Detritivorous fish indirectly reduce insect secondary production in a tropical river

Dominant animals can indirectly regulate population dynamics and energy flow for many other species in an ecosystem by altering habitat structure and resource availability. However, we know little about the degree to which other taxa can compensate for the loss of these dominant species. By removing animals and measuring responses of other animals as secondary production it is possible to assess how much other taxa can perform similar functions of these dominant taxa. Here we tested the response of aquatic insect secondary production to the loss of a dominant detritivorous fish Prochilodus mariae, in a tropical river, Rio Las Marias, Venezuela. Using an impermeable, plastic curtain, we excluded Prochilodus from one half of a 235-m stream reach for 6 weeks. We measured insect production as biomass times empirically measured body mass-specific growth rates for 8 common taxa constituting 59–74% of insect biomass. Removing Prochilodus increased the standing stock of benthic organic sediment. Biomass of the entire assemblage increased 1.7-fold and insect production for 8 taxa tripled upon exclusion of Prochilodus. Two fast-growing mayfly genera, Tricorythodes and Leptohyphes drove the increased secondary production. Despite low biomass of insects, growth rates were among the highest measured for freshwater insects, and these high growth rates in part caused high secondary production. Although insect production was high in the exclusion reach, insects did not compensate for the loss of Prochilodus in terms of consuming organic sediment showing that the capacity of detritivorous fish to process sediments is higher than that for aquatic insects in this tropical river.


INTRODUCTION
In many habitats, animals can alter the physical habitat and resource supply, and therefore indirectly regulate the population dynamics of many other species (Jones et al. 1994, Moore 2006).In freshwater ecosystems, a single species of fish or macroinvertebrate can dramatically alter ecosystem structure and function.Salmon can control nutrient inputs to stream and sediment outputs from their spawning activity (Moore et al. 2007).Detritivorous fishes and shrimps can reduce sediment standing stocks and export in streams and can lengthen carbon transport distances (Taylor et al. 2006).By reducing sediment resources, detritivorous fishes may indirectly reduce populations of other grazers consuming the same resource.In many ecosystems, it is unknown whether an apparently dominant animal is in fact irreplaceable, or whether its loss might be compensated by other groups of animals.An example is grazing caddisflies (Glossosoma) in Michigan streams; following their extirpation by disease, algal resource levels increased despite population increases of several other taxa showing that individually or together, none of the taxa could mitigate for the loss of Glossosoma (Kohler and Wiley 1997).
Secondary production is a valuable means to quantify how much a dominant animal controls energy flow through other animal populations.Secondary production quantifies one component of energy flow through an animal population and thus represents a functional attribute of a population or assemblage.Because small animals grow substantially faster than large ones, they will have higher biomass turnover rates; thus despite their low biomass, they can have high functional importance in food webs.Secondary production allows direct comparisons with other food web and ecosystem fluxes (Benke and Wallace 1997, Hall et al. 2003, Colo ´n-Gaud et al. 2010) and it can be an informative means to measure animal responses in experiments (Wallace et al. 1997, Benke andHuryn 2010).Yet production is more labor intensive because information on life histories must be coupled to estimates of biomass (Benke 1984) which may explain why there are few production studies.Resource availability (e.g., Wallace et al. 1997, Cross et al. 2006) or physical habitat (Huryn andWallace 1987, Chadwick andHuryn 2005) can control secondary production.Yet, the potential for control of secondary production by other groups of animals via predation, competition, or habitat modification is not well understood and has not been tested experimentally despite many studies showing how animal abundance responds to fish predators and competitors (e.g., Forrester 1994, Flecker 1996, Ruetz et al. 2002).
Tropical streams have abundant fishes, shrimps, or tadpoles that can strongly regulate production and biomass of basal resources such as algae and organic sediment (e.g., Flecker 1992a, b, Flecker 1996, Pringle and Hamazaki, 1998, Taylor et al. 2006, Connelly et al. 2008).By depressing benthic organic matter resources, fishes and shrimps can indirectly reduce abun-dances of aquatic insects (Flecker 1996, Pringle and Hamazaki 1998, Rosemond et al. 1998).Relative to large animals, insects appear to play a much smaller role in tropical streams, where they show low biomass (Ramirez andPringle 2006, Ramirez andHernandez-Cruz 2004), production (Ramirez and Pringle 1998), and weak control of ecosystem processes such as litter decomposition (Rosemond et al. 1998, Dobson et al. 2002).However, tropical running waters are warmer than temperate streams, which increases growth rates (Hauer andBenke 1991, Ramirez andPringle 2006) and production relative to biomass.Despite the appearance of low functional significance of insects in tropical streams based on biomass, the potentially fast growth rates may make insects more important than they appear.
Here we tested how a sediment-feeding, detritivorous fish (Prochilodus mariae Eigenmann) controlled insect production in a tropical river.Prochilodus eats organic sediment (Bowen 1984) and algae (Barbarino Duque et al. 1998) and can strongly reduce benthic organic sediment biomass (Flecker 1996).We excluded Prochilodus from a 235-m long section of a tropical river and measured the response of insect production during a six-week period.We hypothesized that following removal of Prochilodus, insect production would increase due to either the accumulation of sediment food resources (Flecker 1996, Taylor et al. 2006) or due to reduced disturbance and intimidation from sediment-feeding fish (Flecker 1992b).

