1We examined how the interaction between the parasitic fungus Valdensia heterodoxa and its host plant Vaccinium myrtillus was affected by nitrogen (N) additions over 5 years in a boreal forest in northern Sweden. To examine whether the N effect was scale-dependent we fertilized different sized plots (1, 10, 100, 1000 and 5000 m2 in area) with N corresponding to 12.5 or 50 kg N ha−1 year−1. We also examined how the N effect varied with the amount of summer precipitation.
2Disease incidence of the parasite increased following N addition and the effect was, on average, stronger in large than in small plots. The effect of plot size was significant for both N addition levels during the final year, but only marginally significant for the entire experimental period. Dispersal distances of ascospores and conidia were short, suggesting that high net emigration probabilities of propagules from small plots could explain the lower disease incidence in such plots and thereby the scale dependence of the disease.
3Disease incidence was also positively correlated with precipitation. High summer precipitation enhanced the N effect, suggesting that precipitation may modify the effects of N deposition on plant–parasite interactions. This may complicate predictions of future effects of N deposition as precipitation patterns are expected to change as a result of climate change.
4Our results suggest that small-scale fertilization experiments may underestimate future large-scale effects of N deposition, and indicate the need for increased awareness of the problems associated with scaling results from experiments using small-scale manipulation of environmental variables.
Climate change and increased deposition of airborne chemical compounds are examples of environmental changes that may affect ecosystems on a continental, or even global, scale. However, due to financial and logistical constraints, much of our knowledge regarding the effects of such large-scale changes stems from experimental studies performed on much smaller spatial scales (Petersen et al. 1999; Gardner et al. 2001; Englund & Cooper 2003). A small but growing body of literature has indicated that observed ecological responses may change with the spatial scale of experimental units (Petersen et al. 1997; Sarnelle 1997; Schindler 1998; Bergström & Englund 2002), leading several authors to question whether the results of small-scale experiments are relevant for natural systems (Carpenter 1996; Schindler 1998; Englund et al. 2001). The range of extrapolation and therefore the potential problem is likely to be greatest in experiments that simulate large-scale environmental changes, making increased knowledge of when and how experimental scale may influence the results important.
Disease incidence or severity of plant pathogens may be scale-dependent. Mixtures of cereal species or strains of species that differ in susceptibility to pathogens may show reduced fungal infection (Burdon & Chilvers 1977; Wolfe 1985; Finckh et al. 2000), particularly when the mixtures are applied to large areas (Wolfe 1985), suggesting that the spread of fungal diseases can be scale-dependent.
The parasitic fungus Valdensia heterodoxa Peyronel has previously been found to play a key role in nitrogen (N)-induced changes in species composition of boreal understorey vegetation (Strengbom et al. 2002). Infection by the parasite has been shown to increase both following N addition (Nordin et al. 1998) and along a gradient in N deposition (Strengbom et al. 2003). High disease incidence results in premature leaf shedding of the dominant understorey species Vaccinium myrtillus L., which increases light availability for other species and enables, for example, the grass Deschampsia flexuosa (L.) Trin. to increase in abundance (Strengbom et al. 2004) as N increases.
Regional climate models for Europe suggest that climate change will lead to altered precipitation patterns, with summer precipitation expected to increase by up to 10% in northern Scandinavia, but to decrease by 30% in southern Scandinavia (Räisänen et al. 2004). Many fungal pathogens are sensitive to humidity, and disease incidence is known to increase following increased precipitation (Sutton 1982; Norvell & Redhead 1994; Clear & Patrick 2000), so that it is important to investigate whether the effects of N deposition may depend on precipitation, and the nature of any interaction. In this study we investigate how the effect of N addition on the incidence of the parasite Va. heterodoxa is altered by experimental scale and how this effect is influenced by variation in summer precipitation. We also examined the scale dependence of dispersal rates of spores and conidia as a possible mechanism underlying the scale dependence of disease incidence.
Material and methods
The study was conducted within the Svartberget Experimental Forest in northern Sweden (64°14′ N, 19°46′ E), 70 km north-west of Umeå. The area is located in the middle boreal zone (Ahti et al. 1968). The experimental site is a late successional Norway spruce (Picea abies L.) forest of Vaccinium–Myrtillus type (Kalliola 1973). The ericaceous shrub V. myrtillus dominates the understorey vegetation and co-dominant species are Vaccinium vitis-idaea L., Linnaea borealis L. and the grass D. flexuosa.
