L. B. Marczak, Department of Forest Sciences, University of British Columbia, 3041-2424 Main Mall, Vancouver, British Columbia, Canada, V6T 1Z4. Tel.: (604)822 8927. Fax: (604)822 9102. E-mail: firstname.lastname@example.org
1Aquatic insects emerging from streams can provide an important energy subsidy to recipient consumers such as riparian web-building spiders. This subsidy has been hypothesized to be of little importance where the primary productivity of the recipient habitat exceeds that of the donor habitat.
2To test this hypothesis, we manipulated emerging stream insect abundance in a productive riparian rainforest in a replicated design using greenhouse-type exclosures, contrasted with unmanipulated stream reaches (four exclosures on two streams).
3Experimental exclosures resulted in a 62·9% decrease in aquatic insect abundance in exclusion reaches compared with control reaches. The overall density of riparian spiders was significantly positively correlated with aquatic insect abundances. Horizontal orb weavers (Tetragnathidae) showed a strong response to aquatic insect reduction – abundance at exclosure sites was 57% lower than at control sites. Several spider families that have not been associated with tracking aquatic insect subsidies also showed significantly decreased abundance when aquatic insects were reduced.
4This result is contrary to predictions of weak subsidy effects where recipient net primary productivity is high. These results suggest that predicting the importance of resource subsidies for food webs requires a focus on the relative abundance of subsidy materials in recipient and donor habitats and not simply on the total flux of energy between systems.
Resource flows between habitats (hereafter referred to as subsidies) can have important implications for food web dynamics in recipient environments (Bastow et al. 2002; Collier, Bury & Gibbs 2002; Sabo & Power 2002b). The movement of energy, nutrients or prey between systems is ubiquitous, although it is not yet clear how strong the effects of these subsidies are in all systems (Polis, Anderson & Holt 1997). In this context it is critical to understand how different characteristics of subsidies (e.g. trophic level of a subsidy, subsidy type, variability in space or time), habitats (e.g. permeability, seasonality of in situ production) or consumers (e.g. trophic level, functional feeding group) might determine the relative importance of cross-habitat subsidies.
Explicit in the definition of resource subsidies has been the idea of recipient benefit. We follow Polis et al. (1997) in defining subsidies as any movement of energy and materials across a habitat boundary that provides benefit to a recipient consumer. Although not implicit or required by this definition, the vast majority of studies have supposed that cross-habitat subsidies move predominantly into less productive systems, and there strongly influence abundance and distribution of consumers (Polis & Hurd 1996; Sears, Holt & Polis 2004; Paetzold, Bernet & Tockner 2006). This premise is based on the assertion that passive dispersal or movement will occur by diffusion from areas of higher to lower density, and that these flows will have strong effects in lower productivity recipient habitats, and by extension negligible effects in productive recipient habitats (Sears et al. 2004). Indeed, examples of strong effects of resource subsidies in low productivity environments are well known, particularly in deserts, dry coastlines and arctic systems (Polis & Hurd 1996). In Polis & Hurd's (1996) island food webs, marine algae and carrion beached on the shores of small, very dry islands supported extremely large populations of detritivorous arthropods that formed greater than 90% of the food available to web-building spider populations. Spiders were 1–2 orders of magnitude greater in abundance than in similar areas without detrital inputs. Marine-derived material has similarly been found to subsidize coyotes along desert coastlines in Baja, California (Rose & Polis 1998). Similarly, emerging aquatic insects have been found to affect the growth and fitness of lizards on dry cobble bars in California (Sabo & Power 2002) and to alter the distribution of ground-dwelling arthropods on low productivity gravel bars in Italy (Collier et al. 2002; Paetzold et al. 2006).
How do consumers respond to subsidies in highly productive recipient habitats? It has been supposed that the effect of subsidies arriving in high productivity habitats will be small in magnitude. However, empirical studies in such habitats are scarce – researchers have understandably focused on systems with strong edges and where the ability to detect a response to subsidies is correspondingly large. The emphasis on studying systems where strong responses to resource subsidies are anticipated has limited our ability to ascertain the variables that determine the strength of responses, or to appreciate the ubiquity of subsidy effects.
