Environmental factors, spatial variation, and specific requirements of Chironomidae in Mediterranean reference streams

Chironomidae spatial distribution was investigated at 63 near-pristine sites in 22 catchments of the Iberian Mediterranean coast. We used partial redundancy analysis to study Chironomidae community responses to a number of environmental factors acting at several spatial scales. The percentage of variation explained by local factors (23.3%) was higher than that explained by geographical (8.5%) or regional factors (8%). Catchment area, longitude, pH, % siliceous rocks in the catchment, and altitude were the best predictors of Chironomidae assemblages. We used a k-means cluster analysis to classified sites into 3 major groups based on Chironomidae assemblages. These groups were explained mainly by longitudinal zonation and geographical position, and were defined as 1) siliceous headwater streams, 2) mid-altitude streams with small catchment areas, and 3) medium-sized calcareous streams. Distinct species assemblages with associated indicator taxa were established for each stream category using IndVal analysis. Species responses to previously identified key environmental variables were determined, and optima and tolerances were established by weighted average regression. Distinct ecological requirements were observed among genera and among species of the same genus. Some genera were restricted to headwater systems (e.g., Diamesa), whereas others (e.g., Eukiefferiella) had wider ecological preferences but with distinct distributions among congenerics. In the present period of climate change, optima and tolerances of species might be a useful tool to predict responses of different species to changes in significant environmental variables, such as temperature and hydrology.

One of the focal points of aquatic community ecology is to identify factors (i.e., habitat, competition) that determine community composition in streams and to study how these factors influence biotic diversity and abundance (Allan 1995). Aquatic macroinvertebrate communities respond to multiple environmental gradients, many of which are scale-related (Vinson and Hawkins 1998). Therefore, communities are shaped by both local-scale processes and broad-scale constraints, such as geology and climate (Menge andOlson 1990, Poff 1997). Several recent studies examined the relationships between freshwater communities and environmental factors measured at distinct spatial scales (Johnson et al. 2007, Mykra et al. 2007). The relative importance given to factors that affected community structure of benthic macroinvertebrates differed among studies, but most authors reported that local-scale factors exerted the greatest influence (e.g., Death andJoy 2004, Sandin andJohnson 2004).
Chironomidae are the most broadly distributed, species-rich, and often the most abundant family of benthic macroinvertebrates in fresh waters (Pinder 1986). They are a heterogeneous group of species with variable responses to environmental gradients (Lencioni and Rossaro 2005). Ecological information on chironomids is still fragmentary, especially for larvae, because species identification is time-consuming and requires sound taxonomic expertise. Chironomidae are widely used in bioassessment as indicators of lake trophic conditions (Saether 1979) and organic pollution in running waters (Orendt 1999). This family also is used in paleolimnological studies for environmental reconstruction (Walker 2001).
Several studies have identified spatial assemblage patterns and significant environmental factors contributing to Chironomidae assemblage structure in tem-perate streams (e.g., Lindegaard andBrodersen 1995, Lencioni andRossaro 2005). One important environmental gradient is longitudinal zonation. Only a few studies have addressed Chironomidae in Mediterranean streams (Gonzá lez et al. 1985, Casas andVílchez-Quero 1993). These studies were conducted along relatively small spatial gradients, were based mostly on pupal exuviae, and were not done exclusively in reference conditions. They reported that altitudinal gradient strongly influenced Chironomidae assemblage composition.
Biological responses to environmental factors can be studied using an autecological approach at the population level (Tokeshi 1999). For example, estimation of optima and tolerances for each species is an excellent way to obtain autecological information on relevant environmental conditions. However, autecological characterization requires species identification because species of the same genus might have different responses to environmental factors (Rossaro et al. 2006). Autecological characterization also requires much data covering a wide range of spatial and temporal variation. Only a few studies report the specific ecological requirements of Chironomidae taxa in near-pristine Mediterranean streams (but see Calle-Martínez and Casas 2006).
According to the Water Framework Directive (European Commission 2000), a prerequisite for effective management of water systems is information on the state of freshwater biodiversity in near-pristine ecosystems. In our study, the ecological requirements of the most frequent chironomid species were analyzed in relation to important environmental gradients present in Mediterranean catchments of the Iberian Peninsula. Sites in reference (or the least-disturbed) condition in middle and lower sections of the catchments that included a range of stream types with different geological, morphological, and physicochemical features (Sá nchez-Montoya et al. 2007) were used for a large-scale examination of Chironomidae assemblages. These related characteristics should affect the composition of Chironomidae assemblages with the result that Chironomidae assemblages should be distinctly different among stream types. The specific aims of our study were to 1) assess the contribution of environmental factors at different spatial scales (geographical, regional, and local) to the structure of Chironomidae assemblages, 2) identify the environmental factors most strongly related to assemblage structure, 3) determine assemblage groups in different Mediterranean reference streams and their representative indicator species, and 4) define the optima and tolerances of Chironomidae taxa to relevant environmental factors influencing assemblage composition.

