Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Benthic macroinvertebrate assemblages in north Swedish streams: environmental relationships

Benthic macroinvertebrate assemblages in north Swedish streams: environmental relationships Malmqyist, B and Maki, M 1994 Benthic macroinvertebrate assemblages in north Swedish streams environmental relationships - Ecography 17 9-16 An analysis of the relationships between lotic macroinvertebrates and environmental vanables was earned out on matenal from 60 nffle sites in streams in northern Sweden The approach involved the use of TWINSPAN classification and canonical correspondence analysis on presence/absence data from two seasons (spnng and autumn) Variables most strongly associated with distribution patterns of assemblages were drainage area, elevation, water quality m terms of alkalinity, colour and phosphate and the presence of macrophytes The significance of affinities of different species to these vanables are discussed The eight clusters resulting from the TWINSPAN analysis could biologically be interpreted as classes of taxa related to stream size, chemical conditions and algae A multiple regression analysis for predicting species nchness using three independent vanables, viz drainage area amount or organic matter, and discharge was constructed The results of the study could be used as a starting point for further work on the community organization of benthic stream assemblages B Malmq\ist and M Maki, Dept of Ammal Ecology, Univ of Umea, S-901 87 Umea, Sweden Introduction Species typically show individual optima along environmental gradients resulting in local and regional differences in distnbutional patterns (e g Brown 1984) For streams, classification and ordination techniques based on species and environmental data have shown the existence of patterns that could be used to typify assemblages and to predict faunal composition at sites on the basis of known values of environmental vanables (e g Armitage et al, 1987, Wnght et al 1989) Especially valuable results are achieved when a limited subset of vanables is extracted which explains a large part of observed variation m species data Previous studies on lotic macroinvertebrate communities in Sweden have essentially been purely descnptive with only limited attempts to quantitatively relate observed patterns with environmental data (e g Ulfstrand 1968, Fnberg et al 1977), Alternatively they have foAccepted 17 November 1993 Copynght © ECOGRAPHY 1994 ISSN 0906-7590 Pnnted in Denmark - all nghts reserved ECOGRAPHY 17 1 (1994) cussed on single types of environmental disturbance, such as pH (Otto and Svensson 1983) and the effects of hydroelectnc regulation (Henncson and Sjoberg 1984), or were carried out in a geographically restncted area (eg Ulfstrand et al 1971) Consequently, there is a need for a broader characterization of the benthic fauna and its relationships with environmental factors in Sweden A general account of Nordic nvers was given by Petersen et al (m press) In the present study we emphasize three aspects 1) classification of sites based on a divisive technique, two-way indicator species analysis (TWINSPAN, Hill 1979), 2) construction of ordination diagrams on the basis of constrained ordination (Canonical correspondence analysis) This enables us to simultaneously examine relations between sites, species and environmental factors, and 3) explonng the possibility of establishing a predictive relationship between species nchness and a subset of environmental factors using multiple regression tween Kiruna and the Norwegian border, all in the Tornealven drainage area, >300 km north of the first set The selected sites are mostly in streams of the orders 1-5, and they do not include sites at elevations above 560 m a s 1 They are all moderately affected by human activities, pnmanly by ditching and logging Recognizably disturbed streams were avoided Field methods Field sampling was performed on two occasions autumn 1990 (18 August - 11 September), and spnng 1991 (26 Apnl - 22 June) Autumn samples were taken at the northern sites pnor to those from southern sites, whereas the reversed order was true for spnng samples This was done to minimize phenological effects on the presence of species A few northern streams were ice-covered until mid June, whtch explains the late sampling there Animal samphng was performed by kick samphng using a deep sweep net (mesh size 0 50 mm) Care was taken to include all possible microhabitats over a representative section of the stream (10-50 m) This means that slow-flowing sections were included when they were present along the margins of fast-flowing nffles The cumulated number of species at a particular site asymptotically approaches the true species number with increasing sampling effort (e g Stout and Vandermeere 1975) The methods for estimating species richness are, however, very sensitive to the magnitude of this effort (Baltanas 1992, Englund and Malmqvist unpubl data) Our efforts were moderate, and therefore merely reflect what will be encountenng dunng approximately one hour of kick sampling, including the sorting of the samples, I e the process of scanning samples placed in a white tray, separating macroinvertebrates from organic matenal until new species no longer are found, and preserving them in 70% ethanol This is a pragmatic solution, which probably is closer to standard inventory technique than to efforts directed at getting all species present, and resulted pnmanly in the inclusion of the most common taxa Since large streams usually contain more microhabitats and therefore demand more time for representative sampling, the time was proportionally longer at such sites compared with the samphng sites in smaller streams Since we were only interested in obtaining presenceabsence information about species our sampling did not attempt to establish any measure of abundance All analyses were earned out on the combined species list from autumn and spnng sampling. At each site the substrate composition, discharge and presence of vegetation were recorded Substrate was estimated as the approximate percentages of boulder (>2(X) mm), rock (50-200 mm), gravel and pebble (2-50 mm) and sand (<2 mm). Further, a representative transect across the stream provided mean values of depth and current speed as well as width and discharge A subECOGRAPHY 17 1 (1994) Fig 1 Map of northern Sweden with the