The experimental area comprised 66 plots in total. We fertilized plots 1, 10, 100, 1000 and 5000 m2 in size with either 12.5 (N1) or 50 kg N ha−1 year−1 (N2) (n = 6 for each N treatment and plot size permutation). The minimum distance between two treatment plots was 20 m (in most cases > 20 m). The N was added as solid NH4NO3 in the form of granules at the onset of the growth period (late May or early June), each year during the period 1996–2000. In addition to the fertilized plots there were six unfertilized control plots.
Valdensia heterodoxa is a parasitic fungus found both in Europe and in North America. It attacks a wide range of species, including V. myrtillus (its primary host in Europe), V. vitis-idaea, L. borealis and Gymnocarpium dryopteris (L.) Newm. (Gjaerum 1970; Norvell & Redhead 1994). Valdensia heterodoxa refers to the imperfect stage while the sexual stage is named Va. heterodoxa Peyronel and belongs to the Sclerotiniaceae (Eriksson 1974). This parasite causes brown spots on the leaves of the host, which may result in premature leaf loss (Norvell & Redhead 1994). Infection, by ascospores, takes place in early summer when fruit bodies (ascocarps) develop from sclerotia that overwintered in the veins of diseased V. myrtillus leaves that were shed the previous summer. During the summer, the fungus is also dispersed asexually by star-shaped conidia. For detailed descriptions of the fungus see Kujala (1946) and Redhead & Perrin (1972). The conidia are larger (∼0.5 mm) than the ascospores (10–12 × 4–5 µm) and are dispersed over a shorter range (Kujala 1946; Redhead & Perrin 1972).
disease incidence of the parasitic fungus
To estimate the effect of N on the incidence of Va. heterodoxa we collected 500 V. myrtillus leaves in August or September each year from 1996 to 2000, along the diagonals of control plots (1000 m2) and all but the smallest (1 m2) N-fertilized plots (where we were concerned that such a collection might have a negative impact on V. myrtillus). We corrected for patches with premature leaf loss caused by the fungus by assigning already shed leaves as diseased. To ensure that we did not include leaf losses caused by natural causes or other types of defoliators by this procedure, we only corrected for areas that showed clear signs of fungal-induced leaf loss, such as lesions on shed leaves or where the few remaining attached leaves were severely diseased. The collected leaves were placed in paper bags, brought to the laboratory and either dried in an oven (40 °C for 24 h) or pressed. We checked 80 of the original 500 leaves from each plot for disease symptoms caused by Va. heterodoxa and the number of infected leaves from each plot was used as an estimate of disease incidence for the plot. During the first year of the experiment, 1996, we scored both disease incidence (proportion diseased leaves) and severity (percentage leaf area with visual lesions). Because we found disease incidence to be well correlated with severity (Pearson's correlation: r = 0.86, P < 0.001) we decided to score only incidence during the following years. To estimate the effect of plot size on incidence of Va. heterodoxa, i.e. scale dependence of the N treatment, we collected 100 leaves from the central m2 of plots of each size (i.e. including the 1-m2 plots) each year from 1998 to 2000 and estimated disease incidence as above.
Data on precipitation for June, July and August during 1996–2000 were obtained from the meteorological station at Vindeln (data published in Väder och vatten, 1996–2000).
scale-dependent dispersal of valdensia heterodoxa
Scale-dependent disease incidence could occur if dispersal of the fungus results in a proportionally greater loss of spores and conidia to the untreated surroundings in small than in large plots, i.e. a higher per capita emigration rate of spores from small than from large plots. Recent studies have shown that the emigration probability is expected to be scale-independent if most dispersal distances exceed the linear dimensions of the plots, but to be scale-dependent if dispersal distances are short compared with plot size (Englund & Hambäck 2004a, 2004b). We measured dispersal distances of spores and conidia in a separate field experiment in a mesic coniferous forest of Vaccinium–Myrtillus type, located on Obbola Island 15 km south of Umeå. We chose this site because Valdensia had not been observed at the site in the two years prior to the experiment (1999 and 2000). Diseased leaves, identifiable by their thickened veins, were collected in October 2000 at a nearby site where they were stored over winter in nylon bags. After snow melt, in early May 2001, we placed the leaves on filter paper in Petri dishes and kept them under moist conditions until immature ascocarps started to emerge. During the last week of May, we placed one diseased leaf with an immature ascocarp on the ground at the mid-point of each of 20 plots (5 m in radius), watering them on days without precipitation in order to prevent the ascocarps from drying out. The first spores were released from mature ascocarps on 3 June, the first disease symptoms were observed on leaves on 13 June and the first mature, still attached conidia on 27 June, when scoring of disease symptoms took place to avoid autoinfection. For each leaf spot observed, we measured the distance to the mid-point of the plot.