Aquatic-terrestrial contrasts have been popular in studies of subsidy effects, at least in part because of the distinct boundary between these habitats. In particular, the study of forest contributions to headwater streams was fruitful long before it was framed in the context of resource subsidies (e.g. Mason & Macdonald 1982; Jackson & Fisher 1986; Richardson 1991). More recent investigations have highlighted the potential contribution of streams to forest habitats in the form of salmon (Wipfli 2005), algal mats (Bastow et al. 2002), and emerging aquatic invertebrates (Sanzone et al. 2003). Recipient consumers for these subsidies include riparian vegetation (nutrient addition from salmon carcasses), invertebrates (decomposer fauna for carcasses, algae and detritus) and spiders, birds, bats and similar organisms (emerging aquatic invertebrates). In keeping with the suppositions above, such work has focused on systems with large contrasts in donor to recipient primary productivity and strongly oligotrophic recipient habitats. In this study, we conducted a manipulative field experiment to test whether the abundance of emerging aquatic insects affected the distribution and abundance of riparian web-building spiders in a highly productive rainforest of coastal British Columbia, Canada. Given the high net primary productivity of the recipient habitat at our study site, we predicted that the response of web-building spiders to aquatic insect subsidy exclusion would be small and limited to those spiders with a strong dependence or specialization on adult aquatic insects.
The field experiment was conducted from May through July of 2004 in two similar headwater streams (Spring Creek and East Creek; Table 1) within the Malcolm Knapp Research Forest (MKRF; 49°18′40″N, 122°32′40″W). The forest is located in the Pacific coastal rainforest of south-western British Columbia. Average mean air temperature ranges from a low of 2 °C in January to a high of 16 °C in July. Precipitation is approximately 2500 mm annually. More than 70% of this precipitation falls between October and March. The coastal temperate rainforests of British Columbia are substantially more productive (above-ground NPP estimated at 1050–1300 g C m−2 year−1 from a model calibrated for the MKRF, Forest Ecosystem Modelling 2006) than the headwater streams that flow through them (production of algal biomass estimated at 3·64 g C m−2 year−1; data adapted from Kiffney, Richardson & Feller 2000). The dominant vegetation surrounding these streams included red alder Alnus rubra and vine maple Acer circinatum with a canopy composed largely of western hemlock Tsuga heterophylla, western redcedar Thuja plicata and Douglas-fir Pseudotsuga menziesii. Understorey vegetation immediately adjacent to the streams consisted of 1–2 m tall shrubs, particularly salmonberry Rubus spectabilis and huckleberry Vaccinium ovatum. The immediate riparian cover of East Creek is more strongly deciduous than Spring Creek, with large numbers of red alder dominating the stream banks.
Table 1. Comparison of physical and hydrological characteristics of East Creek and Spring Creek. Mean ± 1 standard error
Between 16 May and 19 May 2004 two greenhouse-type exclosures were constructed on each of the two streams in the MKRF (Fig. 1a). Exclosures were constructed from transparent plastic sheeting supported by semicircular PVC frames that were anchored to the banks. The edges of the plastic sheeting were partly dug into the stream bank to form a complete seal over the streambed. Each exclosure covered a 50 m long stream reach, and was separated by 50 m long stream reaches that served as controls. Controls and exclosures alternated from the upstream direction (Fig. 1b).
