Response of soil invertebrates to forest canopy inputs

Transcription

1 Pedobiologia 7,127-19, 200 Urban & Fischer Verlag * -i UlOlOffltt Response of soil invertebrates to forest canopy inputs along a productivity gradient Barbara C. Reynolds 1 *, D. A. Crossley, Jr. and Mark D. Hunter Institute of Ecology. University of Georgia 'Current address: Department of Environmental Studies, University of North Carolina at Asheville, Asheville NC , USA Submitted August 8, 2001 Accepted April, 2002 Summary Previous studies have suggested that herbivory in forest canopies can influence forest floor processes such as nutrient cycling and decomposition. We studied the response of litter decomposition to a moisture/productivity gradient with manipulations of the effects of canopy herbivory at the Coweeta Hydrologic Laboratory, North Carolina. Litterbags containing Quercus rubra L. and Acerrubrum L. litter were placed at three vations along the gradient and sampled monthly for two years. Microarthropods, nematodes, and litter mass loss responses to the productivity gradient were measured. The relative abundance of Collembola and three suborders of mites (Oribatida, Mesostigmata and Prostigmata) was compared across the gradient. Mass loss was greater at the middle and high vation sites in both years and was correlated with increased numbers of oribatid mites per gram of litter. The abundance of all the above microarthropods (of which oribatids were the most common) was also greater on the middle and high vation sites and greater on two- year -old litter than on one- year- old litter. Nematode densities were also greater on the older litter. The herbivore inputs study, simulating the effects of canopy herbivory, included frass additions, throughfall additions, greenfall exclusion, total litter exclusion, and controls. Experimental additions of frass to plots on the low and middle vation sites led to an increase in Collembola abundance in litterbags from those plots. Plots with frass and artificial throughfall additions also showed increased numbers of fungal feeding and bacterial feeding nematodes in some months. Numbers of oribatid and prostimatid mites were reduced in litter exclusion plots. Results from these studies suggest not only significant influences of vation on litter decomposition and soil fauna abundance but direct links between canopy herbivory and responses in population densities of forest floor biota. Key words: Canopy herbivores, vation gradient, frass, litter decomposition, microarthropods, nematodes Introduction synthesis in the forest, has become widely recognized (Parker 1989). As access techniques for canopy studies have improved, research has expanded from descrip- tive to more quantitative studies, such as those on Ecological research in the forest canopy has become an integral component of ecosystem ecology. The importance of the canopy as the interface between the atmosphere and the forest, and as the site of most photo- * corresponding author: kreynolds@unca.edu /0/7/ $15.00/0

2 10 Barbara C. Reynolds, D. A. Crossley, Jr. and Mark D. Hunter sticks (Anisimorpha buprestoides (Stoll)), forest tent caterpillars (Malacosoma disstria Hiibner) and spiny oakworm caterpillars (Anisota sp.). Walking sticks and caterpillars were fed ^?. maximum L. and Q. rubra leaves, respectively, both of which are dominant members of the flora at Coweeta (Day et al. 1988; Reynolds & Crossley 1997) Frass was brushed out of plastic rearing cages daily and held at -1.0 C. until shortly before application. All the frass was carefully mixed before the initial weighing. We chose the quantities of frass and throughfall to add to our ments from measurements made at each vation during the previous two years (Reynolds et al. 2000). From 20 frass and throughfall collectors at each vation, we estimated a) average weekly frass deposition during the growing season and b) average monthly N and P concentrations of throughfall. We used these data to double the average frass and throughfall nutrient inputs into our ments and to reflect seasonal variation in inputs. Doubling these inputs is well within the range of natural variation observed previously at Coweeta (Reynolds et al. 2000, and below). Based on ambient input estimates from previous years, the amount of frass applied to each frass ment varied between 2.9 and 1. g (dry weight) per week, depending on the time of year. Artificial throughfall was generated by dissolving NH C1 and KH 2 PO in deionized water. These two ions were used because our previous research at Coweeta indicated a positive correlation between PO and NH throughfall concentrations and herbivore activity in the canopy. NH mass added ranged from 7. to 22.0 mg/ square meter per week and PO mass was 0.7 to 6 mg/ square meter. The same volume of solution, 800 ml, was used on each m 2 plot. Control plots were sprinkled with 800 ml of deionized water. Litter exclusion was chosen as one of our ments to compare the relative effects of herbivore inputs and leaf senescence on soil invertebrates and nutrient dynamics. Greenfall exclusion was used rather than greenfall additions because it was easier than collecting and adding greenfall from additional collectors. Litter and greenfall were removed from the mesh over litter boxes at least once each week. For controls, frass, and throughfall ments, litter which fell on top of the mesh was spread over the litter inside the box at least weekly. Twenty-four litterbags