Study site
Rio Las Marias is a fourth order river located at 225 m elevation in the Andean piedmont of Venezuela.This region has a protracted dry season from December to April, during which river discharge recedes from ca. 0.6 m 3 /s to ,0.1 m 3 /s.The wet season brings frequent storms and stream bed scouring spates that can reduce invertebrate abundance and biomass (Flecker and Feifarek 1994).The river flows over deposited alluvium consisting of cobbles in runs and riffles, and gravel-pebbles in pools.Mean wetted width of the study reach in the dry season was 14 m.Insects dominated the invertebrate assem-blage; although snails, crabs, and shrimp were present, they were relatively rare or completely absent in our invertebrate samples.

Prochilodus manipulation
We conducted a reach-scale removal experiment to test the effect of Prochilodus on material flows through a stream food web.We used a single reach-scale experiment because the effects we wanted to measure (invertebrate demography and production, nitrogen cycling, and carbon cycling) can only be measured at reach scales and not smaller plot scales (Taylor et al. 2006).The experiment was unreplicated because of logistical constraints of installing and maintaining replicates.In January 2005, we separated a 235-m length of Rio Las Marias by burying a plastic curtain longitudinally down the middle of the stream.The middle was defined as the point where 50% of the discharge occurred.The curtain was supported by concrete reinforcing bar every 1 or 2 meters and suspended from wire connecting each bar.On 15 January 2005 we removed a single fish species, Prochilodus mariae, from one side by gill netting and electrofishing, and then we excluded them using a 2.5-cm wire mesh fence placed at the top and bottom (Taylor et al. 2006).We electrofished the reference side of the reach but did not remove any fish.We did not put a fence on the reference side of the curtain because that would have effectively blocked fish migration and movement in this river.This fence allowed all fishes but Prochilodus and four other rarer, large-bodied species (Brycon whitei, Hoplias malabaricus, Hypostomus sp., and Salminus hilarii ) to pass through (Taylor et al. 2006).However, we did not remove any of these rarer, large-bodied fishes; thus, our manipulation only excluded Prochilodus.

Invertebrate and organic matter sampling
We sampled invertebrates from both the exclusion and reference reaches once before and four times after installing the curtain, at 10-15 day intervals.We collected 6 samples from top to bottom in each reach across a range of habitats (one pool, riffle, and run) dominated by cobble and gravel substrates.We collected paired samples on either side of the curtain within 10 m of a fixed point.Samples were collected using a 0.051-m 2 core sampler following Hall et al. (2006).All surficial sediment and twice the volume of water was removed from the core.The sample was elutriated and passed through a 0.25-mm mesh sieve.That same day we separated the samples into .1-mmand 0.25 to 1-mm size fractions; all animals from the .1-mmsize fraction were removed while still alive, and preserved in 95% ethanol.We identified, counted, and measured all invertebrates to the nearest mm in this size fraction.We identified most invertebrates to genus, or in the case of Chironomidae and Baetidae to family.Invertebrates in the 0.25 to 1-mm size fraction were preserved and sorted later in the laboratory under a microscope at 153 magnification.This size fraction was subsampled by suspending the sample in vigorously bubbling water in an Imhoff settling cone and removing a measured volume of the slurry.We repeated subsampling so that we counted, identified, and measured .50individuals.Length-mass regressions were used to estimate ash-free dry mass (AFDM) for each taxon, based on equations in Benke et al. (1999).We only included taxa in biomass estimates in which we found .10 individuals in all 60 samples.We excluded 10 taxa that did not meet this criterion because there were not enough individuals to show any sort of time or treatment response.The number of individuals excluded was 43 (0.4%) of a total of 12,207 individuals counted.
We sampled surficial benthic particulate organic matter by placing a 0.051-m 2 core sampler onto the stream bottom and gently stirring the water to suspend only the surficial organic matter without greatly disturbing the larger mineral sediment.This fraction of the organic sediment is most responsive to Prochilodus exclusion (Flecker 1996).We subsampled a measured volume of the water above the core, filtered it onto a Gelman A/E glass fiber filter, and measured organic matter (ash-free dry mass, AFDM) as mass loss following ignition at 5008C.We converted to areal units by multiplying the concentration of organic matter in the samples (g AFDM/m3 ) by the depth of the water in the core sampler.We collected 6 samples per reach near the stations where we collected invertebrates.

