A POLYETHYLENETHEREPHTHALATE (PET)... 531 A POLYETHYLENETHEREPHTHALATE (PET) DEVICE FOR SAMPLING FRESHWATER

A new device to sample freshwater benthic macroinvertebrates was used in a low and sandy stretch of a Brazilian sub-tropical river (the River Cai, Triunfo, RS) and in one of its small tributaries, Bom Jardim brook (Arroio Bom Jardim). In this study, the effectiveness of this device, a PET sampler, was tested at different sites in the river and the brook throughout the four seasons between 2001-2002. Comparisons were made by PCA and ANOVA, both employing a bootstrap procedure based on similarity matrices. The PET sampler proved to be a reliable tool for detection of seasonal and spatial differences in richness, total abundance of organisms, and Shannon's diversity index in both river and brook and is therefore recommended for use in the monitoring of macroinvertebrate communities in this system.


INTRODUCTION
Water quality monitoring by means of biological indicators has been widely used to verify tendencies towards temporal or spatial alterations (Cairns et al., 1993).The parameters normally used are the abundance of organisms as well as indexes of richness and diversity of the species obtained from surveys of macroinvertebrate fauna (Magurran, 1988;Pontasch et al., 1989;Zmarzly et al., 1994;Arocena, 1996;Veijola et al., 1996), although there is no consensus about which sampling methodologies and indicators are the most suitable for use in different situations (Magurran, 1988;Cairns et al., 1993;Chessman, 1995).In order to overcome surveying problems, particularly with regard to standardization and replication of methodologies, an array of artificial substrates to be colonized by benthic organisms has been proposed and used, particularly in the United States (Rosemberg & Resh, 1982).
Water quality monitoring using macroinvertebrate benthic fauna began in 1997 in the lower course of the River Caí and the Bom Jardim brook, in an area influenced by the Pólo Petroquímico do Sul (the Southern Petrochemical Complex), municipality of Triunfo, RS.The roots of the water jacinth Eichhornia azurea (Sw).Kunth, found along the river banks, had been employed as natural samplers of that fauna (Volkmer-Ribeiro et al., 1984), since dredging techniques had proved inefficient in that area, due to the irregular/sandy river bed.The increasing rarefaction of this macrophyte, noted in 2001-2002 and attributed to intensive commercial extraction of sand in the lower course of the river (FEPAM, 2001), showed a need to design and test an artificial substrate to serve as a basis for a standard method of sampling to be used henceforth both in the river and the brook.Artificial substrates are known to be selective, depending on the materials of which they are made and/or the sites in which they are placed (Rosemberg & Resh, 1993).Hard artificial substrates are usually colonized by sessile or small moving organisms which ultimately results in the exclusion of mud-burrowing or mud-feeding ones.As the area to be surveyed was mostly composed of soft substrate, particular consideration was given to this fact in designing the device.This study describes its construction and reports testing results.

PET sampler assembling
Each sampler consists of two 2 l PET-type, disposable green bottles, sanded inside and out, with the bottoms removed and eliminated.The caps are retained on the bottle necks (Fig. 1A).The walls are then cut into 6 strips, starting from a circular line around the neck to the base.Each strip, measuring about 2 cm in width, is then cut transversally every 2 mm, resulting in a comb-like appearance (Fig. 1B).One bottle is then placed inside the other, with necks opposite, and both are wrapped in a small nylon net that has the opening fixed under the cap of one of the bottles (Fig. 1C and D).Prior to this procedure, a natural filling, consisting of a cellulose network that houses the seeds of the plant Luffa cylindra L. (Fig. 1E), is placed within the recipient formed by the juxtaposition of the two bottles (Fig. 2).Each filling is cut to the appropriate size for fitting inside the recipient.A metal weight of approximately 300 g is fixed to one neck (Fig. 1D).A nylon thread, about 3 m in length, is firmly attached at the other end.The sampler is made of green bottles as, once submerged, they are less conspicuous than transparent bottles, and less prone to the human depredation observed in the area.

