1998. The Journal of Arachnology 26:221-226 THE INFLUENCE OF HABITAT STRUCTURE ON SPIDER DENSITY IN A NO-TILL SOYBEAN AGROECOSYSTEM Robert Andrew Balfour and Ann L. Rypstra: Department of Zoology, Miami University; Oxford, Ohio 45056 and Hamilton, Ohio 45011 USA ABSTRACT. The goal ofthis research was to investigate the relationship between habitat structure and spider density in soybean fields managed under conservation tillage practices. Previous studies suggest that spiders respond to vegetational structure, and fields which are not tilled tend to have greater vege- tational structure due to higher densities ofweeds. Experimental subplots with varying densities ofweeds (High, Medium, and Low weed) were established in soybean plots in southwestern Ohio. By the end of the season significantly more web spiders were found in treatments with higher weed densities. Across the season more than 87% ofthe spiders observed were orb-web weavers and sheet-web weavers. When considered separately, both ofthe common types ofweb spiders had higher densities in areas with higher densities ofweeds. However, the degree to which orb and sheet-web weavers attachedtheirwebs toweeds differed across treatments. Orb-web weavers were more likely to attach their webs to weeds than to soybean plants or ground/ground litterin Medium weed density treatments. Sheet-web weavers weremore likely to use weeds as a web attachment substrate in High weed density treatments. A variety of studies have demonstrated a re- tural and microhabitat diversity provided by lationship between the structural complexity, or these weeds should enhance the web spider vegetational diversity, of a particular area and community and, in turn, may reduce the impact the abundance and/or diversity of web-building of pest insects. In a three year study, Rypstra spiders (Coleboume 1974; Ohve 1980; Robin- & Carter (1995) found a positive correlation son 1981; Rypstra 1986; Gunnarsson between spider density and weed biomass, and 1988,1990; Dobel et al. 1990; Uetz 1991; Ward a negative correlation between spider density & Lubin 1993; Wise 1993; Pettersson 1996). and herbivore damage in a soybean agroeco- Web spiders should be particularly sensitive to sysem. Their study was purely correlative stmctural features of their environment because across fields and years so a more controlled of the specific spatial requirements of web investigation of the manner in which weeds placement (Coleboume 1974; Riechert & Gil- may influence the spider community in a no- lespie 1986; Uetz 1991). Indeed, experiments till agricultural system is warranted. which changed existing features of a habitat, or The goal of this study was specifically to added artificial substrate to which spiders could relate weed density within no-till soybean attach their webs, have demonstrated that spi- fields to the abundance of web building spi- ders positively respond to such stmctural alter- ders. Weed densities were manipulated in a ations (Robinson 1981; Rypstra 1983; McNett replicate design and the web-spider commu- 1995). nity monitored across the season in order to The role spiders play in the food web and test the hypothesis that the stmctural diversity their potential as agents ofbiological control is provided by the weeds enhances spider abun- becoming clearer (Riechert & Lockley 1984; dance. In this way, we can begin to understand Riechert & Bishop 1990; Young & Edwards specifically how the changes in tillage prac- 1990; Carter & Rypstra 1995). Conventional tices implemented by American farmers may managment of agricultural fields leads to a be impacting the other components of the bi- stmcturally homogeneous environment which otic community that live in agroecosystems. may minimize the abundance and diversity of METHODS spiders. However, in recent years, management practices designed to reduce erosion have be- Field work was conducted