Secondary production
We empirically measured growth rates to v www.esajournals.orgcalculate secondary production using the instantaneous growth method for 8 common taxa (Benke 1984).Methods for estimating secondary production that were developed in cold streams are not appropriate in tropical streams because life cycles of insects are short and generations overlap making it difficult to follow a cohort or estimate the length of larval life spans (Ramirez and Pringle 1998).We measured growth rates for 5 insect taxa, Chironomidae (Diptera), Tricorythodes (Ephemeroptera), Baetidae (Ephemeroptera), Thraulodes (Ephemeroptera), and Anacroneuria (Plecoptera), by placing 1-5 measured individuals in a mesh container for 3-7 days.Because chironomids were small enough to pass through the mesh, we placed 3-4 individuals of the same size class in 17 pill bottles with biofilm-coated pebbles and incubated them in the stream for 3 days.We used 212-lm mesh tea infusion containers (Toby Tea Boys) for the three mayfly taxa and incubated them for 5 days following Benke and Jacobi (1986).One individual predatory stonefly, Anacroneuria, was incubated for 7 days in Tea Boys stocked with mayflies and midges.We used 5-12 cages per taxon.Petrophila, a web spinning crambid lepidopteran, did not survive in Tea Boys, so we estimated its growth rate in situ by measuring larval length using calipers while under its silk retreat, incubating for 4 days (n ¼ 13) in a known location, and measuring again.We assumed that Leptohyphes growth rates were similar to that of Tricorythodes based on their similar size and similar growth rates from another warm stream (Jackson and Fisher 1986).For all insects we estimated specific growth rate (1/d) as (lnM t À lnM 0 )t À1 , where M t was average insect mass after t days and M 0 was the average insect mass at the beginning of the experiment.We measured temperature hourly through the study season using an Onset Stowaway thermologger.
For one taxon, Elmidae larvae (Coleoptera), we used the response to the Prochilodus removal as a measure of growth rate because these larvae were so small they escaped through the Tea Boy mesh.Elmid biomass (g AFDM/m 2 ) increased eight-fold through time in the treatment reach following Prochilodus removal (though its biomass averaged two times higher than the reference reach-see Results).We used this increase in biomass (mg/m 2 ) as an estimate of growth rate (1/d) by calculating the slope of the change in ln(elmid biomass) with time before and during the exclusion.This estimate does not account for predatory or emigration losses thus it is likely to underestimate elmid growth rate but it provides a more accurate estimate than if we had used a literature-derived growth rate.
We calculated daily production (mg AFDM m À2 d À1 ) for these taxa as the mean biomass on a particular times instantaneous growth rate.When growth rate varied as a function of insect size, we calculated production for each 1-mm size class of insects within a taxon and summed them.When growth rate was not a function of insect size we multiplied the mean growth rate times mean biomass.We calculated production : biomass ratios (P:B, 1/d) which correspond to the mean specific growth rate of either a population or the assemblage.