investigation area and the 60 study sites Numbers refer to sites named in Fig 2 technique The results of the outlined kinds of analyses have a clear potential in environmental planning, since they provide representations of community types, insight into which factors most importantly influence the invertebrate communities, and allow the construction of models that could predict the outcome of environmental changes Although the value of choosing macroinvertebrates for this kind of developments could be hard to appreciate for several reasons (cf Boon 1988), they may represent the most favourable indicators of the ecological conditions m running waters, reflecting both past and present conditions (Ward 1976) Materials and methods Site selection Sixty riffles were selected for the study (Fig 1) Forty-five of these are situated along a geographical gradient from the coast near UmeS to the mountain area near the Norwegian border. Most of these sites are in the Umealven drainage area A second set of sites were sampled be10 f*ol\ceruropii\ flaMwacuhlHS hoptrta iirammatu a Seri( asloniu personalurn Amphmenuiru borealis Nemoura axictjlans Lepiophlehiu marginaia Nemoura axtctilaris Potamoph\lax lattpennis Fleophthi sp Nemoura ctnere, Lapnopsfi \(htlten Ameleius iruiptnaiits Dturu nun\eni Raeli Lentlru niMra Netiiotira < tnerea Flecirocnemia consperKQ A re sopsy che t\ Atiietetus innpmatus I euctra jw^ca SintHlititn heroiiitn Baetts siihaJpinui Heptagema dalecarlica NettUHira (inerea Die raruna sp I leophtlu ' 20 Kitabacken 52 Yttersjobacken 23 Jovatts&n inbutary 58 Kastbacken 48 Stratbacken Fig 2 The resulting eight groups classified by the TWINSPAN analysis At each level the indicator species' for each division are indicated jective classification of macrophytes, filamentous algae, aquatic mosses, and organic matenal was also ranked along a scale, with five values ranging from not present (1) to very abundant (5) The category "organic matenal" included all kinds of detritus (ranging from wood and leaves to fine particulate matenal) All of these factors were estimated m the autumn only Further information was taken from maps (1 50,000 and 1 100,000), mcl drainage area, stream order, and elevation The coordinates of each site in relation to the national gnd were also estimated Water samples were taken only in the autumn After sampling they were immediately placed in a cool box They were taken to the laboratory (Umea Municipality's Environmental Laboratory) for chemistry analyses on the day of sampling Samples from remote sites were deepfrozen for a maximum of four days prior to analysis Analysis methods Classification was performed using the TWINSPAN technique (Hill 1979) This is a polythetic, divisive methodology that operates in several steps where, at each step, the sites-by-species matrix is divided dichotomously In ECOGRAPHY 17 1 (1994) the analysis we have excluded I) species which were found at less than three sites, 2) two sites with extremely low species nchness, and 3) the largest nver in order to reduce noise in the analysis The analysis followed the default option across the console dialogue and was run for three division levels which resulted in eight TWINSPAN groups In the analysis of community data we applied canonical correspondence analysis, CCA, using the C A N O C O package (ter Braak 1987a, 1990) This method expresses main relations between species and environmental vanables by combining ordinafion and regression (ter Braak 1987b) To extract a reduced vanable set the "Forward selecfion" option was enforced, a procedure in which the vanable that contnbutes most to the vanance is selected in a step-wise procedure and where the significance of each addition of a new vanable is tested in a Monte Carlo permutation test Only taxa represented at a minimum of three sites were included in the CCA Some vanables showed little vanance (nitrate-N and substrate, the latter probably because only nffle habitats were selected) and were therefore excluded from the analysis Multiple linear regression was employed in an attempt to assign a subset of vanables that could be used to Table 1 Means and standard errors (SE) of vanables in each of the eight TWINSPAN groups Macrophytes, moss, filamentous algae, and organic matter are m a five-grade scale from 0 to 4 Vanable 1 Number of sites Stream order Drainage area (km-) Altitude (m a s 1 ) Width (m) Discharge (m' s"') Macrophytes Moss Filamentous algae Organic material 4 mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean 11 III 11 TWlNSPAN-group IV 5 V 7 VI 12 VII 8 VIII Conductivity (mS m"' 1 Alkalinity (mekv L"') ^11 ikj Colour Total P Phosphate P (ixg P L-') Total N ( a s P L^') Nitrate N (Hg P L-') Sulphate (mg L"') Chloride (mg L-') Calcium (mg L"') Magnesium (mg L"') Sodium (mg L'') Potassium (mg L"') North (National gnd) East (National gnd) Current velocity 3 40 0 60 506 370 9 376 44 134 33 2 45 0 78 0 40 0 40 100 044 2 60 0 40 1 80 0 50 6 96 0 07 2 42 0 08 30 6 37 28 08 0 63 30 5 7 1 27 5 0 272 14 Oil 26 5 0 177 0 08 0 72 0 15 2 97 0 07 0 68 Oil SE mean SE mean (m s-') Depth (cm) SE mean SE on 103 2 50 0 36 30 88 467 21 60 16 1 66 0 84 0 16 0 16 166 0 34 0 84 0 30 2 08 0 20 7 25 0 11 3 95 0 19 39 7 57 5 70 0 14 193 35 27 5 0 216 02 0 15 31 1 31 2 94 0 14 0 90 0 07 5 30 0 26 0 80 Oil 0 92 0 05 0 55 0 06 7450 42 3 1546 23 3 0,37 0 05 25 12 f\ 151 Oil 0 62 Oil 7230 31 5 1565 40 8 0 23 004 15 1 1 ECOGRAPHY 17 1 (1994) Table 2 Results of canonical correspondence analysis Species occumng at less than three sites were excluded from the analysis Forward inclusion of vanables was used and from the ten resulting, significant variables, those six with greatest weight were retained in the analysis area, colour, altitude, phosphate-P, macrophytes. and alkalinity Eigenvalues give the importance of an axis on a scale between 0 and 1 Total inertia is total vanance in species data The species-environment correlations scale the strength of the relation between species and environment for the axes The percentage of variance m species data gives the proponion of an axis" eigenvalue in relation to the sum of all unconstrained eigenvalues, whereas that of the species-environment relation gives an axis" contnbution to fitted species values denved from environmental vanables Axes Eigenvalues Species-environment correlations Cumulative percentage variance of species data of species-environment relation Sum of