Niklas Lindberg, Department of Ecology and Environmental Research, Swedish University of Agricultural Sciences, PO Box 7072, SE-750 07 Uppsala, Sweden (fax: + 46 18 67 34 30; e-mail Niklas.Lindberg@eom.slu.se).
1The effects of experimental long-term summer drought and irrigation on soil fauna were studied in a Norway spruce stand in south-western Sweden. The treatments, carried out over 8 and 10 years respectively, were chosen to simulate two scenarios of climate change, involving drier and wetter summers.
2Different microarthropod communities developed in the different treatments. The abundances of enchytraeids, mesostigmatid mites and macroarthropod predators were all lowest in the drought plots. Drought decreased and irrigation increased the abundance and diversity of Oribatida. Drought decreased the abundance of Collembola.
3The dominance structure of Oribatida and Collembola also changed, but less markedly. Drought affected community composition of both groups more than irrigation.
4The study confirms that soil microarthropods can be useful environmental indicators, but their responses did not support the widely held view that deviations from a log-normal dominance structure indicates a stressed community.
5The results also indicate that a drier climate with summer drought will lead to the local extinction of some soil animal species in this region.
Many of the anticipated temperature changes due to emissions of greenhouse gases will disproportionately affect high latitudes (Schlesinger 1997). So far, effects of global change on soil fauna of polar zones have received more attention than effects on the temperate fauna (Strathdee et al. 1995; Hodkinson et al. 1998; Webb et al. 1998), despite the fact that temperate forest soils contain some of the most diverse terrestrial communities on earth (Anderson 1978). Together with enchytraeids (Oligochaeta: Enchytraeidae), collembolans and oribatid mites (Acari: Oribatida) constitute a diverse and numerically important part of the soil decomposer community in these forests (Wallwork 1983; Seastedt 1984; Didden 1993). Temperate soil faunal communities will be affected by alterations in climatic conditions, for example through temperature and moisture (Briones, Ineson & Piearce 1997; Frampton, van den Brink & Gould 2000a,b), but there is a lack of data from forest soils.
Furthermore, an understanding of the possible effects of climate changes on soil fauna at the species level is important because the ecological characteristics of the dominant species in the community may govern important ecosystem functions (Faber & Verhoef 1991; Schwartz et al. 2000). Even species within the same broadly defined functional group can differ in their ecology enough to affect ecosystem processes (Mebes & Filser 1998; Wall & Virginia 1999). Consequently, the prediction of functional aspects of changes in soil faunal diversity requires knowledge about the specific characteristics of the different species (Behan-Pelletier & Newton 1999).
Both decreased and increased precipitation may occur following global change (Melillo, Hall & Ågren 1996; Piervitali, Colacino & Conte 1997; Arnell 1999). The predicted climatic changes for Sweden include increased annual mean precipitation together with an increased mean temperature (SMHI 2000). However, for southern Sweden summer precipitation is predicted to decrease, increasing the risk of summer droughts, while northern regions of Scandinavia are predicted to face moister summers (SMHI 2000). By experimentally simulating frequent summer droughts and, as a contrast, increased precipitation, we aimed to study the responses of the most important mesofaunal decomposer groups (Enchytraeidae, Oribatida, Collembola) in coniferous soils. We examined the effects on species, community composition and diversity. Because changes may be due to the direct effects of moisture or indirect effects of food resources, predation and habitat heterogeneity, we also examined the main predators in the soil food web.
We expected the communities to be less adapted to dry compared with moist conditions due to the naturally high precipitation in south-western Sweden. Our main hypotheses were: (i) long-term drought will reduce soil fauna abundance and diversity; (ii) shifts in the dominance structure of the communities from a log-normal distribution will occur following the change in moisture, at least in the stressed communities in the drought plots; (iii) the species composition of Collembola and Oribatida communities will change towards a dominance of drought-resistant species, while drought-sensitive species will disappear.
Materials and methods
The work was carried out in 1997 in a stand of Norway spruce Picea abies (L.) Karst. at Skogaby, south-western Sweden (56°33′N, 13°13′E), which by then was 33 years old (Persson et al. 2000). The distance to the coast is 16 km and the altitude is 95–115 m above mean sea level. The climate is maritime, with a mean annual precipitation of c. 1150 mm and an annual mean temperature of 7·5 °C. The soil is a poorly developed podzol on a sandy-loamy till layer of > 2 m. The O horizon consists of L, F and H layers, ranging from 5 to 10 cm in thickness, and the F + H layer has a pH (H2O) of about 4·0.
