D ifferential e fficiencies of D ip - net s ampling versus s ampling s urface - floating p upal e xuviae in a B ioDiversity s urvey of c hironomiDae

Relative efficiencies of standard dip-net sampling (SDN) versus collections of surface-floating pu-pal exuviae (SFPE) were determined for detecting Chironomidae at catchment and site scales and at subfamily/tribe-, genus-and species-levels based on simultaneous, equal-effort sampling on a monthly basis for one year during a biodiversity assessment of Bear Run Nature Reserve. Results showed SFPE was more efficient than SDN at catchment scales for detecting both genera and species. At site scales, SDN sampling was more efficient for assessment of a first-order site. No consistent pattern, except for better efficiency of SFPE to detect Orthocladiinae genera, was observed at genus-level for two second-order sites. However, SFPE was consistently more efficient at detecting species of Orthocladiinae, Chironomini and Tany-tarsini at the second order sites. SFPE was more efficient at detecting both genera and species at two third-order sites. The differential efficiencies of the two methods are concluded to be related to stream order and size, substrate size, flow and water velocity, depth and habitat heterogeneity, and differential ability to discriminate species among pupal exuviae specimens versus larval specimens.


Introduction
Standardized dip-net (SDN) sampling for bioassessment of aquatic insects in water quality monitoring programs is a collection method endorsed by the US Environmental Protection Agency (Barbour et al. 1999) and has been widely adopted by state agencies and private corporations.Although widely adopted, sample processing and identification of larval chironomids collected using this method is a time-consuming activity, even when family, subfamily or tribe-level identification are the measurement end-point, and more efficient methods could result in decreased costs to water quality monitoring programs.
By contrast to SDN sampling, collections of surface-floating pupal exuviae (SFPE) have been shown to be a more time-efficient approach to assessing chironomid composition in a variety of stream settings (Anderson and Ferrington 2012, Bouchard and Ferrington 2011, Ferrington et al. 1991, Sealock and Ferrington 2008).However, no comparisons of the results of simultaneous collections using both methods have been made across stream stretches of differing order in catchments with good water quality and high habitat conditions in the United States, so it is not possible to determine how effective collections of SFPE are in resolving chironomid composition.
For this paper, we have re-analyzed historical data from a biodiversity assessment of aquatic invertebrates of the Bear Run Nature Preserve (BRNR) to calculate the comparative efficiencies of detecting chironomids using both SDN and SFPE methods at sites located on first-through third-order stream stretches.Questions we sought to answer were: what are the comparative efficiencies of detecting chironomid compositions at (1) catchment-scale, (2) stream order scale and (3) at differing taxonomic levels (subfamily/tribe, genus and species levels).

Catchment
The Bear Run Nature Reserve is located adjacent to PA Route 381, approximately 3 miles south of Mill Run, Stewart Township and Springfield Township in Fayette County, southwestern Pennsylvania.The BRNR is approximately 5,079 acres, most of which has been protected since 1963.Two streams, Bear Run and Lick Run, and their unnamed springs, seeps and small spring runs drain the BRNR.
The plant community types in the reserve range from regenerating old fields and conifer planta-tions to second-growth deciduous and hemlock forests.Dominant trees include tulip tree, red maple, chestnut oak, sugar maple, and black cherry.American beech is dominant in the upper Bear Run and Lick Run riparian areas.Stands of eastern hemlock occur in the major drainages, especially along Bear Run.Understory forest plants include spicebush, rosebay rhododendron, smooth azalea, witch hazel, and mountain laurel.Wildflowers include moccasin flower, trailing arbutus, painted trillium, Solomon's seal, yellow violet, and sweet violet.

