Environmental Management

, Volume 46, Issue 2, pp 285–301

Evaluating the Response of Biological Assemblages as Potential Indicators for Restoration Measures in an Intermittent Mediterranean River

Authors

    • Centro de Investigação e de Tecnologias Agro-Ambientais e Biológicas (CITAB)Universidade de Trás-os-Montes e Alto Douro
  • Jose Santos
    • Centro de Estudos FlorestaisInstituto Superior de Agronomia
  • Teresa Ferreira
    • Centro de Estudos FlorestaisInstituto Superior de Agronomia
  • Ana Mendes
    • Centro de Estudos FlorestaisInstituto Superior de Agronomia
Article

DOI: 10.1007/s00267-010-9521-3

Cite this article as:
Hughes, S.J., Santos, J., Ferreira, T. et al. Environmental Management (2010) 46: 285. doi:10.1007/s00267-010-9521-3

Abstract

Bioindicators are essential for detecting environmental degradation and for assessing the success of river restoration initiatives. River restoration projects require the identification of environmental and pressure gradients that affect the river system under study and the selection of suitable indicators to assess habitat quality before, during and after restoration. We assessed the response of benthic macroinvertebrates, fish, bird and macrophyte assemblages to environmental and pressure gradients from sites situated upstream and downstream of a cofferdam on the River Odelouca, an intermittent Mediterranean river in southwest Portugal. The Odelouca will be permanently dammed in 2010. Principal Component Analyses (PCA) of environmental and pressure variables revealed that most variance was explained by environmental factors that clearly separated sites upstream and downstream of the partially built cofferdam. The pressure gradient describing physical impacts to the banks and channel as a result of land use change was less distinct. Redundancy Analysis revealed significant levels of explained variance to species distribution patterns in relation to environmental and pressure variables for all 4 biological assemblages. Partial Redundancy analyses revealed high levels of redundancy for pH between groups and that the avifauna was best associated with pressures acting upon the system. Patterns in invertebrates and fish were associated with descriptors of habitat quality, although fish distribution patterns were affected by reduced connectivity. Procrustean and RELATE (Mantel test) analyses gave broadly similar results and supported these findings. We give suggestions on the suitability of key indicator groups such as benthic macroinvertebrates and endemic fish species to assess in stream habitat quality and appropriate restoration measures, such as the release of peak flow patterns that mimic intermittent Mediterranean systems to combat habitat fragmentation and reduced connectivity.

Keywords

River restorationBenthic macroinvertebratesFishBirdsMacrophytesEnvironmental gradientsPressure gradientsMultivariate analysesConcordance

Introduction

Bioindicators have long been used in ecological assessment of surface water quality, which is subject to degradation from impacts ranging from agriculture to industrialisation and urbanisation (Heino and others 2002; Hughes 2005; Karr 1999). They are also used to gauge the success of restoration measures (Gore and others 2001; Kondolf 1995a; Kondolf 1998). Common indicators of lotic condition include benthic macroinvertebrates (Bonada and others 2006; Feld and Hering 2007; Heino 2005; Rundle and others 1992), fish (Godinho and others 2000; Hughes and others 2005; Karr 1981), diatoms (Kelly and others 1998; Round 1991) and macrophytes (Dodkins and others 2005; Ferreira and Aguiar 2006; Ferreira and others 2002). Birds have also been used to a lesser extent to assess the wider riverine landscape such as the riparian gallery (Bryce and others 2002; Jansen and Robertson 2001b; Vaughn and others 2007). To assess ecosystem health, an effective bioindicator should exhibit detectable and measurable levels of change in relation to specific environmental or pressure gradients, ideally starting from reference conditions (Johnson and others 2006a; Karr and Chu 2000; Paavola and others 2006). In this article we use the word “pressure” to describe sources of degradation or impacts upon river systems and the word “concordance” to describe the degree of association between changes in taxonomic composition in a biological assemblage and environmental or pressure gradients (Paavola and others 2006; Sánchez-Montoya and others 2007).