Placing and recovering the PET sampler: the procedures
Samplers were placed in two areas (Fig. 3 (Volkmer-Ribeiro et al., 1984) had found that these were, respectively, the seasons of greater and lesser abundance and richness.To check this information, samples were also taken in summer and autumn at one site in the brook and another in the river.Samplers were placed on the bank, holding one end of the nylon thread and throwing the sampler, then tying the thread to a trunk or a branch of riverside vegetation at water level.Care had to be taken to ensure that the sampler sink, resting horizontally on the river bed, at a depth not overly exposed during the dry season.Occupation of the sampler by sediment and fauna took place over the duration of its stay on the river bottom.
The time estimated for colonization of the artificial substrate was two months.Recovery was done by a researcher positioned in the water by the bank who, after releasing the nylon thread from the substrate to which it was tied, used it to locate the sampler without displacing it.Once located, the device was maneuvered, without raising it, into a plastic bag placed on the river bottom.As soon as the sampler was bagged, it was sealed and lifted from the water.
Out of the river the bag was placed inside a plastic bucket, opened in order to fix the material with 5% formalin, closed again, and transported in the same bucket to the laboratory.The reason for placing the plastic bag in a bucket was to preserve the individual samples, in case the plastic bag leaked, after removal from the river.The contents of each bag, with its sampler, were inverted into a sieve lined with cotton and washed under running water, following which the bottles, as well as the nets and the vegetable fillings, were left to dry for subsequent examination of the attached fauna.

Data analysis
The temporal and spatial patterns in faunal composition and abundance found in the samplers were analyzed using Principal Component Analysis.A sampling unit was considered to be a group of five PET samplers recovered from each point, the data from which were totaled.
We calculated the classically-used parameters for the description of macroinvertebrate communities: abundance of organisms (total number of organisms, independent of taxonomic identity), species richness, and Shannon's Diversity index.The abundance data were previously transformed, using the logarithm (log(|x + 1|)), in order to reduce the effect of the large number of organisms of some species.Differences between the sites were tested using Analysis of Variance, via randomization testing, calculating the sum of the squares (Qb) from a Euclidean distance similarity matrix.This procedure is recommended by Pillar & Orlóci (1996) as an alternative to the data normality requirements of traditional ANOVA.Block delineation, corresponding to the seasons of the year, and two factors (season of the year and areas) were adopted, in order to eliminate variation between the sites of each area from the analysis.The PET samplers obtained on each sampling occasion were taken as replicates.The relationship between the community's descriptive parameters was measured using Correlation Coefficient, with the significance assessed using randomization testing according to Pillar & Orlóci (1996).
The analyses were performed with a MULTIV version 2.1.1 (Pillar, 2000).

RESULTS
The samplers captured a total of 68 macroinvertebrate species, with 36 species common to both river and brook, 10 species found only in the brook, and 22 species found only in the river.The PET sampler also captured one species of small fish occurring in both river and brook.The sampling covered the most representative macroinvertebrate groups, from molluscs to crustaceans and insect larvae, plus sessile groups such as sponges and bryozoa, thereby proving the device successful in enclosure/colonization of mud/sand as well as in affording a hard substrate for fixation of sessile animals such as bryozoan and sponges.Gastropod, crustacean, and insect larvae were the groups with the highest richness, both in river and brook.Turbellaria, the gastropod Heleobia piscium, Oligochaeta, the crustaceans Cyclopidae sp. and Podocopida, Lepidoptera larvae, and Bryozoa were the invertebrates with the highest abundance in the river while, for the brook, the taxa were Heleobia piscium, Oligochaeta, Collembola sp.1, Lepidoptera sp.1, and Elmidae sp.1 (Tables 1 and 2).The richness, the diversity of species, and the abundance of organisms tended to be greater in spring and summer and in the River Caí location (Table 3).The three parameters were correlated, showing that the same patterns can be detected with any one of them (Table 4).Significant differences of richness, abundance, and diversity were detected between the seasons, but not between the two sampled areas (River Caí and Bom Jardim brook) (Table 5).A clear pattern of progressive reduction in the three community descriptors from spring to winter was detected, with significant differences found between the hotter and colder seasons (Table 6).
The ordination of the samples obtained with the PET samplers (Fig. 4) showed a greater variation, particularly apparent in the first axis, in composition and abundance between the seasons of the year.Spring and summer have similar compositions, different from those found in autumn and winter, which also tend to form a group.Limnoperna fortunei, Gastropoda sp., Heleobia piscium, Nematoda sp., Gundlachia concentrica, Hirudinea sp., Cyclopidae sp., Podocopida sp.1, Chlamidotheca sp., Darwinulidae sp., Noteridae sp.1, Noteridae sp.2, and Ceratopogonidae were the taxons that most influenced these patterns.There was great variation until the 13 th axis, although the first two explain 57.6% of total variance, suggesting the absence of a strong association between the species themselves or an area or a particular season of the year.