were placed in each quadrat in late April, Two litterbags from each odd-numbered quadrat were sampled every other month, beginning in late May, 1997, and two litterbags from evennumbered quadrats were sampled the alternate months, beginning in late June, The last litter bag collections were made in January, 1999 (odd-numbered quadrats) and February, 1999 (even-numbered quadrats). Since there were 5 replicates per ment, 1 replicates were collected for odd-numbered quadrat collections, and 12 replicates were collected with even-numbered quadrat collections. Thus, the total number of collection dates for each quadrat box was 12: every other month for two years. Microarthropod and nematode extractions were done as described previously. However, in order to examine possible effects of our ments on various functional groups of nematodes, nematodes were assigned to the following trophic groups, based on the morphology of their stoma (Freckman & Baldwin 1990; Yeates et al. 199): bacterial feeders, fungal feeders, plant feeders, omnivores, predators, and Tylenchida (mostly plant feeders). Litter and greenfall removal were begun in mid- May, 1997, and continued through September, Frass and throughfall additions, done weekly, were begun in May of 1998 and continued through September, Statistical Analyses Gradient Study. Decomposition rates were calculated as % mass remaining in the litterbags after 12 and 2 months in the field, checked for assumptions of normality, and analyzed using ANOVA. Tukey's Studentized Range (HSD) test was used to determine significance of % mass remaining by vation. Microarthropod and nematode abundances, representing collections from February and August of 1997 and 1998, were determined as the mean number of animals/gram litter (n=10 bags for microarthropods, n=5 bags for nematodes). Since the abundance values were not normally distributed, and followed a Poisson distribution (Crawley 199), statistical analysis was performed using Generalized Linear Models (GLIM) (Proc Genmod Procedure, SAS Inst. 1996). Our analyses began with the full models (using all ments) and all the possible interaction terms. The models were then simplified by backstepping (Agresti 1996) until the simplest model with the highest log-likelihood ratio was determined for each analysis. For this study, nematodes were not separated into functional groups. All data were analyzed using SAS 6.12 (1996). Canopy Input Study. For statistical analysis of ment effects, nematode and microarthropod numbers were converted to numbers per g dry weight of litter. Since the data followed a Poisson distribution, effects of the various ments on microarthropod and nematode numbers were again analyzed with a Generalized Linear Model (Crawley 199) using the Proc Genmod Procedure of SAS 6.12 (1996). The most parsimonious model was used as detailed for the gradient study, again following the methods of Agresti (1996). Pedobiologia (200) 7,127-19

3 Effects of canopy herbivores 11 Results Gradient Study Elevation had a significant influence on rates of decomposition. Litter from the low vation site retained more of its initial mass (6% and 51% after 1 and 2 yr respectively) than that from middle (57 % and 2 %) and high vations (55 % and 9 %) during both years of the study. The two higher vation sites were similar in percent mass retained (2 % and 9 % after 2 yr for the mid and high vations respectively). We also calculated % mass remaining on an AFDW basis for subsets of samples (see Methods) and observed Si 0 g!20- f "a oo Elevation p<0.017 Date p < Elevation x Date p < Feb97 Aug 97 Feb 98 Collection Dates Aug 98 Fig. 1. Microarthropods g- 1 dry litter after 2,8,1, and 20 months of decomposition at low (shaded), mid (hollow) and high (black) vations. Each bar is the average of ten litterbags; error bars are ± 1 SE. P values are for a log-likelihood analysis of variance on total numbers per litterbag CA 0 -J i ; f microar i * i r QJ.Q n i o^u-, v I i I rrm I i rrm I I L ; nj 1 :, ; i i i i i i i i i i i i i A i i ti, i i i i i i i su i i IS 0> o> -Q ^o ^5 O "5J 5 ~.2> E -B S O) o 1 «' «<u ^ " fe ^ 2 Q. o «^ 0 5 k 2 Q. g(/) o 0 te 5 2 D. o» o 0 Feb 1997 Aug 1997 Feb1998 Aug1998 Fig. 2. Microarthropods g-' dry litter after 2,8,1, and 20 months of decomposition at low (shaded), mid (hollow) and high (black) vations. Each bar is the number of oribatids, prostigmatids, mesostigmatids or Collembola found in 10 litterbags. Proportions of microarthropods (three mite suborders and collembola) per gram of litter along an vational gradient at Coweeta Hydrologic Laboratory for August 1998, February 1998, August 1997, and February Mesostig=Mesotigmatida, prostig=prostigmatida, oribatid=0ribatida, and collem=collembola Pedobiologia (200) 7,127-19