The European Water Framework Directive (WFD) (European Commission 2000) and the development of lotic monitoring systems in the USA (Barbour and others 1999; Hughes and Peck 2008; Paulsen and others 2008; Stoddard and others 2008) have led to considerable growth in the number of monitoring methods based on the assemblages mentioned above. Bioassessment or restoration measures must consider fundamental changes in aquatic assemblages in relation to environmental quality to assess the success of a restoration or rehabilitation efforts (Kondolf 1995a). Assessment systems that use several biological assemblages (O’Connor and others 2000) consider the concordance of elements of the assemblages with aspects of river environment health to provide a comprehensive ecological image of river health, a concept that is central to the WFD. However, high levels of redundancy can occur in multi assemblage monitoring programmes and relevant information on ecological status can be derived from few or even a single biological assemblage (Resh 2008).

Mediterranean rivers are highly distinct systems, with remarkable but predictable natural cycles of flood and drought that vary in intensity according to levels of annual and interannual rainfall (Bêche and Resh 2007b; Gasith and Resh 1999; Pires and others 2004). This powerful environmental filter determines biological community traits (Bonada and others 2005) and is a strong source of covariance which can detrimentally affect the detection of anthropogenic impacts by biological assemblages. Impacts on Iberian Mediterranean rivers include habitat fragmentation, reduced lateral and longitudinal connectivity and destruction of natural flow regimes as a result of agriculture, forestry, damming, water abstraction and urbanisation (Aguiar and Ferreira 2005; Bonada and others 2005; Hughes and others 2008; Poff 1997).

The Odelouca is an intermittent Mediterranean river (Algarve region, southern Portugal) that, despite human intervention, has intact and floristically diverse riparian galleries along considerable stretches of the river corridor. Two critically endangered endemic fish species (Pires and others 2004) also occur in the Odelouca, namely Iberochondrostoma almacai (Coelho and others 2005) and Squalius aradensis (Coelho and others 1998). Ongoing government initiatives to improve water supply in the Algarve region have authorised completion of a partially constructed dam on the Odelouca by 2010; however compulsory environmental mitigation and compensation measures have been implemented to offset impacts caused by the dam’s construction.

This article assesses the response of four biological assemblages (benthic macroinvertebrates, fish, birds and macrophytes) to environmental and pressure variables in order to identify suitable indicators for monitoring the progress of restoration measures. We aim to answer the following questions: (1) what are the principal environmental and pressure gradients acting upon the Odelouca and how does each biological assemblage respond to them? (2) Does a single group best describe changes in the system under study or is an integrated approach better for assessing restoration measures? (3) Which restoration measures are recommended for the Odelouca?

Methods

Study Area

The Odelouca River (catchment area 511.4 km2) is a medium-sized, low-gradient, intermittent lowland stream running through predominantly schistose areas typical of southern Portugal (Fig. 1). The area’s Mediterranean climate exhibits a predictable seasonal pattern of rainfall (wet season from October to March, dry season from June to September). The Odelouca is relatively slow running, subject to “flashy” spates in the winter, diminishing to unconnected, temporary pools in the dry riverbed during the summer. Catchment topography varies from narrow steep sided valley walls to restricted meander valleys and small floodplains. Greater habitat heterogeneity occurs in less disturbed mid sections of the river corridor and tributaries (temporary side channels, backwaters and riparian galleries). Woody riparian plants comprise Alnus glutinosa (L.) Gaertner, Salix salviifolia Brot. ssp. australis Franco, Nerium oleander L, and Fraxinus angustifolia Vahl. Stands of Tamarix africana Poiret and Nerium oleander L. occur in the lower reaches of the basin.
https://static-content.springer.com/image/art%3A10.1007%2Fs00267-010-9521-3/MediaObjects/267_2010_9521_Fig1_HTML.gif
Fig. 1

Map showing where Portugal is situated and the Odelouca basin with sampling sites. The thick black bar indicates where the cofferdam is situated

Agriculture (extensive citrus groves and low level grazing) has replaced the natural Mediterranean cork-oak woodland vegetation (Quercus suber L.) on the wider floodplain below the partially built dam (a cofferdam and an excavated subterranean flow diversion gallery). Observed impacts include diffuse organic pollution, nutrient enrichment, riparian clearance, bank resectioning, bank reinforcement and reduced longitudinal connectivity from irrigation (pumping from the riverbed and small scale damming). Urbanisation is scant, restricted to two small villages and small agricultural hamlets. Eucalyptus globulus Labill. and Pinus pinaster Aiton plantations are present in parts of the upper basin. Tributaries suffer little physical disturbance but the Monchique stream is affected by organic input from piggeries and Monchique village. The lower reaches of the Monchicão tributary are affected by abstraction for agricultural irrigation. The Ribeira de Carvalho, situated in the upper Odelouca catchment, is far less altered.