We sampled orb-weaving spider abundance by using timed vegetation shake samples (Costello & Daane 1997). We used a 0·3 m2 tray with steep sides and shook vegetation directly over this tray for 20 s. Individual spiders were removed from the tray using an aspirator and stored in 75% ethanol until identified. Orb-weaving spiders were collected in this fashion in the week prior to construction of the exclosures and every week for 10 weeks following construction of the exclosures with four subsamples taken in the middle 15 m of each reach within 2 m of the stream wetted edge, two on each side of the stream (Fig. 1b). Precise collecting locations within a reach varied between weeks. The abundances of flying aquatic and terrestrial insects were estimated each month by sticky trap sampling. Traps were composed of Tanglefoot (The Tanglefoot Company, Grand Rapids, MI, USA) thinly spread on one side of an acetate sheet (each sheet represents a 600 cm2 surface area) and suspended between two garden stakes approximately 1·5–2 m above the ground facing the stream in roughly the same locations as spider shake samples. Sticky trap samples were set for 7 days in the middle of each month (May, June and July), and collected samples were frozen until sorted.
Adult spiders were sorted and identified to species. Insects were identified to Order, or to Family in the case of Diptera, and assigned to either an aquatic or terrestrial group based on published life-history details (Merritt & Cummins 1996). Length and width measurements of insects were determined with an optical micrometer to the nearest 0·01 mm. Biomass estimates of spiders were based on measurements of the length of the tibia–patella of the first pair of walking legs (Higgins 1992). We used published length–mass regression equations to determine biomass of spiders and aquatic and terrestrial insects (Rogers, Hinds & Buschbom 1976; Sample et al. 1993; Sabo, Bastow & Power 2002).
Repeated measures and nested designs involve spatial and temporal autocorrelation that violate assumptions of independence of data points necessary for conventional general-linear modelling (Buckley, Briese & Rees 2003). We used mixed-effects models that can account for the correlated error structures present in our data. By combining both repeated measures and spatially nested random effects in a general linear mixed effects (GLME) model we mitigate problems of nonindependence and pseudoreplication – the combination of random spatial effects and repeated measures made the use of this technique necessary.
The abundance and biomass of flying aquatic and terrestrial insects were each analysed using repeated measures anova with treatment (ambient and reduced insects) and stream (East and Spring Creeks) as main factors, date as the repeated measure (three sampling periods) and sticky trap samples as replicates. As multiple samples from each 50 m stream segment were not independent, these data were combined to expand the total area of habitat sampled (four subsamples in each stream segment combined into one weekly sample, n = 8 per sampling period) for both insect and spider abundances. We also used repeated measures anova with the abundance of all spiders as the response variable, treatment and stream as factors and date (10 sampling periods) as the repeated measure. Only adult spiders were utilized in statistical analyses as the identification of juveniles is less reliable. Subsequent repeated measures anovas were performed separately for the most abundant spider families (five families each representing > 5% of total abundance). We applied a sequential Bonferroni procedure to correct for multiple tests. Stream was considered a random factor in all models as this has the advantage of using fewer degrees of freedom. We used the Satterthwaite approximation to estimate denominator degrees of freedom for both fixed and random effects as recommended by Schabenberger & Pierce (2002). The hypothesized time correlation structure used for these models was heterogeneous autoregressive (i.e. correlation between samples is assumed to decrease as separation in time increases). The assumption of a correct covariance model was examined using a likelihood ratio test (with a χ2 distribution) against models with compound symmetry. Heterogeneous autoregressive was the better covariance model for all insect and spider data. We also used likelihood ratio tests to assess the contribution of the two spatial correlation parameters included as random factors (stream and the stream by treatment interaction). For each model, we first checked that the assumptions of normally distributed data and linearly related fixed-effects means were met by examining residual vs. predicted plots and normal probability plots. All data required ln transformation to meet these assumptions. All analyses were conducted using proc mixed in the statistical package SAS v 9·0 (SAS Institute Inc., Cary, NC, USA).