Study area
The study area ( Fig. 1) covered ;78,560 km 2 of the Iberian Mediterranean coast and included large (e.g., Jú car: 18,136 km 2 ) and small catchments (e.g., Chillar: 54 km 2 ) (Appendix 1). Thermal, pluviometric, and altitudinal gradients were present from north to south and from the mountains to the coast (usually west to east). The annual range in temperature is À2 to 428C, and annual precipitation ranges from 280 to 1000 mm. Strong storms often cause flooding during spring and autumn (MIMAM 2000). The Mediterranean climate has hot dry summers and cool wet winters. Rivers show high seasonality, with high annual and interannual variability in discharge, and frequent and predictable periods of flooding and drying (Gasith and Resh 1999). Limestone and other sedimentary rocks dominate along the coast, although some siliceous areas are present in the Sierra Nevada (south), Montseny, and Pyrenees (north). Sclerophylous and evergreen trees and shrubs are dominant, but deciduous forests are found in some areas.
Only samples taken in spring were considered in our study because streams in Mediterranean zones have a high probability of drying during part of the year (usually in summer). Restricting sampling to spring ensured that water flows and biological assemblages were comparable because when streams are reduced to pools, macroinvertebrate communities (Bonada et al. 2006), and Chironomidae assemblages ) might change. Sixty-three sites in 22 river catchments were sampled during spring 2003 (Fig. 1). The sites ranged widely in altitude (12-1940 m above sea level [asl]) and latitude (from Muga stream in the northeast to Guadiaro in southern Spain) (Appendix 2; see Robles et al. 2004 for a description of the catchments).
Only minimally disturbed sites were used to ensure that they represented near-pristine conditions (i.e., most headwater streams) or least-disturbed sites (most mid-reaches). The site network consisted of very small streams at high altitude to mid-reaches of several medium-sized streams because no minimally disturbed large streams were present in the area. Sites were selected on the basis of 18 criteria used to establish reference conditions in Mediterranean streams (Sá nchez-Montoya et al. 2005). A reference site was identified in relation to features at 3 spatial scales: 1) catchment (e.g., no canalization or water abstraction, natural land uses in catchments .70%), 2) site (e.g., natural riparian vegetation appropriate to the type, absence of point and diffuse pollution source), and 3) instream (e.g., no transversal structures (dams),

248
[Volume 28 T. PUNTÍ ET AL. no sand and gravel extraction). Most selected sites fulfilled the 18 criteria. A few mid-reach sites fit all but one criterion and were retained in the analysis.

Environmental descriptors
The environmental variables (41 variables) were divided into 3 groups based on the spatial scale of the variable: geographical (site coordinates), regional (geological characteristics, catchment area, land use), and local (e.g., water chemistry and habitat) (Appendix 2). Geographical variables were calculated by including all terms of a cubic trend surface regression (i.e., x, y, x 2 , xy, y 2 , x 3 , x 2 y, xy 2 , and y 3 ) with x (latitude) and y (longitude) and using a similar approach to that of Borcard et al. (1992). Use of this geographical component in the analysis allows inclusion of largescale spatial structure in the data set (Meot et al. 1998). The geographical component explains patterns in the species data not shared by any of the other environ-mental variables measured. It is an indirect synthetic descriptor of other unmeasured biological or environmental factors (see Magalhaes et al. 2002, Johnson et al. 2007. Geological characteristics and catchment area were calculated from a digital terrain model (DTM; 30 3 30 m; Centro Geográ fico del Ejército, Ministerio de Defensa, Spain, 2005) and Arc/Info software (version 9.0; Environmental Systems Research Institute, Redlands, California). Classification of catchment land cover was obtained from CORINE LAND COVER (Instituto Geográ fico Nacional, Madrid, Spain;. Local variables included riparian characteristics (Munné et al. 2003) and bedform variables that indicated habitat condition ). Physicochemical variables (e.g., conductivity, pH, temperature, O 2 , and discharge) were measured in situ with portable meters. Water samples were analyzed in the laboratory for alkalinity, Cl -, and SO 4 2following standard procedures (APHA 1992). Other local vari-  ables, such as altitude, stream order, and percentage of dry period were derived from geographical information system (GIS) data available from the database of CEDEX (Centro de Estudios Hidrográ ficos, Spain).