4 12 Barbara C. Reynolds, D. A. Crossley, Jr. and Mark D. Hunter the same pattern of slower mass loss at low vation. Elevation had a significant effect on total microarthropod numbers and on nematode numbers (Fig. 1, ), although the interaction between vation and date for nematode numbers complicates interpretation. Total microarthropod abundance at mid and high vations was as much as twice that at low vation, especially by year two, and increased over time at all vations (Fig. 1). There was no statistically significant interaction between vation and date. The mid vation site had the highest densities of microarthropods in both February collections while the high vation site had the highest densities in both August collections. Oribatida were the most common of the three mite suborders (Fig. 2), except on the high vation site in August 1997, when Mesostigmata were equally abundant. Within a single sampling period, collembolans were usually more abundant on the low vation site except for August 1998, when they were more abundant at high vation where an outbreak of sawflies had occurred recently (Reynolds et al. 2000). Densities of nematodes increased by as much as ten times over 1.5 years (Fig. ). On the low vation site, there was a monotonic increase in nematode density with each date, while on the mid and high vation sites, the August 1997 collection had fewer nematodes than the February 1997 collection (Fig. ) but otherwise increased over the 1.5 year period. By August of 1998, the average total number of nematodes per gram dry weight of litter was close to 600 for all three sites. Canopy Inputs Study Results from our input manipulation study are reported only for the low and mid vation sites, because an outbreak of the sawfly Periclista sp. (Hymenoptera: Tenthredinidae) at high vation added large quantities of frass to all ment boxes, compromising the ments. The consequences of this sawfly outbreak on soil processes are reported elsewhere (Reynolds et al. 2000). As we expected, canopy inputs led to increases in some microarthropods, although there were interactions between date* ment and vation*ment for collembola, and between ment* vation for mesostigmata (Table 1). There was a large increase in numbers of collembolans per g dry weight of litter in litter bags from frass-ed plots during June and August, 1998 (Table ). An increase in numbers also occurred in throughfall ments in August. On the low vation site, collembolan numbers in greenfall exclusion ments were almost twice those in the controls and litter exclusion ment (Table ). On the mid vation site, only frass addition ments appeared to increase numbers of collembola, to twice that of control and other ments. Date and vation also had a strong effect on collembolan numbers (Table 1). In June, collembolan numbers were more than four times greater on the mid than the low vation site. In August, collembolan densities increased at both sites, but the mid vation site still had more collembola than o> 800- > 700- Q.0)600- cfl Q) Elevation p < Date p< Elevation x Date p<0.01 ro o00-j CD -Q E 200 z CD < Feb97 Aug 97 Feb 98 Aug 98 Collection Dates Fig.. Nematodes g- 1 dry litter after 2,8,1, and 20 months of decomposition at low (shaded), mid (hollow), and high (black) vations. Each bar is the average of five litterbags; error bars are ± 1SE. P values are for a log-likelihood analysis of variance on total numbers per litterbag Pedobiologia (200) 7,127-19

5 Effects of canopy herbivores 1 the low vation site. Collembolan densities decreased at both sites in September, but the mid vation site again had more collembolans. There were fewer prostigmata per g of litter in litter exclusion plots, while frass additions appeared to increase prostigmatid numbers slightly (Table ). Densities of prostigmata increased from May through August, and then declined in September. Both oribatid and prostigmatid mites occurred at higher density at mid than at low vation. Oribatid numbers were lowest in litter exclusion plots (Table ). Oribatid densities over time followed a very comparable pattern to those of the prostigmata and there appeared to be more oribatid mites at mid than at low vation. Numbers of mesostigmatid mites showed unexplained interactions between ment and vation and between date and vation (Table 1). Densities appeared to be lower in ment than in control plots at low vation, yet higher following frass addition at mid vation. In August and possibly in June, the mid vation site appeared to have greater numbers of mesostigmatid mites than did the low vation site. Of the six trophic groups of nematodes counted, only bacterial- and fungal-feeding nematodes apparently responded to our ments. The other trophic groups- omnivores, predators, plant feeders, and Tylanchida- were very low in number and are not considered further here. Significant vation*date*ment interactions occurred for both bacterial- and fungalfeeding nematodes (Table 2). At our low vation site, mean densities of bacterial- feeding nematodes were greater than controls in the frass addition ments during June (75 %) and August (120%) (Table ). In the throughfall addition ments for the low vation site, bacterial-feeding nematodes increased over control densities by 75 % in June and 650 % in September. The only response to ments at our mid vation site was a decrease in densities of bacterialfeeding nematodes during May in ed plots compared to controls (Table ). The responses of fungal-feeding nematodes to our ments also varied with season and vation (Table 2). At low vation, frass additions increased the mean densities of fungal-feeding nematodes over controls in May and June (15 % and 1180 % respectively) (Table ). At mid vation, frass-induced increases were restricted to June (70 %). At low vation, throughfall additions increased mean densities of fungal-feeding nematodes in June and September (67 % and 166 %). At mid vation, throughfall-induced increases were restricted to September (510 %) (Table ). Overall, we demonstrated that, for certain taxonomic groups, inputs from canopy herbivores have a greater impact on soil microarthropod densities than do inputs of leaf litter (Table ). For example, frass and throughfall additions had much greater impacts upon Collembola density than did the exclusion of litterfall during the same periods (Table ). We should note, however, that prostigmatid and oribatid mites were more strongly affected by litter exclusion than by herbivore-derived inputs (Table ). Table 1. Microarthropod responses to vation, ment, and time. Data analyzed were total numbers of microarthropods per ten litterbags on each of four dates (February and August of 1997 and 1998). Elevation =, ment =. Treatments were litter exclusion, greenfall exclusion, frass additions, throughfall additions, and control Organism Log-Likelihood Terms Chi-square df Collembola Prostigmata Oribatida Mesostigmata date *date. date* * date date 0.21 date *date * < Pedobiologia (200) 7,127-19