at Miami Uni- come more popular in the United States, even versity’s Ecology Research Center (ERC), lo- though they leadto an increase in weeddensity cated three miles northeast of Oxford, Butler and diversity (Gebhardt et al. 1985). The stmc- County, Ohio USA. In 1995 we randomly se- 221 222 THE JOURNAL OF ARACHNOLOGY lected three of twelve 60 X 70 m (0.42 ha) touching the soybean stems. At two randomly soybean plots, planted in an east/west direc- chosen points along the length of that meter tion separated by 15 m corridors of mowed stick, another meter stick was placed on the grass. Soybeans were planted on 6 June and ground perpendicular to the soybean row. The three herbicides (Roundup® (glyphosate, N- number of weed stems contacting the length of (phosphonomethyl) glycine in the form of its the second meter stick was recorded. Weed ver- isopropylamine salt; 0.96 kg active ingredient/ tical structure was measured by dividing the ha), Dual 8E® (metolachlor; 2.03 kg active subplots into fourquadrats of50 X 50 cmeach. ingredient/ha), and Lorox Plus® (linuron plus We selected two of these quadrats using pairs chlorimuron; 0.67 kg active ingredient/ha)) of random integers between one and four. At were applied pre-soybean emergence on 7 random locations within these chosen quadrats, June. Two herbicides (Poast Plus® (sethox- a meter stick was positioned vertically and the yoim plus dash; 0.23 kg active ingredient/ha) numberofweedleavescontactingits lengthwas and Cobra® (lactofen; 0.20 kg active ingre- recorded. We calculated the vertical structure of dient/ha)) were applied post-soybean emer- the weed community by summing the number gence on 11 July and 12 July, respectively, in of weed leaves contacting the two meter sticks conjunction with a crop oil concentrate/sur- placed perpendicular to the ground. Soybean factant (2.34 kg/ha with Poast Plus® and 1.17 vertical structure was measured at the same two kg/ha with Cobra®). The plots were not tilled points as weed density, by holding a meter stick at any time during the season. vertically in the soybean row and counting the Nine 1.0 X 1.0 m^ subplots were placed number of soybean leaves in contact with the within each ofthe 0.42 ha plots by generating meter stick. We calculated the soybean vertical m coordinates on a 1 grid using a random structure by summing the number of soybean m number table. Each subplot was at least 10 leaves contacting the two meter sticks placed away from any other and assigned to one of perpendicular to the ground. Weed and soybean three weed density treatments: High, Medium, height were calculated using the highest point a and Low weed densities. In subplots desig- weed or soybean leaftouched the vertical meter nated as Low, all weeds were manually re- stick. — moved weekly to maintain low weed struc- Spider census. Web spider density data ture. Subplots initially designated as Medium were collected between 0730-0930 h when or High treatments were reassigned at the end dew increased web visibility. First we of the field season depending on natural weed searched each subplot for spiders on the veg- colonization in each subplot. Subplots with etation and on the ground surface. Then we weed densities between 10-16 stems/m^ were systematically inspected each plant starting at assigned to the Medium treatment and 18-27 the base and moving upward. Each spider stems/ m^ to the High treatment. found was categorized according to its web Data were collected every other week over type. Sheet-webs, (spun by Agelenidae, Lin- a two month period, beginning on 25 July yphiidae), consisted of a horizontal sheet of when the soybeans were in the mid-vegetative silk sometimes bordered by a tangle of silk. stage and ending 16 September, when the soy- Orb-webs, (spun by Tetragnathidae, Aranei- beans had senesced (Teare & Hodges 1994). dae), were two-dimensional and mostly cir- We