flying aquatic and terrestrial insect abundance
The greenhouse cover significantly reduced the abundance and biomass of flying insects of aquatic origin in exclusion reaches relative to control reaches (abundance, F1,5·23 = 15·07, P < 0·01; biomass F1,2·1 = 16·35, P = 0·05; Table 2; Fig. 2a). Adult aquatic insect biomass in the exclusion reaches was 55·9% lower than in control reaches when all dates were combined while adult aquatic insect abundance was 62·9% lower. Flying terrestrial insect abundance and biomass were unaffected by the exclusion treatment (abundance, F1,5·36 = 0·08, P = 0·79; biomass, F1,5·82 = 0·02, P =0·89; Table 2; Fig. 2c). Neither time, stream nor the treatment by stream interaction were significant factors for either aquatic or terrestrial insects, although there was a trend towards overall greater abundances of emerging aquatic invertebrates at East Creek (deciduous canopy). There was no significant change in terrestrial insect abundance or biomass over the sampling period. In control reaches (representing ambient conditions) flying aquatic insects were 6·49 times more abundant than flying terrestrial insects across the entire sampling period and had 4·25 times greater biomass. This pattern did not change over the sampling period; there was no evidence of alternating peaks in abundance between these two prey sources over the 3-month study period (Fig. 2b,d).
Table 2. Repeated measures anovas for mean of abundance and biomass per trap (n = 8) of flying aquatic and terrestrial insects. Contributions of random effects (stream, stream × treatment) to the model were assessed using a χ2 test (d.f. = 1) of the difference in residual log likelihood for the full and reduced model. Significant P-values are highlighted in bold text
Flying aquatic insects
Flying terrestrial insects
F1,5·23 = 15·07
F1,2·1 = 16·35
F1,5·36 = 0·08
F1,5·82 = 0·02
F2,9·99 = 3·46
F2,13·5 = 0·87
F2,10·3 = 0·09
F2,10·9 = 0·15
Time × treatment
F2,9·99 = 4·25
F2,13·5 = 1·00
F2,10·3 = 0·57
F2,10·9 = 0·72
χ2 = 0
χ2 = 0
χ2 = 0
χ2 = 0
Stream × treatment
χ2 = 0
χ2 = 0·9
χ2 = 0
χ2 = 0
There was no effect of future exclosure site, stream or the stream by exclosure site interaction on spider abundance in the week prior to construction of experimental exclosures (effect of future exclosure site: F1,5 = 0·50, P = 0·51). A total of 26 spider species, representing nine families were collected during the experiment (Table 3). The experimental reduction of aquatic insects depressed the overall abundance of spiders adjacent to exclusion reaches (treatment, F1,16·1 = 11·59, P < 0·01). There was no significant effect of time or the interaction of time and treatment. The random effect of stream was not significant across all adult spiders and there was no significant interaction between stream and treatment, although there was a trend towards greater overall spider abundance at East Creek. Five families were present in abundances large enough to merit further analyses (Araneidae, Hahniidae, Linyphiidae, Tetragnathidae, Theridiidae, each representing > 5% of total abundance). Repeated measures anovas for mean abundances of the most common families showed that four of the five families included in the analysis were significantly depressed by the exclusion of aquatic insects (Table 4; Fig. 3). There was no significant interaction between stream and treatment for any of the spider families. Tetragnathids did significantly decrease in abundance over the sampling period (F9,36·2 = 2·22, P = 0·04), while only linyphiids showed a family level difference in abundance between the two streams (P = 0·04). Hahniids and araneids showed the largest magnitude responses to aquatic insect exclusion, being 89·1% and 85·7% lower in exclosure reaches relative to control reaches, while linyphiids and tetragnathids were 50·1% and 42·9% lower, respectively.