Biological sampling
Benthic macroinvertebrates were sampled with the protocol established in the GUADALMED project (Já imez- Cuéllar et al. 2004). This protocol has been used in several benthic studies (e.g., Bonada et al. 2005, Sá nchez-Montoya et al. 2007) and provides a standardized data set. At each site, a multihabitat sample was collected from all available habitats with a kick net (250-lm mesh size). The collected material was placed in trays, and organisms were identified to family level (except for Hydracarina, Oligochaeta, and Ostracoda). Sampling ended when no new taxa were recorded. Samples were preserved in the field using 10% formalin. Chironomidae were sorted in the laboratory. All chironomids collected in our study were larvae, which were sorted, counted, and mounted on slides for identification with high power magnification to the highest taxonomic resolution possible. Larvae were first grouped by morphological appearance (shape of the head capsule, color, body setae, and size) under a stereomicroscope, and all (if ,10 individuals of each morphological type) or part (if .10 individuals of each type) of the larvae in each group were mounted on slides (Pinder 1983). In total, 12,409 larvae were examined (4347 mounted specimens).
We used identification keys and species descriptions selected from the European literature, including Wiederholm (1983), Nocentini (1985), Schmid (1993), and Rieradevall and Brooks (2001). For some genera (e.g., Corynoneura, Micropsectra, and Tanytarsus), the authors' own experiences in the identification of larvae and reference collections were used. In some cases, a chironomid larva could not be identified to species because of small size of individuals (2 nd or 3 rd instars) or difficulty in differentiating some groups (e.g., Orthocladius-Cricotopus) at the larval stage. Therefore, in the final biological matrix, a number of taxonomical levels were mixed. The relative abundance of Chironomidae (percentage of each taxon per sampling site) was calculated and used in multivariate analysis.

Data analysis
Detrended correspondence analysis (DCA) (Hill and Gauch 1980) of taxon relative abundances was done to assess the degree of taxonomic turnover across ecological gradients and to determine the gradient length in the biological data set. The gradient lengths of the first 2 axes were 3.0 and 2.7 standard deviation (SD) units, respectively, indicating that either a linear and unimodal species response model should perform reasonably well (Lepŝ and Ŝmilauer 2003). Methods based on a linear response model were best suited to our data (variance explained by redundancy analysis [RDA] ¼ 64%, variance explained by canonical correspondence analysis [CCA] ¼ 59%). Therefore, RDA was used to examine the relationship between Chironomidae assemblages and the explanatory variables. RDA is a constrained form of the linear ordination method of principal components analysis . All analyses were run on 4 th -root-transformed Chironomidae abundance data. When necessary, environmental variables were log(x) or arcsine(=[x])-transformed to approximate normally distributed random errors (Appendix 2). No consensus exists regarding whether rare taxa should be removed from a data set when multivariate analysis is used (Cao et al. 2001). In our case, taxa occurring in !2 samples and with relative abundance !2% in !1 sample were included in the multivariate analysis to prevent a disproportionate effect of Chironomidae taxa with low occurrence on the results (Gauch 1982). All ordinations were run with CANOCO (version 4.5; Microcomputer Power, Ithaca, New York).
Direct gradient analysis (partial constrained ordination or partition of variance [pRDA]) was used to estimate the fraction of variance in community composition explained by the 3 groups of explanatory variables (geographical, regional, and local). pRDA allows examination of relationships between desired environmental variables and biological variables by removing the effects of known factors of no interest. The same variable can be used both as a covariable and as an environmental variable in different parts of the same analysis. In variation partitioning, covariables are useful for distinguishing the relative contributions of groups of variables to explain species composition .
First, constrained ordinations were run to determine the significant (p , 0.05) environmental variables. Only significant variables were considered as environmental variables in the pRDA. Variables included in the 3 groups (geographical, regional, and local) and the individual effects explained by each variable (k1 or marginal effects) are shown in Appendix 2. A series of pRDAs was run for Chironomidae assemblages (Borcard et al. 1992). pRDAs were carried out in the following steps: 1) RDA with species data and all 3 groups of environmental variables as explanatory variables and no covariables was used to determine the total amount of variation explained (TVE) by the 3 environmental groups; 2) pRDA with 1 of the 3

