2002. The Journal of Arachnology 30:288±297

DO INCREMENTAL INCREASES OF THE HERBICIDE GLYPHOSATE HAVE INDIRECT CONSEQUENCES FOR COMMUNITIES?

James R. Bell: School of Life Sciences, University of Surrey Roehampton, Whitelands College, West Hill, London SW15 3SN, UK

Alison J. Haughton1: Crop and Environment Research Centre, Harper Adams University College, Newport, Shropshire, TF10 8NB, UK

Nigel D. Boatman2: Allerton Research and Educational Trust, Loddington House, Loddington, Leicestershire, LE7 9XE, UK Andrew Wilcox: Crop and Environment Research Centre, Harper Adams University College, Newport, Shropshire, TF10 8NB, UK

ABSTRACT. We examined the indirect effect of the herbicide glyphosate on ®eld margin spider com- munities. Glyphosate was applied to two replicated (n ϭ 8 per treatment) randomized ®eld experiments over two years in 1997±1998. were sampled using a modi®ed garden vac monthly from May± October in the following treatments: 1997 comprised 90g, 180g, & 360g active ingredient (a.i.) glyphosate haϪ1 treatments and an unsprayed control; 1998 comprised 360g, 720g and 1440g a.i. glyphosate haϪ1 treatments and an unsprayed control. We examined the indirect effect of glyphosate on the spider com- munity using DECORANA (DCA), an indirect form of gradient analysis. We subjected DCA-derived Euclidean distances (one a measure of beta diversity and the other a measure of variability), to the scrutiny of a repeated measures ANOVA design. We found that species turnover and cluster variation did not differ signi®cantly between treatments. We attribute the lack of any effect to a large number of common agricultural species which are never eliminated from a habitat, but are instead signi®cantly reduced. Re- duction rather than elimination does not cause the spider communities within these plots to turn over any faster than the control. However, like most other communities, the spider community did turn over and change in structure and composition through the season, regardless of treatment. Using Spearman rank correlations, we found that this within-season species turnover is related to the decline in vegetation height and the increase in percentage dead vegetation cover in the ®eld margin. Keywords: Glyphosate, herbicide, spiders, species turnover, DECORANA, ®eld margins

Field margins play an important agricultural 1991; Feber et al. 1996; Haughton et al. role in providing a refuge for bene®cial in- 1999a,b). This detrimental effect on non-tar- vertebrate predators (e.g., Araneae, some Car- get invertebrates following an experimental abidae; Staphylinidae; Heteroptera) facilitat- spray application of herbicide simulates what ing movements of invertebrates into the crop may happen when spray is allowed to drift (e.g., Duelli et al. 1990). Data from herbicide- onto non-target areas. Drift may be a product treated and untreated cereal headlands suggest of operator error or a sudden increase in wind that butter¯ies, carabid beetles, Auchenorrhy- speed and in either of these situations, the cha and Heteroptera (Hemiptera) are detri- spray is likely to become misplaced and affect mentally affected by spray applications, but to ®eld margins and adjacent semi-natural habi- differing degrees (e.g., Chiverton & Sotherton tats. Recent evidence suggests that spider num- 1 Current address: Department of Entomology and Nematology, IACR-Rothamsted, Harpenden, Hert- bers are likely to be signi®cantly affected even fordshire. AL5 2JQ, UK. by a relatively low rate of herbicide applica- 2 Current address: Central Science Laboratory, tion (Baines et al. 1998; Haughton et al. Sand Hutton,York, YO41 1LZ, UK. 1999c). This effect, even at low rates, may

