OIKOS 36: 147-157. Copenhagen 1981
A model of seston capture by net-spinning caddisflies
Theodore J. Georgian, Jr. and J. Bruce Wallace
Georgian, T. J. Jr. and Wallace, J. B. 1981. A model of seston capture by net-spin- ning caddisflies. - Oikos 36: 147-157.
Six species of net-spinning caddisflies (Trichoptera) coexist in the headwater region of the Tallulah River, a rocky, high-gradient tributary of the Savannah River. These caddisflies feed on the suspended organic matter (seston) captured by their nets. We analyzed these species' nets, microhabitat preferences, and monthly abundances. These data, along with the particle size distributions and monthly abundances of seston, were incorporated into a model of seston capture by these net-spinning cad- disflies. The resulting model permitted us to test the hypothesis that differences in capture net mesh sizes serve to partition, by size, the food available to coexisting net-spinning caddisflies. and thus reduce competition between them. The model predicted annual seston capture by the six species of 1300 g AFDW (ash free dry weight), over 1000 times their annual production. The seston captured per longitudinal meter of stream represents 0.005% of the total seston in transport. Resource overlap coefficients indicated that the instars of the five hydropsychid species do not partition the seston by particle size or food type in a manner that reflects competitive interactions. We hypothesize that these filter feeders are limited by the availability of high-quality food items (primarily drifting animals) rather than by the overall seston supply. Species with highly selective feeding habits must filter large volumes of water to permit them to specialize on rare items. High filtration rates require rapid current velocities, large nets, and coarse net meshes. Instars with coarse-meshed nets showed the largest proportionate declines in seston capture rates when the model was simu- lated using the lower current velocity and smaller seston panicles sizes typical for the larger Savannah River.
T. J. Georgian, Jr. andJ. B. Wallace, Inst. of Ecology and Dept of Entomology, Univ. of Georgia, Athens, GA 30602, USA. Utectb BJUZOB rujeryusoc JJOBTVIO cen> pesatHHKOB cocymecTBOBanH a sepxcehe peKH Taiuiyja, cxajTHcrcM aJ33Korppi.nneHTHCM npHTOKe peat CasaHHa. 3rst pyteBHHKH rorranHcii opraHHtecscHM MaTepnajrM, cycneHSHpoBaHtcM a sane (cecroH) , KoropbB yzBBJMBanH csoea cerao. fti HCcrsEDBanH 3TH earn, sa- 6Hparem>HOCTb a OTHOUSHHH MHKpoeHorcnoB, atoeMaowHyx) 'mcneHHOcnj. 3rn Hapany c JBHHKMI no pawepaM lacrau H esosaeopiHCMy oennsno cec- 3Kj*xes*i a Ntanent. norpeaneHHS cecioHa 3iHn« pineannxawi. wonami noasonjier npoaepmb rancrreay o TCM, ^ro paarortHR paa- Mepoa JKJBV&I cero noaBonsBOT paaranaTb no serawHHe ascrynHyw rmmy, nor- pefinaeMyio oocymecTByioaww ain»« pyneaonjKOB, H, xaraw cCpascM, OHM CHH- HQKIT KOHKVpeHUKKJ MBKITy HHMH. Monenb npescxastiBa&r asterouHoe ncrrpeSTieHHe cacioHa aEcraw 3KD2MH nopsn- Ka 1300 MT (eeasanbHta cyxna aec} , 3 1000 pas eannue HX rcajDBaR nponyx- UHH. KdnHtecTBo cecrona, norpefineHBoro na norowwa MBTP pycna, cocra- smer 0,005% or oCtoero TpaHcnopra cacroHa. Kc® OIKOS 36:2 (1981) 147 1. Introduction '^"/c-*-,?!:^"'"5-'v7v~'^ -'SKssg The role and importance of filter-feeding organisms in the processing of organic matter in stream ecosystems has attracted increasing interest in the past several years. Most research has involved streams which receive their primary energy inputs as allochthonous paniculate detritus (Minshall 1978). These inputs pass from one functional group to another, often as suspended par- ticulate organic matter (seston). Changes in particle size and food quality occur as the detritus is processed (Boling et al. 1975, Anderson and Sedell 1979). Al- though seston customarily refers to all suspended par- ticulates (Runner 1963), we use the term here in a more restricted sense to refer to the organic fraction only. Webster (Webster 1975, Webster and Patten 1979) introduced the term "spiralling" to describe the role of benthic organisms in retaining organic matter and as- sociated nutrients that would otherwise be carried downstream and lost to that segment of the stream sys- Fig. 1. A capture net of IVth instar P. cardis. Note the small tem. The material consumed by filter feeders is either meshes in the central "seam" and the large irregular meshes assimilated (and a fraction retained as secondary pro- around the upper periphery. Scale line = 550 um. duction) or egested and is available for repeated capture and reingestion by other organisms downstream (Wal- lace et al. 1977). The degree to which filter feeders size. These differences in net structure have been inter- retard the downstream transport of fine paniculate preted as a means by which coexisting net-spinning cad- organic matter is postulated to be an important deter- disflies partition the available food resources (Wallace minant of the efficiency with which a stream ecosystem 1975, Malas and Wallace 1977, Wallace et al. 1977), processes its energy and nutrient inputs. and that any competition based on particle size would Net-spinning caddisfly larvae, members of the Hyd- involve instars as well as species. If the hypothesis of ropsychoidea, comprise a major component of the food partitioning by size is correct, competition between filter-feeding fauna of most streams. The larvae spin instars which are adjacent when ranked according to silken capture nets, with mesh opening size and overall mesh opening sizes, regardless of species, should be of a net structure differing from family to family (Wiggins level that would permit coexistence. A test of the 1977, Wallace and Merritt 1980). Hydropsychid larvae hypothesis requires that the capture of various size food spin a net with roughly rectangular meshes, oriented particles by each instar of a group of coexisting net- perpendicular to the current. The size of individual spinning species be known (cf. Pianka 1973). We have mesh openings increases from the base of the net to- analyzed the net structure, microhabitat distribution, ward the periphery (Fig. 1). In all species studied to and monthly abundances of six net-spinning caddisfly date, a new net, with larger mesh openings, is spun by species in a 4th-order stream in the southern Ap- each subsequent larval instar (Sattler 1963, Kaiser palachian Mountains, in conjunction with data on the 1965. Williams and Hynes 1973, Malas and Wallace seasonal availability of seston in the water column. 