Statistical approach
This experiment was unreplicated; hence, any statistical differences between the treatments cannot strictly be ascribed to the manipulation as if it were a randomized, replicated experiment (Fisher 1926).Instead we followed a before-after, control-impact design (Osenberg et al. 1994).
Based on previous research manipulating Prochilodus in replicated, 4-m 2 plots, (Flecker 1996), we have strong predictions as to how insects should respond to removal of these fish.
To test for differences in biomass, production, and organic matter standing stock throughout the experiment we used a permutation test.Within each date we calculated the difference in biomass of 6 paired samples on the reference and exclusion side.Then we summed these differences for the 6 samples on each of 4 dates for insects and 8 dates for organic matter.For the null model, we randomized the 12 samples 10,000 times within each date and summed the differences of 6 random paired values.Then we summed across all dates.We defined the test as significant if the difference in mean biomass from the data exceeded the 5% tail area of that from the null distribution.
To estimate mean and confidence intervals for biomass growth, production, and organic matter standing stock we used a bootstrap analysis because data were non-normally distributed data or contained many zeros for rare insects (Huryn 1996).For each taxon at each date and treatment we sampled with replacement 1000 times from the 6 samples.We summed across all taxa for total biomass.Bootstrap mean was the 50% percentile of the resulting distribution and 90% confidence interval was the region between the 5% and 95% quantiles.To generate mean biomass for each taxon across dates we took the mean of bootstrap means for each date.For production we used the same procedure, except that we multiplied biomass values for each of the 8 taxa by their respective growth rates.For taxa in which growth varied with size (Chironomidae and Thraulodes) we used the size-weighted mean growth rate for each date.
To relate animal size with its growth rate we used standardized major axis regression, because in addition to detecting a linear relation, we wanted to biologically interpret the slope of this symmetrical relationship (Warton et al. 2006).We used R version 2.3.1 (R Core Development Team 2006) for all statistics and calculations including package smatr (Warton et al. 2006) for calculating standardized major axis regressions.

Insect response to Prochilodus exclusion
Rio Las Marias was warm during the dry season, yet diel temperatures varied considerably.Daily temperature averaged 27.58C from 13 January to 2 March.On a daily basis, temperature fluctuated 5-128C, with extremes for instantaneous temperature ranging from 22.5 to 36.38C; however, 95% of the time temperature ranged between 23.5 and 33.08C.
Surficial organic matter standing stock was 2fold higher in the Prochilodus exclusion reach compared to the reference reach (permutation test p , 0.001) (Fig. 1C).Mean values from 25 January to 2 March averaged 16.5 g AFDM/m 2 in the reference and 31.5 g AFDM/m 2 in the exclusion section (Fig. 1C).
During the experiment, biomass of the entire invertebrate assemblage averaged about 1.7-fold higher in the Prochilodus exclusion relative to the reference section (Fig. 1A, Table 1, permutation test p , 0.001).Prior to removing Prochilodus, invertebrate biomass differed by ,10% between reference and exclusion sections.However, differences quickly emerged once Prochilodus was v www.esajournals.orgexcluded from one side, with increased biomass driven largely by two mayfly taxa from the same family, Tricorythodes and Leptohyphes.Total insect biomass increased 441 g AFDM/m 2 during the four sampling dates following exclusion; Tricorythodes and Leptohyphes increased 384 mg AFDM/ m 2 , and thus constituted 87% of the increase (Table 1).Biomass of combined Tricorythodes and Leptohyphes was strongly positively related to surficial organic matter standing stock (Fig. 2).
Taxa varied in their differences in biomass between the exclusion and treatment reaches.In addition to Tricorythodes and Leptohyphes, Elmidae larvae, Marilia, Hexatoma, and Ceratopogonidae had higher biomass in the exclusion reach.(Table 1, Fig. 3).In contrast, some taxa (e.g., Baetidae, Zygoptera, Petrophila, and Polycentropus) had much lower biomass in the reach with no Prochilodus.Among functional feeding groups, gatherers were either higher or similar in the exclusion reach relative to the reference reach.Conversely, scrapers were similar or lower in the exclusion reach (Fig. 3).Predators were either higher, lower, or displayed no clear differences.Along with biomass, abundance was higher in the treatment reach (26,000 ind/ m 2 ) compared with the reference reach (11,200 ind/m 2 ).
Growth rates for insects were high relative to those from temperate streams (Table 2, Fig. 4).Tricorythodes had growth rates averaging 0.28/d and several midge estimates exceeded 0.35/d.Anacroneuria averaged 0.08/d.Only two taxa, Thraulodes and Chironomidae had growth rates vary as a function of body length.Considering all taxa, log 10 growth rate declined linearly as a function of log body mass (Fig. 4).Standardized major axis regression slope was À0.45 for this relationship, and log 10 body mass explained 68% of the variation in log growth rate.
We could measure secondary production for 8 taxa that constituted 59% of total biomass in the reference and 74% in the exclusion section (Table 1).Total secondary production tracked the biomass response, yet secondary production was about 3-fold higher in the exclusion reach vs. the reference reach (Fig. 1B, Table 1).Most of this increase came from Tricorythodes and Lep- Note: Empty cells indicate taxa for which we did not measure secondary production Biomass was significantly higher in exclusion reach.à Biomass was significantly lower in exclusion reach.
v www.esajournals.orgtohyphes whose growth rates were among the highest ever measured; thus the relative increase in production was higher than the 2-fold increase in biomass for the 8 taxa for which we have production estimates (Table 1) (reference 350 mg/ m 2 , exclusion 732 mg/m 2 ).Daily assemblagelevel P:B among these 8 taxa was higher in the exclusion reach (0.23/d) compared to the reference reach (0.16/d) because the greatest response in production was by the fastest growing taxa, Tricorythodes and Leptohyphes.