all canonical eigenvalues Total inertia 3 250 predict species richness at a site The choice of vanables was preceded by a visual inspection of the outcome of a principal component analysis of environmental data followed by testing combinations of those with highest loadings In all analyses, continuous environmental data were loganthm transformed, except for the pH data which already were present in a loganthmic form Results In total, 155 species were identified and used for the analyses, except m the CCA and TWINSPAN analysis where 92 species, occumng at >3 sites (l e 95% of all Lmgr Npict • PO 4-P Saur Halsp* P cons, ^ s vem Ncm " B macu, • , ' • ^ ^ • • Colour S full ^7 7 / Nav Cvil^ Catra ' / • ' Brisi ." K^Ph_Htac L vesp ^ _ "C si!f -• . Sbur ^ sulp L hirt • Bmg Hdale 1 pcit • • Stub sites), were included These numbers did not include chironomids, ohgochaetes, mites, and Pisidium spp, which were left out of the analyses, because they were not identified to the level of species For some genera we were not able to distinguish between similar species, e g in Dicranota, and in Halesus and some other limnephilid caddis larvae, although they were retained in the analysis Six taxa occurred at more than half the sites Rhvacophila nubila Zett 82%, Baetis rhodani Pict 80%, Diura nanseni Kmp 73%, Dicranota spp 70%, Nemoura cinerea Retz 60%, Taentopteryx nebulosa L 57% The high proportion of rare taxa (40%), found at 1-2 sites, m the matenal was only partly (25%) due to the occasional ingression of lentic species from slow-flowing parts of the systems studied The eight groups of sites provided by the TWINSPAN analysis contained two to twelve sites In the first division, sites were selected on the basis of the absence or presence of Polycentropus flavomaculatus Pict, lsoperla grammatica Poda, Sencostoma personatum K & Sp , Amphmemura borealts Mort and Nemoura avicularts Mort In the following divisions another 19 indicator species (sensu Hill 1979) were identified (Fig 2) Group I was charactenzed by high values of colour and phosphate-P, and low altitude Group II shared the low altitude charactenstic of Group I Group III sites were all inland sites in the southern site set Group IV sites had on average the largest drainage areas, whereas those in Alt^ B lapp Cpygm 1 obsc E aurr^ Group V had an above average cover of filamentous algae, and Group VI were at the highest altitude Group VII stream sites were on average much smaller than those in other groups The only two sites m Group VIII had strongly-coloured water, reflecting high humic content, and relatively low pH (6.50 at both sites) Further information on the mean values (back-transformed) of each vanable is given in Table 1 SIX environmental vanables were incorporated into the CCA model, where the first axis had largest explanatory power (Table 2, Fig 3) The vanables were (in order of magnitude) drainage area, water colour, altitude, phos- '-'"^"Alk A mop Area Fig 3 Biplot of species data and environmental vanables Alt = altitude, Alk = alkalinity. Area = drainage area, Mph = macrophytes Only species found at >8 sites are shown in figure ECOGRAPHY 17 1 (1994) Fig 4 TWINSPAN groups (I-VIII) in stream site space as denved from the CCA Note therefore that the sites correspond to species and environmental vanable arrangement in Fig 3 species), have optima in the centre, or have bimodal response curves (ter Braak 1987b) Although the study sites did not form distinct groups, combining the TWINSPAN site groups in CCA space indicated that the sites within each group were related (Fig 4) An overall test of the model showed that the clusters were significantly different (Canonical vanates analysis followed by a Monte Carlo permutation test, ter Braak 1987a,b, F = l 84, p = 0 01) The distinction was most clear between group VII and other groups, whereas among the latter there was always some overlap with neighbour groups Note that the distnbution of sites in Fig 4 IS related to the environmental variables in Fig 3 Thus, group VII sues have small drainage areas, group VI sites are at high altitude, etc The number of taxa, excluding chironomids, water mites, ohgochaetes, and Ptstdium spp , ranged between 3 and 43 (mean ± SD 22 3±9 1) per site Inspection of all environmental vanables in preliminary multiple regression analysis suggested a subset explaining a large part of the vanation Thus, a model based on the present matenal incorporating only drainage area, organic matenal, and discharge explained nearly half of the vanation in species richness (R = 0 46, Table 3) The relationship was lnS = 1 653 -(- 0 218 lnDA -i- 0 178 OM - 0 090 DI where S is species nchness, DA the drainage area in km-, OM the amount of organic matenal ranked by an arbitrary scale 1-5, and DI discharge m m' s"' phate-P, macrophytes and alkalinity In Fig 3 the species with greatest affinity to each of the variables can be identified For example, size related species were Nemurella ptctett Kip , Eeuctra nigra Ol , Plectrocnemta consperm for small size, and Arctopsyche ladogensts Kol and Simuhum tuberosum Lundstr (complex) for large size Similarly, Heptagenia fuscogrisea Retz and two Eeptophlebia species were associated with macrophytes, Baetis lappontcus Bgtss , Capnia pygmaea Zett and C atra Mort with high altitude, and several mayfiies (Ameletus inopinatus Etn , Ephemerella aunvilh Bgtss , Heptagenia dalecarhca Bgtss ) and the snail Eymnaea peregra Mull with high alkalinity, but low phosphate Species near the ongms of the axes are either unrelated to the ordination axes (some of the most widely distnbuted Discussion The now relatively established approach to classifying benthic nver communities by ordination techniques has been successful in showing consistent results (Hildrew Table 3 Regression statistics for a multiple linear regression model for species nchness in North Swedish stream nffles with three independent vanables drainage area (km-), organic matter and discharge (m' s-') R'adj Number of observations Analysis of vanance df Regression 3 Residual 56 Total 59 Intercept Ln discharge Organic matenal Ln drainage area 14 0 461 60 Sum of Squares 7013 7 341 14 354 Mean Square 2 338 0 131 SE 0214 0 048 0 054 0 042 F 17 833 Significance F <0 0001 Coefficients (B) 1653 -0 090 0 178 0218 t Statistic 7712 -2 048 3 278 5 209 p-value <0 0001 0 045 00018 <0 0001 Lower 95% 1 224 -0 177 0 069 0 134 Upper 95% 2 082 -0 002 0 287 0 302 