1977). Each successive net is also larger in total area. These data were combined to provide estimates of par- Mesh opening sizes, then, differ between instars within ticle capture functions for this group of filter feeders. a given hydropsychid species and also between equival- ent instars of different species. Members of the Philopotamidae construct long sac- 2. Methods like nets with an anterior opening facing into the current and extremely fine mesh openings (Wallace and Malas We studied five hydropsychid species and one 1976). Differences in net structure and microhabitat phiiopotamid species which coexist in the headwaters of result in distinct differences in the size and quality of the Tallulah River, a tributary of the Savannah River panicles captured by hydropsychid and phiiopotamid arising in the southern Appalachian Mountains. At the larvae (Williams and Hynes 1973, Malas and Wallace study site, on the Georgia-North Carolina border, the 1977). Phiiopotamid species also spin a new capture net Tallulah River is a 4th-order stream, with an elevation in each larval instar (Wallace and Malas 1976). of 835 m a.s.l., a gradient of 34 m km"1, and a drainage Differences among capture nets suggests that species, basin of 16 km2. The stream flows over a rocky sub- and instars within species, capture and feed upon differ- strate with shallow riffles predominating and small pool ent size particles, based on differences in mesh opening areas interspersed among the riffles. Width ranges from 148 OIKOS36:: (1981) 4 to 7 m, and maximum depth is about 0.6 m. Wallace et Vth-instar larvae, with respect to current velocity and al. (1977) gave a more detailed description of the site. position on rocks, were recorded using the methods of The species studied were: Arctopsyche irrorata Ross, Malas and Wallace (1977). Current velocities were Parapsyche cardis Ross, Symphitopsyche sparna (Ross), measured adjacent to capture nets with a collapsible bag 5. madeodi (Flint), Diplectrona modesta Banks (Hyd- current meter (Gessner 1950). Seston concentrations ropsychidae) and Dolophilodes distinctus (Walker) were determined monthly by filtering and weighing four (Philopotamidae). Two additional unidentified hyd- 2-1 samples of stream water, two from near the surface ropsychid species, Cheumaiopsyche sp. and Hydro- and two from 2 cm above the substrate. Differences psyche sp., represented a combined abundance of well between the two sets of samples were insignificant in below 1.5% of the total hydropsychid larvae in any every month, and the average concentration was used. benthic sample. These two species were omitted from The number and size of diatoms and detritus particles the model because of their extremely infrequent were estimated by filtering 30 ml aliquots of water on occurrence. membrane filters. After clearing the filters and mount- Larvae were collected by handpicking and preserved ing them on slides, particle volumes were measured as along with their accompanying nets and retreats in 70% relative projected area (Wallace et al. 1977) at 200X. ethanol. Larval instar was determined by head capsule Gut analysis was carried out using similar methods. width, measured across the eyes. The capture nets of Detailed seston particle size distributions are not pre- the hydropsychid species were mounted in glycerine and sently available for Tallulah River seston. We incorpo- mesh size measurements were made with a drawing rated data from nearby streams in the Coweeta Hydro- tube and stage micrometer. The sizes of all the indi- logic Laboratory (Gurtz et al. 1980, Wallace unpubl.) vidual meshes in a single net were measured for five into the model where such particle size distributions nets; the mesh opening sizes were normally distributed were required. when the outer row of large, irregular meshes was 3. The model excluded. About 230 nets were measured. The ex- tremely small size of the nets of first and second instars For a generalized filter feeder, the rate of capture of made them difficult to collect and only one or two nets particles of size X, C(X), may be represented as the were collected for the early instars of certain species. product of three functions (Boyd 1976): Because of the increase in mesh size from base to S(X), the abundance of the panicles as a function of periphery, the lengths and widths of a series of the their size, smaller and of the larger meshes of each net were meas- V, the volume filtered in a unit time, and ured. Irregular peripheral meshes were not measured. E(X), the relative capture efficiency of the filter An average mesh opening area was calculated for feeder for particles of size X. each instar by averaging the widths of the small and large mesh series and the lengths of the small and large Tab. 1. Seston particle size distribution curves*. mesh series. The variable X the average mesh opening TO Month Total con- Curve size, was defined as the square root of the product of the centration parameters average mesh length and average width. The variance of 1 Xm was calculated as the arithmetic mean of the vari- (mg I- ) a b r- ance of the mesh lengths and the variance of the mesh (Summer) widths. A closely related variable, X, denoted the dia- Jul ? 065 361 462 -1 796 0 896 meter of a particle which would be caught in a mesh of AUB 1 780 size Xnr The total areas of the capture nets were meas- Sep .. 1 820 ured by drawing and cutting the outline of each net and (Autumn) comparing the relative weights of the cutouts to the Oct .... 0620 ">39 308 -1 693 0 949 weight of a standard area. Capture nets in situ in the Nov 0 538 field are bowed in the downstream direction by the Dec . . . 0.575 current; we measured the outline of the nets' peripheral (Winter) support elements rather than the area of the flattened Jan 1 075 188 929 -1 624 0 908 nets. The former measurement should reflect more ac- Feb . . . 1 "'SO curately the area of the water column intercepted by the Mar 1.240 net. Capture nets of Dolophilodes were examined using (Spring) scanning electron microscopy methods previously de- Aot .... 