288 BELL ET AL.ÐTHE EFFECT OF GLYPHOSATE ON SPIDER COMMUNITIES 289 have detrimental implications for both ®eld Couch-grass (Elymus repens (L.)). Eight rep- margin biodiversity and bio-control programs licates of four treatments, 90 g, 180 g and 360 which use bene®cial spiders to control pests. g a.i. glyphosate haϪ1 (Roundup Biactive, In the context of a rapid increase in worldwide Monsanto, High Wycombe, Berkshire) and an demand for glyphosate when, for example, 70 unsprayed control, were randomly assigned million hectares of land worldwide were along the ®eld margin. Glyphosate was ap- sprayed in 1997 (Woodburn 2000), it is easy plied to these plots at a volume rate of 200 to demonstrate a need for a clearer under- litres haϪ1 and a pressure of 2.5 bar using an standing of the potential effect on spiders and Oxford Precision Sprayer ®tted with ¯at fan other non-target invertebrates. nozzles on 30.v.1997, during dry and calm The properties of glyphosate (N-(phosphon- conditions (24 ЊC, RH 55%, wind speed Ͻ1 omethyl)glycine), an exceedingly effective but msϪ1). slow-acting broad spectrum herbicide, are The 1998 margin was dominated by False well known. The herbicide acts by inhibiting Oat-grass (A. elatius (L.)) and Yorkshire Fog the biochemical pathway in plants (i.e., 5-en- (Holcus lanatus (L.)). Eight replicates of four olpyruvylshikimate 3-phosphate synthase) by treatments were assigned in a randomized blocking the production of aromatic amino ac- block design: 360 g, 720 g and 1440 g a.i. ids thus restricting protein synthesis and pho- glyphosate haϪ1 (Roundup Biactive, Monsan- tosynthetic activity (Bayliss 2000). It is used to, High Wycombe, Berkshire) suspended in in a range of agricultural (particularly cereals), water and an unsprayed control. Glyphosate industrial and domestic situations to control was applied to the plots on the 4.vi.1998 dur- grasses and other weeds (Woodburn 2000). ing dry, calm conditions (17.5 ЊC, RH 84.5%, However, the toxicological effect on inverte- wind speed Ͻ2msϪ1) with the same equip- brates is little understood except to state that ment, volume rate and pressure as that de- the herbicide is classi®ed as `harmless' to spi- scribed in 1997. ders and a range of other common non-target Data collection.ÐSpiders were sampled invertebrates by The Society of Environmen- using a modi®ed garden-vac (g-vac) (Ryobi tal Toxicology and Chemistry (Europe) (Bar- RSV3100E: engine capacity 31 cm3 with a rett et al. 1994) nozzle size 13 cm). The spider samples from In this paper we investigated the within- each experimental plot comprised 10 sub-sam- season effect of incremental levels of gly- ples of 30 second sucks at 1 m intervals along phosate application (six levels from 90±1440 each experimental plot. Sampling was done on g active ingredient (a.i) glyphosate haϪ1 with the central 10 m of each plot to avoid edge a control) on ®eld margin spiders from a com- effects from neighboring treatments: the total munity perspective. sampling area per plot approximated to 0.13 m2. Each sample of spiders was emptied from METHODS the g-vac into a plastic bag, extracted with an Study sites.ÐThe ®eld research was based aspirator into 70% alcohol and then identi®ed at the Allerton Research and Educational Trust to species level. Estate, Loddington, Leicestershire, UK (grid Two measures of vegetation structure were reference: SK 789015). We spread the effort taken. Percentage ground cover of dead veg- of conducting such a large experiment over etation in the experimental plots was recorded two years between 1997±1998. Two ®eld mar- using permanent 0.25 m2 quadrats and, aver- gin study sites were used which were sepa- age vegetation height at ®ve positions (as in rated by a minor road but were no more than a domino-5) within the quadrats was recorded 20 m apart. Both sites had an adjacent dense to the nearest cm. Three quadrats were posi- uncut Hawthorn (Crataegus monogyna Jacq.) tioned at 3 m, 6 m and 9 m within each of and Blackthorn (Prunus spinosa L.) hedge the 32 plots and mean percentage dead vege- running along their length. The experimental tation cover and mean vegetation height. Spi- plot size was the same in both years: ®eld der and vegetation sampling was done month- margins were divided into 32 contiguous ly between May±October inclusive, to plots, each measuring 12 m long ϫ 2 m wide. monitor any changes over the season. The 1997 margin was dominated by False Data analyses.