250
[Volume 28 T. PUNTÍ ET AL. environmental variable groups as explanatory variables and the other 2 groups together as covariables was used to obtain single effects for each group of variables; 3) calculation of the variation shared by several combinations between groups of variables (interaction effects); and 4) calculation of the unexplained proportion of variation (1 -TVE).
RDA with forward selection was run to detect the main environmental variables that could best explain the variability of the analyzed data set. Bonferroniadjusted forward selection was used to reduce redundancy between variables. The significance of each remaining variable was tested with Monte Carlo permutation (9999 permutations, p , 0.05). The significance level was set to a/n for each variable tested to compensate for the number of statistical tests . Environmental variables were chosen only when their addition did not cause any variation inflation factor .20. Pearson correlations among the first 4 canonical axes and environmental variables were used to interpret the meaning of these axes and their significance.
Groups of Chironomidae assemblages were obtained by clustering samples based on their projections onto the first 2 ordination axes with a k-means method (SPSS for Windows, version 10.6; SPSS, Chicago, Illinois). The indicator value method (IndVal) (Dufrêne and Legendre 1997) was applied to determine the most representative Chironomidae taxa among the groups of k-means obtained (PC-ORD for Windows, version 4.20; MjM Software, Glendenen Beach, Oregon; McCune and Mefford 1999). IndVal is based on the comparison of relative abundances and relative frequencies of taxa in different predetermined groups of sites. Each taxon is associated with an indicator value (IV) that varies between 0 and 100, and a p-value obtained by Monte Carlo permutations (9999 runs).
Last, a Weighted Average (WA) regression (C2 programme, CALIBRATE version 1.3; Department of Geography, Newcastle, UK) was used with independent environmental variables to calculate the optima and tolerances of several species of chironomids. This analysis estimates the optimum of an environmental variable for each species based on the average of the values of the variable in sites where taxa are present, weighted by species' relative abundances. WA regression assumes that each taxon has a Gaussian response to an environmental variable; therefore, the species optimum (the mode) and tolerance (standard deviation from the optimum) can be calculated (Birks et al. 1990). WA regression has been widely applied in paleolimnology to infer environmental conditions using optima and tolerances of Chironomidae species (Brodersen and Anderson 2002).

Results
Relative importance of geographical, regional, and local variables In total, 141 taxa of Chironomidae in 73 genera were identified from the 63 sites (Appendix 3). Only 117 taxa had relative abundances .2% and were included in multivariate analyses. TVE was 48.3% for the first RDA (3 groups of environmental variables, no covariables) (Table 1). pRDA showed that the single effect of local variables accounted for 23.3%, whereas the single effects of geographical and regional variables accounted for 8.5 and 8%, respectively, of the total variance (Table 1). Thus, local-scale variables explained substantially more of the among-site variance in community composition than did regional-or geographical-scale variables. The total shared variance of the 3 groups of environmental variables accounted for 4.1%, whereas the total shared variance of regional and local variables accounted for 3.1%, and total shared variance of the geographical and local environmental variables accounted for 1.5% of the total variance. The total shared variance of geographical and regional environmental variables was À0.2%. This negative value indicated that the variance explained by the geographical 3 regional term was substantially lower than the unique variance explained by the geographical and regional variables separately. The single effects of the 3 variable groups accounted for 82.4% and interaction terms accounted for the remaining 17.6% of the TVE.

Best predictors of Chironomidae assemblages
The first 4 axes of the RDA explained 19.2% of the total variation of the 117 Chironomidae taxa in the 63 sites. Five environmental variables were included in the model after applying Bonferroni-corrected forward selection. Catchment area was the 1 st variable selected (6.6% of the total variance), followed by longitude (3.4%), pH (3.4%), altitude (2.9%), and % siliceous rocks in the catchment (2.8%) ( Table 2). These results showed the combination of geographical, regional, and local environmental variables could best explain the variation in among-site differences in Chironomidae assemblages, even though in the pRDA, local variables explained the highest percentage of TVE.
A low percentage of Chironomidae variability was explained by the RDA (Table 2), but canonical axes were significant in relation to the set of variables used (Monte Carlo tests, 999 permutations; F ¼ 1.23, p , 0.01). Based on the 5 significant variables, the 1 st axis explained 8% of the total variability in the species data. This axis was positively correlated with pH and catchment area and negatively correlated with altitude and % siliceous rocks in the catchment (Table 2). It differentiated sites in mainly siliceous headwater streams with lower pH and small catchment area from sites in mid-altitude streams with larger catchment areas and higher pH. The 2 nd axis explained 4.3% of total variability in the species data and was negatively related to longitude, altitude, and catchment area. It differentiated lower altitude sites in the southeast from higher altitude sites in the southwest and northwest where high peaks are found (Fig. 2). Species-environment correlations were high for all axes, despite the low cumulative percentage of variability explained.