1 110 775 195 -1 710 0 91"' scribed (Wallace and Malas 1976). May "> 000 We determined benthic abundances from 3-4 Jun 1 360 monthly 0.25 m* bottom samples, using a fine meshed 3 (ca. 200 urn mesh opening) drift net (Malas and Wallace Fit by least-squares regression to the form: 1977). Larval instars were determined by head capsule Y = aX". width measurements. xViicrodistributions of IVth and See text for further details. OIKOS 36:2 (1981) 149 3.1. Particle abundance Tab. 2. Basic net parameters used in model. Total monthly seston concentrations (mg ash-free dry weight per liter) for the Tallulah River are given in Tab. Species Instar Mean Standard Volume mesh deviation filtered 1. The particle size distribution data used were in the opening (um) (um) (1 d-1) form of the percentage by weight of organic matter in each of five size classes: 0.45-25, 25-43, 43-105, D. distinctus V 2.45 27.143 105—234, and 234—864 urn. For some samples a sixth S. sparna I 39.31 4.12 1.869 class, 864-5000 um, was available. Gurtz et ai. (1980) S. macleodi I 40.80 4.92 2.831 gave further details. The data in this form were biased D. modesta I 42.66 5.17 2.124 S. sparna II 47.12 5.18 4.964 by the unequal ranges of the size classes, which were P. cardis I 48.63 5.98 4.321 chosen arbitrarily. The data were standardized by di- D. modesta II 58.31 6.83 7.432 viding each percentage by the range of its size class, S. macleodi II 63.36 7.36 11.270 yielding the measure "% per um." Use in the model S. sparna III 87.30 12.51 30.982 D. modesta III 88.62 14.19 28.575 demanded that these discrete data be transformed to a A. irrorata I 94.66 10.53 21.120 continuous function. We accomplished the transforma- S. macleodi III 98.59 11.73 46.816 tion by assigning each % per (xm value to the midpoint P. cardis II 113.89 15.62 29.617 of its size class, and fitting a least-square regression of S. macleodi IV 134.64 18.38 167.668 S. sparna IV 141.05 19.14 166.330 the form A. irrorata II 146.50 23.81 96.193 D. modesta IV 152.80 15.20 138.479 f(X) = a X" P. cardis III 161.63 16.78 160.223 to the resulting set of points. This transformation pro- D. modesta V 191.51 21.39 571.740 S. macleodi V 205.39 32.44 216.410 vided the best fit to our data and to seston data from a P. cardis IV 227.25 28.46 1573.003 wide range of North American streams (Sedell et ai. S. sparna V 232.08 28.93 700.068 1978). Four particle size distributions were used, one A. irrorata III 263.93 40.01 1239.476 P. cardis V 302.83 39.80 2663.544 for each season. The curve parameters are listed in Tab. A. irrorata IV 323.95 49.77 2013.974 1. The final measure of seston availability, S(X), is the A. irrorata V 495.50 62.79 7199.850 product of the total seston concentration for a given month and the relative abundance of particles of size X for that season (Fig. 2, solid line). precisely perpendicular to the direction of flow, the vol- ume filtered equals the total area of the net times the 3.2. Volume filtered current velocity in which the net is situated. As noted If we assume that resistance to flow and turbulence above, velocity measurements were possible only for created by a net are not significant, and that the net is IVth and Vth instars (Wallace et al. 1977). Velocity measurements for Vth-instar hydropsychids were re- gressed against the average mesh opening size of each 20 r instar's net (Tab. 2), yielding the relationship 1 l 40C (velocity, cm s' ) = 0.024 Xm - , n = 5, r2 = 0.836. The regression equation was used to estimate the vel- ocity microdistribution of smaller instars. This round- about procedure was required by the absence of a cur- rent meter sensor small enough (1-2 mm) to be placed directly in front of the smaller capture nets. If anything, the velocity values used in this model are probably overestimates, as turbulence and boundary layer effects (Hynes 1970) will reduce the effective flow through the net as compared with the region immediately beside and above the nets where the velocity measurements were made. Also, we were unable to measure the decrease in 100 300 500 velocity of the water flowing through a net caused by SIZE CLASS= jj.m the net's resistance to flow. This velocity drop has not been modeled for the hydropsychid species. We arbit- Fig. 2. Seston concentration curve (solid line). The curve illus- l rarily halved the velocity estimate used for Doio- trated is for July. Concentrations are expressed as ug l~ per um 1 size class. Volume filtered curve (dotted line). Values are philodes V (from 20 to 10 cm s" ) to account for the plotted against Xro the mean mesh size, of each instar. flow resistance created by their extreme fine-meshed Doiophilodes is not included. nets. 150 OIKOS36:2 (1981) The volume filtered, V, was calculated for each instar of the five hydropsychid species and for Vth-instar Dolophilodes (Tab. 2, Fig. 2). The curve of V against Xm is a composite of two functions: the area of the nets, summarized by the regression, 2 5 1 771 (area, cm ) = 1.082 x 10" Xm ' , n = 25, r2 = 0.917, and the velocity function given above. The composite function is 7 3 171 V = 2.60 x 10- Xm - . Wilson (1975) noted that in general the searching capacity of a predator is related to its body size (any linear dimension) by a power function of the form Y = aXb. He suggested that filter feeders, which increase the area of the filtering aperture and also the filtering vel- ocity as they grow larger, will be best described by equations with exponents of approximately 3. The ex- ponent we derived, 3.17, is in reasonably close agree- ment with the theoretical prediction. 3.3. Relative capture efficiency Rubenstein and Koehl (1977) described a number of mechanisms other than simple sieving by which filter feeders may capture particles. Several of their assump- tions may not apply to panicle capture by the Tallulah River hydropsychids. At the velocities we have meas- ured, for example, the Reynolds numbers of particles TOO 300 500 >10 urn in diameter are considerably in excess of 1 and Rubenstein and Koehl's assumption that viscous forces SIZE CLASS = predominate is probably inapplicable. The hydro- Fig. 3. Capture efficiency, illustrated for IVth instar A. ir- psychid species we studied have from 80 to 90% of their rorata. Top: mesh size distribution. Y(Xm), based on mean and total net area open, leading us to assume that there is standard deviation from Tab. 2. Bottom: relative capture effi- ciency curve, E(X). The capture efficiency at Xm = x, indicated not a significant velocity reduction at the net caused by by the star, is proportional to the area shaded above. More filter resistance. This assumption is untested. We sus- generally, E(X) =x J0 Y(X)dx. pect, based