The homogeneous Norway spruce stand was planted as 2-year-old seedlings in 1966, replacing a first generation of Scots pine Pinus sylvestris L. planted in 1913 on former Calluna heathland. Average tree height in 1995 was 15 m and the tree density on the plots after the last thinning in the autumn of 1993 was on average 1420 stems ha−1 (U. Johansson, personal communication). The stand had an average leaf area index (LAI) of 7 and, due to the closed canopy, less than 2% of the incident light penetrated to the ground. No field or shrub layer existed and only some mosses and scattered tussocks of grass were present. Soil conditions and stand type were typical for large areas of managed forest in southern Sweden (Bergholm et al. 1995).
We used plots that were part of a large field experiment established in 1988 and designed to study effects of water and nutrient availability on tree growth (Nilsson 1997). It had a randomized block design with four blocks, each containing one replicate of each treatment. We used three of the seven treatments, namely control (no treatment), drought and irrigation. The irrigation treatment started in 1988 and the drought treatment in 1990. Plot size was 45 × 45 m, except for the drought plots which were 22·5 × 22·5 m. The drought plots had transparent plastic roofs placed 1–1·5 m above ground. The roofs had openings for tree trunks and for maintenance purposes (Fig. 1) but prevented 70% of the throughfall on the plots from reaching the ground during April–September (Alavi 1999). Stem flow comprised only 1% of the throughfall (Alavi 1999). A pipe system led the rainwater from the roofs away from the plots. During winter the roofs were removed and all throughfall could reach the ground, resulting essentially in a summer drought treatment (Fig. 2). Litter falling on the roofs was spread on the ground at roof removal in September–October. During the period of roof coverage approximately 40% of the yearly litterfall was taking place (L.-O. Nilsson, personal communication). In the irrigation plots, water was added during May–September using sprinklers placed 20 cm above the ground, one per 25 m2 of the plot. The plots were irrigated when a 20-mm water storage deficit in the soil had developed but more than 10 mm 24 h−1 was never given (Table 1).
Table 1. Precipitation and additional irrigation on the irrigation plots during the treatment months April–September 1988–97 (J. Bergholm and U. Johansson, unpublished data). Throughfall on the plots was estimated to be 50–60% of the annual precipitation (Alavi 1999)
Number of irrigation days
The plots were sampled on 20 October 1997, 13 days after the roofs had been removed. Arthropods were collected by taking four systematically distributed 100-cm2 × 10-cm deep subsamples of the litter and humus layers from each plot. The sampling points were evenly distributed over the plots at a distance of 7·5 m in from the centre of each plot side. When sampling in drought plots we avoided areas closer than 1 m from tree trunks and gaps in the roof coverage. In the laboratory the subsamples were stored at 4 °C for 12 h before pooling to one sample per layer and plot for extraction in Tullgren dry funnels for 4 days.
High-gradient extractions are sometimes considered more effective in soil microarthropod extraction (Takeda 1979; Van Straalen & Rijninks 1982). To correct for possible inefficiencies of the Tullgren extraction method as regards small microarthropods, additional samples were taken for high-gradient extraction by collecting two 10-cm deep soil cores with a 10 cm2 surface area from each plot, adjacent to two of the 100-cm2 sampling points. Twelve hours later they were divided into 2-cm depth intervals and extracted in Macfadyen high-gradient canister (HGC) extractors (Macfadyen 1961) for 4 days. For calculations of the total abundances of Oribatida, Mesostigmata and Collembola, the HGC method was used. For a limited number of small soft-bodied Oribatida and Collembola this method was also used to calculate the species-specific abundances. However, for calculations of the densities of macroarthropod predators and most species-specific densities of Oribatida and Collembola, data from the Tullgren extractions were used because of the larger numbers of animals in these samples (Table 2 and Appendix). The HGC sampling, with its shortcomings in the form of small soil samples, is not suitable for the larger, sparsely distributed species.