Habitat Assessment

Field data were collected in spring 2005 from 30 sites: 25 sites along the main channel and 5 sites along the tributaries of Ribeira de Carvalho (1 site), Ribeira de Monchique (2 sites) and The Ribeira de Monchicão (2 sites), near their confluences with the Odelouca. Habitat structure, diversity and quality were assessed over a 500 m reach using an adapted version of the River Habitat Survey (RHS; addition of Iberian Peninsula land use categories and plant species). Developed in the UK, RHS records in-stream substrate and flow type, character and modification of the margins, land use, riparian vegetation, predominant habitat features and modifications, together with measurements of stream and bank dimension (Raven and others 1997).

A geographical information system of land use, riparian vegetation quality, conservation state and continuity was created from survey data and aerial photography. Catchment data on geology, climate, altitude, relief, land use, land cover, organic and industrial discharge and the presence of roads were also added (Fernandes and others 2007).

Lotic Communities

All organism group samples were taken within each 500 m RHS reach. Aquatic organisms were sampled using European STAR project methodologies, modified for application in Portuguese lotic systems by the Portuguese Water Institute INAG (http://dqa.inag.pt/dqa2002/port/docs_apoio/nacionais.html). A full taxon list and species codes are given in Appendix 1.

Benthic macroinvertebrate samples were taken along a 50–100 m long stretch at the downstream end of each 500 m reach using an adapted AQEM (Integrated Assessment System for the Ecological Quality of Streams and Rivers throughout Europe using Benthic Macroinvertebrates) multihabitat sampling protocol (AQEM 1999; Hering and others 2004; INAG 2008a). The sampling area covered the greatest possible diversity of habitats in the reach, including (if present) a riffle and areas of deposition. Types and extent of habitats (organic and inorganic substrates) were visually estimated prior to sampling and six 1 m long 25 cm wide sampling units of the most representative habitats were taken (0.25 m × 0.25 m handnet; 500 μm mesh, habitats representing < 5% total cover were excluded). Samples, which were not timed, were proportionally distributed among the identified habitat types. The composite sample was placed in a labelled plastic flask and fixed in situ using 4% formaldehyde. In the laboratory, samples were washed, sieved, sorted and identified using a low-power stereo microscope. All individuals were picked from the samples; sub-sampling was used when more than 200 individuals of a given taxon were present in the sample. Macroinvertebrates were identified to the lowest possible taxonomic level, which was mostly genus or species. Where higher levels of taxonomic resolution were not possible (for example many Diptera, or early instars of some insect groups) the AQEM protocol for taxonomic adjustment was applied (Hering and others 2002).

Fish were sampled using an adapted STAR methodology (INAG 2008b) with an Electracatch International, SAREL model WFC7-HV electrofisher (applying 300 V and maintaining a 3-A output to a 40 cm diameter ring anode). Sampling was carried out (area sampled was 20 times the mean width of the survey reach, minimum length 100 m starting at spot check 1) by walking upstream in a zigzag pattern or sampling from a boat at sites over 1 m deep (Godinho and others 2000). The same person operated the electrofisher across sample sites and two netters picked up stunned fish. Captured fish were held in large plastic containers, identified to species, counted, and immediately returned to the river. No voucher specimens were taken.