ÐThere are many ways of Oat-grass (Arrhenatherum elatius (L.)) and calculating species turnover, but most index- 290 THE JOURNAL OF ARACHNOLOGY type methods which have traditionally been timately, these Euclidean distances indicate used do not take account of the community the level of species stability (i.e., high or low dynamics as they fail to retain all the infor- rates of turnover) between treatments. mation from the species matrix. However, We then calculated the average Euclidean multivariate statistics have been suggested as distance between all points within the same an inclusive technique which does not over- treatment for each of the months from May± simplify the dynamic nature of the community October inclusive (n ϭ 672) using the DCA (Williamson 1987). We introduce a simple axis 1 and 2 scores for each year (i.e., we method of calculating species turnover while generated a distance matrix for each treatment retaining most of the ecological information for each month). This approach would indi- from the data matrix. Using DECORANA cate whether the size of the cluster varied be- (DEtrended CORrespondence ANAlysis or tween treatment. The size of the cluster DCA for short), a widely available package equates to a measure of community variation; designed speci®cally for ecological studies to large clusters have more variation in the spe- avoid distortions caused by either the arch or cies complex than small ones. horseshoe effects (Hill 1994), P values are at- Univariate repeated measures, with date as tached to the turnover rates in DCA-Euclidean the within-subject factor and treatment as the space using ANOVAs. This is an extension of main effect, were used to analyze differences Hill's (1994) approximation of the standard in species turnover (i.e., distance moved in deviation of species turnover along the x-axis. DCA-Euclidean space) and cluster distance Species turnover, which is sometimes referred throughout the season in each year. As a pre- to as beta (␤) diversity, is used here as mean- requisite to using a repeated measures ANO- ing ``the change in the composition of a bio- VA, we logged (x ϩ 1) the data and tested it logical community as a result of either or both for sphericity using Mauchly's W test. Where the immigration and local extinction of spe- signi®cant differences were found between cies'' (Russell et al. 1995). treatments, a Tukey post hoc test was used to First, two binary matrices were created establish the location of the difference. We (1997: 58 ϫ 192; 1998: 59 ϫ 192) which rep- also used Monte Carlo randomization tests to resented the presence/absence of species with- establish whether the null hypothesis, that a in each year. These were analyzed separately pattern is present as purely a chance effect of using DCA (default settings with 26 segments observations in a random order, should be re- used to remove the arch effect), which arrang- jected (Manley 1991). Monte Carlo randomi- es points along the axes on the basis of species composition data. Thus, the further two points zations are useful for veri®ying the signi®- are from one another in the DCA ordinational cance level in an ANOVA design when some space, the more dissimilar the spider com- of the statistical assumptions (i.e., indepen- munities. Once the DCA had produced two dence) may be in question. The level of sig- biplots (1997 & 1998), the axes 1 and 2 scores ni®cance in a Monte Carlo test is expressed were separated into herbicide treatment (i.e., as the percentage of values which are equal control, 90 g, 180 g etc.). Using the DCA axis to, or higher than can be found in a random- 1 and 2 scores, the Euclidean distance for each ized distribution. If the percentage of values replicate within each treatment was calculated that exceed the observed mean square is less for consecutive between-month shifts to estab- than 5%, then this suggests that the null hy- lish the rate of species turnover. For example, pothesis should be rejected. We used 30,000 in 1997 for the ®rst replicate in the 90 g a.i. Monte Carlo randomizations to test this null glyphosate haϪ1 treatment, the following Eu- hypothesis on signi®cant ANOVA test results clidean distances were calculated from axis 1 to check for their validity. and 2 scores: May→June; June→July; In order to interpret the importance to the July→August; August→September; Septem- community of the changes in the ber→October. Thus, there were ®ve measure- vegetation caused by the rate of glyphosate, ments for each of the eight replicates in each Spearman's rank correlation was used to test of the four treatments (n ϭ 160) for both for the strength of association between the years. These distances indicate the seasonal axis scores and vegetation height and per- species turnover in n dimensional space. Ul- centage dead vegetation cover. BELL ET AL.ÐTHE EFFECT OF GLYPHOSATE ON SPIDER COMMUNITIES 291