on the velocities involved, that of the non- sieving mechanisms direct interception and inertial im- paction would predominate. J. Oris, Univ. of Georgia capture by these caddisflies, an assessment of their im- (personal communication), has initiated an analysis of portance remains to be made. the relative contribution of these mechanisms to cad- For a simple sieve, the probability that a particle of disfly filter feeding. diameter X will be caught by the net is proportional to Non-sieving mechanisms may render some hydro- the number of mesh openings in that net which are psychid nets capable of capturing particles much smaller smaller than size X. If the mesh size frequency distribu- than the mesh opening size (on the order of the thick- tion is denoted Y(Xm), then the relative capture effi- ness of the net fibers themselves, ca. 5—30 um). It does ciency of the net for panicles of size X, E(X), is equal to not appear that Tallulah River hydropsychids possess the area under Y(XJ to the left of X (Fig. 3). any anatomical structures suitable for the efficient col- The shape of E(X) is robust against changes in the lection of such small particles from the nets. By com- mesh size frequency distribution, provided that the lat- parison, caddis larvae which filter very fine particles, ter function is approximately normal. Maximum effi- <50 \an in diameter (e.g. Philopotamidae, some Ma- ciency is approached asymptotically as particle size in- cronematinae, and Phylocentropus), possess various creases, and has been set for this model at 100%. Kur- dense brushes of setae on the mouthparts or forelegs tak (1978) compared empirical and calculated seston which serve to rake particles from the net into the oral capture rates of various blackfly larvae and found that opening (Sattler 1963, Wallace and Malas 1976, Wal- actual capture efficiencies were generally less than 1% lace and Merritt 1980). While non-sieving mechanisms of the predicted maximum. Presumably, the flow resis- of filter feeding may make a contribution to seston tance and turbulence effects mentioned above, possibly OIKOS 36:2 (1981) 151 in combination with inefficient transfer of particles from the collecting structures to the mouth, are responsible for the low capture efficiencies determined for blackfly m larvae. No similar data exist for net-spinning caddisfly where pik and pjk are the contributions of the k re- larvae. It should be borne in mind that this model, using source category to the im and f species or instar, re- 100% as the maximum capture efficiency, may spectively, as an estimate of competition for resources overestimate seston capture. between instars. Three possible axes of niche differen- tiation (Schoener 1974) were considered: food particle size, food type, and temporal occurrence of larvae. The overlap in particle size utilization was calculated from 4. Results the seston capture curves of instars adjacent when The rate of seston capture for all twenty-five instars of ranked by average mesh opening size (Tab. 2). Food the five hydropsychid species was simulated on a type preferences were estimated from the per cent of monthly basis. The data base permitted the simulation five food types (animal, vascular plant detritus, diatoms, to be extended only to Vth-instars of Dolophilodes. A filamentous algae, and detritus of unknown origin) seston capture curve was calculated for each instar. Fig. found in the foregut (see Malas and Wallace 1977, 4 gives a typical curve. The ascending left-hand portion Wallace et al. 1977, Benke and Wallace 1980, for of the curve is generated by the relative capture effi- original data). Gut content data were not available for ciency, which rises sharply as particle size increases to- 1st instars. Temporal overlap was calculated for all in- ward the mean mesh opening size of the net, Xm. At stars using monthly abundance data. some point past Xm the effect of the efficiency curve, The calculated overlap coefficients are summarized in which begins to level off, is counterbalanced by the de- Tab. 3. The overlap in panicle size utilization between clining seston concentration curve (Fig. 2, solid line), families is between Vth instar Dolophilodes and the and the capture curve begins to decline. The capture hydropsychid with the smallest mesh opening, 1st instar curves are distinctly asymmetric, with long right-hand 5. spama. The five hydropsychid species show little tails. None of the curves had any appreciable area under feeding differentiation in terms of either food type or their right-hand tails beyond 1000 fim, while only particle size. They do exhibit considerable temporal Dolophilodes had a significant capture rate for panicles segregation. The five Dolophilodes instars, on the other less than 10 urn in diameter, so most of the simulations hand, are present together during much of the year were conducted over a size range of 10 to 1000 um. (Malas and Wallace 1977, Benke and Wallace 1980), Simulations of seston capture by Dolophilodes began at resulting in a high temporal overlap coefficient. We 5-7 um, depending on the season. Below this point, the have not quantified the utilization of particle size for power curve used to describe seston concentration rose Ist-IVth instar Dolophilodes larvae, but Malas and to unrealistic levels (Fig. 2). Wallace (1977) presented data indicating that Dolophilodes instars overlap extensively in both particle size and type ingested. The overlap coefficients between 4.1. Resource partitioning Vth instar Dolophilodes and the five hydropsychid We employed Pianka's (1973) symmetric measure of species range from extremely low (particle size) to in- niche overlap, termediate (food type and temporal occurrence). 