Sample Sites
Five sample sites were defined on first-order (one, referred to as Site 1), second-order (two, referred to as Sites 2 and 5) and third-order (two, referred to as Sites 3 and 4) stretches of stream.At each site, SDN sampling was done by one of us (LCF) along approximately 10 meters of stream and collections of SFPE were taken (WPC) along approximately 60 meters of stream.On each collection date and at each sample site, the time and effort for each method were similar.
Site 1 (first-order) was located where the stream crosses Rhododendron Trail.The stream varied between 0.5 to approximately 1.2 meters wide and up to 15 cm depth.Stream substrates consisted of rounded boulder, intermixed with sand and gravels.Woody debris was also common.The entire stream site consisted of erosional areas, without well-developed pool or glide habitats and slow water velocity which, in July/August, were further reduced to trickle with barely visible water movement.
Site 2 (second-order) was located downstream of Site 1, where the stream crosses under Ridge Trail.The stream varied between 0.8 to approximately 1.2 meters wide and up to 25 cm depth.Stream substrates consisted of smaller amounts of rounded boulder and sand than Site 1, but greater proportion of gravels.Woody debris was also common.The entire stream site consisted of erosional areas, without well-developed pool or glide habitats.Water velocity was intermediate of sites sampled, and flow was nearly constant across sample dates, even during July/August.Site 3 (third-order) was located on Bear Run, where the stream crosses under Arbutus Trail.The stream varied between 2.5 to approximately 5.0 meters wide and was often greater than 1.0 meters depth.Stream substrates were highly heterogeneous, and consisted of gravel, cobble and boulder in well-developed riffles, to sand and finer mud in pools.Woody debris was present, but less common than Sites 1 and 2. Water velocity in riffles was highest of all sites sampled, and flow was nearly constant across sample dates.
Site 4 (third-order) was located on Bear Run, approximately 1.2 kilometers stream-distance downstream of Site 3, where the stream crosses under PA Route 381.The stream varied between approximately 4.0 meters to greater than 5.0 meters wide and up to approximately 0.8 meters depth.Stream substrates were highly heterogeneous, and consisted of gravel, cobble and boulder in well-developed riffles to sand and finer mud in pools.Woody debris was present, but less common than Sites 1, 2 and 3. Water velocity in riffles was less than Site 3 due to less local gradient of slope, and flow was nearly constant across sample dates.
Site 5 (second-order) was located on a tributary of Bear Run that intersects upstream of Site 3 and is upstream of where Hemlock trail crosses this tributary near the south edge of the reserve.This tributary varied between 0.6 to approximately 1.6 meters wide and up to 25 cm depth.Stream substrates consisted of smaller amounts of rounded boulder and sand than Site 1, but greater proportion of gravels similar to substrates at Site 2. Woody debris was also common.The entire stream site consisted of erosional areas, without well-developed pool or glide habitats.Water velocity was intermediate of sites sampled, and flow was nearly constant across sample dates, even during July/August.This site is the only stretch of stream that flowed through unprotected areas of land, approximately 100 meters stream-distance upstream.

Sampling Methods
Dip-net sampling was performed consistent with the method described in Barber et al. (1999), except that on each sample date some marginal leaflitter, submerged or partially submerged wood, and larger boulders were inspected and larval specimens from these substrates were hand-picked and added to the sample in the field.SFPE were collected consistent with the method described in Ferrington et al. (1991).This method targets areas where detritus is accumulated by currents and also accumulations of floating foams in which floating pupal exuviae can become entrained.SDN and SFPE samples were both preserved in the field with 70% Ethanol and returned to lab for sorting under 6-12X magnification.
Samples from each site were collected on the same date, approximately monthly from September 1975 through August 1976.At each sample site on each collection date the timed effort spent sampling was similar, but timed effort varied slightly from winter (less time) to summer.In addition, the SDN samples were strongly oriented to riffle-like conditions and deeper pool habitats were inefficiently sampled.

Specimen Preparation and Identification
Larval and pupal exuviae specimens were dehydrated in 95% Ethanol then slide-mounted in Euparol for identification using primary literature available at the time of collection.Identifications of larvae and pupal exuviae were performed independently and at the end of the project, our identifications were reconciled by sample site and cumulative lists of species were assembled for each sample site and for the entire catchment.Voucher collections were assembled.The larval vouchers are deposited at the University of Minnesota Insect Collection and the pupal exuviae specimens are deposited at the William P. Coffman lab at the La Selva Biological Station in Costa Rica.Nomenclature for all taxa has been updated to reflect current generic and species definitions, and larval vouchers have been reviewed for quality assurance.
Results showed SFPE was more efficient than SDN at catchment scales for detecting both genera and species (Table 1 SDN was always more efficient or equal in efficiency at detecting species within subfamilies or tribes at Site 1 (Table 3); this was also true for detection efficiency at genus-level except for Orthocladiinae, where SFPE was 6.7% more efficient at detecting genera.
For sites on second order stretches of stream, at the genus-level, no consistent trends in efficiencies of either method occurred except for Orthocladiinae, where the SFPE method consistently was more efficient than SDN collections (Table 4).At species-level, however, the SFPE consistently was more efficient for Orthocladiinae, Chironomini and Tanytarsini, but SDN was more efficient for Diamesinae.No consistent pattern was observed for Tanypodinae.
For sites on third order stretches of stream, SFPE consistently was more efficient at detecting both genera and species of Orthocladiinae, Chironomini and Tanytarsini (Table 5).SDN only consistently outperformed SFPE detecting genera and species of Diamesinae.
Table 6 lists the genera that were most species rich across all sample sites and the species detected by each method.The genera Eukiefferiella (9 species), Tanytarsus (9 species), Cricotopus (8 species) and Orthocladius (subgenus Orthocladius) (6 species) were the most species richness, and all species of these genera were detected as SFPE.SDN collections of larvae were less efficient at detecting species in these genera (88.9% to 16.7% detected).Fifty of the genera detected at BRNP were each represented by only a single species and 48 of the genera were detected as SFPE (96%) versus only 25 of the genera detected as larvae using SDN sampling (50%).