A trained field ornithologist carried out bird surveys over eleven days between late May and June to ensure the maximum number of nesting species. Surveys were not carried out in wet or windy conditions. Bird samples were made at three equidistant points (250 m distance between points) including the riparian gallery and surrounding area (Bibby and others 2000). All birds observed or heard were recorded to species over a 10-min period at each point. Records started as soon as observer got into the site, birds leaving the area where recorded as well as birds entering. The distance between the bird and the observer was estimated when the distance was inferior to riparian gallery width. This procedure allowed verification of observer efficacy so that bias was not introduced into the data set. Birds were also surveyed on the flood plain approximately 100 m perpendicular to each point in the riparian gallery.

Macrophyte inventories (INAG 2008c) were carried out along 100 m long reaches starting at the beginning of the 500 m RHS reach, estimating percentage cover of each species in the sample area (minimum percent cover included 0.1%). Species level determinations were made on site; specimens that could not be identified were taken to the Superior Agronomy Institute herbarium (Technical University of Lisbon) and bryophytes to the herbarium of the Botanical Garden (Science Faculty of Lisbon) for identification (epiphytic bryophytes above the splash/humid zone were not included in the surveys).

Data Analyses

Redundant environmental and pressure parameters were removed using the Spearman Rank Correlation analysis method (Feld and Hering 2007; Hughes and others 2009). If two environmental or pressure variables under comparison were highly correlated (threshold value of r ≥ 0.5 or r ≤ −0.5) the variable with the higher mean correlation coefficient was excluded from further analysis.

Principal Components Analyses (PCA) were carried out on retained log (x + 1) transformed environmental and pressure data (binary data and ranked variables were not transformed) to reduce data dimensionality and identify the principal environmental and pressure gradients. From an initial list of 38 environmental variables and 36 pressure variables, a total of 20 environmental and 12 pressure variables were retained (Table 1). The loading value of retained variables (see Table 2) provided an indication of the proportion of their variance with a given PCA component and thus its contribution to distribution patterns in the ordination space. Taxonomic data for each organism group were log (x + 1) transformed (singly occurring specimens and those occurring at fewer than three sampling sites excluded from further analyses) were analyzed via Detrended Correspondence Analysis (DCA) to assess the biological turnover within each data set. Gradient lengths obtained for all organism groups (<3.0 SD units) indicated the use of a linear model, therefore redundancy analysis (RDA) and partial RDA were carried out on each organism group. RDA, a direct gradient analysis, assumes that canonical ordination axes are linear combinations of environmental or pressure variables and provides an eigenvalue, an estimate of variance explained by the “species” and variables used in the analysis.
Table 1

Environmental (E) and pressure (P) variables divided over three spatial scales (basin, reach and habitat) retained for subsequent statistical analyses after analysis for redundancy by comparing average Spearman Correlation Coefficients

Variable and spatial scale

Unit/expression

Data source

Basin

 E

1

Channel form

(CHAN_F)

1 = sinuate, 2 = constrained

GIS

 E

2

Valley form

(VALFRM)

Class 0–4

GIS

 P

1

Urban Area

(URB_A)

% Catchment area

GIS

 P

2

Monocultures

(MONO_A)

% Catchment area

GIS

 P

3

Agriculture

(AGRI_A)

% Catchment area

GIS

Reach

 E

3

Altitude

(ALT)

m.a.s.l

GIS

 E

4

Number of bars

(BARS)

Count

RHS data

 E

5

Average riparian width

(WDTRIP)

Metres

RHS data

 E

6

Subchannels

(SUBCH)

Count

RHS data

 E

7

Land use natural/semi-natural

(LU250_NAT)

Dummy variable

GIS 250 m bankside buffer

 E

8

Land use scrub

(LU250_SCR)

Dummy variable

GIS 250 m bankside buffer

 E

9

Average bank top height

(AVBKTP)

Metres

RHS data

 P

1

Bank reinforced

(BK_RI)

Dummy variable

RHS data

 P

2

Bank embanked

(BK_EM)

Dummy variable

RHS data

 P

3

Land use agriculture

(LU250_AGR)

Dummy variable

GIS 250 m bankside buffer

 P

4

Presence of Ford

(FORD)

Presence/absence

RHS data

 P

5

Tipped debris

(TIP_D)

Presence/absence

RHS Data

 P

6

Land use urban

(LU25 0_UR)

Dummy variable

GIS 250 m bankside buffer

 P

7

Organic point discharge

(ORGP)