Table 1.ÐAraneae species recorded from the ®eld margins at Loddington.

Oonopidae Araneidae Oonops domesticus de Dalmas 1916 Lariniodes cornutus (Clerck 1757) Gnaphosidae Araniella opistographa (Kulczynski 19505) Micaria pulicaria (Sundevall 1832) Clubionidae Ceratinella brevipes (Westring 1851) Ceratinella scabrosa (Cambridge 1871) Clubiona reclusa Cambridge 1863 Walckenaeria acuminata Blackwall 1833 Clubiona lutescens Westring 1851 Walckenaeria nudipalpis (Westring 1851) Clubiona compta Koch 1839 Walckenaeria unicornis Cambridge 1861 Zoridae Walckenaeria cuspidata (Blackwall 1833) Zora spinimana (Sundevall 1833) Dicymbium nigrum (Blackwall 1834) Entelecara erythropus (Westring 1851) Thomisidae Dismodicus bifrons (Blackwall 1841) Xysticus cristatus (Clerck 1757) Gonatium rubens (Blackwall 1833) Ozyptila praticola (Koch 1837) Maso sundevalli (Westring 1851) Philodromidae Pocadicnemis juncea Locket & Millidge 1953 Philodromus dispar Walckenaer 1826 Oedothorax fuscus (Blackwall 1834) Philodromus cespitum Walckenaer 1802 Oedothorax retusus (Westring 1851) Philodromus collinus Koch 1835 Cnephalocotes obscurus (Blackwall 1834) Tibellus oblongus Walckenaer 1802 Monocephalus fuscipes (Blackwall 1836) Gongylidiellum vivum (Cambridge 1875) Salticidae Micrargus herbigradus (Blackwall 1854) Euophrys frontalis (Walckenaer 1802) Micrargus subaequalis (Westring 1851) Lycosidae Erigonella hiemalis (Blackwall 1841) Pardosa palustris (L. 1758) Savignya frontata (Blackwall 1833) Pardosa pullata (Clerck 1757) Diplocephalus latiforns (Cambridge 1863) Pardosa prativaga (Koch 1870) Diplocephalus connatus Bertkau 1889 Pardosa amentata (Clerck 1757) Araeoncus humilis (Blackwall 1841) Pardosa nigriceps (Thorell 1856) Panamomops sulciforns (Wider 1834) Alopecosa pulverulenta (Clerck 1757) Erigone dentipalpis (Wider 1834) Trochosa ruricola (Degeer 1778) Erigone atra (Blackwall 1833) Trochosa terricola Thorell 1856 Porrhomma microphthalmum (Cambridge 1871) Meioneta rurestris (Koch 1836) Pisauridae Meioneta saxatilis (Blackwall 1844) Pisaura mirabilis (Clerck 1757) Syedra gracilis (Menge 1869) Mimetidae Centromerus sylvaticus (Blackwall 1841) (Blackwall 1833) cambridgei Kulczynski 1911 Bathyphantes gracilis (Blackwall 1841) Ero furcata (Villers 1789) Bathyphantes parvulus (Westring 1851) Theridiidae Diplostyla concolor (Wider 1834) Episinus angulatus (Blackwall 1836) Poeciloneta globosa (Wider 1841) Theridion bimaculatum (L. 1767) Stemonyphantes lineatus (L. 1758) Enoplognatha ovata (Clerck 1757) Lephthyphantes tenuis (Blackwall 1852) Robertus lividus (Blackwall 1836) Lepthyphantes mengei Kulczynski 1887 Pholcomma gibbum (Westring 1851) Lepthyphantes ericaeus (Blackwall 1853) Lepthyphantes pallidus (Cambridge 1871) Tetragnathidae Lepthyphantes insignis Cambridge 1913 Tetragnatha extensa (L. 1758) Neriene clathrata (Sundevall 1830) Tetragnatha montana Simon 1874 Microlinyphia pusilla (Sundevall 1830) Pachygnatha clercki Sundevall 1823 Pachygnatha degeeri Sundevall 1830 Meta segmentata (Clerck 1757) Meta mengei (Blackwall 1869) 292 THE JOURNAL OF ARACHNOLOGY

Figure 1.ÐDCA biplot of 1997 ®eld margin spider samples. (Note: plots indicating symbols as treat- ments rather than months are not shown because no clear separation can be made between different herbicide rates).