DISCUSSION Prochilodus reduced insect production
Insect biomass and production increased rapidly following Prochilodus removal showing that removing this dominant detritivore increased organic carbon flow through the insect assemblage.This response by insects mirrored responses of invertebrate densities from replicated 4-m 2 cage exclusions of Prochilodus (Flecker 1996), providing strong evidence that the responses of invertebrates in our large-scale experiment were caused by Prochilodus.Flecker (1996) observed a 3-fold increase in insect density, which closely corresponded to a 2.3-fold increase in density for all taxa observed here.Tricorythodes, Ceratopogonidae and Elmidae increased in abundance in Flecker (1996) as did their biomass in this study.Chironomidae did not respond to our Prochilodus removal, whereas they increased in the smaller cage studies (Flecker 1996).Despite the difference in spatial scale of exclusions (2-m long cages vs. 235-m long stream reach), insect responses were largely similar between these two experiments, showing that a dominant sedimentfeeding fish can regulate insect secondary production.
Increased insect abundance and biomass in the exclusion reach was unlikely to be caused by higher immigration relative to emigration that one might find in a plot-scale study (Cooper et al. 1990).The experimental reach was large, which minimized the effects of immigration, and insects grew fast.As reach size increases, the population of insects inside the reach increases linearly while the immigration to the reach stays the same; thus turnover of individuals due to immigration will decline as reach size increases.We examined whether immigration drove the increase in mayflies by comparing mayfly immigration to their benthic density.Average drift density of mayflies in Rio Las Marias was about 15 individuals/m 3 (Flecker 1992a), multiplied by daily discharge in early February (2600 m 3 /d) gave a daily immigration rate of 39,000 ind./d or scaled to the reach area, 23 ind.m À2 d À1 .Mayfly abundance in the treatment reach ranged from 3000-13,000 ind./m 2 as the experiment progressed.Immigration divided by benthic density gives a turnover rate of 0.002-0.008/d,which is 40-150 times lower than the measured mayfly biomass turnover rate of 0.28/d.
The mechanism for increased biomass and production is likely due to increased population growth or survivorship.With growth rates for Tricorythodes at 28%/d, and assuming a cohort P:B ¼ 5 (Waters 1977), then generation time for Tricorythodes is about 17 d, which would have allowed almost 3 generations during our study.We cannot test that growth was faster because we used the same growth rate estimates for both the reference and the treatment reaches.In addition, There are two potential mechanisms for the increase of insect production following Prochilodus exclusion.One is the release of resource competition due to increases in benthic organic matter biomass, and the rate at which these stocks turn over (Taylor et al. 2006).Prochilodus feeds heavily on organic sediment; gut contents analyses show only fine sediment and algae (Bowen 1984, Barbarino Duque et al. 1998).The treatment without Prochilodus had 2 times more organic matter, which translated into 3 times higher insect production; this finding agrees with higher biomass of leaf litter increasing insect Fig. 3. Insect response varied among taxa and depended upon functional feeding group.Y axis is the log 10 response ratio, i.e., log 10 (mean biomass in exclusion/mean biomass in reference).Error bars are 90% bootstrap confidence intervals.Taxa are grouped according to functional-feeding group and ordered according to relative response within groups.
Table 2. Mean and bootstrap 90% confidence intervals of insect growth rates used to estimate secondary production.For two taxa growth rate (1/d) varied as a function of insect total length (mm) which approximated a linear model, growth rate  et al. 1997et al. , Entrekin et al. 2007)).Tricorythodes responded strongly to the manipulation because it grows fast and specializes on detritus in sediment-rich habitats (McCullough et al. 1979).
The second mechanism is that reduction in physical disturbance by Prochilodus increased Tricorythodes populations (Flecker 1992b).The response of individual taxa within functional feeding groups suggests that the first mechanism (sediment removal by Prochilodus limiting insect production) is the most likely.Collector-gatherer taxa had either higher or similar biomass in the exclusion reach.Scraper taxa on the other hand had either similar or lower biomass in the exclusion reach, suggesting that Prochilodus did not hinder them and that the mechanism for any declines in the treatment reach could be from increased resource availability on sediment-covered rocks where they forage.For example, the silk retreats made by Petrophila get buried by sediment following fish exclusion, likely impeding their foraging and respiration (A. S. Flecker, personal observation).This mechanism contrasts with fishes, such as trout, that potentially may reduce insect secondary production via direct consumption (Huryn 1996), although no experimental studies have shown predators depressing secondary production in streams.
Few studies have examined the production response of insects following the extirpation of dominant animals, and these studies show a much smaller response (Colo ´n-Gaud et al. 2010).Secondary production increased much more following Prochilodus removal than that for tadpole extirpations in tropical mountain streams (Colo ´n-Gaud et al. 2010).Tadpoles are a dominant animal grazer in some highland streams, and extirpation by disease increases standing stock of algae and production : respiration ratio v www.esajournals.org in streams (Connelly et al. 2008).However production of the entire invertebrate assemblage did not increase (Colo ´n-Gaud et al. 2010), possibly because the biomass of tadpoles (0.12 g AFDM/m 2 , Colo ´n-Gaud et al. 2010) was 50 times smaller than that for Prochilodus in run habitats of Rio Las Marias (6.5 g DM/m 2 , McIntyre et al. 2008), yet biomass of invertebrates was only twice as high in Rio Las Marias relative to invertebrates in Colo ´n- Gaud et al. (2010).The biomass disparity between fish and insect in Rio Las Marias combined with high growth rates may allow a strong, sudden increase in insect biomass and production.

