999 (0·93%, monoculture; 0·49%, three-way; 0·28%, five-way). This was also true for S. × dasyclados (F = 67·32, d.f. = 1, 8, P < 0·001) (0·27%, monoculture; 0·08%, three-way; 0·06%, five-way). There were significant differences in disease levels along the length of stem for both willow genotypes between monocultures and three-way and five-way mixtures (F = 2·44, d.f. = 4, 1396, P < 0·05 ‘Bowles Hybrid’; F = 3·33, d.f. = 4, 1408, P < 0·01 S × dasyclados;Figs 5 and 6). This was also found for S. viminalis‘Bowles Hybrid’ between regular and random configurations (F = 4·65, d.f. = 8, 1396, P < 0·001; Fig. 5).

Figure 5.

Rust severity, as a percentage of the monoculture, in 1999 on five leaves per stem (sampled equidistant down stems) of Salix viminalis‘Bowles Hybrid’ plants in mixtures of three or five willow genotypes planted in regular rows of genotypes or randomly. The mean severity (% severity +0·1) for the monoculture is presented as a histogram (right-hand y-axis).

Figure 6.

Rust severity, as a percentage of the monoculture, in 1999 on five leaves per stem (sampled equidistant down stems) of Salix × dasyclados plants in mixtures of three or five willow genotypes planted in regular rows of genotypes or randomly, as a percentage of the monoculture. The mean severity (% severity +0·1) for the monoculture is presented as a histogram (right-hand y-axis).

Spatial distribution in plots

From the 1998 data, there was, generally, a weak regular or random distribution for the number of leaves per plant, with three plots having a significantly aggregated distribution and one plot a regular distribution (Table 1). Thus, the plots within the plantation were relatively homogeneous with regard to the number of leaves per plant. The index of aggregation (Ia) for beetle damage was significant for six of the 15 plots, showing that beetle damage in these plots was spatially aggregated. These plots comprised two of the three-way random designs and one plot from each of the other planting designs. The remaining nine plots had random distribution for beetle damage. The index of association (It) between the distribution of beetle damage and the number of leaves per plant was significantly positive for three of the 15 plots (A1, It = 1·03, Pt < 0·06; G3, It = 1·03, Pt < 0·10; I1, It = 1·04, Pt < 0·10), the remaining showing no association. In 1999, the Ia for beetle damage on S. viminalis‘Bowles Hybrid’ was below zero (i.e. regular) for six of the 15 plots (Table 2), one of these significant. Three of the remaining plots had a significantly aggregated distribution. The distribution of feeding damage on S. × dasyclados appeared to be more regular, although only two plots showed a significant Ia.

Table 1.  Statistics of 1998 spatial distribution of rust incidence, willow beetle-feeding damage and leaves per plant of Salix viminalis‘Bowles Hybrid’ when grown as a monoculture or as a component of a mixture containing three or five willow varieties planted in regular rows of varieties or randomly. Ia and Pa are measures of index and probability of aggregation, respectively
Planting designRustBeetleNo. leaves plant−1
  • *


  • **


  • ***

    Significant spatial distribution within a plot at P < 0·05, 0·025 and 0·005, respectively.

Monoculture 10·890·221·220·101·080·25
Monoculture 21·390·018**1·370·026*0·940·41
Monoculture 31·000·431·060·280·800·13
Three-way regular 11·570·004***1·190·131·180·14
Three-way regular 21·390·024**1·420·015**1·320·005***
Three-way regular 31·550·005***1·110·221·090·25
Three-way random 11·350·032*1·150·161·040·33
Three-way random 21·390·034*1·330·004***1·510·009**
Three-way random 31·230·0881·360·026*0·770·040*
Five-way regular 11·400·034*1·250·101·350·048*
Five-way regular 21·490·017**0·850·211·030·38
Five-way regular 30·900·321·350·033*1·110·26
Five-way random 11·140·191·120·211·010·41
Five-way random 21·880·001***1·420·025**1·070·29
Five-way random 31·560·004***1·290·070·890·27
Table 2.  Statistics of 1999 spatial distribution for rust and willow beetle-feeding damage on Salix viminalis‘Bowles Hybrid’ and S. × dasyclados plants when grown as a monoculture or as a component of a mixture containing three or five willow varieties planted in regular rows of varieties or randomly. Ia and Pa are measures of index and probability of aggregation, respectively
Planting designRust damageBeetle damage
‘Bowles Hybrid’dasyclados‘Bowles Hybrid’dasyclados
  • *


  • **


  • ***

    Significant spatial distribution within a plot at P < 0·05, 0·025 and 0·005, respectively.