Dispersal of conidia was studied in 15 plots (3 m in radius) in July 2001. On 3 July we sampled diseased leaves with well-developed conidia and placed each leaf on a black piece of paper in a Petri dish. We used a moistened pin needle to transfer a single conidium to a healthy, moistened leaf at the mid-point of each plot. We used a hand lens to check that the transfer had been successful: conidia developed and dispersed in 12 out of 15 plots. We measured the distance from each new infection to the inoculation point on 17 July, when the first of the new leaf spots started to develop mature conidia.
The overall effect of N addition on disease incidence was tested with repeated-measures anova. We used disease incidence across plot sizes of 10–5000 m2 as the dependent variable and level of N addition and year as factors. In cases of heterosphericity (identified by Mauchly's test of sphericity), degrees of freedom were adjusted by using Huynh-Feldt's epsilon within the SPSS 11.00 routine.
We tested the statistical significance of the effects of precipitation and N addition on disease incidence with ancova using N level as factor and annual summer precipitation (June to August) and year as covariates. Precipitation data were measured on a yearly basis, which suggests that calculating mean incidence per year (pooled over 10- to 5000-m2 plots) leads to an appropriate level of replication. As we were also interested in the interaction between N addition and precipitation, we choose to use the annual average of disease incidence from each N treatment (0, 12.5 and 50 kg N ha−1 year−1) as a dependent variable. We arcsin transformed the incidence data (Zar 1996) prior to analyses but, because statistical interactions can result from inappropriate transformations, we repeated this analysis on untransformed data, with the same outcome.
To test for the effect of N and spatial scale on the incidence of Va. heterodoxa we used repeated-measures anova. We used data on disease incidence from the central 1 m2 of plots of all sizes from the three years 1998–2000 as the dependent variable and plot size and N level as factors. As no spatial scale of treatment can be assigned to control plots, we excluded these plots from this analysis, i.e. the N effect describes the difference in response between the two levels of added N. To test if the effect of spatial scale differed between years we used one-way anova with disease level as the dependent variable. Each year and N level was tested separately.
In all years, disease incidence of the parasitic fungus on V. myrtillus leaves was higher in the N-fertilized plots analysed (i.e. excluding plots of 1 m2) than in unfertilized plots (repeated-measures anova: F1,63 = 17.07, P < 0.001), and the effect of N fertilization increased over the experimental period (repeated-measures anova N × time interaction: F5.62, 176.99 = 2.61, P = 0.022; Fig. 1). There was a large variation in disease incidence between years (repeated-measures anova: F2.81, 176.99 = 30.13, P < 0.001; Fig. 1) and the ancova using annual summer precipitation (June, July and August) as a covariate and annual disease incidence (pooled across the 10- to 5000-m2 plots from all N treatments) as the dependent variable showed that there was a significant positive association between summer precipitation, which also varied substantially over the experimental period (1996: 174 mm; 1997: 193 mm; 1998: 372 mm; 1999; 145 mm; 2000: 342 mm), and disease incidence (Fig. 2, Table 1). The relationship was stronger in fertilized plots than in unfertilized control plots, causing a significant interaction between precipitation and N fertilization (Fig. 2, Table 1) and suggesting that summer precipitation modifies the effect of N on disease incidence. Figure 1 shows that the effect of precipitation may be partly confounded with a temporal trend. Adding time as a covariate in the ancova showed that this temporal trend was highly significant (P < 0.001), but it did not alter the significance of the precipitation × N interaction.
