*Present address and correspondence: Enviros Consulting, 20–23 Greville Street, London EC1N 8SS, UK (fax +44 2074302210; e-mail email@example.com).
1Annual meadow grass Poa annua is the most problematic weed within sports turf in temperate climates. It is so abundant that herbicides cannot be used against it because almost total loss of the sward would occur. Arbuscular mycorrhizal (AM) fungi can be used as biological control agents of P. annua, acting to reduce its growth while increasing that of desirable perennial grasses. However, natural levels of AM fungi in amenity turf are very low. Sports turf is characterized by high fungicide usage, so this study tested the hypothesis that levels of toxic elements (derived from historical fungicide applications) and/or organic fungicides are related to the low mycorrhizal abundance observed.
2Records of fungicide applications to putting greens at four golf courses in south-east England were collected for the period 1993–2000. Soils from all 18 putting greens on each course were sampled during spring 2000. Levels of arsenic, cadmium, copper and lead were recorded, as these elements formed part of the fungicidal compounds applied in the past. Of the organic fungicides in current use, chlorothalonil, fenarimol and iprodione were found to comprise more than 70% of all compounds used and soil levels of these three chemicals were also measured. Levels of AM colonization of P. annua were recorded at every putting green. The rates of AM colonization of desirable grasses could not be obtained because these were too rare in all the swards examined.
3There was little evidence that the abundance of AM fungi, as measured by arbuscular colonization of roots, was affected by the presence of any of the chemicals. Levels of all elements were below the ambient levels for UK soils.
4Chlorothalonil, fenarimol and iprodione were applied to a single putting green over a 6-month period but no effects on AM colonization were found.
5Synthesis and applications. We have demonstrated that the low levels of AM fungi in putting greens are unlikely to be a consequence of excessive fungicide application. Levels of compounds applied in the last 20 years are very low and modern fungicides do not reduce existing AM colonization when applied to turf. Therefore, if AM fungi are added to sports turf to control P. annua, their effectiveness will not be compromised by current or past fungicide use.
One of the keys to successful sports turf management in the UK and other temperate areas of the world is control of the weed grass Poa annua L. (annual meadow grass or annual bluegrass) (Adams & Gibbs 1994). For example, P. annua infests virtually every golf green in the UK and in many cases it can represent the majority of grass cover in the sward (Mann 2004). Although selective herbicides have been used as control agents (Johnson 1982), application can result in almost total loss of the sward. This is unacceptable on golf putting greens. Poa annua is a particular problem in fine turf such as putting greens because it is susceptible to drought, so large amounts of irrigation water must be applied to maintain the sward in a playable condition. In addition, it is susceptible to disease, notably microdochium patch [causative organism Microdochium nivale (Fr.) Samuels & I.C. Hallett] and anthracnose [basal rot, causative organism Colletotrichum graminicola (Ces.) Wilson]. In order to maintain fine turf on golf putting greens that contain high proportions of P. Annua, fungicide application is necessary in order to protect against disease (Perris & Evans 1996).
Currently, 16 fungicides containing seven different active ingredients are approved for use on turf grass in the UK (Whitehead 2000). In a recent survey, it was found that golf putting greens receive five times the amount of active ingredient of all pesticides per unit area per annum than is applied to a cereal crop (Garthwaite 1996). Such a level of pesticide application, combined with a high frequency of irrigation, means that there is the potential for leaching of chemicals and contamination of groundwater (Odanaka et al. 1994). Despite this theoretical risk, a recent review found little evidence for groundwater contamination because chemicals persist in the soil profile or are degraded to less harmful substances by microbial action (Cohen et al. 1999).
There are several reasons why turf management on golf courses needs to become less dependant on chemical inputs and more ecologically based. First, reliance on any one chemical cannot be sustained because the target organism is likely to develop resistance. Indeed, fungicide resistance in M. nivale, the most prevalent and damaging pathogen of sport turf, has been reported in the USA (Vargas 1994). Secondly, many chemicals have non-target effects on beneficial organisms, for example fungicides applied to turf can reduce bacterial populations in the soil (Yang et al. 2000). Thirdly, the current pesticides review programme in the European Union (EU) may mean that a number of compounds will be lost to UK growers in the future, including products used by turf managers (Wood 2001). Over the years, pesticide strategies on turf grass have changed from the use of inorganic compounds, such as mercuric chloride, cadmium chloride, Malachite/Bordeaux mixture (containing copper sulphate), Paris Green (containing copper acetoarsenite) and lead arsenate (Greenfield 1962), to the array of modern, less persistent organic pesticides. However, the number of approved compounds is likely to diminish in the future. Biological approaches to turf management are needed to replace chemical control, but these need to be compatible with pesticide residues derived from past or present usage.