Monoculture 11·500·009**0·850·171·520·01**1·160·15
Monoculture 21·830·001***0·810·120·890·300·950·47
Monoculture 31·390·023**0·970·490·980·470·960·47
Three-way regular 11·010·420·950·450·790·120·940·42
Three-way regular 21·450·021**0·900·341·170·160·750·04*
Three-way regular 31·100·261·130·201·390·04*1·330·04*
Three-way random 10·850·251·330·060·710·04*  
Three-way random 21·330·0760·850·211·620·005***0·710·02**
Three-way random 31·010·430·820·151·130·210·820·18
Five-way regular 10·970·500·920·330·850·160·910·36
Five-way regular 21·260·0751·280·0571·070·290·930·39
Five-way regular 30·810·0680·970·491·100·251·050·33
Five-way random 10·940·421·220·120·930·360·880·25
Five-way random 21·110·240·830·131·270·0931·160·18
Five-way random 31·140·211·120·201·100·270·960·43

In 1998, Ia for rust damage was significantly aggregated for nine of the 12 (75%) mixture plots, yet for only one of the three (33%) monoculture plots (Table 1). Thus, the distribution of rust damage within the monocultures was more often random or regular than aggregated, whereas it was aggregated in most mixtures. In 1999, the spatial pattern on S. viminalis‘Bowles Hybrid’ again showed differences between plantation designs, all three monocultures but only one mixture (three-way regular) having a significantly aggregated distribution (Table 2). Rust on S. × dasyclados was randomly distributed in all plots.

Rust–beetle interactions

In 1998, correlation between the degree of rust severity and beetle damage was calculated both from the plant means and at the different positions along stems. All plots showed negative correlations between mean beetle and rust damage, significant (P < 0·01) in 12 of the 15 plots. Correlations ranged from r = −0·40 for the monoculture to −0·04 for the three-way regular mixture (Fig. 7). All position-wise correlations (n = 120) were also found to be negative, yet only significant in approximately half of the section comparisons, with no obvious pattern between position of rust severity vs. beetle damage along the stem (Table 3). Rust severity data from leaves two to five were combined for this analysis as they had similar damage levels. Most positional correlations, however, did occur in the monoculture and the three-way random and five-way regular mixtures. Compared with their alternative spatial configuration, these two mixtures had somewhat lower levels of beetle damage, yet similar rust damage, along stems (Figs 1 and 3). In 1999, however, there were very few significant position-wise correlations (24/375 S. viminalis‘Bowles Hybrid’, 18/375 S. ×dasyclados) between the degree of beetle damage and rust severity on either willow genotype. This may have been a factor of the reduced number of trees sampled compared with the numbers in 1998.

Figure 7.

Scatter plots of the 1998 mean rust and beetle-feeding damage per tree of Salix viminalis‘Bowles Hybrid’ grown in monocultures or as a component of a mixture of three willow genotypes planted in rows of genotypes. Regression lines are of rust vs. beetle (a) and beetle vs. rust (b).

Table 3.  Correlations between the beetle damage in each quarter along willow stems (1st = top, 4th = basal) and the rust damage on the top (1st) and remaining four (2nd–5th) leaves in 1998. Salix viminalis‘Bowles Hybrid’ was grown as a monoculture or as a component of a mixture of three or five genotypes with random or regular spatial configurations
Planting designQuarter of stem for beetle damageLeaf number for rust damage
Block 1Block 2Block 3
  • *

    Significant interactions at P < 0·05.

Three-way regular1st−0·14−0·26*−0·13−0·09−0·11−0·13
Three-way random1st−0·15−0·20*−0·18*−0·10−0·17*−0·24*
Five-way regular1st−0·19−0·23*−0·07−0·09−0·41*−0·40*
Five-way random1st−0·07−0·20−0·34*−0·26*−0·27*−0·11
 3rd−0·14−0·33*−0·22−0·18−0·08  0·00

The index of association between the distribution of beetle and rust (> 1·0% severity) damage was significantly or strongly dissociated for three of the 15 plots (A3, It = 0·96, Pt < 0·10; F2, It = 0·92, Pt < 0·05; G3, It = 0·92, Pt < 0·10), the remainder showing no association. Only one plot showed a dissociation between beetle and rust when lower levels of rust damage were incorporated in the analysis (F2, It = 0·95, Pt < 0·10).


The concept that mixtures can be developed to produce a degree of functional diversity that limits pathogen or pest expansion (Andow 1991; Finckh & Wolfe 1998) has been utilized primarily in annual crops, with little work involving long-term perennial crops (Aluja et al. 1997; Docherty & Leather 1997; McCracken & Dawson 1998). Response to uniform or mixed willow plantings of different host genotypes by each member of the complex of pests and pathogens likely to attack this crop will determine the success of mixtures. It is known that the mode of dispersal for willow rust and beetles is different, consequently the functional diversity of a common cropping system may be different for each damaging agent. However, the present study shows the beneficial suppressive effect of mixtures against both organisms, as reflected in lower beetle feeding and rust damage compared with that found in the monocultures. The over-riding factor influencing beetle and rust damage was willow genotype numbers, with no consistent significant effect observed with structural design. However, the trend was for decreased overall damage when the spatial configuration was random rather than regular.