0.4- Tab. 3. Resource overlao coefficients* Comparison Resource type Particle Food Time size type I 0.2 Hydropsychidae 0.956±0.017 0.760±0.027 0.387±0.018 n = 24 n = 190 n = 300 Between families 0.112" 0.699±0.035 0.515=0.027 n = 1 n = 60 n = 125 Philopotamidae 0.997±0.001 0.912±0.012 n = 3 n = 10 TOO 300 500 SIZE CLASS' jj.m a Calculated after Pianka 1973. All values are mean ± 1 stan- dard error. b Fig. 4. Seston capture curve for IVth instar A. irrorata. Xm is Based on Vth instar Dolophilodes distinctus and 1st instar 5. the mean mesh opening size. spama only. 152 O1KOS 36:2 (1981) Two generalizations emerge from the overlap coeffi- 200 r 2OO r cients presented here: (1) As has been noted previously (Edington 1968, Williams and Hynes 1973, Malas and Wallace 1977), the hydropsychid species differ signific- A, IRRORATA P.CARDiS antly from the philopotamid Dolophilodes in the size, and somewhat in the type of food particles utilized; (2) Differences in food size and type do not appear to pro- 1OO 1OO vide sufficient niche differentiation to explain the coexistence of the hydropsychid species if they are com- peting for food. Three lines of evidence lead us to reject the original hypothesis of resource partitioning by parti- cle size. First, theoretical considerations (MacArthur and Levins 1967) suggest that coexistence is impossible I II II! IV V I II III IV V when total niche overlap exceeds roughly 0.5. Over 90% of the particle size overlap coefficients between adjacently ranked instars (Tab. 2) are greater than 0.9, and all are greater than 0.5. Second, the hydropsychid too 100 instars do not evenly divide the particle size resource spectrum. When compared to a neutral model (e.g. S. SPARNA D.MODESTA 50 50 Caswell 1976) generated by Monte Carlo simulations, the mean mesh sizes listed in Tab. 2 were significantly less evenly distributed than would have been expected had their values been randomly distributed. Finally, I II ill iv v I II III IV V temporal partitioning, while of the level predicted to INSTAR NO. INSTAR NO. permit coexistence, appears to be unrelated to the way in which the instars utilize either particle size or food Fig. 5. Seston capture rates predicted by the model, using the Tallulah River (solid line) and the Savannah River (dotted type. Schoener (1974) recommended that niche com- line) seston concentration curves. Capture rates are mg d~' per plementarity (i.e. the tendency for species which individual. overlap extensively in one niche dimension to be well separated on some other dimension) be viewed as evi- dence for resource partitioning due to competition. dieted seston capture rate for Vth instar Dolophilodes at Among the hydropsychids, however, instars with similar its period of maximum abundance (summer) is 50 mg particle size utilization did not necessarily have low per day. Capture rates per individual, combined with temporal overlaps: regression of particle size overlap x abundance data, provide an estimate of the total seston temporal overlap gave a correlation of r = -0.101, n = capture, on a daily basis, during each month (Fig. 6). 24, P > 0.99. The total annual capture, using the average of The overlap coefficients between the two arcto- November and January for the missing December sam- psychinine species (Arctopsyche and Parapsyche) may ple, is 985 g m~2. Inclusion of Dolophilodes would ele- provide the only exception to the results discussed above. Parapsyche has feeding habits very similar to those of Arctopsyche, but develops later in the year than co A m Arctopsyche. The coefficient of temporal overlap be- co "v-"\ tween instars of these two species is low (X = 0.355) - ~ -6 o and temporal overlap is negatively correlated with par- / o ticle size overlap (r = -0.829, n = 9, P < 0.01). These / \ * < n o > O N CAPTURE D two species may be an example of the pattern noted by 2 day- ' i— i Hutchinson (1959), who proposed that the larger of the \ P 2 s ^ y 1—i 3 species in a guild should breed first, and therefore CO Cn _ •" _j ^ ** L maintain' the size differential throughout the period of <^ "-•-' *^ 'vi immature development. ~l ^ O | 4.2. Seston capture rates MONTH The area under each instar's seston capture curve is its Fig, 6. Total seston capture by the five hydropsychid species, predicted rate of seston capture, in milligrams per day. by month (bars). The peaks in July and September are caused by the high seston concentration (dotted line) in those months. These rates appear in Fig. 5 (solid lines) for the four The peak in February is generated by increased numbers of most abundant hydropsychid species. The rates for 5. late-instar larvae with high seston capture rates. The decline madeodi closely resemble those for 5. sparna. The pre- from March through May is caused by emergence. 11 OIKOS 36:2 (1981) 153 vate the annual total to about 1300 g m~2. Capture by for hydropsychids in some larger streams (Edington early instar Dolophilodes may have been underesti- 1965, Hildrew and Edington 1979, Cudney and Wal- mated. lace 1980). The maximum hydropsychid abundances Benke and Wallace (1980) measured annual produc- observed (five species combined) were 250 individuals tion by these six species as 1.0 g m~2. Calculating from per m2 during the summer when early instars predomi- data on food quality and assimilation efficiencies, they nated, and ranged from 25-75 larvae per m3 during the estimated that 5.33 g m~2 per year of ingestion would rest of the year. Cudney and Wallace (1980), by com- support the observed level of production. Conversely, parison, measured annual average densities of net- 20-30 metric tons of paniculate organic matter pass spinning caddisfly larvae in the Savannah River as high over 1 m2 of the Tallulah River in a year. The model as 22600 per m2 of substrate surface area in the most suggests, then, that these species of net-spinning cad- favorable habitats. disflies are unlikely to be limited by the quantity of food What then is the significance of the range of mesh available in the Tallulah River, since they are predicted sizes found in the caddisfly nets there? We suggest that to capture roughly 200 x the material required for in- there are advantages to be found in the operation of gestion, while only depleting the available seston supply either a fine-meshed net in low current velocities or a by about 0.005% per longitudinal meter of stream. coarser-meshed net in rapid current. For the sake of McCullough et al. (1979) suggested that filter-feeder simplicity, we will distinguish these two extremes as dis- populations may be self-limiting since they reduce the tinct "strategies" (i.e. discrete, alternate sets of adapta- energy content of the seston they ingest and then egest. tions which permit survival in a given environment), Benke and Wallace (1980) found that the net-spin- while recognizing that a broad range of intermediate ners in the Tallulah River increase the detritus portion patterns exist. A fine-meshed net is more efficient at of the seston. Oswood (1980) also found evidence that filtration than a coarser-meshed net and the smaller the main effect of filter feeders below a