Count

GIS

Habitat

 E

10

pH

(PH)

Sorensen scale

Fish sampling site

 E

11

Dissolved oxygen

(DO)

mg/l

Fish sampling site

 E

12

Water velocity

(WVEL)

m s−1

Macroinvertebrate sampling site

 E

13

Conductivity

(COND)

μS/cm

Macroinvertebrate sampling site

 E

14

Boulder/stone substrate

(BOLSTONB)

Visual estimate

Macroinvertebrate sampling site

 E

15

Boulder/stone substrate

(BOLSTONF)

Visual estimate

Fish sampling site

 E

16

Depth

(DEPTB)

Metres

Macroinvertebrate sampling site

 E

17

Depth

(DEPTF)

Metres

Fish sampling site

 E

18

Water temperature

(W_TEMPF)

° Celsius

Fish sampling site

 E

19

Sand/silt substrate

(SASIC)

Visual estimate

Fish sampling site

 E

20

Gravel substrate

(GRAV)

Visual estimate

Fish sampling site

 P

8

Bank modification

(BNK_MOD)

Visual estimate

RHS/sample site

 P

9

Channel modification

(CH_MOD)

Presence of features such as culverts, weirs or sluices

RHS/sample site

 P

10

Banktop landuse agriculture

(BT_AG10)

Class

10 m bankside buffer

 P

11

Banktop landuse pasture

(BT_RP10)

Class

10 m bankside buffer

 P

12

Banktop landuse forestry

(BT_FR10)

Class

10 m bankside buffer

Table 2

PCA eigenvalues and loadings (axes 1–3) for retained environmental and pressure variables divided over three spatial scales of habitat (H), reach (R) and basin

 

Axis 1

Axis 2

Axis 3

Environmental variables

 Eigenvalue

 

0.358

0.264

0.117

 BARS

R

−0.385

−0.108

0.779

 WDTRIP

R

0.103

−0.449

0.599

 LU250_NA

R

−0.041

0.506

−0.311

 WVELB

H

0.72

0.683

0.058

 DEPTB

H

0.467

0.083

−0.574

 pHF

H

0.814

0.537

−0.199

 DOF

H

0.512

−0.032

−0.173

 BOLSTONF

H

−0.236

−0.006

0.504

Pressure variables

 Eigenvalue

 

0.271

0.182

0.117

 BK_RI

H

0.629

−0.12

0.314

 BK_EM

H

0.64

0.582

0.002

 LU250_UR

R

0.288

−0.067

0.767

 LU250_AG

R

0.561

−0.315

0.122

 BNK_MOD

R

0.79

0.2

−0.454

 BT_AG10

H

0.428

0.459

0.575

 BT_RP10

H

0.532

0.73

−0.071

The loadings of variables with values ≥ 0.5 are given in bold text. Acronyms are defined in Table 1

A forward selection, (cut off point of P > 0.1) was used to retain non-redundant subsets of environmental and pressure variables to explain taxonomic distribution patterns (Aguiar and Ferreira 2005). A Monte-Carlo permutation test (999 permutations) was run on the first axis eigenvalue and the “trace” (sum of all canonical eigenvalues) to test the significance of the environmental and pressure effects under analysis.

Partial Redundancy Analyses (Liu 1997) was used to determine decomposition of variance between environmental and pressure variables. Sets of forward selected environmental or pressure variables were used as covariables to discern the distribution of unique, shared and unexplained variance for each biological assemblage. All analyses were carried out with the software packages PRIMER (Version 6.1.8), STATISTICA (version 6), and CANOCO (version 4.5 for Windows).