The comparison between plantation designs for the pattern of damage distribution along stems and within plots showed differences for the two organisms, reflecting their passive and active dispersal. There was a strong positive correlation in rust severity between leaves along stems, with a similar disease distribution between designs (but at reduced severity) as the number of host genotypes increased. For beetles, there were weak positive correlations in damage between leaves in alternate sections along stems, with relatively more damage on leaves in the monoculture compared with the mixtures, especially towards the shoot tip.

The strong positive correlation in rust severity between leaves along stems can indicate a spread of inoculum from neighbouring leaves on the plant. Parker et al. (1995) suggest that the spread of rust within an individual plant would be a reflection of what might occur within a monoculture in that susceptible host tissue is always in close proximity. Thus, in a monoculture rust can disperse radially to neighbouring trees, resulting in spatially aggregated distributions of incidence and severity (Parker, Arnold & Royle 1996). At a later stage in the epidemic, there would no longer be discernible aggregations (as our results showed for 1998) as more plants became infected. The occurrence of rust was later in 1999, thus at the time of rust assessment disease development would probably have been at an earlier stage, with the resulting aggregated distribution patterns found in monocultures in that year.

In mixed plantings, rust spores landing on resistant plants would not propagate; thus the spread of disease through a mixture of willow genotypes would be slowed down relative to the monoculture, as was suggested for other crops (Chin & Wolfe 1984). Therefore, in 1999, as rust in the mixtures would most likely be at an even earlier stage of development than in the monoculture, the non-aggregated spatial distribution reflects the initial random spore deposition and germination pattern. With time, the rust would spread and form discernible aggregations, as shown in our 1998 assessments. Thus, the difference in the spatial distribution of rust between the monocultures and the mixtures for both assessment years is probably due to a delay in epidemic development in the mixed plantings.

Differences in beetle damage along stems, especially between plantation designs, can indicate a movement of beetles from trees or, as suggested by Power (1988) for insects in heterogeneous plantings, an emigration from mixtures over time. Because the willow beetle is very mobile and discriminates between host-plants within a mixture (Peacock, Herrick & Brain 1999), time would be lost looking for suitable plants, especially in plots with several different willow genotypes and less regular spatial configurations (Peacock & Herrick 2000). These mobile aggregations of adults would result in a distribution of damaged plants throughout the plots. The spatial patterns of damage incidence in the present study were primarily regular or random, indicating beetles were not consistently choosing adjacent trees. There was also no evidence that beetles were choosing trees on the basis of leaf number, as there were very few associations between beetle damage and leaves per tree. Beetles may be responding to other factors, such as conspecifics or kairomones, as shown previously (Peacock, Lewis & Herrick 2001). This plant choice behaviour makes predictions of insect movement, especially in relation to configuration of genotypes within a plantation, very difficult.

The sampling of insect and rust damage on the same stems allowed for an assessment of interactions between the two causal agents. In the first year of this study, there were significant negative correlations between the extent of rust and beetle damage on stems. It is unclear what agent was responsible for these interactions as there were no consistent correlations between rust and beetle damage by position along stems. Only three of the 15 plots showed spatial dissociation, suggesting that beetles were not choosing or avoiding areas of rust infection. These dissociations occurred when rust damage was greater than 1%. Therefore any negative effects of rust on beetle feeding may be noticeable only at relatively high rust severity. As willow beetles tend to be present on host-plants before rust infections take place, the pattern and degree of leaf damage by beetles may affect subsequent rust infection. This occurs in Rumex spp., as feeding by the chrysomelid beetle Gastrophysa viridula causes a significant reduction in the infection of the rust Uromyces rumicis due to the induction of a resistance response throughout the undamaged portion of the damaged leaf (Hatcher et al. 1994).

In conclusion, the results shown here indicate the usefulness of deploying coppiced willow mixtures to reduce the impact of both rust and beetles within the same willow cropping systems. The spatial distribution of rust throughout the willow plantings differed from the distribution of beetles, reflecting their different modes of dispersal. Nevertheless, surveys of rust and beetle damage must consider the potential differential aggregations for both agents. However, it has been shown that both pest organisms can be assessed simultaneously. Results also suggest that spatial design of willow plantations affects rust distribution, and hence development, to a greater extent than it does beetle distribution. Therefore, this must be taken as a major criterion in selection of mixture configuration. Thus, further work is required to determine the optimum design and deployment for various willow genotype mixtures in renewable energy plantations.


Many thanks to Sarah Herrick and Emma Silversides for help with the statistics, David Bohan for assistance with SADIE and Angela Karp and David Glen for helpful comments on the manuscript. The Ministry of Agriculture, Fisheries and Food sponsored this work. IACR receives grant-aided support from the Biotechnology and Biological Sciences Research Council of the United Kingdom.

Received 8 December 2000; revision received 21 June 2001