lake outlet was particles which it can capture comprise by far the largest to alter seston quality rather than quantity. Our results fraction of the seston in a stream like the Tallulah. Such confirm that the hydropsychids may effect the quality of a net, however, creates considerable resistance to flow seston by selectively removing animal material, but are and is limited to slower currents (Wallace et al. 1977). unlikely to retain major quantities of the organic mat- A coarse-meshed net can compensate for its low effi- erial being swept downstream. Filter feeders utilizing ciency and the relative scarcity of the particles it can very fine particles (<25 um) may have a more signific- capture by a dramatically increased filtration volume ant impact on nutrient spiralling, as they have access to (Tab. 1). The increase in filtration volume may be the most abundant class of particles, and are found at achieved with a larger net area as well as by locating the low velocities in which non-sieving mechanisms of filt- net in a rapid current. ration may be important. We did not include other filter We propose that the hydropsychid species, with the feeders such as Simuliidae, several Chironomidae, and possible exception of Diplectrona, are limited not by the the mayfly Isonychia (Wallace and O'Hop 1979). Our total quantity of food available, but by the supply of estimate of capture may be compared with the data on high quality food (Maciolek 1966), primarily drifting ingestion rate obtained for Hydropsyche occidentals animals. Animal material is generally scarce in the ses- and several species of Simulium in Deep Creek, Idaho ton (Maciolek and Tunzi 1968, Malas and Wallace (McCullough et al. 1979), which were calculated to 1977, Wallace et al. 1977), except below lake outlets consume 0.011% of the seston per longitudinal meter of (Chutter 1963, Cushing 1963, Kubicek 1970). Animals stream (McCullough pers. comm.). The ingestion rate generally drift on the average less than 50 m before of the Tallulah River net-spinning caddisflies would be returning to the substrate (Elliott 1971). O'Hop (pers. several times lower than the predicted capture rate. comm.) studied diel drift in two 2nd-order streams. He found fewer than 2 animals, on the average, in 1 m3 of water, including such small organisms as nematodes, 5. Discussion copepods, and ostracods. Animal material comprised 5.1. Limiting factors less than 0.25%, by weight, of the total seston. The hydropsychids in the Tallulah River are predo- Our model provides no support for the hypothesis that minantly dependent on drifting animals as a food food is a limiting resource partitioned among instars or source. When their gut contents are corrected for as- species. Benke and Wallace (1980) estimated that the similation efficiencies on various food types, 86.5% of Tallulah River net-spinners ingest a minimal quantity of their production may be attributed to animal material, food in a year when compared with the seston available. with only 5.1% attributed to detritus, which is typically These two calculations, based on essentially indepen- the largest component of the total seston (Benke and dent sets of assumptions, together support the conclu- Wallace 1980). The percent attributable to animal sion that the total food supply is not a limiting factor for material is 93.3% for Arctopsyche. These figures are the Tallulah River net-spinners. Nor does it seem likely undoubtably underestimates of the importance of ani- that net-spinning sites are limiting, as has been reported mal material, since some of the detritus probably rep- 154 OIKOS 36:2 (1981) resents the gut contents of prey organisms and it is likely lulah, accounting for 33 and 24% of the total seston that animal material is assimilated more rapidly than capture, respectively. other food and therefore is underrepresented in the gut (Cummins 1973). 5.2. Longitudinal distribution The following factors explain why there is no net- spinner in the Tallulah River with meshes larger than A wide variety of factors have been used to explain the those of Vth instar Arctopsyche: Animal material is rare sequential longitudinal distribution of filter feeders in seston, and must be regenerated over short reaches to along a stream (Edington 1968, Gordon and Wallace compensate for the rapid rate at which drifting animals 1975, Badcock 1976, Wiggins and Mackay 1978, Mac- return to the substrate (Elliott 1971). Hence it is not kay 1979, Alstad 1980) as well as their coexistence unlikely that Arctopsyche larvae remove most of the within a given segment of the stream continuum (Mil- drifting animals from the water column and are limited drew and Edington 1979, Wallace and Merritt 1980). by the availability of high-quality food. A net-spinner Some of the factors emphasized have been stream size with even larger net meshes would have to occupy sub- (order), substrate, current velocity, temperature and stantially more rapid currents to filter sufficiently large particle availability (see Hynes 1970, Gordon and volumes of water. But Arctopsyche is found in velocities Wallace 1975, Wallace and Merritt 1980, for reviews). as high as 160 cm s"1 (Wallace et al. 1977), and more Alstad (1980) has shown, for hydropsychids in Utah, rapid currents are not generally available. that the species with the largest meshed net is the first to Instars with smaller-meshed nets may be forced to be found at high elevations, and the sequential order in settle for the animals small enough to pass through the which other species appear downstream is almost per- coarser nets, supplementing their diets with lower qual- fectly correlated with their decreasing mesh sizes. We ity foods. It seems likely that, even for quantities of food hypothesize that the mesh sizes, current1" preferences, with comparable caloric content and assimilability, and food type selectivities of net-spinning caddisflies in there will be an advantage to the selection of the largest a stream are determined by the available current and particles available (and manageable), since these re- seston particle size distributions. Food preferences are quire fewer manipulations per unit volume or weight linked to current: in order to specialize on rare, high- and thus permit more rapid ingestion (McLachlan et al. quality food items, a larva must filter sufficiently large 1978). Wilson (1975) has explored other adaptive ad- volumes of water to