Both Procrustean analysis (PROTEST software package - PROcrustean randomization Test: http://www.zoo.utoronto.ca/jackson/pro1.html) and the RELATE software package (PRIMER) were used to evaluate the degree of concordance between each biological assemblage and environmental and pressure data. Procrustean analyses superimpose, scale and rotate one data matrix upon the other until an optimal fit is found (Heino and others 2004; Jackson and Harvey 1993; Peres-Neto and Jackson 2001). Estimated residuals between original values and the derived best fit solution give the m2 statistic. A low m2 statistic indicates a good level of correspondence between data matrices (Paavola and others 2006).We used the PROTEST analysis package to assess the degree of concordance between the sample site scores of the first 3 axes of the PCAs run for hierarchically organized environmental and pressure matrices and the sample site scores derived from the first 3 axes of Canonical analyses (CA) run for each biological assemblage. The PROTEST permutation procedure (999 permutations) was used to assess the statistical significance of the Procrustean fit between the two matrices (Paavola and others 2006; Peres-Neto and Jackson 2001).

The Mantel test (the RELATE procedure in the PRIMER software package) was used to compare similarity matrices for each biological assemblage (Bray Curtis similarity) with similarity matrices of environmental and pressure data (Euclidean distance). Significance was assessed using a permutation procedure (999 permutations) applied to Spearman ranked transformed data. Peres-Neto and Jackson (2001) suggest that Procrustean analysis is more effective than the Mantel test for assessing concordance, since the former can be used on raw data or derived ordination solutions while the latter is based upon measures of distance or similarity.

Results

Environmental and Pressure Gradients

For clarity, only variables with loadings ≥0.5 on axes 1, 2 or 3 are described (Feld and Hering 2007). The environmental PCA had higher overall levels of explained variance compared to the pressure PCA, illustrated by the eigenvalues in particular of the first two axes (Table 2). The variables with the highest loadings along axis 1 of the environmental PCA were water velocity (WVELB), pH (PHF) and dissolved oxygen (DOF). The strong negative loadings of WVELB and PHF indicated the strong separation of sites upstream and downstream of the cofferdam. The highest, principally negative, loadings along axes 1 and 2 were water velocity (WVEL_B, −0.72 and −0.683, respectively) and pH (PH, −0.814 and 0.537, respectively). Habitats in the middle and upper reaches were characterised by higher water velocity levels while lower lying sites had higher pH, temperature and dissolved oxygen levels (DO_F). The higher downstream dissolved oxygen and pH levels can be partly attributed to the considerable daytime photosynthetic activity of dense stands of macrophytes and filamentous algae recorded at these sites. The water velocity and pH gradients were also evident along axis 2; however the presence of reaches with natural land use (LU250_NA) characterised sites in the middle and upper sections of the Odelouca. Parameters with high loadings on axis 3 were related to habitat quality. Extremely long side bars (BARS; over 100 m in length), probably associated with the reduced flow and resulting deposition of substrates (loading 0.779) tended to be associated with downstream sites, however smaller discrete side bars and mid channel bars were recorded at upstream sites as well. Sites with greater average riparian width (WDTRIP) tended to occur at sites upstream of the cofferdam. These sites also tended to be deeper (DEPTB, loading −0.574) with coarser boulder stone substrates (BOLSTONF, loading −0.504).

Several pressure PCA parameters revealed the strong gradient of physical disturbance of the river habitat and adjacent changes in land use at several spatial scales along axis one, namely bank reinforcement (BK_RI, loading 0.629) and embankment (BK_EM, loading 0.640), agricultural landuse (LU250_AG, loading 0.561), bank modification (BK_MOD, loading 0.790) and pasture (BT_RP10 loading 0.532). Sites suffering from these impacts were in the lower reaches of the Odelouca, mostly below the cofferdam. On axis 2, bank embankment (BK_EM, loading 0.582), and pastoral land use (BT_RP10, loading −0.730) further separated these sites. Axis 3 loadings were related to urbanisation and agricultural land use changes (LU250_UR, LU250_AG).

Biological Assemblages

Organism abundance and diversity varied considerably among the different assemblages. Benthic macroinvertebrates were the most abundant organisms (total 40,346 individuals were sorted and identified) but the second most diverse group (34 species, 22 genera, 2 families, 1 sub family and one order were identified following taxonomic adjustment). The most abundant and frequently occurring order was the Diptera, in particular chironomids and simuliids.

Fish were the second most abundant (1,336 individuals) but least diverse assemblage (11 species were collected but only 8 species were subject to statistical analyses), reflecting the depauperate but endemic nature of Mediterranean fish assemblages. The most abundant species was the introduced alien, Gambusia holbrooki Girard. Anguilla anguilla L., was the most widely distributed fish species, occurring at 79% of the sample sites.