6 1 Barbara C. Reynolds, D. A. Crossley, Jr. and Mark D. Hunter Table 2. Nematode responses to vation, time, and ment. Data analyzed were total numbers of nematodes per five litterbags collected on each of four dates (February and August of 1997 and 1998). Elevation =, ment =. Treatments were litter exclusion, greenfall exclusion, frass additions, throughfall additions, and control Organism Bacterialfeeding Nematodes Log-Likelihood Terms date *date * date* *date* Chi-square df P Fungalfeeding Nematodes 520. date *date * date* *date* Table. A comparison of the relative importance of herbivore-derived and litter inputs on the densities of soil microarthropods Microarthropod and Treatment Density (+/-1SE) per g DW of litter Control Density (+/- 1SE) per g DW of litter Treatment Percent Difference Collembola June 2.6 (8.68) 60.2 (28.26) +155 Collembola June litter excluded 2.6 (8.68) 15.5(7.87) -5 Collembola August.(5.17) (21.22) +225 Collembola August.(5.17) 51.98(10.21) +51 Collembola August litter excluded.(5.17) 5.10(1.1) +1 Collembola Ele greenfall excluded 11.76(.18) 22. (7.6) +90 Prostigmata 1.1(2.) 17.22(.07) +1 Prostigmata litter excluded 1.1(2.) 7.81 (1.2) -1 Oribatida (25.01) 168.6(.76) + 10 Oribatida litter excluded (25.01) (16.58) -0 Pedobiologia (200) 7,127-19

7 Effects of canopy herbivores 15 Table. A comparison of the relative importance of herbivore-derived inputs on the densities of soil nematodes. (BF = bacterial feeders, FF = fungal feeders) Elevation Mid Mid Nematode and Treatment BFJune BF August BFJune BF September FF May FFJune FFJune FF September FF September Density (+/- 1SE) per g DW of litter Control 85.6(19.1) 2.9 (2.8) 85.6(19.1) 16.(91.) 5.1(2.1) 7.0(1.) 19.5(9.) 6. (2.6) 11.(6.2) Density (+/-1SE) per g DW of litter Treatment Percent Difference 06.5(15.) (98.9) (15.8) (1620.7) (9.2) (62) (55.6) (20.8) (7.) +510 Discussion Gradient Study After two years in the field, the average percent mass remaining for low vation litter was significantly greater than mass remaining at mid and high vations. This pattern is similar to that found by Hoover and Crossley (1995) in a one-year decomposition study using three types of litter in single-species litterbags on sites adjacent to those that we used. Thus our first prediction was confirmed. However, litter in the present study decomposed more quickly than litter in the 1995 study. This is probably due, at least in part, to the use of mixed red maple and red oak in our litterbags. Seastedt (198) hypothesized that nutrients rased from more rapidly-decaying species (red maple in our study) could fertilize and thus stimulate decomposition on adjacent, more recalcitrant species (red oak, in this case). High rates of decomposition for mixed species of litter have previously been reported from Coweeta (Cromack 197), where percent mass remaining of mixed litter after one year was less than or equal to the percent mass remaining of two oak species (Quercus alba L. and Quercus prinus L.) used in single and mixed species litterbags. Our findings are of added interest, since the types of litter used by Hoover and Crossley (Q. prinus, Liriodendron tulipifera L., and R. maximum) were different from those (ft n ijirsj on /"l A vr/tit*** wt\ 11 co/^ if» tv» i t-vfo o ai-i < citn/-lt7 Thus we can state that many species of litter decompose more quickly at higher vation sites at Coweeta. We think that higher moisture levels on higher vation sites (Swift et al. 1988), combined with somewhat cooler temperatures (Hoover & Crossley 1995), provide an environment more favorable for fungal decomposers. The greater abundance at higher vations of collembola and mites (Fig. 2), many of which are fungivores and comminuters, suggests that fungal activity and direct effects of microarthropods on decomposition are greater at higher vations. Hansen (1997) found many endophagous oribatid species, mites whose juveniles burrow into plant material, at a high vation site adjacent to ours. In the second year of that study, presumably when the oribatid populations had had a chance to respond to oak litter ments, Hansen found densities of endophagous oribatids positively correlated to the proportion of oak leaves in the mixed litter layer and that oak leaves decomposed more quickly in the second year than did other litter types. We also predicted that microarthropods would be most abundant on the mid vation site, based on data Pedobiologia (200) 7,127-19