combined data for the first month (two cular, with radii extending from the hub to the sampling dates between 25 July-23 August) periphery. Any tangle or damaged webs we and refer to it as Early season. Similarly, we encountered were placed in a separate cate- combined data for the second month (two gory. We also recorded the specific substrate sampling dates between 23 August-16 Sep- to which each web was attached: ground/ tember) and refer to it as Late season. Early ground litter (plant debris), soybean, weeds, and Late season each consisted of two plant or some combination of the three. and two spider census samples. A mean value RESULTS for each plot and treatment was calculated for — both Early and Late seasons. — Quantification of plant structure. Ap- Quantification of plant structure. ^Weed proximately seven species of weeds invaded density was measured by placing a meter stick the no-till soybean plots in 1995 (Table 1). on the ground parallel to the soybean row. Weed density per m^ and weed vertical struc- BALFOUR & RYPSTRA—HABITAT STRUCTURE AND SPIDER DENSITY 223 — Table 1. Common and scientific names of the 0.05). If the data are uncoupled so each sub- most abundant weed species invading no-till soy- plot and treatment are included, there is a sig- bean fields of southwestern Ohio in 1995. nificant correlation between the number of spiders and number of weed stems counted in Common name Scientific name subplots in the Late season {P = 0.356, P = Giant ragweed Ambrosia trifida 0.001). Common ragweed Ambrosia artemisifolia Orb-web weavers comprised 44% of the Foxtail (grass) Setaria sp. spiders censused both in the Early and Late Common milkweed Asclepias cyriaca seasons. The dominant orb-spinner in this sys- Fescue (grass) Festuca elatior Ivy Convolvulus sepium tem was Glenognatha foxii (McCook 1894) Canadian thistle Cirsium arvense (Araneae, Tetragnathidae). In the Early sea- son, the mean number oforb-webs perm^ was not different among treatments; however, by ture were significantly different among the the Late season there was a significant treat- three treatments in both time periods (Table ment effect (Table 3). Pairwise comparisons a2)n.dWseoyebdeahneihgehitg,htsodiydbenaont dviefrfteircabletswtreuecntutrhee, msuogrgeestorHbi-gwheawveeresd wsuhbepnlotcsohmapdarseidgnitfoicaLnotlwy three treatments i—n either season (Table 2). weed subplots (DNMR, P < 0.05). Orb-web Spider census. Although in the Early sea- weavers attached their webs to different sub- = son weed density had no effect on the density strates in different treatments (x^ 16.74, df of spiders per m^ (Table 3), there was a sig- = 4, P < 0.005, Fig. 1). More orb-webs were nificant effectofweed density on spider abun- attached to weeds in Medium weed density dance in the Late season (Table 3). Pairwise treatments than in High or Low weed density comparisons of the late season data suggest treatments. that there were significantly more spiders in Sheet-web weavers comprised 43% of the High weed subplots than in Low weed sub- spiders censused in both the Early and Late plots where weeds were removed (Duncan’s seasons. Meioneta micaria (Emerton 1882) New Multiple Range Test (DNMR), P < (Araneae, Linyphiidae) was the dominant — Table 2. Summary of vegetation structure within the experimental treatments (mean ± SE). Experi- mental treatments included High weed density (High), Medium weed density (Medium), and Low weed density (Low). High Medium Low ANOVA results Weed density per m^ Early season 22.5 ± 1.2 11.8 ± 0.9 3.7 ± 0.1 F = 105.7, df^ 2, P < 0.05 Late season 21.3 ± 4.5 16.0 ± 1.8 8.0 ± 0.8 F = 5.48, df= 2, P < 0.05 Weed vertical struc- ture (sum leafnumber) Early season 42.0 ± 7.5 31.7 ± 3.7 5.7 ± 5.7 F = 10.07, df= 2, P < 0.05 Late season 53.5 ± 2.8 33.8 ± 7.0 0 F = 38.46, df= 2, P < 0.05 Soy vertical structure (sum leaf number) Early season 24.8 ± 3.6 22.2 ± 2.8 23.9 ± 1.0 F = 0.228, df= 2, P > 0.05 Late season 19.8 ± 2.1 20.3 ± 2.5 21.3 ± 2.0 F = 0.110, df= 2,P > 0.05 Soy height (cm) Early season 61.0 ± 3.4 61.8 ± 4.1 60.9 ± 4.2 F = 0.016, df= 2,P > 0.05 Late season 85.5 ± 1.8 80.7 ± 2.2 81.5 ± 1.9 F = 1.799, df= 2, P > 0.05 Weed height (cm) Early season 47.0 ± 6.1 44.5 ± 7.8 27.0 ± 2.0 F = 1.480, df= 2, P > 0.05 Late season 68.4 ± 8.1 48.2 ± 8.4 66.5 ± 17.0 F = 1.532, df= 2, P > 0.05 224 THE JOURNAL OF ARACHNOLOGY — Table 3. Summary of the total number of webs, subsequently broken down into sheet webs and orb webs, within the three weed density treatments (mean ± SE). High Medium Low ANOVA results Total number of webs per m^ Early season 3.6 ± 0.5 4.7 ± 0.5 2.2 ± 0.7 F = 3.564, df= 2, P > 0.05 Late season 9.4 ± 0.6 7.1 ± 0.5 4.9 ± 1.3 F = 5.914, 2, P < 0.05 Number of sheet webs per m^ Early season 1.6 ± 0.2 2.3 ± 0.7 1.3 ± 0.3 F = 0.994, df= 2, P > 0.05 Late season 5.0 ± 0.2 2.6 ± 0.4 2.1 ± 0.8 F = 7.226, df= 2, P < 0.05 Number of orb webs per m^ Early season 1.6 ± 0.6 2.1 ± 0.9 0.8 ± 0.4 F = 0.924, df= 2, P > 0.05 Late season 4.0 ± 0.5 3.0 ± 0.4 1.5 ± 0.5 F = 6.664, df= 2, P < 0.05 sheet-web spinner in the fields. As was the DISCUSSION case for orb-web spiders, we found no treat- The manipulation of weed density clearly ment effect on sheet-web weavers until the affects the spider density in no-till soybean late season (Table 3). At that time, High weed agroecosystems. We presume this relationship subplots had significantly more sheet-weavers was due to differences in web support struc- than Low weed subplots (DNMR, P < 0.05). tures and/or the availability ofappropriate mi- Sheet-web weavers also utilized different web crohabitats. Increased structural complexity attachment sites as weed density changed (x^ has previously been correlated with spider - 14.91, df= 4, P < 0.005, Fig. 2). Unlike abundance and diversity (Greenstone 1984; orb-weavers, sheet-weavers were more likely Rypstra 1986). Likewise, the addition of ar- to attach their webs to weeds in High weed tificial web support structures has repeatedly subplots than in Low or Medium weed sub- resulted in an increase in web-spiders (Rob- plots. inson 1981; Rypstra 1983; McNett 1995). 0.81 Weed 0.81 Weed Soybean Soybean Ground/GroundLitter 06- . u u oa o& o o u u Ph Ph High Medium Low Treatment Treatment — — Figure 1. The proportion of orb-webs attached Figure 2. The proportion of sheet-webs at- to each substrate (weed, soybean, ground/ground tached to each substrate (weed, soybean, ground/ litter) within each weed density treatment (High, ground litter) within each weed density treatment Medium, and Low). (High, Medium, and Low). BALFOUR & RYPSTRA—HABITAT STRUCTURE AND SPIDER DENSITY 225 Here we attempted to be as realistic as pos- The differences we observed in web substrate sible by monitoring the effects ofnaturalplant usageinresponseto weeddensitybetweensheet invaders on web-spiders in an economically and orb-web weavers is intriguing and deserves important habitat. These spiders rarely at- further investigation. tached their webs tojust one substrate as they Spiders are important generalist predators would be forced to if we had used artificial in terrestrial systems and no-till soybean constructs to alter the structural complexity agroecosystems are an increasingly important available to them. Most of them used a com- terrestrial habitat in the United States (Geb- & bination of available plants, ground litter, and hardt et al. 1983). Rypstra Carter (1995) dirt as web substrates (Figs. 1, 2). demonstrated that spider density was positive- Difference in weed abundance not only ly correlated with weed biomass across years changes the structural complexity of the en- in conventionally tilled soybean fields. Typi- vironment but also ameliorates the microhab- cally, areductionin tillage leads to anincrease itat under the vegetation; especially near