RESULTS ain. Between 58 and 59 species were recorded The spider community.ÐIn both 1997 in total in 1997 and 1998 respectively, 44 of and 1998, the adult spider community was which were recorded in less than 10% of the dominated by linyphiids (Table 1). The three total number of samples for both years, 28 of species that occured most frequently (percent- which were singletons. The total number of age of all the samples in 1997 and 1998 re- spiders recorded in the study was 46,393. spectively) were Lepthyphantes ericaeus Voucher specimens were deposited with Lec- (Blackwall 1853) (77.1%; 73.9%), Lepthy- eister Museum through John Crocker, the phantes tenuis (Blackwall 1852) (69.3%; county recorder for Leicestershire. 84.3%) and Bathyphantes gracilis (Blackwall DCA biplots.ÐNo separation along axis 1 1841) (49.5%; 43.2%), all of which are con- or 2 is apparent on either of the 1997 or 1998 sidered common ®eld margin spiders in Brit- biplots (Figs. 1, 2). Instead, all samples form BELL ET AL.ÐTHE EFFECT OF GLYPHOSATE ON SPIDER COMMUNITIES 293

Figure 2.ÐDCA biplot of 1998 ®eld margin spider samples. (Note: plots indicating symbols as treat- ments rather than months are not shown because no clear separation can be made between different herbicide rates). a homogenous cluster, with a tendency for on axis 2 where, apart from a lack of a sig- samples within the same month to be aligned. ni®cant result in 1997 with vegetation height, In the 1997 biplot (Fig. 1), axis 1 scores de- axis 2 is negatively correlated with vegetation cline as samples move in ordinational space height, but positively correlated with percent- through the season. This effect is less discern- age dead vegetation cover in both years. able in the 1998 biplot (Fig. 2), although sam- DCA-Euclidean species turnover.ÐSpe- ples within each month do aggregate to some cies turnover analysis revealed a signi®cant extent. The relationship between axis 1 scores treatment effect in 1997 (F3,28 3.63 P ϭ 0.024) and vegetation height and percentage dead but no such effect in 1998 (F3,28 0.24 P ϭ vegetation cover indicates signi®cant correla- 0.867). Sign®cant test results were compared tions (Table 2); axis 1 is positively correlated with 30,000 Monte Carlo randomizations and with vegetation height, but negatively corre- their respective observed mean squares. For lated with percentage dead vegetation cover in the 1997 ANOVA result (i.e., F3,28 3.63 P ϭ both years. These relationships are reversed 0.024), only 0.060% of the randomizations ex- 294 THE JOURNAL OF ARACHNOLOGY

Table 2.ÐSpearman rank correlations between vegetation characteristics and DCA axis scores for ®eld margin spiders.