ECOGRAPHY 17 1 (1994) and Townsend 1987, Ormerod and Edwards 1987, Noms and Georges 1992) In Bntain, where the technique was employed early, factors foremost in explaining benthic community structure were substrate, pH-related factors (alkalinity, hardness), and components relating to nver size (distance from source, discharge, summer temperature, slope, stream order) The present study suggests that factors like size of drainage area, alkalinity and water colour are also important in northern Sweden In addition, altitude significantly contributed to the vanation, although this to some extent also is related to nver size in that large nvers are missing at high altitudes, although small ones are represented from all elevations Altitude also reflects temperature The presence of macrophytes and concentration of phosphate phosphorus further strengthened the pattern River ecology has few general theones The River Continuum Concept (Vannote et al 1980) is one of those, and It has caused extensive debate since its generation One of the most attractive ideas lies in its coupling between the trophic structure of animal communities and vanation in energy inputs (Hildrew and Townsend 1987) This notion is supported by the present study, where animal communities did not show distinct groupings but rather a continuous pattern where nver-size-related factors were of pnme importance Most strikingly, the assemblages found in very small streams deviated from the larger ones, suggesting that these were functionally different Such a contention is, for example, upheld by the strong connection of small streams to the npanan vegetation, through shading and the high litter input per unit area (Cummins 1975) The strong relationship found between drainage area and species nchness was interesting Drainage area was well correlated with factors such as discharge (r = 0 79), stream order (r = 0 89), current velocity (r = 0 54), width (r = 0 83), and depth (r = 0 54) Species nchness is known to increase with such size related measures (for freshwater invertebrates see e g Strayer 1983, Bronmark et al 1984, Bronmark 1985, Malmqvist etal 1992) However, the explanations for the relationship are not clear Hart and Horwitz (1991) reviewing alternative models claimed that most studies in fact were not designed such that the underlying causes could be revealed The same holds true for this study, although one factor, the extinction rate, may be of particular interest in dealing with drainage areas, the best of the size-related factors In addition to the fact that large drainage areas certainly are likely to contain more habitats than small ones, extinction rates should be lower within a large drainage area This could be expected since vanous parts of a drainage area are linked together Survival in and dispersal from a higher number of tnbutanes acting as refuges in large systems would then be possible The fact that detntal matenal is significant in the energy budget of streams in forested regions and thereby positively influences species nchness is hardly surpnsing In contrast, it is less obvious why discharge has a negative relationship with nchness ECOGRAPHY 17 1 (1994) when drainage area is accounted for It may imply that run-off or precipitation regime is of importance The value of individual species as indicators of disturbance could be useful as a shortcut to classifying affected sites However, species evolved in natural habitats, not in habitats disturbed by man Therefore, the presence of particular species does not necessarily indicate disturbance For example, species which typically are associated with pollution do occur naturally in habitats where organic debns is deposited in slow-flowing stream sections and backwaters, one good example is Asellus aquaticus L Neither could the lack of a species be used as evidence for disturbance, because this could not only be the consequence of a true absence at a site, but also be due to failure of sampling the species Despite these problems there remains a lot of information of a site based on its species composition This information value should be particularly strong when the entire community IS considered rather than separate species Biotic interactions were not considered in the present study Such interactions are likely to be important locally, especially where abiotic disturbance is low or moderate (Peckarsky 1983), although, their importance for shaping the species structure is probably limited over entire riffles or streams Few studies, if any, have conclusively demonstrated the complete removal of taxa by predators or competitors m stream benthos, whereas they cenainly may affect abundances (eg McAuliffe 1984, Reice 1991) Abiotic factors, including the ones studied here, seem to be the more lnfiuential One abiotic factor not directly measured in this study, although It probably had a major effect on the presence of different taxa, was temperature (cf Sweeney and Vannote 1978, Vannote and Sweeney 1980) It is probable, however, that temperature was related to some of the variables considered in this study, e g elevation and measures of nver size. Although only two transects across north Sweden may seem to poorly represent lotic nffie habitats over the entire area, we believe that in this study the mosi common types present have been identified It must be remembered, however, that larger nvers were omitted as were high altitude habitats In addition, highly specialised habitats such as springs and those combining summer drought and winter bottom ice (Johansson and Nilsson in press) were not included Our conclusion is that assemblages of macroinvertebrates m stream riffles in northern Sweden show gradually changed structure with environmental factors, the most important of which are the S12X of drainage area, elevation, water quality in terms of alkalinity, water colour and phosphate Species richness was also most strongly related to size of drainage area, and to a lesser degree, to organic matter, and discharge. Acknowledgements - A N Nilsson and C Otto offered valuable cnncism on an earlier version of this manuscnpt Financial support was received from the Swedish Environmental Protection Agency 15 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Ecography Wiley

Benthic macroinvertebrate assemblages in north Swedish streams: environmental relationships

Ecography , Volume 17 (1) – Mar 1, 1994

Loading next page...