It has been shown recently that arbuscular mycorrhizal (AM) fungi may have potential for reducing the amount of P. annua in putting green turf (Gange 1998; Gange, Lindsay & Ellis 1999). These fungi generally form mutualistic associations with about 70% of vascular plants (Hodge 2000), but for most plants there exists a continuum of responses to fungal colonization, from positive (i.e. beneficial) to negative (i.e. antagonistic) (Gange & Ayres 1999). In fine turf, it appears that AM fungi are antagonistic to the growth of P. annua, while being beneficial to the growth of desirable grasses such as Agrostis spp. (Gemma et al. 1997; Gange 1998). The mechanism is thought to be one in which carbon outflows to the mycorrhiza exceed nutrient inflows in P. annua, thus resulting in a net reduction in plant growth (Gange, Lindsay & Ellis 1999). AM fungi therefore have the potential to be an important part of an integrated control programme for P. annua. However, for this to be successful management techniques must be developed to encourage and sustain high levels of these fungi in sports turf to sustain sufficient coverage of desirable grass species.
There is a potential conflict in the use of turf fungicides and AM fungi, because many chemicals in common use have been shown to reduce AM abundance in ecological experiments (e.g. chlorothalonil, Venedikian et al. 1999; iprodione, Gange, Brown & Farmer 1990). Furthermore, if lack of pesticide leaching from golf greens is indicative of chemical retention within the soil profile (Cohen et al. 1999; Armbrust 2001), it is also important to determine if this is a cause of the relative scarcity of these fungi in turf soil. Certainly, levels of AM fungi in putting greens are considerably lower than those of less intensively managed areas (Koske, Gemma & Jackson 1997a; Gange, Lindsay & Ellis 1999). This could be due to persistent elements, such as cadmium, lead and arsenic, derived from compounds applied many years ago, or from frequent applications of modern organic fungicides. However, to date no study has examined whether application of fungicides to fine turf affects the colonization of grass roots by AM fungi.
In this study we described recent fungicide application patterns at four golf courses in southern England to determine which chemicals are most commonly applied. In addition, we measured levels of arsenic (As), cadmium (Cd), copper (Cu) and lead (Pb) and three of the most commonly used organic pesticides in putting green soils, and performed regression analysis of these data against the abundance of AM fungi. Clearly, it is not desirable or relevant to conduct experiments involving historically applied compounds such as lead arsenate or cadmium chloride. Therefore, we took an observational approach to identify relations between element levels and AM colonization. We also conducted experiments with modern fungicides where three of the most commonly used chemicals were applied to a working golf green and AM abundance levels were measured over a 6-month period. This experiment was designed to test the hypothesis that fungicide application will reduce AM colonization in sports turf, given that one of the most commonly used chemicals (iprodione) has been shown to do so in a natural plant community (Gange, Brown & Farmer 1990).
Materials and methods
Four 18-hole golf courses were selected for the study, referred to as A, B, C and D. Courses A, B and C opened in 1896, 1894 and 1928, respectively, and were therefore of similar age. Each of these three courses had greens in which the root zone was constructed using local soil. This gave pH ranges of 5·6–7·9 (course A), 6·6–7·4 (course B) and 4·5–6·8 (course C). Course D was of more recent construction (1991) and the root zone in the greens comprised a mixture of 80% sand and 20% milled peat, known as the United States Golf Association (USGA) specification (Bengeyfield 1989). Bicarbonate-extractable P concentrations were 19·4 ± 1·2 µg g−1, 26·3 ± 3·6 µg g−1, 24·7 ± 5·4 µg g−1 and 17·9 ± 2·3 µg g−1, respectively. Greens were mown daily in summer to a height between 4 and 5 mm, and all were equipped with automatic irrigation systems.
fungicide application records
Applications of all fungicides (whether for fungal or earthworm control) were recorded per year from 1993 to 2000 to provide information on the relative amount of use of each. These data showed that chlorothalonil, iprodione and fenarimol were the most widely used chemicals, and so these three were selected for soil analyses (see below). Differences in the mean number of applications per year between courses were examined with one-factor anova, to determine if the extent of fungicide use varied between courses.