insure a dependable supply of those vantages of feeding on larger particles than the next items. Equally, the rapid-current, high-selectivity nearest competitor. strategy is dependent on the availability of large parti- Dolophilodes represents the other extreme strategy. cles, as described by the seston particle size distribution. Its nets are extremely fine-meshed and are located at The precise mesh size at which the increased filtration depositional velocities where large particles will be rare. rate compensates for reduced efficiency and diminishing It is not surprising that its feeding mechanisms rule out particle abundance depends on the particle size dis- selectivity by particle size or type (Wallace and Malas tributions and current velocities present in a stream. To 1976, Wiggins 1977). The panicles utilized by Dolo- illustrate this point, we repeated our simulation, using philodes, roughly 1 to 25 urn in diameter, are the most panicle size curves and current regimes characteristic of abundant and also the most seasonally dependable of all coastal plain portions of the Savannah River. The the seston (Sedell et al. 1978, Naiman and Sedeil 1979, Savannah River there is a large river, averaging ap- Gurtz et al. 1980) implying that food is not a factor proximately 4 m in depth and 150 m in width. Discharge limiting this species. Dolophilodes occupies a very varies from 100 to over 600 m3 s'1 (Cudney and Wal- specific microhabitat, the undersides of rocks (Maias lace 1980). The total seston concentrations (1.0—3.5 mg and Wallace 1977), and in considerable densities. The 1~\ mean = 2.1), were similar to those used for the maximum standing crop of Dolophilodes larvae meas- Tailulah River simulation but the panicle size distribu- ured was 445 per irr in July. These were not predomin- tions had larger intercepts (557-2410, mean = 888) antly early instars; 60% were IV and Vth instars. In the and steeper slopes (2.0-2.5, mean = 2.1), representing Tailulah River, where many of the large rocks are par- the dominance of very small particles. The percentage tially buried in the substrate, the microhabitat available (by weight) of particles in the smallest size class, to Dolophilodes may be limiting. 0.45-25 um, ranged from 75 to 84% (mean = 81.0). It appears, then, that there are two different Additionally, a maximum current velocity of 80 cm s~l strategies available to net-spinning caddisflies: (1) was imposed (Cudney pers. comm.). The five hydro- selective feeding on large, rare, high-quality panicles, psychid species present in the Tallulah River showed especially animals, with a resulting requirement for suf- marked decreases in seston capture rates when ficiently rapid currents and large-meshed nets; (2) Non- •'moved" to the Savannah River. Combined seston selective feeding on the abundant very fine panicles, capture by all instars of the five species fell by 80%. largely detritus. It is significant that the two extreme Inspection of the capture rates by species and instar examples of these strategies, Arctopsyche and (Fig. 5, dashed line) indicates that the later instars of the Dolophilodes, dominate the net-spinners in the Tal- largest species, Arctopsyche and Parapsyche, suffered 11- OIKOS36:2 (1981) 155 the largest proportionate declines in seston capture. The Boling, R. H.. Jr.. Goodman. E. D.. Van Sickle. J. A.. Zimmer. arctopsychines, with their extremely coarse-meshed J.~O.. Cummins. K. W.. Petersen, R. C. and Reice. S. R. 1975. Toward a model of detritus processing in a woodland nets, are unable to survive in large, low-gradient rivers stream. - Ecology 56: 141-151. due to the combined effects of the absence of very rapid Boyd. C. M. 1976. Selection of panicle sizes by filter-feeding currents and the relative scarcity of large seston pani- copepods: a plea for reason. - Limnoi. Oceanoar. 21: cles. 175-180. Caswell. H. 1976. Community structure: a neutral model Our hypothesis may explain the longitudinal dis- analysis. - Ecol. Monogr. 46: 327-354. tribution noted by Gordon and Wallace (1975), as well Chutter. F. M. 1963. Hydrobiological studies on the Vaal as the similar pattern observed by Alstad (1980). In River in the Vereeniging area. Pan I. Introduction, water contrast to the distribution shown by the hydropsychids, chemistry and biological studies of the fauna of habitats other than muddy bottom sediments. - Hydrobiologia 21: the Philopotamidae. with extremely fine-meshed nets, 1-65. are found from the headwater streams to the tidal re- Cudney, M. D. and Wallace. J. B. 1980. Life cycles, microdis- gions of the Savannah River basin, although genera and tribution, and production dynamics of six species of net- species change. Because the fine particles on which they spinning caddisflies in a large southeastern (USA) river. - Hoiarctic Ecol. 3: 169-182" feed are extremely abundant even in headwater Cummins, K. W. 1973. Trophic relations of aquatic insects. - streams, the species pursuing a low-velocity, non-selec- Ann. Rev. Ent. 18: 183-206. tive strategy have access, as a group, to the entire basin. Gushing, C E., Jr. 1963. Filter-feeding insect distribution and Gradient may function as a son of "driving variable," pianktonic food in the Montreal River. - Am. Fish. Soc. Trans. 92: 216-219. determining both the longitudinal sequences of net- Edington, J. M. 1965. The effect of water flow on populations spinners and their coexistence in a given region in sev- of net-spinning Trichoptera. - Mitt. Int. Ver. Theor. eral ways: (1) As a direct determinant of current velo- Angew. LimnoT. 13: 40—18. city; (2) Through stream power, which has been - 1968. Habitat preference in net-spinning caddis larvae with special reference to the influence of water velocity. - suggested as having a major impact on seston quantity J. Anim. Ecoi. 37: 675-692. and retention (Sedell et al. 1978); (3) Through its rela- Elliott. J. M. 1971. The distances travelled by drifting inver- tion to stream order, which (on the continuum tebrates in a Lake District stream. - Oecologia (Berl.) 6: hypothesis, Vannote et al. 1980) controls the processing 350-379. Gessner, F. 1950. Die okologische Bedeutung der Stro- of detritus and hence the size and quality of seston; and mungsgeschwindigkeit fliessender Gewasser und ihre finally, (4) By the effect of current and discharge on Messung auf klemstem Raum. - Arch. Hydrobiol. 