Table 2. Mean densities m−2 (SE) of the faunal groups studied in the Skogaby experiment (n = 4). HGC, high-gradient canister extractor; T, Tullgren funnel; B, Baermann funnel. P-values of treatment effect according to two-way anova, d.f. = 2,6. Within each group, numbers with different letters are significantly different (Tukey's studentized range test, P < 0·05)
The extracted animals were preserved in 70% ethanol. Identification and enumeration were done under a binocular microscope, or under a conventional transmission microscope with a 40–400 magnification for the smaller animals. All Collembola and all adult stages of Oribatida were determined to species.
Mesostigmatid mites (Acari: Mesostigmata) were counted as a group due to time constraints, and included Uropodina, Zerconina, Parasitina and Dermanyssina (Walter & Proctor 1999). Macroarthropod predators were also grouped together and included Chilopoda, Pseudoscorpiones, Araneae, Opiliones, Formicidae, Carabidae, Staphylinidae and larvae of Cantharidae, Elateridae and Empididae. Other predators, i.e. predaceous prostigmatid mites (Acari: Prostigmata), were very few and were not counted. Large litter decomposers such as Diplopoda, Isopoda and Gastropoda were almost entirely absent at the site and none was found in our samples. Earthworms (Oligochaeta: Lumbricidae) were also scarce and only a few specimens of Dendrobaena octaedra Sav. were found.
Enchytraeids were collected by taking four systematically distributed soil cores of 4·5 cm diameter to a depth of 15 cm from each of the 12 plots. Each core was divided into three subcores, each 5 cm high. The enchytraeids were extracted from each subcore in modified Baermann funnels (O’Connor 1962) and counted alive. All enchytraeids were determined, whenever possible, to species.
Oribatid mites were identified using Willmann (1931), Gilyarov & Krivolutsky (1975) and special keys for Oppioidea by Woas (1986) and the genus Carabodes by Sellnick & Forsslund (1953). Species determination of the Oribatida family Brachychthoniidae was judged to be too time-consuming, but two different taxa were easily identified: Liochthonius sp. and Brachychthoniidae sp. (see Appendix). Due to the difficulties concerning identification and species definition in the family Suctobelbidae (Horak 1997), we classified all specimens as Suctobelbella spp. although it was possible to identify two species in the material (Appendix). Cases where species identification remained uncertain are marked with ‘cf.’ in Appendix. The Oribatida nomenclature largely follows Niemi, Karppinen & Uusitalo (1997).
To detect significant treatment effects, two-way analysis of variance (anova) was carried out with block and treatment as fixed factors. The experiment was treated as a randomized block design and the treatment categories used were drought (D), irrigation (I) and control (C). Significant treatment effects according to the anova were investigated further using Tukey's studentized range test (P < 0·05). Normality and independence of residuals were checked visually and by using the Shapiro-Wilks’ test of normality. All analyses of abundances were done with ln(x + 1) transformed data.
To enable comparisons with the literature, and to reduce the risks associated with using only one statistic (Huston 1994), we measured faunal diversity in several ways. For the Collembola and Oribatida we calculated the following measures at the community level: total number of individuals, total number of species (species richness), the Shannon–Wiener diversity index H′ (Magurran 1988), Fisher's alpha diversity index (Magurran 1988; Krebs 1989) and the evenness index J′ (Krebs 1989).
Widely used in ecological studies, H′ is largely independent of sample size and is insensitive to the presence of rare species (Anderson 1978). Fisher's α was calculated using software from Krebs (1989). It is insensitive to sample size and has been recommended as a standard diversity index (Magurran 1988; Rosenzweig 1995). The evenness index was used as a measurement of the evenness of the relative species’ abundances.
The number of species in a sample is a function of both sample size and the underlying species–abundance distribution. To overcome the effect of sample size (number of individuals) on species number S, species richness was standardized using rarefaction (Krebs 1989). In this procedure, all faunal samples from the same treatment were pooled together to give the treatment-specific species–abundance distribution for Collembola and Oribatida, respectively. Differences between the treatments were examined by rarefying each sample to the size of the smallest one. For rarefaction analyses and for testing the observed species–abundance distribution against the log-normal and log-series distributions, software from Krebs (1989) was used. Octave classes were used on the log-abundance axis. Goodness-of-fit tests were done using the Kolmogorov–Smirnov one-sample test (Zar 1984).