Two soil cores (5 cm deep × 2·5 cm in diameter) were taken from random positions on each of the 18 greens on each course. Duplicate cores were also taken at random from two greens per course to assess the sampling precision of both element and organic analysis. A reference (background) soil untreated with fungicides was sampled in the same way from the edge of one green per course.
Each core of soil was split into two halves. One half was stored for organic fungicide analysis (see below) while the other subsample, required for analysis of copper, cadmium, lead and arsenic, was placed in an oven at 50 °C for 24 h, before being crushed and passed through a 500-µm mesh sieve. One gram (±0·01 g) was then leached with 10 mL of a 20% HNO3 solution on a hot plate at 55 °C for 30 min. Once cooled, the samples were filtered through a medium-grade filter paper (Whatman® grade 2) into 20-mL flasks, made up to 20 mL with distilled water and placed into labelled plastic tubes.
This sample preparation offered several advantages. Because arsenic analysis was required, any option involving a vigorous reaction or heating over 60 °C had to be avoided. This meant that fusion or full hydrofluoric (HF) acid digestions could not be performed. Samples also contained a large amount of organic matter (e.g. roots) that a simple acid leach could cope with. Finally, a mild acid leach is more representative of the bioavailable fraction of the soil matrix, which was of interest in this study. However, acid leaches suffer poorer precision and repeatability/reproducibility compared with full digestions or fusions (Potts 1989). To assess detection limits, a blank was associated with each batch (one per course) and sample preparation duplicates were made in order to assess the precision of this step. Certified reference materials were not analysed as their composition is determined after full digestion only. Therefore bias errors could not be assessed.
Samples were analysed by inductively coupled plasma–atomic emission spectrometry (ICP-AES; Perkin Elmer Optima 3300R2 ICP-AES unit, Perkin Elmer AS-91 autosampler and Winlab 32 Software) for lead, cadmium and copper, and by inductively coupled plasma–mass spectrometry (ICP-MS; Perkin Elmer SCIEX Elan 5000 ICP-MS unit (Perkin Elmer, Boston, MA, USA) and Winlab 32 Software) for arsenic. Drift was monitored during ICP-AES by running a standard composed of most major and trace elements found in soils, for every tenth sample that was analysed. The samples obtained from the initial sample preparation and used for ICP-AES analysis were diluted a further 50-fold and analysed by ICP-MS. Drift was monitored by running a standard solution containing 10 p.p.b. arsenic. A classic standard calibration approach was used for both techniques, and some samples were analysed twice to assess instrumental precision.
The remaining subsamples were stored at −20 °C prior to analysis for the organic fungicides chlorothalonil, iprodione and fenarimol. Redistilled ethyl acetate was used as the extracting solvent, following Fucci, Ciaravolo & Mazza (1995). The frozen soil samples were placed on trays and left to defrost and dry at room temperature, until their mass remained unchanged. Ten grams of soil from each were tipped into a flask and extracted twice with 8 mL of ethyl acetate (5 min in an ultrasonic bath). The extracts were combined after filtration through a medium-grade filter paper (Whatman grade 2). The solvent was then evaporated until dryness under a stream of nitrogen, and 1 mL ethyl acetate was added back into the flask. After agitation, the extract was ready to be analysed.
Recoveries were assessed by spiking each sample with a mixture of 0·1 mg mL−1 chlorothalonil, iprodione and fenarimol. These were analysed and the fungicide peak integrations (minus those for unspiked samples) were matched against the peak integrations for the 0·1 mg mL−1 standards. Results within ± 10% of the mean recovery for one site were accepted, while others were discarded and replaced by spiking and analysing another similar subsample.
Stock solutions of 1 mg ml−1 were prepared by weighing 50 mg each of pure fungicides and diluting to volume in 50-mL volumetric flasks with ethyl acetate (purity > 99%). New stock solutions were prepared every 3 weeks. Solutions containing 0·1 mg mL−1 were regularly made up from the stock solutions by dilution with ethyl acetate. The latter were analysed by GC-MSD (SCAN mode). All the compounds were identified on the chromatograms.