8 16 Barbara C. Reynolds, D. A. Crossley, Jr. and Mark D. Hunter from Hoover and Crossley (1995). Our prediction was valid only for February, 1997 and possibly February, 1998 (Fig. 2). In August of both years, the high vation site had the greatest numbers of microarthropods. This difference between the two studies is probably the result of differences in data analysis. Hoover and Crossley averaged their microarthropod numbers over one year for each vation, while our data are specific for a particular month. Our gradient study on decomposition found faster decomposition on the mid and high vation sites, with no significant difference between the mid and high sites. This suggests greater numbers and/or activity on the part of soil fauna in these higher vation sites, and is consistent with our findings of more microarthropods on the two higher vation sites by 1998 (Fig. 1, 2). Microarthropod abundance, in terms of average numbers of microarthropods per gram of litter, was lowest in February of 1997, probably because colonization of the litter by fungi and microarthropods was still taking place two months after setting out the bags (Fig. 1). Numbers of microarthropods increased over time, presumably as microarthropod decomposers continued to colonize the litter and as decomposing fungi increased and provided more food resources for fungivores. Oribatida were the predominant suborder of mites in this study, except for August of 1997 when Mesostigmata were similar in abundance at the highest vation (Fig. 2). Mesostigmata may respond positively to warm weather, for they were more abundant on the other two sites in August than in either February collection (Fig. 2). Prostigmata were always present, although below 10 % in proportion except for February, 1998, when they comprised from 1% to 2%. Prostigmata were fewer in numbers than either Mesostigmata or Oribatida. Perhaps that is because many of the Prostigsmata are predators on other mites. As one would expect for predators tracking a prey population, Prostigmata densities and proportions increased by August, 1998 (Fig. 2). Collembola seem to prefer cooler temperatures, or at least are more active than mites at cooler temperatures, since collembolans were more abundant in both February collections. Although the significant interaction between vation and date for nematode densities (Table 2) makes it meaningless to consider vation effects averaged across time, or time effects averaged across vation, some intriguing comparisons can be made between microarthropod and nematode densities by date along the vation gradient. For example, in August, 1997, microarthropod densities appeared to be significantly greater in the high vation site compared to low and mid vations. However, nematode densities were apparently much lower at mid and high vation sites than at the low vation site. We think this difference stems from contrasting life history characteristics. Our records of throughfall amounts (unpublished data) on these sites for two weeks prior to the August litterbag collection, show that, on average, 0 ml (appx ") of throughfall was collected on the low vation site. The mid site had 2.5 ml and the high site had 50 ml. The greater amount of moisture measured on the high vation site, albeit small in volume, might explain the greater number of microarthropods counted on this site (Fig. 1). Nematodes, which primarily inhabit water films in soil (Coleman & Crossley 1996), seemed unresponsive to this difference in moisture, and probably had become either anhydrobiotic (an inactive state which can be immediately reversed by reduction of the environmental stress, Freckman et al. 1987) or had migrated from the dry litter to the soil at the upper vation sites. Perhaps their greater numbers on the low site are a rapid response to rainfall which occurred only on the low site the night before the litter bags were collected (Reynolds, personal observation). Nematodes are known to have a short response time to changes in environmental conditions (Freckman et al. 1987). Canopy Inputs Study In 1998, an outbreak of sawflies, Periclista sp. (Hymenoptera: Tenthredinidae: Blennocampinae) at the highest vation caused an estimated 0 % defoliation within a few weeks of bud-break. This led to a tripling of frass inputs in May (compared to the lower vation sites) plus a five-fold increase in soil solution nitrate compared to earlier and later time periods (Reynolds et al. 2000). Therefore it was difficult to separate the effects of the sawfly inputs from vation effects on microarthropods at the high vation site. However, experimental additions of frass at low and mid vation sites resulted in a doubling of collembolan numbers (Table ). Since collembola are primarily fungivores (Coleman & Crossley 1996), it is reasonable to suggest that increased nutrients from frass inputs led to an increase in fungal biomass which could, in turn, support greater numbers of collembola. It is also interesting to note that collembola were most abundant at the high vation site in August of 1998 in our gradient study, three months after the sawfly outbreak and its natural frass additions (Fig. 2). This natural experiment also provides support for frass inputs stimulating increases in collembolan densities. As we found in our Gradient Study (Fig. 2), oribatid and prostigmatid mite plus collembola numbers were greater at the mid vation site than at low vation in our Canopy Inputs Study. Again, this suggests that fungivorous microarthropods are tracking food resources Pedobiologia (200) 7,127-19