To determine if the species assemblages differed between the treatments, we analysed the community composition of Collembola and Oribatida in the Tullgren samples by principal component analysis (PCA). The sample scores along the first three principal components were used in further statistical analyses. Species with less than seven individuals in total or occurring in less than three of the 12 plots were down-weighted by excluding them from the PCA, making the number of included Oribatida species 27 and Collembola species 19. Statistical tests used the SAS package, version 6·12 (SAS Institute 1996).
effects on total abundance
The abundance of Oribatida was significantly higher in the irrigated plots (317 000 individuals m−2; Table 2) and the control plots (220 000 individuals m−2) than in the drought plots (70 200 individuals m−2). The numbers of Collembola per m2 were highest in the irrigated plots, on average 66 000 by high-gradient extraction, and only 12 800 in the drought plots (Table 2). The control plots were intermediate and not significantly different from the two other treatments. The same treatment gradient was evident for both Oribatida and Collembola in the samples extracted with Tullgren funnels, in spite of the lower extraction efficiency (data not shown).
Similarly, Mesostigmata abundance was significantly lower in the drought plots and highest in the irrigated plots (Table 2). The abundance of macroarthropod predators was also significantly lower in the drought plots than in the other two treatments (Table 2). Furthermore, enchytraeid abundance was affected by the treatments, with significantly higher densities in the irrigated plots (on average 97 400 individuals m−2; Table 2) and the control plots (36 600 individuals m−2) than in the drought plots (3300 individuals m−2).
effects on species diversity of collembola and oribatida
In all, 45 species of Oribatida and 25 species of Collembola were found (Appendix). No additional species were found in the HGC samples compared with the Tullgren samples. The number of oribatid mite species unique to only one of the treatments was 1 for drought, 0 for control and 10 for irrigation. The corresponding numbers for Collembola species were 5 for drought, 1 for control and 0 for irrigation (Appendix).
The number of Oribatida species was lowest in the drought treatment and highest in the irrigation treatment, and the drought plots differed significantly from the other treatments (Table 3). In samples of the same sampling area, only 20 species on average were found in the drought plots, while 26 and 31 species were found in the control and irrigated plots, respectively. The number of Collembola species was significantly lower in the drought plots than the control plots (Table 3), while it was intermediate in irrigated plots and not significantly different from the other treatments.
Table 3. Goodness-of-fit tests of species–abundance distributions and diversity (measured by α, H′, J′, S and S500; see text) of the Oribatida and Collembola communities in different treatments (n = 4). Numbers (SE) based on extraction of 400-cm2 soil samples in Tullgren funnels. Fits to log-normal and log-series distributions were done using the Kolmogorov–Smirnov one-sample test (see text). F- and P-values of treatment effect according to two-way anova, d.f. = 2,6 (NS = P > 0·05). S500 indicates expected species number at a sample size of 500 individuals. Within each group and test, values with different letters are significantly different (Tukey's studentized range test, P < 0·05)
Shannon- Wiener H′
Mean species number S
P < 0·001
P < 0·001
When correcting for sample size by rarefaction, the rarefaction curves showed a decrease in species richness in the drought plots for both Oribatida and Collembola (Fig. 3a,b). Furthermore, the two groups differed in that the Collembola were richest in the control plots, while the Oribatida were richest in the irrigated plots. Species richness for all the three treatments differed for the Oribatida at S= 500 (Table 3), while in the Collembola the irrigated and drought plots had lower diversity than the control plots, but the two treatments partly overlapped for some sample sizes (Fig. 3b).
Evenness was negatively affected by drought in Oribatida (Table 3). The Shannon–Wiener index of diversity was similarly affected by the moisture treatments, the drought plots having the lowest diversity of Oribatida (Table 3) and the irrigation plots the highest. For both indices Collembola showed the lowest values in drought plots but differences were not significant (Table 3).
Fisher's α diversity index also showed a clear treatment effect for the Oribatida, with a significantly higher value for the irrigated plots than the other two treatments (Table 3). However, there was no significant treatment effect on Collembola.
The species–abundance distributions did not differ from log-normal in any case (P > 0·05; Table 3). A significant deviation from the log-series distribution was found for Oribatida in control (dmax= 15, k= 9, n= 35, P < 0·001; Table 3) and irrigated (dmax= 19, k= 11, n= 43, P < 0·001) plots. In these cases, for the large number of individuals found in the samples, the number of species was too low to fit the log-series distribution.