A Hewlett Packard 5890 Ga Chromatograph (GC) coupled with Hewlett Packard 5970 MSD (Agilent Technologies UK Ltd, Stockport, Cheshire, UK) was used for analysis and the GC column was of medium polarity (SE-54, 30 m, 0·25 mm internal diameter). The GC temperature programme (initial temperature 60 °C for 5 min, gradient of 10 °C min−1, final temperature of 300 °C for 5 min) remained the same over the whole period of analysis. Injection of 2 µL was manual (split-less for 1–5 min). Under SCAN mode, the range of mass detection was scanned from 50 to 350 atomic mass units (amu).
Once the chromatogram peaks of chlorothalonil, iprodione and fenarimol and their related mass spectra were clearly defined, a single ion monitoring (SIM) programme was edited by choosing three specific amu values for each compounds. This technique dramatically improves detection and, when dealing with real samples, the signal is much less likely to suffer interferences from other compounds (from the matrix or contamination). Table 1 shows the retention times as well as the selected amu values for each compound analysed.
Table 1. Organic fungicides retention times and selected atomic mass units (amu) for analysis by GC-MS
Retention time (min)
A classical calibration was performed by analysing 100, 10, 1 and 0·1 µg mL−1 chlorothalonil, iprodione and fenarimol and mixed solutions in the GC-MSD. A linear regression between peak areas and concentrations (both log-transformed) gave the equation of the calibration curve, which was accepted when the regression coefficient was higher than 0·985. The calibration curve was regularly checked by running a 10-µg mL−1 standard mix solution.
arbuscular mycorrhizal colonization
To obtain AM colonization data, all 18 greens from all four courses were sampled on four occasions (March, June, September and December) in 1999 and 2000. On each sampling occasion, three cores, measuring 2·5 cm diameter × 5 cm deep, were removed from random positions on each green. Roots of the dominant grass P. annua were removed from each and washed free of soil. Slide preparations of roots were examined at ×200 using a Zeiss Axiophott epifluorescence microscope (Carl Zeiss MicroImaging Inc., Thornwood, NY, USA) fitted with a UV lamp and filters giving a transmission of 455–490 nm blue, to reveal arbuscules (Gange et al. 1999), which were recorded with the cross-hair eye-piece method of McGonigle et al. (1990). At least 200 intersections were recorded per slide, to give a measure of percentage root length colonized (% RLC). Mean mycorrhizal levels per green were calculated as the average of all 24 (three cores × four dates × two years) values obtained. Such a calculation takes into account inter- and intraseasonal variation and is as accurate a measure of AM occurrence in each green as is possible.
A randomized block experiment, consisting of 40 × 1 m2 plots, each separated by 2 m, was laid out on a practice putting green at course C. The sward comprised 83%P. annua, 14%Agrostis stolonifera L. and 3%Festuca spp. Plots were allocated at random within blocks to one of four treatments: (i) control (no fungicide applied); (ii) application of chlorothalonil; (iii) application of iprodione; (iv) application of fenarimol. Each fungicide was diluted with an appropriate volume of water to produce application rates identical to those recommended for sports turf. These were chlorothalonil 15000 g a.i. ha−1, iprodione 5000 g a.i. ha−1 and fenarimol 780 g a.i. ha−1. Each plot received 50 mL of diluted product, equivalent to the application rate of 500 L ha−1 recommended for each. The first application took place on 2 July 1999, with a second application on 12 August. The green had not received any fungicide in the 6 months prior to the experiment and no other compounds were applied during the time of study. Levels of the three compounds used were below detectable limits in the soil.
Immediately before the first application, three 2·5-cm diameter × 5-cm deep soil cores were taken from each plot. The hole left behind was filled with sterilized top dressing (a mixture of 90% sand and 10% loam soil). Roots of P. annua were extracted from each core, washed free of soil, and arbuscules within them were revealed by autofluorescence microscopy (Gange et al. 1999). Percentage root length colonized by arbuscules was again recorded with the cross-hair eye-piece method of McGonigle et al. (1990), on at least 200 intersections per slide. Root samples were taken on 12 September, 15 November and 10 January 2000.