Table 3. List of families and species of vegetation-dwelling spiders collected at Spring Creek and East Creek in the Malcolm Knapp Research Forest indicating relative percentages (%) of abundance
% of total abundance (adults)
Vertical orb webs
Active hunters (with silk retreat)
Active hunters (no web)
Active hunters (no web)
Horizontal orb webs
Hackled orb webs
Table 4. Results of separate repeated measures anovas for mean abundance of adult spiders in five families each representing > 5% of total spider abundance. Contributions of random effects (stream and stream by treatment) to the model were assessed using a χ2 test (d.f. = 1) of the difference in residual log likelihood between the full and reduced model. Significant P-values are highlighted in bold text. All P-values were adjusted using a sequential Bonferroni procedure
F1,18·9 = 9·38
F1,26·9 = 17·45
F1,15·4 = 4·53
F1,19·8 = 4·69
F1,17·9 = 1·20
F9,34·4 = 0·82
F9,33·4 = 1·28
F9,37·7 = 2·04
F9,36·2 = 2·22
F9,32·9 = 1·75
Time × treatment
F9,34·4 = 0·75
F9,33·4 = 1·00
F9,37·7 = 0·82
F9,36·2 = 0·83
F9,32·9 = 0·58
χ2 = 0
χ2 = 0·7
χ2 = 4·1
χ2 = 1·0
χ2 = 2·2
Stream × treatment
χ2 = 0·1
χ2 = 0·7
χ2 = 3·8
χ2 = 0·1
χ2 = 0·1
This study demonstrates that predators in highly productive terrestrial habitats can respond strongly to trophic subsidies. Most previous studies have focused on the case where local production is largely absent and consumers are obligately dependent on allochthonous contributions. For example, Paetzold et al. (2006) noted that large differences in productivity, such as occurred at their river cobble bar study site, should result in greater transfers of energy from donor to recipient habitats (via incorporation of subsidies by recipient habitat dwelling consumers). They found that ground-dwelling arthropods showed a substantial numerical response to the exclusion or addition of drifting and emerging aquatic invertebrate subsidies. Polis & Hurd (1995) found that, in most years allochthonous marine inputs controlled the dynamics of web-building spiders on dry Gulf of California islands – allowing large populations of consumers to persist despite terrestrial primary productivity fluctuations. These and other empirical results have led to an assumption that the effects of subsidies will only be significant when recipient primary productivity is lower than that of the donor system. Our results are the first to test that assumption in a highly productive terrestrial setting. Although yearly aquatic NPP was substantially lower than terrestrial NPP, the average emerging aquatic insect abundance was 5·9 times higher than terrestrial insects. This indicates that assumptions about subsidy movements based purely on ratios of NPP may be misleading. Large subsidies (relative to terrestrial production) of emerging aquatic insects occurred despite low stream NPP. This is perhaps not entirely surprising given the well-known relationship between secondary production in these headwater streams and detrital inputs (from the surrounding forest). In these systems it appears that detrital inputs from the forest to streams is driving growth and development of aquatic invertebrates, which feeds energy back into the surrounding forest. Within the stream environment, these detrital inputs may be converted to insect biomass at a substantially higher rate than would occur for the equivalent material on the forest floor (Shurin, Gruner & Hillebrand 2006). Streams may therefore be important bioreactors for converting relatively recalcitrant forest litter into energy sources that are available to higher trophic levels in terrestrial settings.
Results from our system showed that spiders in diverse families, with widely divergent web morphologies and capture techniques, tracked aquatic insect subsidies for some portion of their diet. In their study of the effects of aquatic insect exclusion on riparian spiders in Horonai, Japan, Kato et al. (2003) found that horizontal orb weavers were noticeably affected while vertical orb-weaving and sheet-web weaving spiders were not. They attributed this result to the different feeding strategies implied by the web morphologies and thus prey preferences of these distinctive families. Sheetweb weavers (Linyphiidae) have a high investment in three-dimensional web structures that make it costly to move in order to track spatially and temporally ephemeral resources. Vertical orb weavers (Araneidae) construct large vertical webs that are structurally more suited to catching larger, faster-flying terrestrial prey (Olive 1982; Foelix 1996). In contrast, the large open webs and nonsticky silk of horizontal orb weavers are often viewed as adapted to the capture of small or weakly flying insects (Olive 1982) and represent a lower investment that may promote resource tracking. The insect exclusion in this study generated a response in a larger number of spider groups, including those that are not associated with resource tracking of aquatically derived insects. This unexpected pattern of response by nonspecialists on aquatic resources suggests that, in the riparian habitats of the MKRF, adult aquatic insects play a disproportionately large role in determining the distribution of many web-building spiders. This may be particularly true when subsidy resources form a large fraction of total available resources or more broadly when the ratio of subsidy to equivalent local resources is greater than 1. The high abundance of aquatic insects evidently supports groups of spiders with a broader range of capture techniques than may be occurring where aquatic insects form a much smaller proportion of the overall prey base.