 
/lp/wiley/benthic-macroinvertebrate-assemblages-in-north-swedish-streams-0JOmj4hUtM

References (31)

Publisher
Wiley
Copyright
Copyright © 1994 Wiley Subscription Services, Inc., A Wiley Company
ISSN
0906-7590
eISSN
1600-0587
DOI
10.1111/j.1600-0587.1994.tb00072.x
Publisher site
See Article on Publisher Site

Abstract

Malmqyist, B and Maki, M 1994 Benthic macroinvertebrate assemblages in north Swedish streams environmental relationships - Ecography 17 9-16 An analysis of the relationships between lotic macroinvertebrates and environmental vanables was earned out on matenal from 60 nffle sites in streams in northern Sweden The approach involved the use of TWINSPAN classification and canonical correspondence analysis on presence/absence data from two seasons (spnng and autumn) Variables most strongly associated with distribution patterns of assemblages were drainage area, elevation, water quality m terms of alkalinity, colour and phosphate and the presence of macrophytes The significance of affinities of different species to these vanables are discussed The eight clusters resulting from the TWINSPAN analysis could biologically be interpreted as classes of taxa related to stream size, chemical conditions and algae A multiple regression analysis for predicting species nchness using three independent vanables, viz drainage area amount or organic matter, and discharge was constructed The results of the study could be used as a starting point for further work on the community organization of benthic stream assemblages B Malmq\ist and M Maki, Dept of Ammal Ecology, Univ of Umea, S-901 87 Umea, Sweden Introduction Species typically show individual optima along environmental gradients resulting in local and regional differences in distnbutional patterns (e g Brown 1984) For streams, classification and ordination techniques based on species and environmental data have shown the existence of patterns that could be used to typify assemblages and to predict faunal composition at sites on the basis of known values of environmental vanables (e g Armitage et al, 1987, Wnght et al 1989) Especially valuable results are achieved when a limited subset of vanables is extracted which explains a large part of observed variation m species data Previous studies on lotic macroinvertebrate communities in Sweden have essentially been purely descnptive with only limited attempts to quantitatively relate observed patterns with environmental data (e g Ulfstrand 1968, Fnberg et al 1977), Alternatively they have foAccepted 17 November 1993 Copynght © ECOGRAPHY 1994 ISSN 0906-7590 Pnnted in Denmark - all nghts reserved ECOGRAPHY 17 1 (1994) cussed on single types of environmental disturbance, such as pH (Otto and Svensson 1983) and the effects of hydroelectnc regulation (Henncson and Sjoberg 1984), or were carried out in a geographically restncted area (eg Ulfstrand et al 1971) Consequently, there is a need for a broader characterization of the benthic fauna and its relationships with environmental factors in Sweden A general account of Nordic nvers was given by Petersen et al (m press) In the present study we emphasize three aspects 1) classification of sites based on a divisive technique, two-way indicator species analysis (TWINSPAN, Hill 1979), 2) construction of ordination diagrams on the basis of constrained ordination (Canonical correspondence analysis) This enables us to simultaneously examine relations between sites, species and environmental factors, and 3) explonng the possibility of establishing a predictive relationship between species nchness and a subset of environmental factors using multiple regression tween Kiruna and the Norwegian border, all in the Tornealven drainage area, >300 km north of the first set The selected sites are mostly in streams of the orders 1-5, and they do not include sites at elevations above 560 m a s 1 They are all moderately affected by human activities, pnmanly by ditching and logging Recognizably disturbed streams were avoided Field methods Field sampling was performed on two occasions autumn 1990 (18 August - 11 September), and spnng 1991 (26 Apnl - 22 June) Autumn samples were taken at the northern sites pnor to those from southern sites, whereas the reversed order was true for spnng samples This was done to minimize phenological effects on the presence of species A few northern streams were ice-covered until mid June, whtch explains the late sampling there Animal samphng was performed by kick samphng using a deep sweep net (mesh size 0 50 mm) Care was taken to include all possible microhabitats over a representative section of the stream (10-50 m) This means that slow-flowing sections were included when they were present along the margins of fast-flowing nffles The cumulated number of species at a particular site asymptotically approaches the true species number with increasing sampling effort (e g Stout and Vandermeere 1975) The methods for estimating species richness are, however, very sensitive to the magnitude of this effort (Baltanas 1992, Englund and Malmqvist unpubl data) Our efforts were moderate, and therefore merely reflect what will be encountenng dunng approximately one hour of kick sampling, including the sorting of the samples, I e the process of scanning samples placed in a white tray, separating macroinvertebrates from organic matenal until new species no longer are found, and preserving them in 70% ethanol This is a pragmatic solution, which probably is closer to standard inventory technique than to efforts directed at getting all species present, and resulted pnmanly in the inclusion of the most common taxa Since large streams usually contain more microhabitats and therefore demand more time for representative sampling, the time was proportionally longer at such sites compared with the samphng sites in smaller streams Since we were only interested in obtaining presenceabsence information about species our sampling did not attempt to establish any measure of abundance All analyses were earned out on the combined species list from autumn and spnng sampling. At each site the substrate composition, discharge and presence of vegetation were recorded Substrate was estimated as the approximate percentages of boulder (>2(X) mm), rock (50-200 mm), gravel and pebble (2-50 mm) and sand (<2 mm). Further, a representative transect across the stream provided mean values of depth and current speed as