Andorra is a small landlocked country between France and Spain (Fig. 1). Its altitude ranges between 840 m and 2946 m. The Andorran hydrographic network is organized around two main rivers located on the Spanish slope: the Riu Valira d'Orient and the Riu Valira Del Nord (Fig. 1). The first crosses Andorra from east to west, the second from north to south. They join in Andorra la Vieilla, the main city of Andorra, and become the Riu Gran Valira, a tributary of the Rio Segre (Ebro's catchment). Twelve out of the 13 sampling sites were located on this hydrographic unit. The site of Pas de la Case is atypical, because located on the Pyrenees French slope on the Riu Arieja and downstream the first wastewater treatment plant built in Andorra. Geographical and morphological characteristics of the studied sites are given in Table 1 with additional information about the type of anthropogenic impact affecting each site. Sites located in the upper course of the streams are exposed to ski activities. Most of the sites located in the mid and low courses of the streams are exposed to the discharge of domestic sewages (cities, ski resorts and camping areas) and to mechanical disturbances caused by the development of the tourism infrastructures.

Nine physicochemical variables (temperature, pH, conductivity, COD, BOD, nitrate, nitrite, ammonium and phosphate) were measured weekly from September 1998 to December 1999 by the Medi-Ambient Department of Andorra. Except for temperature, measured in situ, the other physicochemical analyses were conducted at the laboratory according to a normalized protocol.

The benthic macroinvertebrates were sampled in autumn 1998, winter 1998/1999, summer 1999 and autumn 1999 in each sampling sites. At each site, eight subsamples were taken using a Surber net (200 μm mesh size, sampling area of 1/20 m²), according to the microhabitat diversity). All the collected specimens were counted and the family level was retained for this study.

Three approaches were combined in order to analyse the following data.

(1) An analysis of the inter-date inertia (between-analysis [12]) was conducted on both physicochemical and biological datasets. This analysis allowed us to focus on differences between sampling dates without excluding the spatial dimension [13,14]. The matrix of physicochemical parameters was built using the data recorded in the four weeks prior to the invertebrate sampling, in order to detect the modifications of the structure of the faunal assemblages due to prior alteration of the water quality. The four single records were averaged. The final matrix was of 52 samples by nine parameters. The matrix of benthic macroinvertebrates was built using densities of taxa. Taxa represented by fewer than 10 specimens recorded over the study were previously removed from the matrix in order to overcome the strong dependence of ordination methods on single outlier species [15]. Data were standardized by total sample density in order to allow comparability between samples. The working matrix was of 416 samples by 43 taxa. Between-analyses required a pre-ordination of the sites by an ordination method [13,14]. A standardized principal component analysis (sPCA) was performed on the physicochemical variables and a correspondence analysis (CoA) was conducted on the fauna variables.

(2) An analysis of the community structure based on the taxa richness, on the Fisher [16], Shannon [17], Simpson [18], and Margalef [19] diversity indices, and on the Shannon equitability index [17]. A functional feeding group analysis was also performed. The taxa were classified into five functional feeding groups: predators, filtering collectors, gathering collectors, shredders, and scrapers, according to the Cummins's categorisation [20,21].

(3) The third approach was to calculate two biological quality indices expressed in quality classes, i.e. the French indice biologique général normalisé (IBGN) [22], used in routine by the French water agencies in the monitoring programs, and the modified Biological Monitoring Working Party (BMWP′) [23], adapted for Spanish rivers.