Differences between golf courses in metal and organic fungicide concentrations were examined with one-factor anova, after checking for normality and homogeneity of variances. Relations between AM fungal colonization of P. annua and each chemical were examined with linear regression. In all cases, mean values per green were used as replicates. The effects of fungicides on AM colonization in the manipulative experiment were examined with a repeated-measures anova, employing block, fungicides and dates as main effects. In this case, data were subject to the angular transformation to meet the assumptions of normality (Zar 1996).
fungicide application patterns
The proportion of all fungicides used at each course represented by chlorothalonil, iprodione and fenarimol varied between courses (Fig. 1a). At course C, these three compounds comprised 95% of all applications, while at course D this figure was 71%. Furthermore, the proportion of each chemical varied, with the most dramatic difference seen with chlorothalonil. On course A, 42% of all applications were chlorothalonil, while the comparable figure for course D was only 4·5%. Carbendazim (mainly used for earthworm control) was the most common constituent of the ‘others’ category but was never more than 8% of total usage at any course.
There were significant differences between courses in the mean number of applications per annum of chlorothalonil and iprodione (Fig. 1b). For the former compound, course D applied significantly less than both courses A and B (F3,25 = 6·7, P < 0·01), while for the latter there was a significant difference between courses A and C (F3,25 = 3·1, P < 0·05). No significant differences were found in the frequency of application of fenarimol between courses (Fig. 1b).
element and organic fungicide analysis
The detection limits (limits of determination) for the four elements in this study were 40·5 µg Cu kg−1 soil, 4·11 µg Cd kg−1, 65·5 µg Pb kg−1 and 67 µg As kg−1. There were significant differences between courses in the levels of copper (F3,68 = 26·6, P < 0·001), cadmium (F3,68 = 33·9, P < 0·001), lead (F3,68 = 12·1, P < 0·001) and arsenic (F3,68 = 7·2, P < 0·001) (Fig. 2). For the latter three elements, the newly constructed course D had much lower levels than the other three courses. However, this did not occur with copper, where levels in course D were similar to those in the older course A (Fig. 2a).
In the majority of greens, the levels of all elements were very low. However, in courses A, B and C, the maximum level of lead recorded (29·9 mg kg−1, 25·9 mg kg−1 and 70·3 mg kg−1) exceeded the UK mean of 20 mg kg−1. Course C had particularly high levels of lead in some greens, leading to the highest mean value overall (Fig. 2c), although this figure was not significantly different from the mean of course B.
There were highly significant positive correlations between levels of lead and arsenic (all P < 0·001) in courses A, B and C, suggesting that the source of these elements was lead arsenate (PbHAsO4). In courses A and B there were also significant but weak correlations (P < 0·05) between copper and arsenic levels, suggesting that application of Paris Green (copper acetoarsenite) was responsible for the levels of these elements. The background element levels on course A were higher than the average levels on the same course. This was probably because soil used to repair greens was used to represent the background. It is not inconceivable that this soil was once part of an old green on which inorganic based mixtures were applied.
The detection limits for the three fungicides were 1·88 µg chlorothalonil kg−1 soil, 242·81 µg iprodione kg−1 and 27·43 µg fenarimol kg−1. Highly significant differences were found in chlorothalonil (F3,68 = 4·2, P < 0·01), iprodione (F3,68 = 5·2, P < 0·01) and fenarimol (F3,68 = 64·6, P < 0·001) levels between courses (Fig. 3). Course B had the highest levels of chlorothalonil and iprodione (Fig. 3a,b), while course D had much higher levels of fenarimol than the other courses (Fig. 3c). These data were only a weak reflection of fungicide applications (Fig. 1b), for course B only differed from D in the application of chlorothalonil while there were no differences in applications of fenarimol.
pesticides and arbuscular mycorrhizas
The mean levels of %RLC (± 1 SE) of AM colonization were: course A, 7·8 ± 1·1; course B, 11·4 ± 1·2; course C, 13·5 ± 0·8; course D, 18·2 ± 2·2. A summary of the relations between AM fungal colonization and element concentrations is given in Table 2. Only one of the 16 regressions was significant, where a negative relation was found between cadmium and AM fungi in course C. Meanwhile, a similar picture was evident in the relations between organic fungicide levels per green and AM colonization (Table 2). Only two significant negative relations were found, for chlorothalonil in course C and fenarimol in course A. Therefore, there was little evidence that AM levels were lower in greens where element or organic fungicide levels were relatively high.