Table 1. Results of the ancova using N level as factor, annual summer precipitation (June to August) and year as covariates, and the annual average of disease incidence from each N treatment (0, 12.5 and 50 kg N ha−1 year−1) during the 5 years as dependent variable
1.43 × 10−3
3.10 × 10−2
N × precipitation
2.15 × 10−2
N × year
3.11 × 10−4
2.10 × 10−3
The scale dependence of the N treatment was tested on the disease data collected from the central m2 of all plot sizes over the period 1998–2000. Plots that received the higher N dose showed significantly higher disease incidence than those receiving the lower N dose (Fig. 3, Table 2). The effect of plot size (scale dependence) varied over the experimental period (significant plot size by time interaction, Table 2). Testing the effect of plot size for each year and N dose independently (one-way anova) revealed significant effects of plot size (scale dependence) only in the last year of the experiment, both for the N1 (F4,25 = 2.78, P = 0.049) and the N2 (F4,25 = 3.27, P = 0.020) treatment. The lack of scale dependence in 1998 and 1999 explains why the main effect of plot size was only marginally significant (P = 0.079, Table 2) over three years (Fig. 3).
Table 2. Result of the repeated-measures anova on the disease incidence of Valdensia heterodoxa for the study period 1998–2000
N × plot size
2.21 × 10−2
Year × N
Year × plot size
Year × plot size × N
7.03 × 10−2
4.79 × 10−2
In the separate study examining dispersal ability of the fungus, we found that both ascospore (sexual) and conidia (asexual) dispersal distances were short. The majority of ascospores (91%) and conidia (100%) were found within 1 m from the source (Fig. 4), which indicates that the emigration probability of the fungus is indeed scale-dependent over the range of plot sizes covered in the fertilization experiment. Hence, a lower disease incidence in the small plots may be explained by a high net emigration rate of dispersal propagules.
We found that disease incidence of the parasitic fungus Va. heterodoxa was affected by both N addition and high summer precipitation. Increased disease incidence following N addition is in accordance with results from earlier fertilization experiments from boreal forests (Nordin et al. 1998; Strengbom et al. 2001) and with correlative data from a natural gradient in N deposition (Strengbom et al. 2003). The N effect was, however, stronger during summers with high precipitation, consistent with the observation that many fungal pathogens show better growth and infection success under wet or humid conditions (Sutton 1982; Norvell & Redhead 1994). We cannot separate whether higher incidence during wetter years was due to high precipitation per se or to high humidity, with which precipitation is probably well correlated, but our findings do suggest that the effect of increased N input can be modified by altered precipitation patterns. Because climate models (Räisänen et al. 2004) predict that precipitation patterns will change differently over different parts of Europe, it may be important to consider this interaction when evaluating the effects of increased N input on plant–pathogen interactions.
The finding that incidence of Va. heterodoxa was higher in the larger fertilized plots shows that the N-induced responses of a plant community can be spatially scale-dependent. We did not use enclosures, or any other modifications of the system, and the mechanism generating the scale effect could not therefore be considered as an artefact of the experimental protocol. Rather, we expect that the mechanism involved is general – in the sense that it may also cause scale-dependent responses in natural patches of different size. A similar type of scale dependence of disease is known from agricultural systems where mixtures of cereal species or strains (genotypes) of the same species may reduce disease problems (Burdon & Chilvers 1982; Wolfe 1985; Finckh et al. 1999, 2000), especially if the mixture is applied to larger areas rather than to small experimental plots (Wolfe 1985).
One possible mechanism for the scale dependence observed in the present study could be higher heterogeneity in large than in small plots. A scale-dependent response can be caused by scale-dependent heterogeneity in a driving variable if the response is a non-linear function of this variable (Englund & Cooper 2003; Bergström & Englund 2004). The spread of an infection is an inherently non-linear process. At low to moderate disease levels it is expected that the spread of the disease is nearly exponential, while the rate of spread must decline at high incidence levels, because most of the potential infection sites are already occupied. Thus, we expect that increased heterogeneity can increase the rate of spread when the incidence is low. For example, we observed that, under control conditions, Va. heterodoxa is more frequent in moist and shady patches and, if such patches are hot spots for fruit body production (sexual dispersal), spread would be more rapid in large plots; these larger plots would be more likely to contain both moist and dryer areas, i.e. to be more heterogeneous. However, the expected effect of increased heterogeneity at high incidence levels is to decrease the rate of spread, which means that the scale dependence should level off and reverse over time. As we recorded high incidence values and a scale dependence that increased rather than decreased over time, it seems that our data do not support this mechanism. However, it is difficult to know if the assumption that almost all potential sites for new establishment of the disease are occupied was fulfilled. Nevertheless, as disease incidence and severity were well correlated during the first year of the experiment, it seems reasonable to assume that high incidence of the disease indicates high severity and that the probability of new infections under such circumstances should be limited.