A visual examination of the rank–abundance curves for both Oribatida and Collembola in the drought plots indicated that the subdominant species showed a decreased abundance relative to the dominant ones in this treatment (Fig. 4a,b). The dominant oribatids (irrespective of species identity) were not more abundant in the drought or irrigation treatments than in the control. However, due to a mass occurrence in one sample, the most dominant collembolan species in the drought treatment (Xenylla brevicauda) was more numerous than the dominant species in the other treatments (Mesaphorura macrochaeta). Species of the genus Xenylla are known for their aggregated distribution (Hopkin 1997). In addition, the curve describing the Oribatida community in the irrigated plots had a ‘tail’ of rare species that did not occur in the other treatments.
We examined the variation in rank abundances between treatments in more detail in two ways. First, we calculated Kendall's coefficient of concordance W (a multisample rank correlation statistic ranging from 0 to 1, where 1 indicates complete similarity in relative abundances between samples; Bengtsson 1994). We used the Tullgren samples (which were also used for Fig. 4a,b) and deleted higher taxa than genera and taxa with < 3 individuals sampled (cf. Bengtsson 1994). For both Oribatida and Collembola, relative abundances were significantly similar between treatments (Oribatida: W= 0·72, χ2 = 82·3, d.f. = 38, P < 0·001; Collembola: W = 0·56, χ2 = 37·0, d.f. = 22, P < 0·05). Enchytraeid relative abundances were also fairly similar between treatments, but not significantly so (W = 0·81, χ2 = 7·3, d.f. = 3, NS but P < 0·1). This indicated that the treatments did not result in large changes in relative abundances. However, the changes in soil fauna composition due to the treatments in the present study were larger than the year-to-year changes on single mature forest sites found by Bengtsson (1994; average W= 0·80) and the changes between treatments and between years found in a whole-tree harvesting experiment (Bengtsson, Persson & Lundkvist 1997; W-values > 0·80).
Secondly, we examined how many of the species showed large shifts in rank between the treatments, considering a rank shift > 10 as being large. Of the 38 Oribatida species, 20 exhibited large rank shifts between the treatments, the largest being a shift of 34 for Zygoribatula exilis (Appendix). For the less species-rich Collembola, nine of 23 species showed large rank shifts, the largest being 16·5 for Folsomia quadrioculata (Appendix).
Taken together, our results suggest that the rank abundance distributions and relative abundances were fairly similar between the treatments in our study (Table 3; W-tests above). However, for both Oribatida and Collembola, there was a number of species that changed substantially in relative abundance between the treatments, indicating that some species indeed shifted between being rare and common along the moisture gradient from drought to irrigation.
effects on species diversity of enchytraeidae
Only five enchytraeid species were found in the samples (Appendix). Of these, Cognettia sphagnetorum made up 66% of the sampled 3486 individuals. Of the remaining species, Achaeta sp. was the most abundant at 33%. No differences in species number between the treatments could be detected (F = 2·64, d.f. = 2,6, P= 0·15). Because of the low number of individuals belonging to other species than C. sphagnetorum and Achaeta sp., we made no further analyses of species richness and diversity.
species-level effects on collembola and oribatida
In total, the abundance of 14 species of Oribatida and 10 species of Collembola showed significant treatment effects (Appendix). Of these, nine Oribatida species and six Collembola had significantly lower densities in the drought plots than in the control plots, while four species of Oribatida and two of Collembola responded positively to irrigation compared with the control plots.
To ensure that the extraction method chosen for the abundance calculations did not have a large impact on the species-wise results, we compared data from both extraction methods. Of the 14 oribatid species, six (Atropacarus striculus, Rhysotritia sp., Nothrus silvestris, Nanhermannia coronata, Suctobelbella spp. and Parachipteria punctata) showed the same significant effects for both methods. Seven Collembola species showed significant treatment effects in the Tullgren samples (P < 0·05) but not in the HGC samples. Of the three small Collembola showing significant treatment effects in the HGC samples, Micraphorura absoloni also showed the same result in the Tullgren samples. However, for both Oribatida and Collembola, HGC samples often contained too few individuals of larger species to get significant results.
The community composition of Oribatida and Collembola differed between the treatments. The PCA plot showed a very distinct group consisting of the communities in the drought plots, while the communities in the control and irrigated plots differed less (Fig. 5a). The first three PCA axes could explain 62% of the total variation in the data set, of which the PC1 axis alone accounted for 37%. On the first three axes anovas showed a significant treatment effect along PC1 (F = 104·74, d.f. = 2,6, P < 0·001) and PC2 (F = 6·40, d.f. = 2,6, P < 0·05). The effect of block was not significant.