Table 2. Summary of results from linear regression analyses, testing for relationships between element and organic fungicide levels and arbuscular mycorrhizal colonization in putting greens at four golf courses. In all cases, n= 18
Figure 4 depicts the levels of arbuscular colonization of P. annua in the manipulative experiment. Natural levels of colonization were about 8% of the root system colonized in early July. In control plots, there was an increase in this figure during autumn, rising to about 12%. Levels then decreased during the autumn and into winter. No significant effects of any fungicide could be found over the course of this experiment (F3,36 = 2·27, P > 0·05). Root biomass of P. annua and A. stolonifera showed a very similar pattern to that of AM colonization (data not shown) and there was no effect of fungicide on root production in either grass species.
AM fungi have great potential in aiding the establishment and growth of perennial grasses in golf putting greens (Gemma et al. 1997). Furthermore, evidence suggests that the abundance of these fungi in fine turf is negatively related to that of P. annua (Gange 1998) and that addition of AM inoculum to an established green may result in a decrease in the abundance of this weed (Gange, Lindsay & Ellis 1999). If this biological management strategy is to be successful, it is important that AM fungi form part of an integrated programme of turf management in which their presence is compatible with the array of chemicals currently in use. Studies have shown that AM colonization levels of grasses in putting greens are considerably lower than in more natural situations (Koske, Gemma & Jackson 1997a; Gange, Lindsay & Ellis 1999), although a surprising number of species have been found in sports turf (Koske, Gemma & Jackson 1997b). The results from the current study indicate that fungicidal chemicals do not appear to have an adverse effect on levels of AM colonization in fine turf.
In the past, lead arsenate and copper acetoarsenite (Paris Green) were widely used as insecticides on turf, while cadmium chloride and various salts of mercury and copper were used to control fungi (Greenfield 1962). Lead arsenate was probably used on turf until the late 1960s, and the immobility of lead in the soil, coupled with the generally undisturbed nature of putting green soils, led us to hypothesize that levels of lead and arsenic would be relatively high in the older putting greens. Although there were positive correlations between lead and arsenic and between copper and arsenic levels in the older courses, indicating the source of these elements, we found that levels of As, Cd, Cu and Pb were only a little above ambient levels reported elsewhere for UK soils (Lepp 1981). The metal extraction procedure used in this work probably leads to a lower metal recovery (especially for non-volatile copper) than from a more established total digestion method such as hydrofluoric/perchloric acid. The latter procedure is more likely to be used for measuring non-volatile element concentrations, although some guidance methods for the determination of elements (including non-volatiles) in environmental samples involve weaker digestions, such as 20% HNO3 or aqua-regia (US Environmental Protection Agency 600/R-94-111 method). However, none of the levels of any of the four elements in any green were likely to result in phytotoxicity (Carbonell et al. 1998; Das, Samantaray & Rout 1997; Aksoy, Hale & Dixon 1999).
Our data show that there is no evidence to suggest that levels of heavy metals in putting green soils from past fungicide applications are likely to account for the low abundance of AM fungi. However, if metal-tolerant fungi are a feature of golf greens, future work involving the inoculation of greens with AM fungi must use isolates that have been obtained from turf grass. Indeed, this is the only ecologically realistic approach to AM manipulation (Read 2002) and may explain why a previous experiment, in which non-turf grass isolates were used, was only a partial success (Gange, Lindsay & Ellis 1999).
The three most commonly applied fungicides in this study were fenarimol, chlorothalonil and iprodione. Among the ‘others’ category, carbendazim (for earthworm control) was the most prevalent but only contributed to 8% of the total fungicide use at maximum. This chemical can reduce AM root colonization (Venedikian et al. 1999), but there are also studies where it has been shown to be ineffective (Schweiger, Spliid & Jakobsen 2001). We consider that in no case was application of carbendazim or other compounds frequent enough to have been a likely confounding factor in any relationship between AM levels and fungicides. Chlorothalonil and iprodione have been used in a number of ecological experiments, where they have been found to be effective in reducing the abundance of AM fungi (Gange, Brown & Farmer 1990; Aziz, Habte & Yuen 1991; Wan, Rahe & Watts 1998). We found very little evidence that AM fungal abundance was related to the levels of any fungicide in the soil and, overall, levels of the three fungicides were very low. This may be because the root zone of a putting green is generally free-draining and thus chemicals are rapidly leached through it. However, previous studies indicate that this is unlikely to be a common occurrence (Cohen et al. 1999). Instead, it is more likely that the chemicals are degraded by microbial action (Mercadier, Vega & Bastide 1997; Armbrust 2001) and the most probable site where this occurs is the thatch layer in a golf green (Sigler et al. 2000). This layer of undecomposed plant material occurs at the soil surface and in some cases can be several centimetres deep. A moderate thatch layer is required, to improve ball bounce and putting quality (Perris & Evans 1996), but too much is regarded as problematic, as it prevents the ingress of fertilizers or even water to the root zone and encourages disease outbreaks.