Chironomidae assemblages
The first 2 canonical axes were used in the classification of sites by k-means clustering because they included the maximum variability expressed by environmental variables (Table 2). As a result, 3 groups of sites with distinct Chironomidae assemblages were identified: 1) siliceous headwater streams, 2) midaltitude streams with small basin areas (mixed siliceous and calcareous), and 3) medium-sized calcareous streams (Fig. 2).

Optima and tolerances
We assumed that taxa would be most abundant in streams with values of environmental variables near their optima. Altitude and catchment area were the major environmental gradients relevant in our study area, but optima and tolerance were calculated for many other factors. Optima and tolerances of altitude and surface catchment area are presented for the 59 most frequent taxa collected in the study area and that occurred in !10 samples (Figs 3, 4). Of the selected taxa, Heleniella ornaticollis, Diamesa sp. A, and Eukiefferiella brevicalcar had the highest optimum for altitude (.1000 m), whereas Phaenopsectra spp., Virgatanytarsus spp., Cricotopus group sylvestris, and Paramerina spp. were restricted to lower altitudes (,500 m) (Fig. 3). In general, taxa that had lower optima for catchment area (,200 km 2 ), such as Stempellinella spp., Corynoneura lobata, and Paratrissocladius excerptus, had narrow values of tolerances for this variable, indicating that these taxa were restricted to small catchments (Fig. 4). In contrast, taxa with higher optima values for catchment area (.800 km 2 ), such as Orthocladius rivulorum, Microtendipes pedellus group, and Virgatanytarsus spp., had wider tolerances and preferred midreaches (Fig. 4).

FIG. 2. Ordination biplot (redundancy analysis) of Chironomidae assemblages in 63 streams in Mediterranean Spain.
Environmental variables were selected using forward selection and Monte Carlo permutation tests. Chironomidae taxa are represented by stars, but taxa that are indicators for a site group are shown with standardized codes for species names (Schnell et al. 1999; see Appendix 3 for codes). Numbers indicate groups of sites identified by k-means clustering. Hand-drawn shapes show the distribution limits among the 3 groups of sites for the first 2 ordination axes (X1 and X2). Group 1 sites were in siliceous headwater streams, group 2 sites were in mid-altitude streams with small catchment areas, and group 3 streams were medium-sized calcareous streams.

2009] 253 CHIRONOMIDAE IN MEDITERRANEAN REFERENCE STREAMS
Optima and tolerance values for some of the environmental variables identified previously in the RDA as relevant for chironomid assemblage composition (altitude, % siliceous rocks, catchment area, pH, temperature, and discharge) provided information regarding niche specificity of some congeneric species (Table 4). In the genus Corynoneura, C. lobata and C. scutellata groups occurred in headwater (mid-high altitudes), mainly siliceous streams with low temperatures. Corynoneura coronata occurred in streams at intermediate altitudes with higher temperature, discharge, and percentage of carbonates and had a wide tolerance for catchment area. The 6 Eukiefferiella taxa had variable optima and tolerances. Eukiefferiella   TABLE 3. Indicator values (IVs) of Chironomidae taxa for site group. Sites were classified by k-means analysis of Chironomidae assemblages into 3 groups (see Data analysis for explanation). Group 1 sites were in siliceous headwater streams, group 2 sites were in mid-altitude streams with small catchment areas, and group 3 streams were medium-sized calcareous streams.  3. Optima (modes) and tolerances (error bars) for altitude for the 59 most frequent taxa (occurred in !10 sites) in streams in Mediterranean Spain. Taxa are arranged on the y-axis in order of increasing optima for altitude. Taxa are listed by standardized codes for species names (Schnell et al. 1999; see Appendix 3 for codes).