9 Effects of canopy herbivores 17 which are probably more abundant at the higher vation sites. As indicated in our Gradient Study, vation had a significant influence on decomposition and on numbers of soil microfauna (Fig. 1, ). It comes as no surprise, then, that vation by itself or as an interaction term had a significant influence on the numbers of all soil microarthropods and the two groups of nematodes studied. An increase in microarthropod density is apparent from May through August for collembola, prostigmata, and oribatida (Fig. 2). A similar pattern is present for control groups of both fungal-feeding and bacterial-feeding nematodes, although only at the low vation site. We assume that the microarthropods are tracking food resources, which are increasing over time. Decreases in numbers of all these taxa by September, except for bacterial feeding nematodes, support our contention that many of these microarthropods are fungivorous and strongly suggests that fungi were less abundant in September. However, interpretation of nematode response is obscured by the interactions among vation, date, and ment (Table 2). The impact of spatial variation between different vations is apparent in these results. Prostigmatid and oribatid mite numbers were reduced in our litter exclusion ments, but only prostigmatida showed a strong increase in response to herbivore inputs (Table ). It is not surprising that oribatids were few in number in the 1998 frass and throughfall ments, since these two ments were initiated in May of Oribatids reproduce rather slowly compared to other microarthropods (or nematodes), having only one or two generations per year (Coleman & Crossley 1996). Prostigmata and mesostigmata have higher fecundity; some may complete their life cycle in less than a week (Behan-Pelletier & Newton 1999). In contrast to the limited responses of other microarthropods to our ments, collembolan abundances showed increases to frass ments in June and August and to throughfall in August (Table ). We ascribe this quick response by collembola in part to their potential for massive population increases (Hopkin 1997). Collembola abundances also responded to litterfall exclusion, both positively and negatively, but the extent of this response was much less than the response to frass (June and August) and throughfall (August) (Table ). The strong response shown by collembola to ments simulating canopy herbivore activity indicates that the effects of frass inputs from canopy herbivores can, at times, be greater than the effects of litter inputs, over the time scales studied here. On the low vation site, numbers of mesostigmatida were apparently reduced in the litter exclusion plots, similar to the oribatida and prostimatida (Table ). It is possible that harsh conditions of litter exclusion bags, especially dryness, were ameliorated at the mid vation (wetter) plots, thus explaining the interaction between this ment and vation. Mesostigmatid numbers may have increased for frass ments on the mid-vation site, as did collembola numbers. The mesostigmatida in soil are almost all predators (Coleman & Crossley 1996), especially on nematodes (Walter & Proctor 1999). It seems likely that the mesostigmatida were feeding on fungivorous nematodes which had responded to the frass inputs -as evidenced by the increased fungal-feeding nematodes observed in June and September at the mid-vation site and the increase in collembola numbers. Again showing a similar response to collembola, mesostigmatid numbers were greater on the mid vation site compared to the low site in June and August. We think this is probably due to the influence of frass additions on fungal-feeding nematodes (Table ). The responses of nematodes to our ments highlight the importance of spatial variability in ecological interactions (Hunter & Price 1992). Bacterial-feeding nematodes responded positively to herbivore-derived inputs at the low vation site only and appeared unresponsive at mid vation (Table ). Fungal-feeding nematodes increased in response to ments at both vations, but those responses were more frequent at low vation than at mid vation (Table ). Elevation has been shown to have large effects on other ecological interactions in the Coweeta drainage basin (Hoover & Crossley 1995; Reynolds & Crossley 1997). Phenological delay may explain, in part, the differences in ment responses between vations. Warmer temperatures, leading to increased biological activity, occur first at lower vations. Nematode activity is known to be constrained by low temperature (Freckman & Baldwin 1990) and, as would be expected, responses of fungal-feeding nematodes to our ments at mid vation occurred later than those at low vation. In contrast to our findings, Maraun et al. (2001) observed no significant response to nutrient amendments by soil fauna in litter of a beechwood forest in Germany. Differences in experimental methods may be responsible for some of these differences in results; Maraun et al. sampled once at the close of their experiment. In our study, significant differences in response were associated with the date of sampling (Tables 1, 2). Also, the soil in Germany was limestone-based and contained a high density of earthworms (Maraun et al. 2001), whereas Coweeta soils are on granite bedrock (Swank and Crossley 1988) and have few earthworms Reynolds, personal observation). Any forest ecosystem study encompassing only three years cannot hope to include all the variation Pedobiologia (200) 7,127-19