A total of 292 birds were surveyed represented by36 species (27 species were used in statistical analyses). The most frequent and abundant species was Sylvia atricapilla L, recorded at 90% of the sampling sites.

Plant biomass was harder to estimate due to the percentage cover estimate applied. However, a total of 71 species level taxa were described. Most frequently occurring species included Mentha suaveolens Ehrh Oenanthe crocata L, Rubusulmifolius Scott (all three species were recorded at 83.3% of sample sites) and Salix salviifolia Brot. (recorded at 90% of sampling sites)

RDA Results

Extracted first axes and the trace for RDA ordinations were highly statistically significant for all assemblages (Monte Carlo test 999 permutations; Table 3). The highest levels of variance described by the axis 1 eigenvalues were for the invertebrate and bird biological assemblages. Strong environmental/pressure gradients were evident along axes 1 and 2 of the macroinvertebrate RDA (Fig. 2a). The longitudinal pH (intra-set correlation r = −0.89) and altitude gradient (intra-set correlation r = 0.59) along axis 1 separated more disturbed downstream sites with higher pH values and macrophyte growth from less disturbed upstream sites. The axis two gradient separated faster flowing upstream sites (WVEL, intra-set correlation r = −0.45) from sites in agricultural areas (LU250_AGR intra-set correlation −0.87). More abundant taxa found in the lower reaches (higher pH axis 1), included Baëtis sp, Chironomidae Dugesia sp, Gyraulus sp and Simulium (Simulium) sp. Taxa occurring more frequently at the other end of this gradient were Capnioneura mitis Despax tribe Tanytarsini, and Chironomus (Chironomus) plumosus-Gr. Taxa strongly associated with surrounding agricultural land use (axis 2) were Stylaria lacustris L. (r = −0.82 P < 0.05) and Procloeon sp (r = −0.58 P < 0.05) while taxa associated with increased water velocity and altitude at the other end of axis 2 were Dicronata (Dicronata) sp, Atherix sp and Onchychogomphus forcipatus.
Table 3

Results of RDA and Monte Carlo permutations (F statistic, 999 permutations) for testing the significance of environmental and pressure variables in relation to distribution patterns of benthic macroinvertebrates, fish, birds and macrophytes

 

1st canonical axis

Sum of all canonical axes (trace)

Eigenvalue

F ratio 1st axis

P 1st axis

Eigenvalue

F

P

Axis 1

Axis 2

Axis 3

Invertebrates

0.215

0.067

0.038

6.867

0.001***

0.353

3.406

0.001***

Cumulative percentage variance of species-environment data

61.1

80.0

90.9

     

Fish

0.199

0.179

0.064

5.731

0.009**

0.466

4.018

0.001***

Cumulative percentage variance of species-environment data

42.8

81.2

95.0

     

Birds

0.166

0.064

0.039

4.971

0.001***

0.284

2.476

0.001***

Cumulative percentage variance of species-environment data

58.4

81.1

94.7

     

Macrophytes

0.157

0.077

0.049

4.665

0.001***

0.303

2.712

0.001***

Cumulative percentage variance of species-environment data

52.0

77.6

93.6

     
https://static-content.springer.com/image/art%3A10.1007%2Fs00267-010-9521-3/MediaObjects/267_2010_9521_Fig2_HTML.gif
Fig. 2

RDA ordinations of environmental and pressure variables and a macroinvertebrate taxa, b fish taxa, c bird taxa and d macrophytes. For clarity, only taxa explaining more than 15% explained variance are shown. Biplots in the left hand column describe environmental and pressure variable distribution patterns while biplots in the right hand column describe taxa distribution patterns in the ordination space. Full species names and corresponding codes are given in Appendix 1