well as width and discharge A subECOGRAPHY 17 1 (1994) Fig 1 Map of northern Sweden with the investigation area and the 60 study sites Numbers refer to sites named in Fig 2 technique The results of the outlined kinds of analyses have a clear potential in environmental planning, since they provide representations of community types, insight into which factors most importantly influence the invertebrate communities, and allow the construction of models that could predict the outcome of environmental changes Although the value of choosing macroinvertebrates for this kind of developments could be hard to appreciate for several reasons (cf Boon 1988), they may represent the most favourable indicators of the ecological conditions m running waters, reflecting both past and present conditions (Ward 1976) Materials and methods Site selection Sixty riffles were selected for the study (Fig 1) Forty-five of these are situated along a geographical gradient from the coast near UmeS to the mountain area near the Norwegian border. Most of these sites are in the Umealven drainage area A second set of sites were sampled be10 f*ol\ceruropii\ flaMwacuhlHS hoptrta iirammatu a Seri( asloniu personalurn Amphmenuiru borealis Nemoura axictjlans Lepiophlehiu marginaia Nemoura axtctilaris Potamoph\lax lattpennis Fleophthi sp Nemoura ctnere, Lapnopsfi \(htlten Ameleius iruiptnaiits Dturu nun\eni Raeli Lentlru niMra Netiiotira < tnerea Flecirocnemia consperKQ A re sopsy che t\ Atiietetus innpmatus I euctra jw^ca SintHlititn heroiiitn Baetts siihaJpinui Heptagema dalecarlica NettUHira (inerea Die raruna sp I leophtlu ' 20 Kitabacken 52 Yttersjobacken 23 Jovatts&n inbutary 58 Kastbacken 48 Stratbacken Fig 2 The resulting eight groups classified by the TWINSPAN analysis At each level the indicator species' for each division are indicated jective classification of macrophytes, filamentous algae, aquatic mosses, and organic matenal was also ranked along a scale, with five values ranging from not present (1) to very abundant (5) The category "organic matenal" included all kinds of detritus (ranging from wood and leaves to fine particulate matenal) All of these factors were estimated m the autumn only Further information was taken from maps (1 50,000 and 1 100,000), mcl drainage area, stream order, and elevation The coordinates of each site in relation to the national gnd were also estimated Water samples were taken only in the autumn After sampling they were immediately placed in a cool box They were taken to the laboratory (Umea Municipality's Environmental Laboratory) for chemistry analyses on the day of sampling Samples from remote sites were deepfrozen for a maximum of four days prior to analysis Analysis methods Classification was performed using the TWINSPAN technique (Hill 1979) This is a polythetic, divisive methodology that operates in several steps where, at each step, the sites-by-species matrix is divided dichotomously In ECOGRAPHY 17 1 (1994) the analysis we have excluded I) species which were found at less than three sites, 2) two sites with extremely low species nchness, and 3) the largest nver in order to reduce noise in the analysis The analysis followed the default option across the console dialogue and was run for three division levels which resulted in eight TWINSPAN groups In the analysis of community data we applied canonical correspondence analysis, CCA, using the C A N O C O package (ter Braak 1987a, 1990) This method expresses main relations between species and environmental vanables by combining ordinafion and regression (ter Braak 1987b) To extract a reduced vanable set the "Forward selecfion" option was enforced, a procedure in which the vanable that contnbutes most to the vanance is selected in a step-wise procedure and where the significance of each addition of a new vanable is tested in a Monte Carlo permutation test Only taxa represented at a minimum of three sites were included in the CCA Some vanables showed little vanance (nitrate-N and substrate, the latter probably because only nffle habitats were selected) and were therefore excluded from the analysis Multiple linear regression was employed in an attempt to assign a subset of vanables that could be used to Table 1 Means and standard errors (SE) of vanables in each of the eight TWINSPAN groups Macrophytes, moss, filamentous algae, and organic matter are m a five-grade scale from 0 to 4 Vanable 1 Number of sites Stream order Drainage area (km-) Altitude (m a s 1 ) Width (m) Discharge (m' s"') Macrophytes Moss Filamentous algae Organic material 4 mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean SE mean 11 III 11 TWlNSPAN-group IV 5 V 7 VI 12 VII 8 VIII Conductivity (mS m"' 1 Alkalinity (mekv L"') ^11 ikj Colour Total P Phosphate P (ixg P L-') Total N ( a s P L^') Nitrate N (Hg P L-') Sulphate (mg L"') Chloride (mg L-') Calcium (mg L"') Magnesium (mg L"') Sodium (mg L'') Potassium (mg L"') North (National gnd) East (National gnd) Current velocity 3 40 0 60 506 370 9 376 44 134 33 2 45 0 78 0 40 0 40 100 044 2 60 0 40 1 80 0 50 6 96 0 07 2 42 0 08 30 6 37 28 08 0 63 30 5 7 1 27 5 0 272 14 Oil 26 5 0 177 0 08 0 72 0 15 2 97 0 07 0 68 Oil SE mean SE mean (m s-') Depth (cm) SE mean SE on 103 2 50 0 36 30 88 467 21 60 16 1 66 0 84 0 16 0 16 166 0 34 0 84 0 30 2 08 0 20 7 25 0 11 3 95 0 19 39 7 57 5 70 0 14 193 35 27 5 0 216 02 0 15 31 1 31 2 94 0 14 0 90 0 07 5 30 0 26 0 80 Oil 0 92 0 05 0 55 0 06 7450 42 3 1546 23 3 0,37 0 05 25 12 f\ 151 Oil 0 62 Oil 7230 31 5 1565 40 8 0 23 004 15 1 1 ECOGRAPHY 17 1 (1994) Table 2 Results of canonical correspondence analysis Species occumng at less than three sites were excluded from the analysis Forward inclusion of vanables was used and from the ten resulting, significant variables, those six with greatest weight were retained in the analysis area, colour, altitude, phosphate-P, macrophytes. and alkalinity Eigenvalues give the importance of an axis on a scale between 0 and 1 Total inertia is total vanance in species data The species-environment correlations scale the strength of the relation between species and environment for the axes The percentage of variance m species data gives the proponion of an axis" eigenvalue in relation to the sum of all unconstrained eigenvalues, whereas that of the species-environment relation gives an axis" contnbution to fitted species values denved from environmental vanables Axes Eigenvalues