the in weeds (Gebhardt et al. 1983). In this study, ground surface. Most of the spiders surveyed we demonstrated that, within one year, weed were small (< 2 mm), and the majority ofthe density in no-till soybean fields influenced webs were constructed on the lower third of spiderabundance. These datacontributeto our the vegetation. Small spiders are more prone understanding of how shifts in agricultural to dehydration than larger spiders due to their practices may affect the spider community relatively high surface area to volume ratios which may have larger implications for the (Pulz 1987). Building webs lower in the veg- productivity of the agroecosystem. etation where there is increased hunfidity re- In the process of censusing for spiders it sults in less direct exposure to sunlight, re- was necessary to disturb the vegetation within ducing the chance of dehydration. Also, a the subplots. The greater the vegetational spider’s ability to build an efficient capture structure within a subplot the greater the dis- web is maximized at certain thermal condi- turbance caused by the close visual inspection tions (Barghusen et al. 1997), which may be of the plants and soil surface. Therefore some present lower in the vegetation. Web destruc- spiders present in the subplots were probably tion by wind is another factor affecting web overlooked due to web destruction. Since dis- site tenacity (Hodge 1987), The bases of turbance is related to the amount of vegeta- plants provide sturdy support for web attach- tion, sampling error should have resulted in ment and are less affected by wind. our values of spider density being underesti- If the high spider density in the presence of mates in the Medium and High treatment sub- weeds was due to an increase in web supports, plots. Therefore any effects we report as sig- then one would predict that the spiders would nificant would only be more striking ifwe had be more apt to use weeds for web attachment been able to find every spider. in the weedierplots. In ourplots, spiders tended Web spider density is increased by weed to use the soil surface less and use weeds more density presumably due to an increase in as weed density increased (Figs. 1,2). Although structural complexity. The close relationship orb-web weavers used weeds to a high degree we observed between weed density and spider at Medium weed densities, they reduced their density helps to explain the observed relation- usage ofthis substratumin the Highweedplots. ship between weedbiomass and spiderdensity It may be thatorb-web weavers, who have very Rypstra & Carter (1995) found across three specific requirements forappropriate webplace- seasons. Ourworkoffers agreaterunderstand- ment, were responding more to microhabitat ing of how spider communities interact with changes in the High weed treatments than to the plant communities around them. It also of- structural features. Once they estabhshed them- fers us further insight into habitat selection by selves in the plot, theregular spacing oftherow spiders and gives us a greater understanding of soybean plants may have offered a greater of the animal community in agroecosystems. number of open spaces suitable for their planar ACKNOWLEDGMENTS webs. In a field study such as this, it is difficult to uncouple the relative role of structural com- We would like to thank S.D. Marshall, A.B. plexity and microhabitat in producing the ob- Cady, and S.E Walker for comments and sug- served differences in web spider abundance. gestions on drafts of this paper, Tamie Beltz 226 THE JOURNAL OF ARACHNOLOGY for her assistance in data collection, and J.R. Olive, C.W. 1980. Foraging specializations in orb- Dobyns for his assistance in spider identifi- weaving spiders. Ecology, 61:1133-1144. cation. We would also like to thank D.M. Pa- Pettersson, R.B. 1996. Effect of forestry on the vuk and R.F. Stander for all their work culti- abundance and diversity of arboreal spiders in the boreal spruceforest. Ecography, 19:221-228. vating the soybeans. Voucher specimens were Pulz, R. 1987. Thermal and water relations. In deposited in Miami University’s Hefner Zo- Ecophysiology of Spiders. (W. Nentwig, ed.). ology Museum; Oxford, Ohio USA. This re- Springer-Verlag, Berlin. search was funded by a Grant-in-aid of Re- Riechert, S.E. & L. Bishop. 