The two inter-dates analyses, performed respectively on the physicochemical and invertebrate datasets, similarly classified the 13 studied sites into four groups: group 1, Castellar–Grau Roig–Arinsal; group 2, Cortinada–El Tarter–Massana; group 3, Canillo–Pont San Antoni–Encamp; group 4, Escaldes–Margineda–Frontera–Pas de la Case (Fig. 2A and B). Only the site El Tarter shifted from group 2 to group 1 and the site Escaldes from group 4 to group 3 in the ordination based on the abiotic variables by comparison to the ordination based on the biotic variables. In both analyses, the total amount of variance explained by the two first axes was more than 70% with more than 55% explained by the first axis and less than 20% by the second axis. Groups 1 and 2 (upstream sites) were separated from group 3 and 4 (mid course and downstream sites) along the F1 axis and groups 1 and 4 were separated from groups 2 and 3 by the second axis (Fig. 2A and B). The projection of the physicochemical variables on the factorial plan of the sPCA showed that the values of all the parameters were higher in the downstream sites (Fig. 2C). Sites from the mid-course of the streams (group 3) exhibited the higher values of organic disturbance (BOD, COD) as well as of reduced forms of nitrogen compounds (NO2− and NH4+). Sites from the lower course (group 4) exhibited the higher values of conductivity, nitrate and phosphate contents (Fig. 2C). The inter-seasons analysis separated on the F1 axis autumn samples from winter and summer samples with 45% of explained variance. Samples from autumn 1998 and summer 1999 were separated on the F2 axis to samples from autumn and winter 99, with 41% of explained variance (Fig. 2D).

The analysis of the physicochemical data showed increasing values for all the parameters from upstream sites to downstream sites (Table 2).

Mean values of conductivity ranged from 49 to 69 μS cm⁻¹ in sites from group 1 and from 110 to 268 μS cm⁻¹ in sites from group 4. The same pattern was observed for the organic compounds with for example mean values of BOD ranging from 0.2 to 0.7 mg l⁻¹ O₂ in sites from group 1 and from 1 to 4 mg l⁻¹ O₂ in sites from group 4. For the nitrogen compounds, means values of ammonium ranged, for example, from 66 to 147 μg l⁻¹ in sites of group 1, from 54 to 373 μg l⁻¹ in sites of group 2, from 430 to 553 μg l⁻¹ in sites of group 3, and from 525 to 67 700 μg l⁻¹ in sites of group 4. Similarly, mean values of phosphates ranged from 0.12 to 0.17 mg l⁻¹ in sites from group 1, 0.19 to 0.37 mg l⁻¹ in sites from group 2, 0.34 to 0.82 mg l⁻¹ in sites from group 3 and 0.43 to 0.97 mg l⁻¹ in sites from group 4. The standard errors on the mean (SEM), calculated on the weekly measurements of the physicochemical variables, were higher for all the parameters in the sites from group 4, especially in Pas de la Case (Table 2). Sites of group 3 exhibited the highest SEM values for conductivity, BOD and ammonium, followed by those of groups 2 and 1.

Values of taxa richness, diversity and equitability indices decreased downstream (Table 3). The sites located upstream, i.e. Grau Roig, Castellar, and Arinsal, sheltered respectively 25, 22 and 27 families, whereas only 19 families were recorded at Frontera and only 12 below the wastewater treatment plant of Pas de la Case. The sites of group 1 had values of indices of diversity ranging between 0.14 and 4.53; between 0.42 and 0.74 for the indices of structure, between 13 and 15 for IBGN index and between 73 and 139 for the BMWP′ index. The values for the sites of the group 2 were ranging between 0.12 and 4.05 for indices of diversity, between 0.19 and 0.79 for the indices of structure, between 11 and 16 for the IBGN index, and between 105 and 126 for the BMWP′ index. The values for the sites of the third group were ranging between 0.06 and 2.94 for indices of diversity, between 0.02 and 0.48 for the indices of structure, between 2 and 14 for the IBGN index and between 16 and 188 for the BMWP′ index. The values for the sites of the group 4 were ranging between 0.02 and 1.53 for indices of diversity, between 0.01 and 0.20 for the indices of structure, between 2 and 11 for the IBGN index, and between 7 and 62 for the BMWP′ index.

Regarding the seasonal variability of the faunistic indices, no seasonal trends were observed for the sites located in upper (group 1) and extreme low course (Frontera and Pas de la Case) of the streams, while diversity and equitability indices were lower in summer and winter in the sites located in the middle course (group 3 and Massana in group 2) and in the low course (Margineda and Escaldes) of the streams (Table 3). Pollutant-sensitive taxa, such as Athericidae and Rhyacophilidae, were present at the upstream sites: Grau Roig, El Tarter, Castellar, Cortinada and Arinsal, and disappeared completely in the downstream sites (Table 4). The Plecoptera Perlidae, Perlodidae, Leuctridae and Nemouridae, and the Trichoptera Brachycentridae followed this pattern too.