Vegetation height Percentage dead vegetation cover

Year Axis rs P rs P 1997 1 0.434 Ͻ0.001 Ϫ0.452 Ͻ0.001 2 Ϫ0.753 0.299 0.185 0.010 1998 1 0.477 Ͻ0.001 Ϫ0.314 Ͻ0.001 2 Ϫ0.183 0.011 0.218 0.002 ceeded the observed mean squareÐstrong ev- this was not related to an incremental increase idence that observed test results could not be in the rate of glyphosate application. One ma- generated randomly. jor factor which in¯uenced the spider com- The differences between treatments in 1997 munity composition over the season was the were due to signi®cantly higher rates of turn- profound effect of differences in spider phe- over in the 90 g a.i. glyphosate treatments nology. There is often a marked difference be- when compared all others (Table 3). Overall, tween the composition of the spider commu- the general trend was for species turnover to nity in the spring compared to autumn. July decline slightly over the season in 1997 (Fig. and August are the transition months when 3). The trend in 1998 was much more variable spring species disappear (e.g., Pardosa spe- with treatments apparently acting indepen- cies) and autumn species emerge (e.g., Gon- dently of one another (Fig. 4). atium rubens (Blackwall 1833)). However, DCA-Euclidean cluster size.ÐCluster size most of the abundant spiders occurred did not differ signi®cantly over the season be- throughout the whole sampling period; either tween treatment for 1997 (F3,108 1.17 P ϭ low in number in spring and more abundant 0.327) or 1998 (F3,108 1.45 P ϭ 0.233). The in autumn (e.g., Bathyphantes gracilis (Black- general trend was that cluster size did get wall 1841)) or the reverse (e.g., Pocadicnemis smaller over the season in 1997 (Fig. 5) but juncea Locket & Millidge 1953). in 1998, it was more variable with size de- Overall, the response of the spider com- clining towards the middle of the summer and munity was, on ®rst inspection, very different increasing approaching the end (Fig. 6). from that of various single species responses DISCUSSION observed in other Loddington research. At rel- DCA biplots and related tests.ÐConsid- atively low rates of glyphosate application ering all the repeated measures ANOVAs, no (360 g a.i. haϪ1), a signi®cant decline in num- consistent monotonic relationship between bers of L. tenuis was detected (Haughton et glyphosate application rate and species turn- al. 2001a; Haughton et al. 1999c). Similarly, over or cluster size could be found. The only another linyphiid, G. rubens was reduced by signi®cant result that was detected in the 90 even lower rates of glyphosate application g a.i. glyphosate haϪ1 treatment, but this does (180 g a.i. haϪ1) and evidence from the same not compare with the effects measured at experiment suggests that total spider abun- higher treatments and thus must be considered dance, and the number of web-spinners are a rogue effect. Spider communities did change also detrimentally affected (Haughton et al. in many ways over the season (Figs. 1±6), but 1999c).

Table 3.ÐTukey P values for differences between mean DCA-Euclidean distances for ®eld margin spiders in the 90 g a.i. glyphosate haϪ1 treatment when compared with all other individual treatments.

90 g a.i. 180 g a.i. 360 g a.i. Control glyphosate haϪ1 glyphosate haϪ1 glyphosate haϪ1 Mean 73.32 141.10 92.80 93.62 P 0.002 Ð 0.037 0.040 BELL ET AL.ÐTHE EFFECT OF GLYPHOSATE ON SPIDER COMMUNITIES 295

Figure 3.ÐThe mean (n ϭ 160) seasonal spider Figure 5.ÐThe mean (n ϭ 672) seasonal decline species turnover between glyphosate treatments in in spider community DCA-Euclidean cluster size 1997 showing moderate variability within the sea- between glyphosate treatments in 1997 showing the son and the lack of a single trend. (Note: this graph trend of variable decline in cluster size through the should not be used to located differences between season. (Note: this graph should not be used to lo- treatments, please refer to statistics in Table 2). cated differences between treatments, please refer to statistics in the text).

The question arises then, why are the mea- sures of species turnover and cluster size not fects. The great majority of spiders from the showing a similar trend? The answer lies with Loddington species list (78%) have a propen- the way in which spiders respond, even to the sity to balloon, either in their immature phase highest rate of glyphosate application. As de- (e.g., Pardosa species) or throughout their tailed earlier from previous experiments at lives (i.e., most Linyphiidae and some Theri- Loddington, although spiders were shown to diidae and Tetragnathidae) (e.g., Duffey, be signi®cantly reduced, they were never 1956; Meijer, 1977). The ability to balloon completely eliminated (see Haughton et al. will undoubtedly have affected the rate at 1999c; Haughton et al. 2001a). This constant which the different spider communities turn species presence was apparent in the lack of over, causing the differences between treat- any signi®cant differences in either the spe- ments to be blurred by a constant aerial fallout cies turnover or cluster size results, suggesting or emmigration. that there was little variation in the composi- Exposure to herbicides reduces weed cover tion of the spider community, whatever the and diversity (e.g., de Snoo & van der Poll herbicide treatment. This lack of variation 1999), which in turn reduces the habitat qual- may indicate that an active aerial spider com- munity is ``blanketing out'' any treatment ef-