1990. Prey control by search from Sigma Xi, The Scientific an assemblage ofgeneralist predators: spiders in Research Society, Miami University’s Under- garden test systems. Ecology, 71:1441-1450. & graduate Summer Scholars Program, Miami Riechert, S.E. R. Gillespie. 1986. Habitatchoice and utilization in web building spiders. Pp. 23- University’s Department of Zoology, and the 49. In Spiders: Webs, Behavior, and Evolution. Hamilton Campus of Miami University. (W. Shear, ed.). Stanford Univ. Press, Stanford, California. LITERATURE CITED & Riechert, S.E. T. Lockley. 1984. Spiders as bi- Barghusen, L.E., D.L. Claussen, M.S. Anderson & ological control agents. Ann, Rev. Entomol., 29: 299-320. A.J. Bailer. 1997. The effects oftemperature on the web-building behaviour of the common Robinson, J.V. 1981. The effect of architectural house spider, Achaearanea tepidariorum. Funct. variation in habitat on a spider community: an experimental field study. Ecology, 62:73-80. CarEtecro,L,RE1.1:&4-1A0..L. Rypstra. 1995. Top-down ef- Rypstra, A.L. & RE. Carter. 1995. The web spider community ofsoybean agroecosystems in south- fects in soybean agroecosystems: spider density western Ohio. J. ArachnoL, 23:135-144. affects herbivore damage. Oikos, 72:433-439. Rypstra, A.L. 1986. Web spiders in temperate and Coleboum, RH. 1974. The influence of habitat tropical forests: relative abundance and environ- structure on the distribution of Araneus diade- mental correlates. American Midi. Nat., 115:42- matus Clerck. J. Anim. EcoL, 43:401-409. 51. Dobel, H.G., R.F. Denno &J.A. Coddington. 1990. Rypstra, A.L, 1983. The importance of food and Spider (Araneae) community structure in an in- space in limiting web-spider densities; a test us- tertidal salt marsh: effects ofvegetation structure ing field enclosures. Oecologia, 59:312-316, and tidal flooding. Environ. Entomol., 19:1356- Teare, LD. & H.F. Hodges. 1994. Soybeanecology 1370. and physiology. In Handbook of soybean insect & Gebhardt, M.R., T.C. Daniel, E.E. Schweizer pests (L.G. Higley Sl D.J. Boethel, eds.). Ento- R.R. Allmaras. 1985. Conservation tillage. Sci- mol. Soc. America; Landham, Maryland, ence, 230:625-630. Uetz, G.W 1991. Habitat structure and spider for- Greenstone, M.H. 1984. Determinants ofweb spi- aging. Pp. 325-348, In Habitat structure, the der species diversity: vegetation structural diver- physical arrangement of objects in space (S.S. sity vs, prey availability. Oecologia,62:299-304. Bell, E.D. McCoy & H.R. Mushinsky, eds.). Gunnarsson, B. 1988. Spruce-living spiders and Chapman & Hall, New York, forest decline; the importance of needle loss. Ward, D. & Y. Lubin. 1993. Habitat selection and Biol. Conserv., 43:309-319. the life history of a desert spider, Stegodyphus Gunnarsson, B. 1990. Vegetation structure and the lineatus (Eresidae). J. Anim. Behav., 62:353- abundance and size distribution of spruce-living 363. spiders. J. Anim. EcoL, 59:743-752. Wise, D.H. 1993. Spiders in ecological webs. Hodge, M.A. 1987. Factors influencing web site YouCnagm,brOi.dRge&UGn.iBv.. EPrdewsasr,dCsa.mb1r9i9d0.ge,SpEidnegrlsanidn.the residence time of the orb weaving spider Mi- UnitedStates fieldcropsandtheirpotentialeffect crothena gracilis. Psyche, 94:363-371. on crop pests. J. ArachnoL, 18:1-27. McNett, B. 1995. Habitat Selection in the Silver Garden SpiderArgiope trifasciata. Unpubl. mas- Manuscript received 11 October 1996, revised 25 ter’s thesis. Miami University, Oxford, Ohio. September 1997.