Species-environment correlations Cumulative percentage variance of species data of species-environment relation Sum of all canonical eigenvalues Total inertia 3 250 predict species richness at a site The choice of vanables was preceded by a visual inspection of the outcome of a principal component analysis of environmental data followed by testing combinations of those with highest loadings In all analyses, continuous environmental data were loganthm transformed, except for the pH data which already were present in a loganthmic form Results In total, 155 species were identified and used for the analyses, except m the CCA and TWINSPAN analysis where 92 species, occumng at >3 sites (l e 95% of all Lmgr Npict • PO 4-P Saur Halsp* P cons, ^ s vem Ncm " B macu, • , ' • ^ ^ • • Colour S full ^7 7 / Nav Cvil^ Catra ' / • ' Brisi ." K^Ph_Htac L vesp ^ _ "C si!f -• . Sbur ^ sulp L hirt • Bmg Hdale 1 pcit • • Stub sites), were included These numbers did not include chironomids, ohgochaetes, mites, and Pisidium spp, which were left out of the analyses, because they were not identified to the level of species For some genera we were not able to distinguish between similar species, e g in Dicranota, and in Halesus and some other limnephilid caddis larvae, although they were retained in the analysis Six taxa occurred at more than half the sites Rhvacophila nubila Zett 82%, Baetis rhodani Pict 80%, Diura nanseni Kmp 73%, Dicranota spp 70%, Nemoura cinerea Retz 60%, Taentopteryx nebulosa L 57% The high proportion of rare taxa (40%), found at 1-2 sites, m the matenal was only partly (25%) due to the occasional ingression of lentic species from slow-flowing parts of the systems studied The eight groups of sites provided by the TWINSPAN analysis contained two to twelve sites In the first division, sites were selected on the basis of the absence or presence of Polycentropus flavomaculatus Pict, lsoperla grammatica Poda, Sencostoma personatum K & Sp , Amphmemura borealts Mort and Nemoura avicularts Mort In the following divisions another 19 indicator species (sensu Hill 1979) were identified (Fig 2) Group I was charactenzed by high values of colour and phosphate-P, and low altitude Group II shared the low altitude charactenstic of Group I Group III sites were all inland sites in the southern site set Group IV sites had on average the largest drainage areas, whereas those in Alt^ B lapp Cpygm 1 obsc E aurr^ Group V had an above average cover of filamentous algae, and Group VI were at the highest altitude Group VII stream sites were on average much smaller than those in other groups The only two sites m Group VIII had strongly-coloured water, reflecting high humic content, and relatively low pH (6.50 at both sites) Further information on the mean values (back-transformed) of each vanable is given in Table 1 SIX environmental vanables were incorporated into the CCA model, where the first axis had largest explanatory power (Table 2, Fig 3) The vanables were (in order of magnitude) drainage area, water colour, altitude, phos- '-'"^"Alk A mop Area Fig 3 Biplot of species data and environmental vanables Alt = altitude, Alk = alkalinity. Area = drainage area, Mph = macrophytes Only species found at >8 sites are shown in figure ECOGRAPHY 17 1 (1994) Fig 4 TWINSPAN groups (I-VIII) in stream site space as denved from the CCA Note therefore that the sites correspond to species and environmental vanable arrangement in Fig 3 species), have optima in the centre, or have bimodal response curves (ter Braak 1987b) Although the study sites did not form distinct groups, combining the TWINSPAN site groups in CCA space indicated that the sites within each group were related (Fig 4) An overall test of the model showed that the clusters were significantly different (Canonical vanates analysis followed by a Monte Carlo permutation test, ter Braak 1987a,b, F = l 84, p = 0 01) The distinction was most clear between group VII and other groups, whereas among the latter there was always some overlap with neighbour groups Note that the distnbution of sites in Fig 4 IS related to the environmental variables in Fig 3 Thus, group VII sues have small drainage areas, group VI sites are at high altitude, etc The number of taxa, excluding chironomids, water mites, ohgochaetes, and Ptstdium spp , ranged between 3 and 43 (mean ± SD 22 3±9 1) per site Inspection of all environmental vanables in preliminary multiple regression analysis suggested a subset explaining a large part of the vanation Thus, a model based on the present matenal incorporating only drainage area, organic matenal, and discharge explained nearly half of the vanation in species richness (R = 0 46, Table 3) The relationship was lnS = 1 653 -(- 0 218 lnDA -i- 0 178 OM - 0 090 DI where S is species nchness, DA the drainage area in km-, OM the amount of organic matenal ranked by an arbitrary scale 1-5, and DI discharge m m' s"' phate-P, macrophytes and alkalinity In Fig 3 the species with greatest affinity to each of the variables can be identified For example, size related species were Nemurella ptctett Kip , Eeuctra nigra Ol , Plectrocnemta consperm for small size, and Arctopsyche ladogensts Kol and Simuhum tuberosum Lundstr (complex) for large size Similarly, Heptagenia fuscogrisea Retz and two Eeptophlebia species were associated with macrophytes, Baetis lappontcus Bgtss , Capnia pygmaea Zett and C atra Mort with high altitude, and several mayfiies (Ameletus inopinatus Etn , Ephemerella aunvilh Bgtss , Heptagenia dalecarhca Bgtss ) and the snail Eymnaea peregra Mull with high alkalinity, but low phosphate Species near the ongms of the axes are either unrelated to the ordination axes (some of the most widely distnbuted Discussion The now relatively established approach to classifying benthic nver communities by ordination techniques has been successful in showing consistent results (Hildrew Table 3 Regression statistics for a multiple linear regression model for species nchness in North Swedish stream nffles with three independent vanables drainage area (km-), organic matter and discharge (m' s-') R'adj Number of observations Analysis of vanance df Regression 3 Residual 56 Total 59 Intercept Ln discharge Organic matenal Ln drainage area 14 0 461 60 Sum of Squares 7013 7 341 14 354 Mean Square 2 338 0 131 SE 0214 0 048 0 054 0 042 F 17 833 Significance F <0 0001 Coefficients (B) 1653 -0 090 0 178 0218 t Statistic 7712 -2 