Recent studies have emphasized the importance of alternating periods of productivity between habitats, creating seasonally reciprocal flows of energy (Nakano & Murakami 2001; Takimoto, Iwata & Murakami 2002; Kato et al. 2003). At the interface between streams and forests, the seasonal emergence of aquatic insects may be temporally offset from the time of maximum secondary terrestrial productivity – influencing the distribution and abundance of generalist consumers such as riparian spiders (Kato et al. 2003) and birds (Nakano & Murakami 2001; Uesugi 2002). Terrestrial invertebrate production should be greatest during the spring and early summer as deciduous trees leaf out, coniferous trees put on new growth and understorey vegetation is flush. While previous studies have shown strong variability in the relative availability of aquatic and terrestrial insect prey, this pattern was not detectable in the 3 months (spring through mid-summer) of this experiment. In our study, availability of emerging aquatic insect prey was always greater than terrestrial insect abundance in the riparian forest. It seems probable that, in the highly productive rainforests of the Pacific Northwest, emerging aquatic insects are more abundant than flying terrestrial invertebrates at most times in the year. However, this production and export of material to the riparian forest is itself a consequence of terrestrial detrital inputs. The high secondary productivity of headwater streams in this region supports a diverse group of invertebrate predators. This secondary productivity is itself created by the higher trophic efficiency within streams, based on terrestrially derived materials indicating that for small headwater streams and their adjacent forests, the land–water boundary is particularly porous.
In our system, existing theory predicted that responses to subsidy exclusion would be weak, or constrained to family groups known to specialize on aquatic insects as a subsidy resource. Contrary to this expectation, the overall abundance of several spider families decreased with subsidy exclusion, including families not thought to be particularly sensitive trackers of aquatic insect abundance. This suggests that even in highly productive settings, specific subsidy types can have important effects on fauna. Although allochthonous resources may indeed contribute close to 100% of productivity in some habitats (e.g. headwater streams, caves, snowfields, islands, etc.; Vanni et al. 2004), the importance of subsidies appears nearly as great in habitats with substantial in situ primary productivity, such as that described here.
This study provides direct evidence that the local distribution of multiple families of riparian orb weavers can be controlled by changes in inputs of emerging aquatic insects, particularly when that input to adjacent terrestrial habitats is sufficiently large compared with equivalent terrestrial resources (terrestrial insects). The high relative abundance of aquatic insects in headwater, temperate rainforest streams may impact riparian food web dynamics particularly where terrestrial insect production is limited, even though riparian forest primary productivity is relatively high. Subsidy effects appear to be largest when they subsidize a system with comparable resources that are at low levels – the pool of labile or available carbon is often not equivalent to the overall contrast in donor and recipient habitat primary productivity, particularly when the focal consumer is a predator. The development of predictions about where subsidies are likely to produce the greatest effects in recipient habitats will require more specific studies that examine the nature of the subsidy relative to the nature of available resources in the recipient habitat.
We thank Yixin Zhang, Deirdre Leard, Conan Phelan, Tatiana Lee, Trent Hoover and other members of the Stream and Riparian Research group (StaRR) for assistance in the field and lab. Ross Thompson and Rebecca Best provided statistical guidance. The authors acknowledge funding assistance from the Natural Sciences and Engineering Research Council (Canada) and the Forest Sciences Program (British Columbia).