Figure 6.ÐThe mean (n ϭ 672) seasonal decline in spider community DCA-Euclidean cluster size Figure 4.ÐThe mean (n ϭ 160) seasonal spider between glyphosate treatments in 1998 showing the species turnover between glyphosate treatments in trend of variable decline in cluster size through the 1998 showing the large variability within the season season until August±October when clusters become and the lack of a single trend. (Note: this graph larger. (Note: this graph should not be used to lo- should not be used to located differences between cated differences between treatments, please refer treatments, please refer to statistics in the text). to statistics in the text). 296 THE JOURNAL OF ARACHNOLOGY ity for the spiders, and consequently their commodating the experiments on the estate. numbers decline. Recently, using the spider L. We also thank Paul Jepson (Oregon State Uni- tenuis, we demonstrated that at relatively low versity) for intial experimental advice. Fran- rates of glyphosate application in the ®eld cesca Tencalla (MonsantoÐBelgium) has (360 g a.i. haϪ1), a signi®cant decline in num- been extremely helpful throughout the period bers of L. tenuis was detected (Haughton et of this research. JRB would like to thank al. 2001a). The herbicide was not shown to Monsanto for funding conference trip travel act as an insecticide in direct toxicity tests, expenses. even at very high rates of glyphosate appli- cation (2160 g a.i. haϪ1), but indirectly LITERATURE CITED through habitat degradation: poisson regres- Baines, M., C. Hambler, P.J. Johnson, D.W. Mac- sion showed that numbers of L. tenuis were donald & H. Smith. 1998. The effects of arable related (non-linear) to vegetation height and ®eld margin management on the abundance and percentage dead vegetation cover (Haughton species richness of the Araneae (spiders). Ecog- et al. 2001a,b). The same two variables, veg- raphy 21:74±86. etation height and percentage dead vegetation Barrett, K.L., N. Grandy, E.C. Harrison, S. Hassan cover, were also found to be signi®cantly re- & P. Oomen. 1994. Guidance Document On Reg- lated to the spider community along axes 1 ulatory Testing Procedures For Pesticides With and 2 in this experiment (Table 2). It is un- Non-Target . Society of Environmen- deniable that vegetation structure has a pro- tal Toxiciology and Chemistry, Europe. found effect on spiders (Uetz 1991). Various Bayliss, A.D. 2000. Why glyphosate is a global her- studies have demonstrated a relationship be- bicide: strengths, weaknesses and prospects. Pest Management Science 56:299±308. tween spiders and vegetation height (e.g., DoÈ- Chiverton, P.A. & N.W. Sotherton. 1991. The ef- bel et al., 1990; Rushton & Eyre 1992) and fects on bene®cial arthropods of the exclusion of selection of dead vegetation by spiders has herbicides from cereal crop edges. Journal of Ap- also been noted (e.g., Duffey 1962; Gibson et plied Ecology 28:1027±1039. al. 1992). However, vegetation height and per- de Snoo, G.R. & R.J. van der Poll. 1999. Effect centage dead vegetation cover were only cor- of herbicide drift on adjacent boundary vege- related with the change in the spider com- tation. Agriculture, Ecosystems and Environ- munity over the season, not with the ment 73:1±6. difference between treatments as there were DoÈbel, H.G., R.F. Denno & J.A. Coddington. 1990. no cluster divisions by treatment in either of Spider (Araneae) community structure in an in- the DCA ordinations. tertidal salt marsh: Effects of vegetation structure Implications of incremental increases of and tidal ¯ooding. Environmental Entomology 19:1356±1370. glyphosate.ÐBased on our previous research, Duelli, P., M. Studer, I. Marchand & S. Jakob. 1990. incremental increases of glyphosate have Population movements of arthropods between caused a signi®cant reduction in numbers at natural and cultivated areas. Biological Conser- the species level, between guilds and for total vation 54:193±207. numbers. From this perspective, the spider Duffey, E. 1956. Aerial dispersal in a known spider community is detrimentally affected by appli- population. Journal of Animal Ecology 25:85± cations of this herbicide. However, in terms of 111. effect on the composition of the spider com- Duffey, E. 1962. A population study of spiders in munity, the answer is less clear; species were limestone grassland: The ®eld-layer fauna. Oikos not eliminated from herbicide treated margins 13:15±33. and thus there would seem less cause for con- Feber, R.E., H. Smith, & D.W. Macdonald. 1996. cern. However, perpetual use of glyphosate at The effects on butter¯y abundance of the man- high rates during the growing season may agement of uncropped edges of arable ®elds. Journal of Appied Ecology 33:1191±1205. contribute to a sustained reduction in species, Gibson, C.W.D., C. Hambler, & V.K. Brown. 1992. particularly for those species which are not r- Changes in spider (Araneae) assemblages in re- stategists. lation to succession and grazing management. ACKNOWLEDGMENTS Journal of Applied Ecology 29:132±142. Haughton, A.J., J.R. Bell, S. Gates, P.J. Johnson, We thank all the staff at Loddington, es- D.W. Macdonald, F.H. Tattersall, & B.H. Hart. pecially Phil Jarvis (farm manager), for ac- 1999a. Methods of increasing invertebrate abun- BELL ET AL.ÐTHE EFFECT OF GLYPHOSATE ON SPIDER COMMUNITIES 297