The species showing significant species-wise differences between treatments (Appendix) could be found to the right and in the cluster at the lower right of the PCA plot, with the exception of those found in higher abundance in drought plots (Zygoribatula exilis to the left in Fig. 5b).
Enchytraeid, oribatid mite and Collembola abundances were reduced by long-term summer drought compared with the control plots. Furthermore, even if the decreased abundances were accounted for, drought resulted in a lower diversity of Oribatida. Although our study cannot estimate the magnitude of the effects of large-scale global warming on these soil organisms, it indicates that changes in precipitation may have significant effects on soil fauna, with potential consequences for ecosystem functions.
In accordance with our prediction, the effects of the drought treatment on community and rank–abundance structure of microarthropod decomposers were larger than irrigation effects (Figs 4a,b and 5a). However, the measurements of soil moisture during 1992–94 (Fig. 2) showed that drought plots on average differed more from control plots than did irrigated plots. Therefore we cannot conclude from our study that in general drought effects will be larger than effects of increased moisture levels.
The intensities of our treatments were probably not unrealistic compared with natural conditions. June 1992 was very dry in south-west Sweden (SMHI 1992) and consequently soil moisture content on control plots was very close to drought plots at that time (Fig. 2). Further, occasions when soil moisture differences were slight between irrigated and control plots in the treatment seasons of 1992–94 (Fig. 2) were all in connection with rainy months (data not shown). Therefore we are confident that treatment intensities were realistic, but that frequency and longevity of the treatments were increased compared with current conditions.
The PCA showed that different communities developed in the different moisture treatments. The pattern in Fig. 5a,b indicates that the PC1 axis represents a moisture gradient. Briones, Ineson & Piearce (1997) also found that soil fauna community composition was largely determined by temperature–moisture gradients in a moorland ecosystem. Frampton, van den Brink & Gould (2000a,b) showed that spring irrigation affected most farmland Collembola species positively, while spring drought had a negative effect.
The results are most unlikely to be extraction or sampling artefacts, because abundance differences were consistent between the HGC-extracted and Tullgren-extracted samples. The limited number of soil samples per plot means that we mainly detected large and consistent differences in abundance and diversity, i.e. real biological differences between treatments.
By taking our samples > 7·5 m in from the plot edges, we avoided edge effects. Because the plots were situated within a growing forest, edge effects on light penetration, wind, temperature, etc., were minor. In any case, they would homogenize the communities between treatments, not create differences.
Differences in tree density or litterfall between the treatments plots are not a likely cause of the patterns found. The average stem density on the plots did not differ between treatments, nor did the average litterfall amounts between 1989 and 1996 (L.-O. Nilsson, personal communication). Only 40% of the litterfall normally takes place during the time of the year when roofs were present, and if the change in litterfall dynamics was a problem we would particularly expect early colonizers among Collembola and Oribatida to be negatively affected by the reduced litterfall during the 5 months preceding the soil sampling. However, most Oribatida are intermediate or late colonizers of litter (after 6–12 months) (Hågvar & Kjøndal 1981; Hasegawa & Takeda 1996; Hasegawa 1997; Irmler 2000), and among the Collembola early colonizers include species and genera like Entomobrya corticalis, Orchesella flavescens, Lepidocyrtus cyaneus and Xenylla sp., taxa that were present in or even characteristic of the drought plots.
It is also unlikely that the roofs in themselves affected dispersal of soil fauna negatively and thereby contributed to the drought effect. Mobile tree-climbing Collembola in our plots (mainly Orchesella and Entomobrya spp.) were mostly found in the roof-covered drought plots (Appendix), contrary to this suggestion. Additionally, arboreal and air-dispersed species only comprised a very small part of the oribatids and predatory arthropods found in the plots.
All sampling cores had organic layers exceeding 8 cm in depth and did not differ markedly from each other. Therefore, effects of differences in organic layer thickness could be excluded. The results from the HGC extraction showed no differences in faunal abundances below 4 cm depth.