Discussion
Comprehensive surveys of Chironomidae biodiversity across stream orders or at catchment scales require substantial sampling effort in seasonal environments such as those that occur at latitudes similar to that of the Bear Run Nature Reserve, and are best achieved using multiple field methods targeting all developing life stages.The results of this Total taxa Taxa as larvae % of total Taxa as SFPE % of total Sample site totals, genera Tanypodinae   project compare favorably to a more intensive field approach used by Coffman (1973) to study composition and phenology in Linesville Creek (134 species versus 143 species).However, the difference in species detected in the two studies suggests that more intensive sampling effort would result in higher species richness at BRNR.
Our field design, which incorporated monthly sampling events, is important in terms of interpreting our results.A concern related to the SFPE method is that it only detects taxa that are emerging at the time of collection, and in highly seasonal environments there is expected to be high temporal variability in species emergence (see Coffman 1989, Coffman andde la Rosa 1998).Consequently, a single annual SDN sample would likely detect more taxa as larvae than as pupal exuviae.
However, this pattern was not seen by Ferrington et al. (1991) in two organically enriched urban streams in Kansas, where there was strong congruence between larvae detected with SDN samples and pupal exuviae from SFPE samples; on the other hand SFPE was more efficient at detecting species as water quality conditions improved and species richness increased across sites.These results need to be interpreted cautiously, however, because the water quality and habitat conditions of the two streams assessed by Ferrington et al. (1991) were much poorer than exists at BRNR, and the species rich communities at Sites 2-5 may have much greater temporal emergence heterogeneity than the more enrichment-tolerant species of the streams in Kansas.
Given that budgetary limitations influence sample designs, the disparity in efficiencies of the two methods at the catchment-scale suggest that an al-ternative design consisting of more field sampling events using SFPE could be a more cost-efficient approach to assessing biodiversity than always collecting and processing the more time-consuming and costly SDN samples.However, the patterns of relative efficiency across sample sites with differing stream order suggest that there could be an optimal allocation of field effort employing the two methods among streams of differing orders, size and discharge in a catchment.
The higher efficiency of SDN at Site 1 reflects better coverage of microhabitats in small streams versus larger streams.This also strongly suggests that an equal-effort approach using SDN across streams of differing size will result in decreasing efficiency detecting chironomid taxa in larger streams.By contrast, the low efficiency of SFPE at Site 1 was likely due to difficulty using a pan to dip among large boulders in areas with minimal water depth.In addition, the slower water velocity and reduction to trickle-like flow conditions during summer also decreases the natural accumulating effect on floating pupal exuviae of currents in larger, deeper and faster flowing streams.Use of a drift net or smaller aquarium-size net that would fit into tighter spaces between rocks may have resulted in higher efficiency of collection of SFPE at this site.
A drift net, however, would likely have needed to be in place for a 24-hour period or longer to filter a sufficient amount of stream flow, which would have added substantively to field efforts and sampling costs.
At BRNR, SFPE begins to outperform SDN in the transition from second order to third order sample sites characterized by well-developed pool/riffle conditions and widths exceeding about three meters.Our results for sample sites on second order stretches suggest that factors other than just size, flow and depth influence the efficiency of SFPE.These factors could include the size and space between larger boulders and the amount of woody debris extending into the water and creating conditions for back-flow and accumulation of floating detritus and pupal exuviae.
In addition to field sampling design, stream order and habitat conditions, the ability to resolve species is another variable that needs to be considered when comparing the efficiencies of the two methods.All genera detected in this study are welldefined and recognizable in both larval and pupal stages.Consequently, it can be argued that the differences in efficiencies at catchment-scale detection of genera are related to sampling error because of reduced efficiencies of SDN related to stream size, or difficulties sampling deeper pools, or specialized microhabitats of some larvae.However, at species-levels both sampling error and inability to resolve species-level difference among larvae are both sources of variability that decrease efficiency of SDN.The differential ability to resolve species is especially important in species-rich genera.Examples in this study include the genera Tanytarsus, Cricotopus, Orthocladius, Corynoneura, Micropsectra, Parakiefferiella and Microtendipes, where species level identifications of pupal exuviae are more readily achieved compared to identification of larvae.
The SDN sampling in this project was supplemented with targeted hand picking of marginal accumulations of wetted leaves, submerged and partially submerged wood of varying stages of decomposition, and larger boulders.Consequently, we detected larvae of some semi-aquatic species and xylophagic taxa that otherwise might have been undetected with SDN samples from coarser gravel/cobble substrates in riffles.Larvae from the targeted hand-picking efforts were included in the totals for SDN, likely increasing the efficiencies at the higher stream order sites.Despite this targeted effort, SFPE strongly out-performed our efforts aimed at detecting genera and species using larvae and whole pupae.SFPE also was slightly more effective at detecting phoretic taxa compared to the SDN technique.

Table 3 .
Number of taxa (genera, species)detected by subfamily or tribe and the relative detection percentages for larval collections versus collections of SFPE at Site 1, the first order stretch of stream.

Table 4 :
Number of taxa (genera, species)detected by subfamily or tribe and the relative detection percentages for collections of larvae versus SFPE at Sites 2 and 5, the second order stretches of stream.

Table 5 :
Number of taxa (genera, species)detected by subfamily or tribe and the relative detection percentages for collections of larvae versus SFPE at Sites 3 and 4, the third order stretches of stream.

Table 6 .
Number of species detected by genus and the relative detection percentages for larval collections versus collections of SFPE.