43: constant and catastrophic drift (Waters 1965), in- 159-165. fluencing the number of animals drifting and the dis- Gordon, A. E. and Wallace, J. B. 1975. Distribution of the family Hydropsychidae (Trichoptera) in the Savannah tance they drift (Elliott 1971). Temperature tolerances River basin of North Carolina. South Carolina and Geor- and specialized microhabitat distributions may be sec- gia. - Hydrobiologia 46: 405—123. ondary, or proximate adaptations permitting species to Gurtz. M. E"., Webster. J. R. and Wallace. J. B. 1980. Seston inhabit the portions of the stream continuum (Vannote dynamics in southern Appalachian streams: effects of clearcutting. - Can. J. Fish. Aquatic Sci. 37: 624-631. et al. 1980) and even those biomes (Wiggins and Mac- Hildrew, A. G. and Edington, J. VI. 1979. Factors facilitating kay 1978) which offer the appropriate current and ses- the coexistence of hydropsychid caddis larvae (Trichopt- ton distributions. era) in the same river system. - J. Amm. Ecoi. 48: 557-576. Hutchinson, G. E. 1959. Homage to Santa Rosalia, or why are there so many kinds of animals'? - Am. Nat. 93: 145—159. Acknowledgements - We thank J. O'Hop, D. Malas and J. Oris Hynes. H. B. N.' 1970. The Ecology of Running Waters. - for assistance. T. Curfney, Drs J. R. Webster and J. Meyer Univ. Toronto Press, Toronto. provided valuable constructive criticism on earlier drafts of the Kaiser. P. 1965. Uber Netzbau und Stromungssinn bei den manuscript. This work was supported by Grant No. DEB Larven der Gatung Hydropsyche Pict. - Int. Rev. Ges. 78-03143 from the National Science Foundation (USA). Hydrobiol. 50: 169-224. Kubicek, F. 1970. On the drift of a brook running through a pond. - Vestnik Ceskoslovenske Spoiecnosti Zooloaicke 34: 219-226. Kurtak. D. C. 1978. Efficiency of filter feeding of biackfly larvae. - Can. L Zool. 56: '1608-1623. MacArthur, R. H. and Levins. R. 1967. The limiting similarity, References convergence and divergence of coexisting species. - Am. Alstad, D. N. 1980. Comparative biology of the common Utah Nat. 101: 377-385. Hydropsychidae. - Airier. Midi. Mat. 103: 167-174. Maciolek, J. A. and Tunzi. M. G. 1968. Microseston dynamics Anderson. N. H. and Sedell. J. R. 1979. Detritus processing by in a simple Sierra Nevada lake-stream system. - Ecology macroinvenebrates in stream ecosystems. - Ann. Rev. 49: 60-75. Ent. 24: 351-377. Mackay. R. J. 1979. Life history patterns of some species of Badcock. R. M. 1976. The distribution of Hydropsychidae in Hvdropsyche (Trichoptera: Hydropsychidae) in southern Great Britain. - Proc. First Int. Symp. Trichoptera, pp. Ontario.'-Can. J. Zooi. 57: 963-975. 49-58. Malas, D. and Wallace. J. B. 1977. Strategies for coexistence in Benke. A. C. and Wallace, J. B. 1980. Trophic basis of pro- three species of net-spinning caddisflies (Trichoptera) in duction among net-spinning caddisflies in a southern Ap- second-order southern Appalachian streams. - Can. J. palachian stream. - Ecology 61: 108—118. Zool. 55: 1829-1840. 156 OIKOS36:2 C98I) McCulIough, D. A., Minshail, G. W. and Gushing, C. E. 1979. Wallace, J. B. 1975. Food partitioning in net-spinning Bioenergetics of lotic filter-feeding insects Simulium spp. Trichoptera larvae: Hydropsyche vemdaris, Cheumato- and Hydropsy die occidentalis and their function in con- psyche eirona, and Macronema zebratum. — Ann. Ent. Soc. trolling organic transport in streams. - Ecology 60: Amer. 68: 463-472. 585-596. - and Malas, D. 1976. The significance of the elongate rec- McLachlan, A. J., Brennan. A. and Wotton, R. S. 1978. Pani- tangular mesh found in capture nets of fine panicle filter cle size and chironomid (Diptera) food in an upland river. feeding Trichoptera larvae. — Arch. Hydrobiol. 77: -Oikos31: 247-252. 205-212. Minshail, G. W. 1978. Autotrophy in stream ecosystems. - - and Merritt, R. W. 1980. Filter-feeding ecology of aquatic BioScience 28: 767-771. insects. - Ann. Rev. Ent. 25: 103-132. Naiman, R. J. and Sedell, J. R. 1979. Characterization of par- - and O'Hop, J. 1979. Fine particle suspension-feeding ticulate organic matter transported by some Cascade capabilities of Isonychia spp. - Ann. Ent. Soc. Amer. 72: Mountain streams. - J. Fish. Res. Bd Can. 36: 17-31. 353-357. Oswood, M. W. 1980. Abundance patterns of filter-feeding - , Webster, J. R. and Woodall, W. R. 1977. The role of filter caddisflies and seston in a Montana (USA) lake outlet. - feeders in flowing waters. - Arch. Hydrobiol. 79: Hydrobiologia 63: 177-83. 506-532. Pianka, E. R. 1973. The structure of lizard communities. - Waters, T. F. 1965. Interpretation of invertebrate drift in Ann. Rev. Ecol. System. 4: 53-74. streams. - Ecology 46: 327-334. Rubenstein, D. I. and Koehl, M. A. R. 1977. The mechanisms Webster, J. R. 1975. Analysis of potassium and calcium of filter feeding: some theoretical considerations. - Am. dynamics in stream ecosystems on three southern Ap- Nat. 111:981-994. palachian watersheds of contrasting vegetation. - Unpubl. Ruttner, F. 1963. Fundamentals of Limnology. D. G. Frey and Ph.D. Thesis, Univ. Georgia, Athens. F. E. J. Frey, transl. - Univ. Toronto Press, Toronto. - and Patten, B. C. 1979. Effects of watershed perturbation Sattler, W. 1963. Uber den Korperbau, die Okologie und on stream potassium and calcium dynamics. - Ecol. Ethologie der Larve und Puppe von Macronema Pict.. ein Monogr. 49: 51-72. als Larve sich von "Mikro-Drift" emahrendes Trichoptera Wiggins, G. B. 1977. Larvae of the North American Caddisfly aus dem Amazonasgebiet. - Arch. Hydrobiol. 59: 26-60. Genera (Trichoptera). - Univ. Toronto Press, Toronto. Schoener, T. W. 1974. Resource partitioning in ecological - and Mackay. R. J. 1978. Some relationships between sys- communities. -Science 185: 27-39. tematics and trophic ecology in Nearctic aquatic insects, Sedell, J. R., Naiman, R. J., Cummins, K. W., Minshail, G. W., with special reference to Trichoptera. - Ecology 59: and Vannote, R. L. 1978. Transport of paniculate organic 1211-1220. matter in streams as a function of physical processes. - Williams, N. E. and H. B. N. Hynes. 1973. Microdistribution Verh. Int. Verein. Limnol. 20: 1366-1375. and feeding of the net-spinning caddisflies of a Canadian Vannote, R. L.. Cummins, K. W.. Minshail, G. W., Sedell, J. stream. - Oikos 24: 73-S4. R. and Gushing, C. E. 1980. The river continuum concept. Wilson. D. S. 1975. The adequacy of body size as a niche - Can. J. Fish. Aquatic Sci. 37: 130-137. difference. - Am. Nat. 109: 769-784. OIKOS 36:2(1981) 157