dance within ®eld margins. Aspects of Applied Carlo Methods In Biology. Chapman and Hall, Biology 54:163±170. London. Haughton A.J., A. Wilcox & K. Chaney. 1999b. Meijer, J. 1977. The immigration of spiders (Ara- The effects of different rates of glyphosate on neida) into a new polder. Ecological Entomology non-target invertebrates in ®eld margins. Aspects 2:81±90. of Applied Biology 54:185±190. Rushton, S.P., M.D. Eyre. 1992. Grassland spider Haughton, A.J., J.R. Bell, N.D. Boatman & A. Wil- habitats in north-east England. Journal of Bio- geography 19:99±108. cox. 1999c. The effects of different rates of the Russell, G., J.M. Diamond, S.L. Pimms, and T.M. herbicide glyphosate on spiders in arable ®eld Reed. 1995. A century of turnover: community margins. Journal of Arachnology 27:249±254. dynamics at three timescales. Journal of Animal Haughton, A.J., J.R. Bell, N.D. Boatman & A. Wil- Ecology 64:628±641. cox. 2001a. The effect of the herbicide glyphos- Uetz, G.W. 1991. Habitat structure and spider for- ate on non-target spiders. II. Indirect effects on aging. Pp. 325±348. In Habitat Structure (S.S. Lepthyphantes tenuis in ®eld margins. Pest Man- Bell, E.D. McCoy & H.R. Mushinsky, eds.). agement Science 57:1037±1042. Chapman and Hall, London. Haughton, A.J., J.R. Bell, A. Wilcox & N.D. Boat- Williamson, M. 1987. Are communities ever stable? man. 2001b. The effect of the herbicide gly- Pp. 353±371. In Colonization, Succession And phosate on non-target spiders. I. Direct effects on Stability. (A.J. Gray, M.J. Crawley & P.J. Ed- Lepthyphantes tenuis under laboratory condi- wards, eds). 26th Symposium of the British Eco- tions. Pest Management Science 57:1033±1036. logical Society, Blackwell Scienti®c Publica- Hill, M.O. 1994. DECORANA and TWINSPAN, tions, Oxford, Woodburn, A.T. 2000. Glyphosate: production, for ordination and classi®cation of multivariate pricing and use worldwide. Pest Management species data: a new edition, together with sup- Science 56:309±312. porting programs, in FORTRAN 77. Institute of Terrestrial Ecology. Huntingdon, UK. Manuscript received 15 July 2001, revised 1 April Manly, B.F.J. 1991. Randomization And Monte 2002.