Another possible explanation for the observed scale dependence is that some of the added N was taken up by plants growing outside the plots or transported away by surface runoff, affecting N availability close to the edges of the plots. As the edge zone makes up a larger proportion of the total area in small plots than in large plots, this could lead to a difference in mean host plant N concentration and, thus, to a difference in disease incidence. This kind of perimeter-to-area effect is likely to occur in almost all experiments conducted in small plots that simulate changes in factors such as N deposition, CO2 level or temperature. A scale-dependent N leakage can be expected to increase the difference in N concentration between small and large plots over time and thereby to affect the disease incidence, which would be consistent with our result: a scale effect emerging during the last year of the experiment. Given that the effect on the vegetation often is sharply restricted to the experimental plot in fertilization experiments from boreal forests (Tamm 1991; Strengbom et al. 2001), and that clear differences in N concentration of plant tissue have been observed even when plot sizes are small (Nordin et al. 1998; Graglia et al. 2001), we suspect that this kind of perimeter-to-area effect is relatively weak and may not have contributed significantly to scale dependence in this study.
A third mechanism that may cause a scale dependence is scale-dependent dispersal of ascospores and conidia, i.e. that a greater proportion of the propagules produced in a fertilized plot will be lost to the unfertilized surroundings when plots are smaller (Englund 1997). Fertilization is known to increase host plant susceptibility and disease incidence (Strengbom et al. 2002), but if spore net emigration is higher for small than for large plots, fertilization of small plots will result in a less pronounced fertilization effect than fertilization of large plots. This corresponds to dilution of inoculum due to increased distance between suitable host plants, thought to be an important mechanism generating disease reduction in cereal mixtures (Burdon & Chilvers 1977; Wolfe 1985). However, this mechanism requires that most dispersal distances are shorter than the linear dimensions of the treatment plots (Englund & Hambäck 2004a, 2004b) because longer dispersal distances would lead to high emigration irrespective of plot size. Although fungal spores in general are capable of long-distance dispersal, we found ascospore dispersal by V. heterodoxa to be rather restricted, with the majority of spores ending up within 1 m from the source, and even shorter dispersal distances for conidia. Thus, our dispersal data are in agreement with the hypothesis that a larger proportional loss of propagules in small plots led to lower disease incidence.
A recent analysis (Hambäck & Englund 2005) shows that the scaling of random small-scale movements can cause a positive area–density relationship only if the studied patches are sources. For our system this would mean that there should be a net outflow of spores from fertilized plots. This seems reasonable, given that disease incidence was consistently lower in control than in fertilized plots. It is possible that several years of N fertilization are needed before the difference in inoculum production, or plant susceptibility between fertilized and unfertilized areas, becomes large enough to allow scale effects to be detected, which may explain why we only observed a significant scale effect during the final year of the experiment.
In summary, our study shows that both spatial scale and precipitation may modify the effect of N addition on boreal understorey vegetation. Of the mechanisms we have considered, it seems that a scale-dependent dispersal of the fungus is the most likely explanation for the observed scale dependence in disease incidence. The findings from the present study may have implications for our ability to use field experiments to predict the large-scale and long-term effects of increased N deposition. As the parasite has previously been found to play a key role in N-induced vegetation changes, altered precipitation patterns may result in higher or lower rates of vegetation change in response to N, depending on how the precipitation changes. The finding that the effect of N addition may be scale-dependent suggests that the potential bias caused by the choice of experimental scale should be considered before generalization is made from experiments based on small plots.
We would like to thank Hanna Oscarssson, Jens Ingerlund, Anders Hedefalk and Hans-Göran Nilsson for skilful assistance during field and laboratory work, and two anonymous referees for valuable comments on an earlier version of this paper. This study was financially supported through grants from the Swedish Environmental Protection Agency (to L.E.) and from the ASTA programme financially supported by MISTRA (Swedish Foundation for Strategic Environmental Research) (to L.E.).