High biomass turnover in Rio Las Marias
Insect production response to the loss of Prochilodus was very fast.High rates of biomass turnover enabled insect biomass to increase quickly in the experimental reach.Growth rates of chironomids in Rio Las Marias were about as fast as those measured from a subtropical river in Georgia (Hauer and Benke 1991).Mayflies in the genera Leptohyphes and Tricorythodes had growth rates among the fastest ever measured.Only those from Sycamore Creek, a Sonoran desert stream (0.34/d) exceeded those from Rio Las Marias (0.28/d) (Jackson and Fisher 1986).Tricorythodes in Costa Rica had median larval lifespans of 76-86 d (Jackson and Sweeney 1995).Given a cohort P:B of 5 (Waters 1977) yields a daily growth rate of these Costa Rican mayflies of 0.06/ d, which was about 4 times lower than our measured growth rates for these taxa.Tricorythodes growth rates in other rivers vary considerably.Tricorythodes atratus in the much colder headwaters of the Mississippi River had a P:B (0.071/d) (Hall et al. 1980), which is similar to that from Costa Rica.In Deep Creek, Washington, USA (188C), instantaneous growth rate for Tricorythodes was 0.13/d (McCullough et al. 1979), and in a warm subtropical river (26.78C) growth, measured using the same method as here, was 0.16/d (Benke and Jacobi 1986).Based on the large geographical variation among growth rates that is unexplained by temperature, empirical estimates of growth are needed to accurately measure secondary production in tropical habitats.
Body size also determined biomass turnover rates.As expected, insect growth rate declined with body mass (Fig. 4); however, there was much variance in this relationship.The slope was significantly steeper than the À0.25 slope between growth rate and body size predicted based on theoretical predictions (Brown et al. 2004), probably because scaling within taxa (Class Insecta in this case) does not necessarily represent scaling across the many taxa and broad body size ranges that are used to generate the general scaling relationships (e.g., Glazier 2006).Nonetheless, this relationship shows that high turnover rates of invertebrates in Rio Las Marias were in part due to their small size.