048 3 278 5 209 p-value <0 0001 0 045 00018 <0 0001 Lower 95% 1 224 -0 177 0 069 0 134 Upper 95% 2 082 -0 002 0 287 0 302 ECOGRAPHY 17 1 (1994) and Townsend 1987, Ormerod and Edwards 1987, Noms and Georges 1992) In Bntain, where the technique was employed early, factors foremost in explaining benthic community structure were substrate, pH-related factors (alkalinity, hardness), and components relating to nver size (distance from source, discharge, summer temperature, slope, stream order) The present study suggests that factors like size of drainage area, alkalinity and water colour are also important in northern Sweden In addition, altitude significantly contributed to the vanation, although this to some extent also is related to nver size in that large nvers are missing at high altitudes, although small ones are represented from all elevations Altitude also reflects temperature The presence of macrophytes and concentration of phosphate phosphorus further strengthened the pattern River ecology has few general theones The River Continuum Concept (Vannote et al 1980) is one of those, and It has caused extensive debate since its generation One of the most attractive ideas lies in its coupling between the trophic structure of animal communities and vanation in energy inputs (Hildrew and Townsend 1987) This notion is supported by the present study, where animal communities did not show distinct groupings but rather a continuous pattern where nver-size-related factors were of pnme importance Most strikingly, the assemblages found in very small streams deviated from the larger ones, suggesting that these were functionally different Such a contention is, for example, upheld by the strong connection of small streams to the npanan vegetation, through shading and the high litter input per unit area (Cummins 1975) The strong relationship found between drainage area and species nchness was interesting Drainage area was well correlated with factors such as discharge (r = 0 79), stream order (r = 0 89), current velocity (r = 0 54), width (r = 0 83), and depth (r = 0 54) Species nchness is known to increase with such size related measures (for freshwater invertebrates see e g Strayer 1983, Bronmark et al 1984, Bronmark 1985, Malmqvist etal 1992) However, the explanations for the relationship are not clear Hart and Horwitz (1991) reviewing alternative models claimed that most studies in fact were not designed such that the underlying causes could be revealed The same holds true for this study, although one factor, the extinction rate, may be of particular interest in dealing with drainage areas, the best of the size-related factors In addition to the fact that large drainage areas certainly are likely to contain more habitats than small ones, extinction rates should be lower within a large drainage area This could be expected since vanous parts of a drainage area are linked together Survival in and dispersal from a higher number of tnbutanes acting as refuges in large systems would then be possible The fact that detntal matenal is significant in the energy budget of streams in forested regions and thereby positively influences species nchness is hardly surpnsing In contrast, it is less obvious why discharge has a negative relationship with nchness ECOGRAPHY 17 1 (1994) when drainage area is accounted for It may imply that run-off or precipitation regime is of importance The value of individual species as indicators of disturbance could be useful as a shortcut to classifying affected sites However, species evolved in natural habitats, not in habitats disturbed by man Therefore, the presence of particular species does not necessarily indicate disturbance For example, species which typically are associated with pollution do occur naturally in habitats where organic debns is deposited in slow-flowing stream sections and backwaters, one good example is Asellus aquaticus L Neither could the lack of a species be used as evidence for disturbance, because this could not only be the consequence of a true absence at a site, but also be due to failure of sampling the species Despite these problems there remains a lot of information of a site based on its species composition This information value should be particularly strong when the entire community IS considered rather than separate species Biotic interactions were not considered in the present study Such interactions are likely to be important locally, especially where abiotic disturbance is low or moderate (Peckarsky 1983), although, their importance for shaping the species structure is probably limited over entire riffles or streams Few studies, if any, have conclusively demonstrated the complete removal of taxa by predators or competitors m stream benthos, whereas they cenainly may affect abundances (eg McAuliffe 1984, Reice 1991) Abiotic factors, including the ones studied here, seem to be the more lnfiuential One abiotic factor not directly measured in this study, although It probably had a major effect on the presence of different taxa, was temperature (cf Sweeney and Vannote 1978, Vannote and Sweeney 1980) It is probable, however, that temperature was related to some of the variables considered in this study, e g elevation and measures of nver size. Although only two transects across north Sweden may seem to poorly represent lotic nffie habitats over the entire area, we believe that in this study the mosi common types present have been identified It must be remembered, however, that larger nvers were omitted as were high altitude habitats In addition, highly specialised habitats such as springs and those combining summer drought and winter bottom ice (Johansson and Nilsson in press) were not included Our conclusion is that assemblages of macroinvertebrates m stream riffles in northern Sweden show gradually changed structure with environmental factors, the most important of which are the S12X of drainage area, elevation, water quality in terms of alkalinity, water colour and phosphate Species richness was also most strongly related to size of drainage area, and to a lesser degree, to organic matter, and discharge. Acknowledgements - A N Nilsson and C Otto offered valuable cnncism on an earlier version of this manuscnpt Financial support was received from the Swedish Environmental Protection Agency 15

Journal

EcographyWiley

Published: Mar 1, 1994

There are no references for this article.