Data on microclimatic changes resulting from the roofs would have been informative. The roofs may have reduced temperature fluctuations. However, the dense forest (LAI = 7), the gaps in roof coverage (Fig. 1) and the possibility of having air circulating underneath roofs from the sides probably reduced the roof effect on temperature and light conditions.
The observed differences between treatments are unlikely to be unique to the year studied (1997). Preliminary samples from October 1996 showed a clear treatment effect on Enchytraeidae in the same direction as 1997. Similarly, additional samples taken from the same plots in June and October 1996 and extracted in Tullgren funnels, showed a similar pattern for Oribatida and Collembola as reported here (data available from the correspondence author on request).
effects on abundance and diversity
The differences in diversity found in our study cannot be explained by variation in sample size, i.e. the number of individuals sampled, as rarefaction showed differences in the number of species even after adjustment for sample size. Because the drought and irrigation treatments had been applied annually for 8 and 10 years, respectively, the treatments were not short-term disturbances simply excluding species with low tolerance to, for example, drought. Long-term differences between the treatments in environmental conditions, resources and animal community structure may possibly have affected the diversity pattern.
Several explanations can be rejected. The differences in predator abundance followed the differences in prey abundance (Table 2), and hence general predator pressure was unlikely to have caused differences in prey diversity between the treatments. Most of the mesostigmatid mites encountered, as well as the Araneae, Coleoptera and Diptera families found, are generalist predators. We find it unlikely that differences in predation pressure were a major factor for the diversity patterns, although we cannot exclude this possibility.
Vegetation structure can be important for the distribution of Collembola (Petersen 1995) and Oribatida (Steiner 1995), but we avoided the few vegetated patches of mosses and grasses when sampling. Furthermore, the Collembola species specific to the drought treatment (Appendix) were corticolous and/or surface-living species and not associated with plant cover. Hence, variation in plant cover cannot have had effects on diversity in this experiment.
The general pattern for all the faunal groups studied was a decrease in abundance with drought and an increase with irrigation. Enchytraeids and several collembolan and oribatid mite species are sensitive to drought (for Enchytraeidae, Lundkvist 1982; Didden 1993; Briones, Ineson & Piearce 1997; for Collembola, Verhoef & Witteveen 1980; Vegter 1983; Frampton, van den Brink & Gould 2000a,b; for Oribatida, Siepel 1996). The collembolans that decreased or disappeared in the drought plots were mostly living at intermediate litter layers, while the five species of Collembola only found in the drought plots were all tree-climbing and surface-dwelling ones. Pore size and dehydration resistance may have excluded some species in the drought plots, by affecting either the adult, juvenile or egg stages. Most Oribatida species that differed in abundance between treatments were larger species. These were probably unable to reach refugia with moist microhabitats deeper in the soil during dry conditions, resulting in lower abundance and diversity in the drought plots.
Spatial and temporal variability of soil conditions and, hence, the number of available niches and the possibility to specialize on particular conditions, may have been different between the treatments. Soil arthropod diversity is often positively correlated with habitat diversity (Anderson & Hall 1977; Anderson 1978; Hansen & Coleman 1998). Irrigated plots may have had lower microhabitat diversity due to the constant conditions provided by the irrigation treatment. On the other hand, stable and predictable environments, such as the irrigated plots, may also allow species to specialize and, hence, increase diversity (May 1975). The enchytraeids and insect larvae might have been important in creating suitable microhabitats for mesofauna in the control and irrigated plots. Whether microhabitat heterogeneity or environmental stability was more important for the patterns seen we cannot say.
Soil animal communities are regarded as being diverse because of an excess of food or a large number of available microhabitats (Anderson 1975a,b; Giller 1996). In our case, both factors may have decreased the possibilities of niche partitioning for the mesofauna in the drought plots. In addition, direct effects of drought may have excluded several species. However, more detailed experiments and observations of interactions between, for example, fungi and fungivores are needed to examine the relative importance of the processes that led to the treatment effects we found.
Overall, Collembola showed fewer changes in diversity than Oribatida. Steiner (1995) found that changes in microarthropod species composition between sites were mainly dependent on oribatids. One reason may be a lower dispersal ability of oribatids compared with Collembola (MacLean et al. 1977; Hopkin 1997) or a lower reproductive rate (MacLean et al. 1977). Collembola may also be generally less sensitive to disturbances, as has been shown for metal pollutants (Van Straalen, Schobben & de Goede 1989).