The energetic role of insects in Rio Las Marias
A striking attribute of Rio Las Marias, like many tropical streams, is the sheer amount of fish biomass relative to insects.Biomass of fishes at the site was quite high, 44.2 g wet mass/m 2 , which corresponded to an estimated standing stock of 10 g dry mass/m 2 , assuming fish are 22% dry mass (McIntyre et al. 2008).Thus, fish biomass was 16 times higher than the 0.6 g AFDM/m 2 of insect biomass in the reference reach.However, turnover of insect biomass driven by warm temperatures and their small size (Fig. 4) makes insects functionally much more important than they first appear (Moulton et al. 2004).Using the trophic basis of production method (Benke and Wallace 1997) and assuming a net production efficiency of 0.4 and assimilation efficiency of 0.2 (mid way between that for leaf detritus and diatoms), insect consumption rates of organic matter would be 0.7 (reference) to 2.0 (exclusion) g AFDM m À2 d À1 (Benke and Wallace 1997).These calculations are conservative because we only included the 8 taxa for which we measured production.Consumption rates in the exclusion reach were only 3 times lower than that for the hyper-abundant New Zealand mud snail (Potamopyrgus antipodarum) in a geothermal stream (Hall et al. 2003) despite having 50-fold lower biomass than the snails.Consumption rates of organic matter were quite high and represented a large fraction of gross primary production (1.3 g AFDM m À2 d À1 , standard deviation 0.34) and ecosystem respiration (2.4 g AFDM m À2 d À1 , standard deviation 0.78) (Taylor et al. 2006).Thus, despite having a very low biomass relative to temperate streams, the functional significance of insects in Rio Las Marias may be quite large.There is little question about the importance of even one or two fish species on Rio Las Marias ecosystem structure and function (Flecker 1996, Flecker and Taylor 2004, Taylor et al. 2006), but insects likely process large amounts of organic matter relative to their biomass because of temperature-and size-driven controls on metabolism and growth.
Despite high production and turnover rates, insects did not compensate for the removal of Prochilodus in terms of regulating organic sediment standing stock.As has been repeatedly shown in Rio Las Marias (e.g., Flecker 1992b, 1996, Taylor et al. 2006), sediment biomass increased when Prochilodus was removed, despite a tripling of insect production for 8 taxa, demonstrating that insects did not fully compensate for the loss of Prochilodus during our 6-week experiment.If insect production did not increase it is possible that the sediment response following Prochilodus removal would have been much greater.Prochilodus and other sediment feeding fishes can control organic matter processing in tropical streams (Flecker 1996, Taylor et al. 2006, Winemiller et al. 2006) and we show that they also can modify organic matter flow through insects.Given that Prochilodus is commercially harvested with declining populations (Barbarino Duque et al. 1998;Taylor et al. 2006), streams may respond with higher insect production and a larger role for insect-driven processing of organic matter.

ACKNOWLEDGMENTS
Kate Behn assisted with growth rate measurements and field sampling and Angela Detweiler persevered through the Petrophila.Amber Ulseth, Mayme Grant, Erin Richmond, Stephanie Juice, and Madeleine Fairbairn helped build the curtain, processed organic matter samples, and sorted bugs.Alex Buerkle assisted with statistics.Wyatt Cross commented on an earlier draft of this manuscript.The Figueredo family provided housing and support and M. and C. Perez allowed access to Rio Las Marias.Permits for scientific collection of fish were granted by the Venezuela Ministry of Agriculture.The National Science Foundation supported this research through grants DEB-0319593 and DEB-0321471.

Fig. 1 .
Fig. 1. (A) Total invertebrate assemblage biomass was higher following Prochilodus exclusion.The arrow labeled ''Prochilodus exclusion'' shows the date when we fenced Prochilodus from the exclusion reach.(B) Mean daily secondary production for seven taxa increased following Prochilodus exclusion.(C) Surficial organic matter standing stock increased following Prochilodus exclusion.For all 3 panels, hollow points are reference reach, solid points are the exclusion reach and error bars are 90% bootstrap confidence intervals.AFDM is ash-free dry mass.

Fig. 4 .
Fig. 4. Insect specific growth rates declined with body size.Points are means of 3-6 individuals in cages, or individual animals in the case of Anacroneuria and Petrophila.Line is a standardized major axis regression with the equation log 10 growth rate (1/d) ¼ À0.45 3 log 10 body mass (mg) À 1.39.The 95% confidence interval of the slope is À0.39,À0.52 and r 2 ¼ 0.68.

Table 1 .
Biomass and production of insect taxa in Rio Las Marias in the 235-m long experimental sections with (Reference) and without (Exclusion) Prochilodus.Exclusion only includes the four sampling dates after Prochilodus was removed from the exclusion reach.