A progressive reduction in the density of some taxa along the longitudinal gradient until their disappearance in the downstream sites was also observed. This was the case for the Baetidae, which were present in the upstream reaches with a density of more than 4080 ind m⁻² in Grau Roig, Castellar, Arinsal, El Tarter, Cortinada, a density from 520 to 21 040 ind m⁻² in the mid-elevation sites (Massana, Canillo, Encamp), of less than 6860 ind m⁻² in the downstream sites (Pont San Antoni, Escaldes) and disappeared at Margineda, Frontera and Pas de la Case. The same pattern was observed to a lesser extent for Chironomidae, as well as for the Heptageniidae, Simuliidae, Limoniidae, and Thaumaleidae, which almost disappeared in the Riu Gran Valira (Table 4). In contrast, some taxa increased in density from the upstream to the downstream sites, however disappeared in the highly disturbed sites. Psychodidae typically followed this pattern, with respectively densities of 4 and 8 ind m⁻² in Grau Roig and Cortinada, and 24 and 34 ind m⁻² in Encamp and Pont San Antoni (Table 4). In the highly polluted sites Pas de la Case and Frontera, densities decreased again (14 ind m⁻² at Frontera and 2 ind m⁻² at Pas de la Case). Strongly disturbed sites such as Margineda, Frontera and Pas de la Case were numerically dominated by Oligochaeta (respectively 11 175, 16 192 and 12 908 ind m⁻² – annual density) and, to a lesser extent, by Nematodes and Glossiphoniidae (Table 4).

The functional feeding group analysis showed a modification of the trophic structure of the communities along the longitudinal gradient. In the upper sites, Grau Roig, El Tarter, Castellar, Cortinada, and Arinsal, the five trophic categories were represented, while in the mid-course sites, Canillo, Encamp, Escaldes, Margineda, Massana and Pont San Antoni, predators and filtering collectors disappeared (Fig. 3). In the downstream sites, Pas de la Case and Frontera, all the trophic categories disappeared, except gathering collectors, largely represented by Oligochaeta, whose proportion increased downstream parallel to the degradation of the water quality. They dominated the communities at Pont San Antoni, Escaldes, Margineda, Frontera and Pas de la Case where they represented more than 95% of the invertebrates (Fig. 3).

Replacement of pollutant-sensitive taxa by pollutant-tolerant taxa within a trophic group was also observed. This was the case in the Rius d'Arinsal and Del Nord at the mid-elevation sites (cites ones) where, for example, Perlodidae, Rhyacophilidae, and Planariidae (pollutant-sensitive predators) were progressively replaced downstream by Glossiphoniidae (pollutant-tolerant predators) (Fig. 3, Table 4) – except accidents related to the drift. No seasonal changes were observed in the trophic structure of the benthic macroinvertebrate assemblages of the studied sites.

The calculation of the IBGN and BMWP′ scores led to classify the studied sites in three groups differing in their seasonal patterns of biological quality (Table 3): (1) sites with a good biological quality throughout the year (Class 1 of IBGN – Class I of BMWP′), namely Grau Roig, El Tarter, Castellar, Cortinada, and Arinsal; (2) sites whose biological quality remained bad throughout the year (classes 3 and HC of IBGN, classes IV and V of BMWP), namely Escaldes, Frontera and Pas de la Case; (3) sites whose biological quality changed seasonally (Canillo, Encamp, Massana, Pont San Antoni, Margineda).

This pattern was similar to that observed for the diversity indices with sites located in the upper course of the streams (groups 1 and 2) exhibiting a good biological quality throughout the year, sites located in the middle section of the streams (group 3, Massana and Margineda) exhibiting seasonal variations of the biological quality and sites located downstream (group 4) exhibiting a bad biological quality throughout the year.