254
[Volume 28 T. PUNTÍ ET AL. brevicalcar and E. coerulescens was found mostly at higher altitudes, mainly in siliceous catchments. Eukiefferiella brevicalcar was restricted to fast-flowing streams but had wide tolerances for catchment area, whereas E. devonica and E. minor-fittkaui larvae inhabited mid-altitude, not exclusively calcareous streams. Eukiefferiella gracei and E. ilkleyensis were clearly differentiated from other Eukiefferiella taxa by their distribution in relatively low-altitude streams with a higher percentage of carbonates and higher temperature, discharge, and catchment area. In the genus Rheocricotopus, R. effusus were found in headwaters of siliceous streams and had a wide tolerance for catchment area, whereas R. fuscipes was more restricted to small, mid-altitude, mineralized, and slow-flowing streams. Rheocricotopus chalybeatus group was found in fast-flowing lower altitude streams that were mainly calcareous with variable catchment areas.

Scale-dependent effects on community composition
Establishing the effects of coarse-scale and local environmental factors on species distribution is a prerequisite for a comprehensive understanding of processes that determine structural and functional features of stream communities (Sandin and Johnson 2004). Several factors, such as dispersal capacity, historical effects, climatic constraints, and spatial variation in local environmental conditions, determine the structure of biological communities (Minshall 1988, Bonada et al. 2005. Our study examined Chironomidae distributions across the Mediterranean region of the Iberian Peninsula, over a large area with strong environmental gradients and allowed analysis of the contribution of environmental factors structuring Chironomidae communities in near-pristine streams. Our finding that local environmental variables explained the highest amount of variance (23.3%) in Chironomidae community structure is consistent with the results of a number of previous studies of groups of organisms, such as benthic diatoms (Soininen et al. 2004), macroinvertebrates (Death andJoy 2004, Mykra et al. 2007), fish (Magalhaes et al. 2002), and macrophytes (Johnson et al. 2007). In contrast, other authors have reported that large-scale factors are the best predictors of stream communities (Richards et al. 1996, Urban et al. 2006). These disagreements regarding the importance of local or large-scale variables in stream communities might also result from differences FIG. 4. Optima (modes) and tolerances (error bars) for catchment area for the 59 most frequent taxa (occurred in !10 sites) in streams in Mediterranean Spain. Taxa are arranged on the y-axis in order of increasing optima for catchment area. Taxa are listed by standardized codes for species names (Schnell et al. 1999; see Appendix 3 for codes).

2009] 255 CHIRONOMIDAE IN MEDITERRANEAN REFERENCE STREAMS
in the study design or may be caused by criteria used in the classification of variables at distinct spatial scales (Sandin and Johnson 2004).
The percentage of variation in Chironomidae assemblages explained by geographical variables in our study was considerable (8.5%) and was similar to the percentage of variation explained by regional variables (8%). The geographical pattern in distributions might reflect historical and climatic factors that are largely independent of present-day environmental variables (Sandin and Johnson 2000). However, a low percentage of the explained variability (4.1%) was described by the interaction of the 3 explanatory variable groups. Thus, the groups of variables used in our study were less related among themselves in comparison with other studies (Sandin and Johnson 2004). TVE (48.3%) also was higher in our study than in other studies. For example, TVE was 24.8% in a study of caddisfly communities in the streams that we studied (Bonada et al. 2005). In contrast, the relatively high percentage of unexplained variation (51.7%) in our study is typical of noisy data sets with many taxa and many 0 values (Borcard et al. 1992). Unexplained variation could be the result of unmeasured variables, such as species interactions, food resources, dispersal, sampling vari-ability, or measurement errors. Overall, our results show that different groups of variables act at local and regional spatial scales to affect community composition (habitat filters sensu Poff 1997).