10 18 Barbara C. Reynolds, D. A. Crossley, Jr. and Mark D. Hunter which occurs in a forest due to differences in weather (especially different combinations of precipitation and temperature), and fluctuations in populations of herbivores and their predators. However, we have shown that spatial variation, in terms of vation differences, can have a significant influence on rates of decomposition and numbers of soil microfauna. We have also demonstrated that there can be significant effects from the influence of canopy herbivore-derived inputs on forest floor invertebrates even over this short timescale - indicating that there are strong connections between the canopy and forest floor. Longer-term studies on the effects of canopy inputs such as frass and throughfall may disclose even more interactions among these inputs, soil fauna, and soil processes such as decomposition, nutrient cycling and soil respiration. Our results add to the growing evidence that forest canopies and soils are linked through the process of herbivory (Chew 197; Mattson & Addy 1975; Kitchell et al. 1979; Swank et al. 1981; Seastedt & Crossley 198; Grace 1986; Hollinger 1986; Schowalter et al. 1991; Risley & Crossley 1992; Lovett & Ruesink 1995; Reynolds et al. 2000; Reynolds and Hunter 2001; Hunter 2001). We have shown that decomposition and microarthropod abundances are greater at high vation than at low vation at the Coweeta Hydrologic Laboratory. This supports previous work by Hoover and Crossley (1995) using different litter species. In addition, we observed increases in four groups of soil fauna (collembola, prostigmatid mites, bacterial-feeding nematodes, and fungal-feeding nematodes) following the addition of herbivore-derived inputs to the forest floor. For collembola, at least, the addition of insect frass had a strong, positive effect on their abundance, suggesting that herbivore inputs from the forest canopy play a role in the spatial and temporal dynamics of a microarthropod that has important effects on soil processes (Verhoef & Meintser 1991; Hopkin 1997). Acknowledgments. The authors wish to thank Erich Tilgner (University of Georgia) for supplying walking sticks and Dylan Parry (Michigan State University) for providing us with eggs of forest tent caterpillars. Thanks to Christien Ettema for instructions on nematode identification. Many University of Georgia undergraduates were instrumental in field and laboratory chores. We thank Richard Warzinski, Kimberly Deal, Kimberly Holton, Jason Leathers, Abigail Feinstein, Alex Reynolds, Mathew Wood, Robert Zinzan and Scott Connelly. Members of the Hunter lab group- Rebecca Klaper, Rebecca Forkner, Kathrine Kearns, Alissa Salmore and Michael Madritch - assisted in many aspects of this investigation. Two anonymous reviewers provided many helpful suggestions. Research was supported by the University of Georgia (Graduate School University-Wide Assistantship to BCR), NSF Grant DEB to the Coweeta LTER program, NSF Grant DEB to MDH, M. man, and T. Schowalter, and the Andrew W. Mellon Foundation. References Agresti, A. (1996) An Introduction to Categorical Data Analysis. John Wiley and Sons, New York. Behan-Pelletier, V., Newton, G. ( 1999) Linking soil biodiversity and ecosystem function - The taxonomic dilemma. Bioscience 9, Bolstad, P. V., Vose, J. M., McNulty, S. G. (2001) Forest productivity, leaf area, and terrain in southern Appalachian deciduous forests. Forest Science 7, 19^27. Chew, R. M. (197) Consumers as regulators of ecosystems: An alternative to energetics. The Ohio Journal of Science 7, Coleman, D.C., Crossley, D.A. Jr. (1996) Fundamentals of Soil Ecology. Academic Press, San Diego. Crawley, M. J. (199) GLIM for Ecologists. Blackwell Scientific Inc., Oxford. Cromack, K. Jr. (197) Litter production and decomposition in a mixed hardwood watershed and a white pine watershed at Coweeta Hydrologic Station, North Carolina. Ph. D. thesis. University of Georgia. Athens, GA. Crossley, Jr. D.A., Hoglund, M.P. (1962) A litter-bag method for the study of microarthropods inhabiting leaf litter. Ecology, Day, F. P. Jr, Phillips, D. L., Monk, C. D. (1988) Forest communities and patterns. In: Swank, W. T., Crossley, D. A. Jr. (eds) Ecological Studies, Vol. 66: Forest Hydrology and Ecology at Coweeta. Springer-Verlag, New York. pp Eshleman, K.N., Morgan, R.P. II, Webb, J.R., Deviney, F. A., Galloway, J.N. (1998) Temporal patterns of nitrogen leakage from mid-appalachian forested watersheds: Role of insect defoliation. Water Resources Research, Ettema, Ch. (1998) Soil nematode diversity: Species coexistence and ecosystem function. Journal of Nematology 0, Freckman, D. W., Whitford, W. G., Steinberger, Y. (1987) Effect of irrigation on nematode population dynamics and activity in desert soils. Biology and Fertility of Soils, -10. Freckman, D. W, Baldwin, J.G. (1990) Nematoda. In: Dindal, D.L. (ed) Soil Biology Guide. John Wiley and Sons, New York. pp Grace, J. R. (1986) The influence of gypsy moth on the composition and nutrient content of litter fall in a Pennsylvania oak forest. Forest Science 2, Hansen, R. (1997) Effects of habitat heterogeneity and composition on the abundance, diversity and functional impact of the oribatid mite assemblage in leaf litter. Ph. D. thesis, University of Georgia, Athens, GA. Hollinger, D. Y. (1986) Herbivory and the cycling of nitrogen and phosphorus in isolated California oak trees. Oecologia 70, Pedobiologia (200) 7,127-19