The most common species e.g. Brachychthoniidae sp. and M. macrochaeta. were similar in all treatments, with the exception of the oribatids Zygoribatula exilis and Porobelba spinosa and collembolans of the genera Xenylla, Orchesella and Entomobrya, which were among those exhibiting large rank shifts between treatments. In the drought plots, the subdominant species were less abundant relative to the dominant ones than in the other treatments. The evenness index also decreased in the drought plots. The drought treatment may have ‘stressed’ the arthropod community more than the irrigation treatment, for example because the fauna was less adapted to dry conditions in this moist region of Sweden.
The use of describing dominance structure in communities by the log-normal distribution has been criticized for conceptual reasons (Dewdney 1997). Nevertheless, in the light of recent discussions (Basset et al. 1998; Hill & Hamer 1998; Watt 1998) it is still interesting to test whether a deviation from log-normality can indicate disturbance in soil communities (Hågvar 1994; Van Straalen 1998). In our case, as found in other temperate soil communities (Giller 1996), the species–abundance distributions all fitted the log-normal distribution and no major change in distribution characteristics from the treatments was indicated. This does not accord with our third hypothesis. An 8-year summer drought should indeed be considered a major disturbance, for example because it caused large decreases in abundance (Table 2). Therefore our results cast some doubt on using deviations from log-normal distributions as an indicator of disturbance in the soil communities studied.
soil animal communities as indicators
This study supported our first hypothesis that a long-term decrease in summer precipitation has significant consequences for soil fauna abundance and diversity in temperate coniferous forests. However, all faunal groups were not affected to the same degree. Oribatid diversity decreased more than Collembola diversity.
Different communities developed under the different moisture conditions. Data in the literature (Fjellberg 1980; Verhoef & Witteveen 1980; Steiner 1995) indicate that the community in the drought plots included more drought-tolerant species, such as Zygoribatula exilis and Entomobrya spp., in accordance with our second hypothesis. The irrigated plots also had distinct communities. Therefore an increased summer precipitation is also likely to cause changes in the soil community. The implications of these changes for ecosystem functions such as nutrient cycling rates are so far unclear.
Log-normality of the species–abundance distribution did not seem to be a good indicator of undisturbed soil communities under the studied conditions. However, detecting moisture-induced changes in soil fauna composition by field sampling seems uncomplicated, as the majority of the microarthropod species affected by the treatments were larger surface-living forms that are easily sampled.
Our results add further to the body of knowledge regarding the suitability of oribatid mites as bioindicators (Weigmann & Kratz 1987; Hogervorst, Verhoef & Van Straalen 1993; Webb et al. 1998). Soil-dwelling oribatid mite communities on different continents have few species in common, but show strong similarities on a higher taxonomic level for similar habitat types. This suggests that habitat preferences might be expressed at the family level (Osler & Beattie 1999). This, in turn, indicates that drought effects on oribatid families at our site may be applicable for other oribatid communities in similar habitats.
However, we want to caution against uncritically using small-scale studies such as the present one to infer changes due to large-scale processes such as climate change. On a large geographical scale, the possibilities of recolonizing disturbed areas may be far fewer because the distances between refugia will be greater. In our case plot size was probably small enough to enable many species to spread into the plots from the surroundings. This is unlikely to be the case when large-scale climatic changes occur.
We would like to thank Peter Gjelstrup and Arne Fjellberg for assistance with arthropod determinations, Birgitta Vegerfors-Persson for statistical advice and Shelagh Green for a linguistic revision. Ulf Johansson, Johan Bergholm and Marina Östergren provided valuable additional data from the site. This work was funded by the European Commission (ENV4-CT95-0027) as part of the GLOBIS project and by the Swedish Environmental Protection Agency. A grant from the Swedish Natural Science Research Council enabled further insight into the taxonomy and determination of Oribatida.
Appendix. Densities m−2 (SE) of Oribatida, Collembola and Enchytraeidae species at Skogaby. D, drought; C, control; I, irrigation. P-values of treatment effect according to two-way anova. Relative abundance rank shifts > 10 between treatments are indicated. Methods: HGC, high-gradient canister extractor; T, Tullgren funnel; B, Baermann funnel. Because we found clear overall differences in species composition between the treatments in the PCA (see text), the use of a Bonferroni test to correct for the large number of species-wise tests was considered to be unnecessarily rigorous.