DISCUSSION
The characterization of macroinvetebrate communities is highly influenced by decisions about sampling design (Beisel et al., 1998;Larsen & Herlihy, 1998), parameters calculated (Solimini et al., 2000;Thompson & Townsend, 2000;Brown, 2001) and sampling devices employed (Bartsch et al., 1998).Several studies have pointed out that all methods of macroinvertebrate collection are selective (Muzaffar & Colbo, 2002;Humphries et al., 1997).Choice of a device should be based on the specific objectives of a particular monitoring program and pilot studies measuring the efficiency of the selected device in pattern detection (Carter & Resh, 2001).The main aim in monitoring the structure of the macroinvertebrate community of the River Caí and the Bom Jardim brook, in the area influenced by the Southern Petrochemical Complex in Triunfo, RS, is the detection of fluctuations between years and sites.The detection of any reduction in abundance of organisms, species richness, or diversity is particularly important as it may indicate alterations occurring in environmental conditions (Magurran, 1988;Pontasch et al., 1989;Cairns et al., 1993).The PET sampler proved, first of all, capable of yielding a broad spectrum of macroinvertebrate benthic fauna and, second, to be sensitive in the detection of patterns and identification of differences between the seasons of the year and the two surveyed areas.
The use of artificial samplers is still open to debate.Some studies have found them efficient (Benoit et al., 1998) but others have demonstrated that direct sampling captures a more diversified fauna (Casey & Kendall, 1997).The correlation between species richness, total abundance of organisms, and Shannon's index obtained with the PET sampler is an advantage not offered by every method.This property makes the new device suitable for use in different strategies of data analysis in river systems under study.The faunal composition sampled in the present case is comparable to that found in Volkmer- Ribeiro et al. (l984) in the same river and to that of the survey performed by Sampóns (1988) for similar antropically-modified lotic environments in the area around Buenos Aires.The patterns also compare to those found in the literature (Hynes, 1970;Whitton, 1975;Petts & Calow, 1996).The clear seasonal pattern found with the new sampler is characteristic of macroinvertebrate communities sampled using different methods and in different locations (Gratwicke, 1998), including in Brazilian systems (Melo & Froehlich, 2001).The fact that the River Caí system had previously been shown to present the maximum count for most taxons in spring (Volkmer-Ribeiro et al., 1984) also reinforces the utility of the new sampler.
No clear difference in richness, abundance, and Shannon's index was found between the two areas sampled because the variation between the seasons of the year is of much greater importance.On the other hand, the variation in composition between these two areas was detected by the multivariate procedure, which reflects the differences in habitat preferences (Baptista et al., 2001).
The introduction of the artificial sampler along the length of the River Caí under examination allowed, for the first time, a viable statistical comparison between the different sampling sites and periods of the year.This sampler also creates the possibility of standardization of sampling, which is advantageous for monitoring purposes, over long stretches of river systems where natural habitats vary (Rosemberg & Resh, 1982).The PET samplers represent a reliable, low-cost alternative for monitoring the macrobenthos of the studied river system.The technique is particularly promising in situations where the river characteristics inhibit the use of other techniques.
the River Caí and the Bom Jardim brook obtained in the artificial samplers, in the period 2001-2002.N = total number of organisms, S = species richness, H' = Shannon's diversity index.

Fig. 4 -
Fig. 4 -Dispersion diagram of the first two PCA axes of the macrobenthic samples from the River Caí (rc) and the Bom Jardim stream (bj) with PET samplers in the period 2001-2002.a = autumn, w = winter, s = spring, su = summer.