Variables that affect Chironomidae composition and distribution
Little information is available on environmental factors and mechanisms that regulate assemblage composition and distribution of Chironomidae taxa in Mediterranean streams (Calle-Martínez and Casas 2006. Our data indicate that longitudinal zonation is the strongest environmental gradient underlying distribution patterns in Iberian Mediterranean streams. Geographical position was next in importance and was closely related to community patterns along the secondary axes of the RDA. pH also was an important driver of community assembly and is directly related to other regional variables, such as catchment geology. This pattern is consistent with the findings of other authors, who have demonstrated that Chironomidae composition changes along the river continuum, in association with altitude, stream order, and channel width (Ward andWilliams 1986, Lindegaard andBrodersen 1995).
Altitudinal gradients affect distributions of other organisms in the Mediterranean region and in other parts of the world (Coffman 1989, Casas andVílchez-Quero 1993), and altitude strongly influenced Chironomidae assemblages in our study. For example, Chironomidae assemblages in headwater siliceous streams in the Pyrenees and the Sierra Nevada were similar despite the geographical distances between these mountains. Differences in altitude can result in considerable differences in local climate and other physical conditions, thereby affecting assemblage structure. However, molecular taxonomic techniques might show that populations of the same morphological species (such as Diamesa or Eukiefferiella) that are separated by great distances (Pyrenees and Sierra Nevada) actually differ. Future studies based on molecular taxonomic techniques might help clarify the importance of mountain isolation.

Chironomidae as indicators for reference condition
Chironomids have many adaptations for dispersal and colonization (Armitage 1995), but many species have regionally restricted distributions and ecological preferences. Our data show that 3 distinct Chironomidae assemblages provided a broadly meaningful ecological interpretation for reference conditions in Mediterranean streams.
Indicators for headwater streams (group 1) were a diverse group of taxa. Several taxa, such as T. bavaricacalvescens, H. ornaticollis, and R. effusus, typically are associated with low-temperature torrential mountain streams. These taxa, and Brillia bifida and P. excerptus, occur in the Sierra Nevada (Casas and Vílchez-Quero 1993) and Pyrenees (Prat et al. 1983. They are representative of headwater systems but are not restricted to upper altitudes. In contrast, Diamesa is regarded as a characteristic genus with a narrow ecological niche. The genus consists mainly of coldstenothermal species (Maiolini and Lencioni 2001) that inhabit siliceous headwater streams. However, even in this cold-stenothermal genus, differences in optima and tolerances were observed at the species level. Diamesa zernyi-thienemanni group and Diamesa hamaticornis were found in headwater streams at lower altitudes and were not restricted to siliceous geology, whereas Pseudodiamesa branickii and Diamesa bertrami are typical of nonglacial alpine streams (Lods-Crozet et al. 2001). Diamesinae maintain relatively dense populations at mean water temperatures of ;58C (Maiolini and Lencioni 2001). Our results indicate that many Diamesa species have a higher temperature optima than are reported for alpine streams, but pH optima similar to those described by Rossaro et al. (2006).
Most species of Eukiefferiella were widely distributed along the altitudinal gradient, but E. brevicalcar, E. devonica, and E. coerulescens were found at higher altitudes, and E. brevicalcar was an indicator for group 1 streams. Casas and Vílchez-Quero (1993) analyzed the altitudinal distribution of Chironomidae in the Sierra Nevada Mountains and found that Eukiefferiella was one of the richest and most numerically dominant genera in headwater streams. Many of our headwater streams were at relatively low altitudes, and optima and tolerances for altitude for some Eukiefferiella taxa were lower than values reported in other studies Vinc¸on 1991, Casas andVílchez-Quero 1993).
Assemblages in mid-altitude streams (group 2) were characterized by more ubiquitous species with short life cycles. Many of these species were tolerant of slowflow conditions and were warm-water adapted. Chironominae and Tanypodinae were dominant in this group (García and Laville 2000). Chironominae are more abundant when water temperature increases (Maiolini and Lencioni 2001). For example, most Polypedilum species recorded were found mid-altitude mountain and foothill streams.
Assemblages in medium-sized calcareous streams (group 3) were characterized by Orthocladius and Cricotopus, which are tolerant and opportunistic genera generally associated with mineralized waters (Calle-Martínez and Casas 2006). Their presence or absence was not related to a well-defined range of environmental variables.
Our data confirm the importance of species-level identification to provide information about the ecological requirements of chironomids in reference streams. Our results are consistent with the observation that species belonging to the same genus often have clearly different ecological niches. However, large data sets are required to determine species autoecology when optima and tolerances are obtained from field data because a weak sampling effort might not define the full range of conditions in which some species exist. Mediterranean streams are strongly seasonal. Thus, samples from all seasons are required to obtain large data sets that integrate space and time and include intra-and interannual variability. Predicted climate changes might increase the number of ephemeral streams and decrease cold-water habitats (Rossaro et al. 2006). Therefore, a better understanding of the ecological requirements of chironomids in Mediterranean regions will help us understand the potential consequences of climate change in these highly diverse ecosystems.