11 Effects of canopy herbivores 19 Hoover, C. M., Crossley, D. A. Jr. (1995) Leaf litter decomposition and microarthropod abundance along an altitudinal gradient. In: Collins, H.P., Robertson, G.P., Klug, M.J. (eds) The significance and regulation of soil biodiversity. Kluwer Academic Publishers, Netherlands, pp Hopkin, S.P. (1997) Biology of the Springtails (Insecta, Collembola). Oxford University Press, Oxford. Hunter, M.D. (2001) Insect population dynamics meets ecosystem ecology: effects of herbivory on soil nutrient dynamics. Agricultural and Forest Entomology, Hunter, M.D., Price, P. (1992) Playing chutes and ladders: Heterogeneity and the relative roles of bottom-up and topdown forces in natural communities. Ecology 7, Kareiva, P. (199) Space: the final frontier for ecological theory. Ecology 75, 1. Kitchell, J.F., O'Neill, R.V., Webb, D., Gallepp, G.W., Bartell, S.M., Koonce, J.F., Ausmus, B.S. (1979) Consumer regulation of nutrient cycling. Bioscience 29,28-. Knoepp, J.D., Swank, W.T. (1998) Rates of nitrogen mineralization across an vation and vegetation gradient in the southern Appalachians. Plant and Soil 20, Louda, S. M., Rodman, J. E. (198) Ecological patterns in the glucosinolate content of a native mustard, Cardamine cordifolia, in the Rocky Mountains. Journal of Chemical Ecology 9, Lovett, G. M., Ruesink, A. E. (1995) Carbon and nitrogen mineralization from decomposing Gypsy moth frass. Oecologia 10, man, M., Wittman, P.K. (1996) Forest Canopies: Methods, Hypotheses, and future directions. Annual Review of Ecology and Systematics 27, Lussenhop, J. (1992) Mechanisms of microarthropod-microbial interactions in soil. In: Begon, M., Fitter, A. H. (eds) Advances in Ecological Research 2,1-. Mallow, D., Crossley, D.A. Jr. (198) Evaluation of five techniques for recovering postlarval stages of chiggers (Acarine:Trombiculidae) from soil habitats. Journal of Economic Entomology 77, Maraun, M., Alphei, J., Beste, P., Bonkowski, M., Buryn, R., Migge, S., Peter, M., Schaefer, M., Scheu, S. (2001) Indirect effects of carbon and nutrient amendments on the soil meso- and microfauna of a beechwood. Biology Fertility Soils, Mattson, W.J., Addy, N.D. (1975) Phytophagous insects as regulators of forest primary production. Science 190, Moore, J.C., Walter, D.E., Hunt, H.W. (1988) Arthropod regulation of micro- and mesobiota in below-ground detrital food webs. Annual Review of Entomology, Parker, G. (1989) Vertical profile and canopy organization in a mixed deciduous forest. Vegetatio 85, Reynolds, B.C., Crossley, D.A. Jr. (1995) Use of a canopy walkway for collecting arthropods and assessing leaf area removed. Selbyana 16, Reynolds, B.C., Crossley, D.A. Jr. (1997) Spatial variation in herbivory by forest canopy arthropods along an vation gradient. Environmental Entomology 26, Reynolds, B.C., Hunter, M.D., Crossley, D.A. Jr. ( 2000) Effects of canopy herbivory on nutrient cycling in a northern hardwood forest in western North Carolina. Selbyana 21,7-78. Reynolds, B. C., Hunter, M. D. (2001) Responses of soil respiration, soil nutrients, and litter decomposition to inputs from canopy herbivores. Soil Biology and Biochemistry, Risley, L.S., Crossley, D.A. Jr. (1988) Herbivore-caused greenfall in the southern Appalachians. Ecology 69, Risley, L. S., Crossley, D. A. Jr. (1992) Contribution of herbivore-caused greenfall to litterfall N flux in several southern Appalachian forested watersheds. American Midland Naturalist 129, SAS version 6.12 for Windows TSO20. (1996) Sas Institute, Inc. Gary, NC. Schowalter, T.D., Sabin, T.E., Stafford, S.G., Sexton, J.M. (1991) Phytophage effects on primary production, nutrient turnover, and litter decomposition of young Douglas fir in western Oregon. Forest Ecology and Management 2, Seastedt, T.R. (198) The role of microarthropods in decomposition and mineralization processes. Annual Review of Entomology 29, 25^6. Seastedt, T.R., Crossley, D.A.Jr. (198) The influence of arthropods on ecosystems. Bioscience, Swank, W.T., Waide, J.B., Crossley, D.A.Jr., Todd, R.L. (1981) Insect defoliation enhances nitrate export from forest ecosystems. Oecologia 51, Swank, W.T., Crossley, D.A.Jr. (eds) (1988) Ecological Studies, Vol. 66, Forest hydrology and ecology at Coweeta. Springer-Verlag, New York. Swift, L.W.Jr., Cunningham, G.B., Douglass, J.E. (1988) Climatology and hydrology. In: Swank, W.T., Crossley, D.A.Jr. (eds) (1988) Ecological Studies, Vol. 66. Forest hydrology and ecology at Coweeta. Springer-Verlag, New York. Swift, M. J., Heal, O.W., Anderson, J.M. (1979) Decomposition in terrestrial ecosystems. University of California Press, Berky. Verhoef, H. A., Meintser, S. (1991) The role of soil arthropods in nutrient flow and the impact of atmospheric deposition. In: Veeresh, G. K., Rajagopal, D., Viraktamath, C. A. (eds) Management and Conservation of Soil Fauna. Oxford and IBH Pub. Co., New Delhi. Wall, D.H., Moore, J.C. ( 1999) Interactions underground: Soil biodiversity, mutualism and ecosystem processes. Bioscience 9, Walter, D., Proctor, H. (1999) Mites: Ecology, Evolution, and Behavior. CABI Publishing, New York. Yeates, G. W. (1998) Feeding in free-living soil nematodes: A functional approach. In: Perry, R. N., Wright, D. J. (eds) The physiology and biochemistry of free-living and plantparasitic nematodes. CABI Publishing, New York. Yeates, G.W, Bongers, T., de Goede, R.G.M., Freckman, D. W, Georgieva, S. S. (199) Feeding habits in soil nematode families and genera-an outline for soil ecologists. Journal of Nematology 25, Pedobiologia (200) 7,127-19