Generally, there was little indication that frequency of pesticide application was related to the amount of chemicals in the root zone, providing further evidence that chemicals may be entrapped and degraded within the thatch layer before they reach the roots. These data show the difference in pesticide usage by course managers, and it was encouraging to see that all four courses used a variety of chemicals, in an attempt to lessen the chances of resistance to any one chemical occurring. Concentrations of fenarimol were highest in the newest course, in which the root zone was composed predominantly of sand rather than soil. It is known that retention of fenarimol in a soil is positively related to the cation exchange capacity (CEC) (Wehtje, Walker & Shaw 2000) and this may be why levels were higher in this course, as inorganic products designed to increase the CEC had been added to the root zone.
The experiment failed to show any effects of chlorothalonil, iprodione or fenarimol on AM colonization levels. It was conducted over a 6-month period, which should have been sufficient for any detrimental effect of iprodione or chlorothalonil on root colonization levels to become apparent (Gange, Brown & Farmer 1990; Aziz, Habte & Yuen 1991). Given the efficacy of these chemicals in reducing AM levels in other experiments (above), this is perhaps a surprising result. It is possible that AM fungi in putting greens are resistant to fungicides, although to our knowledge such a phenomenon has never been recorded. Indeed, it seems unlikely for fenarimol, which is a relatively recent addition to the pesticide market. However, this is not the first time that fungicides have failed to reduce AM abundance. Sukarno, Smith & Scott (1993) have attributed this to the way in which colonization levels have been assessed. For example, they have shown that some fungicides affect root production, thus biasing results, while in other cases the amount of living fungal structures may be reduced, a fact not revealed by conventional staining methods. In our study we could detect no effect of fungicides on root production of either P. annua or A. stolonifera, and chose to record arbuscules only. As these are relatively transient structures, we believe that if the fungicide had a detrimental effect on the symbiosis it would have been revealed by this method over the course of a 6-month period. Instead, we suggest that the degradation of chemicals within the thatch layer meant that insufficient pesticide reached the roots and thus had no effect on AM colonization (Negre et al. 1997).
The data reported here do not provide any evidence that toxic chemicals are the reason for the relatively low levels of AM fungi in putting green soils, and we suggest two alternative explanations. The first concerns soil P levels, which can be high in some putting greens (Baker, Binns & Cook 1997). Indeed, it is well known that AM fungi tend to become less abundant and less functional at high soil P (Demiranda & Harris 1994). However, if this is so, it still does not explain why such a relatively high diversity of AM species has been recorded in turf grass soils (Koske, Gemma & Jackson 1997b). There has been no study of the occurrence of AM fungi in fine turf in relation to soil P status and we suggest that this is an important area that needs to be addressed.
The second feature of fine turf is the intensity of mowing throughout the year, which in a golf green results in a height of about 4·5 mm for most of the growing season (Perris & Evans 1996). Although occasional mowing may have little effect on AM colonization (Eom et al. 1999; Smilauer 2001), such an intensity of foliage removal is certain to affect the capacity of the plant to direct carbon to the mycorrhiza (Jakobsen, Smith & Smith 2002). Therefore, in any future manipulation of AM populations in turf, factors that affect the carbon economy of the plant will also need to be considered.
We thank the Centre for Chemical Sciences and the Geology Department at Royal Holloway (in particular Professor Mike Cooke and Dr Sarah James) for providing help and instruments for the chemical analysis. We are grateful to the four course managers for allowing us access to their courses and to Dow AgroSciences for funding the manipulative part of this study. K. Hagley was funded by the Natural Environment Research Council.