Axis 1 of the fish taxa ordination plot also described a similar pH gradient (Fig. 2b; intraset correlation −0.828) but also a habitat complexity gradient (intraset correlation ALT = 0.63, BARS = 0.21). A shorter gradient related to channel modification and riparian gallery width was evident along axis 2 (intraset correlation CH MOD = −0.23, WDTRIP = −0.29). Threatened native species S. aradensis and C. almacai were clearly associated with upstream sites with greater in-channel habitat diversity and negatively associated with downstream sites suffering modification, including reduced flow and the physical barrier of the cofferdam. Anguilla anguilla L. was more abundant at lower lying sites but widespread in the study area. Both alien (Gambusia holbrooki Girard and Lepomis gibossus L.) and native species (Atherina boyeri Risso, Barbus sclateri Günther and Cobitis paludica de Buen) were associated with channel modification and riparian gallery width. The presence of the native species B. sclateri with alien species reflects its preference for the pool-like conditions (Magalhães and others 2002), that predominate in the immediate area of the cofferdam. G. holbrooki, comprising just over 50% of the total catch, occurred exclusively at sites with reduced flow close to the cofferdam but where riparian galleries were still present.

Distinct bird species groups occurred in the ordination space (Fig. 2c). A longitudinal gradient from higher lying to lower lying degraded sites was discernible (axis 1 intraset correlations pH r = 0.75, BK_RI r = 0.79; axis 2 intraset correlation ALT r = 0.63, LU250_AG r = 0.58). Species clearly associated with bankside disturbance (reinforcement) and reduced riparian cover at lower lying sites were Ardea cinerea L, Egretta arzetta L. and Passer domesticus L. Woodland species such as Phylloscopus brehmii von Homeyer, Fringilla coelobs L, Parus major L, and Troglodytes troglodytes L. were negatively associated with areas affected by agriculture. A third distinct group of predominantly woodland/scrub species was also associated with less impacted upstream sites. Species significantly correlated with axis 2 of the plot were Luscinia megarhynchos Brehm Oriolusoriolus L. and Sylvia melanocephala Gmelin.

Distinct plant species distribution patterns were related to longitudinal hydric regime patterns and habitat degradation resulting from changes in land use (Fig. 2d). The pH gradient (pH axis 1 intraset correlation r = 0.86, axis 2 intraset correlation r = −0.42), separated lower lying sites on the wider valley floor in agricultural areas from other main channel sites (LU250_AG axis 1 intraset correlation r = 0.34, axis 2 intraset correlation r = −0.58) and also narrow tributary sites with monocultures present in the area beyond the riparian zone (MONO_A axis 1 intraset correlation r = 0.58, axis 2 intraset correlation r = 0.58). Woody riparian species, tolerant of wetter conditions and occasional flooding, such as Alnus glutinosa (L.) Gaertner, Fraxinus alnus P. Mill, Viola riviniana (Rchb.) and Prunella vulgaris L. subspecies vulgaris, were associated with tributaries. Main channel sites affected by agriculture (bottom right hand side of the ordination biplot) were occupied by hardy woody shrubs tolerant of both aridity and disturbance such as Tamerix africana Poiret, Fraxinus angustifolia Vahl and Juncus bulbosus L. Degraded lower lying sites were characterised by the in-channel presence of Apium nodiflorum (L.), Lemna minor L., Bidens frondosa L. and Cyperus eragrostis Lam on the bankside. Species at less disturbed sites further upstream on the main channel included woody shrubs such as Rubus ulmifolius Scott, Crataegus monogyna Jacq and Festuca arundinacea Schreber.

Partial RDA

Distinct patterns of decomposition of variance by pRDA (Fig. 3) occurred across the biological assemblages despite the universally high levels of unexplained variance (ranging from 53.4% for fish assemblages to 71.6% for birds) and very low levels of shared variance (environmental and pressure combined) for all groups (the highest value of just 3.9% attributable to bird assemblages). Levels of variance uniquely attributable to environmental parameters were consistently higher than those uniquely attributable to pressure variables for all biotic assemblages. This was particularly notable for fish assemblages (over 40% variance derived from the forward selected variables BARS, WDTRIP, PH and ALT, followed by the benthic macroinvertebrates (29.2%, WVEL, PH and ALT), macrophytes (17.7%, WVEL and PH) and birds (17.4%, variables PH and ALT). The pRDA forward selection results highlight the high level of redundancy between all assemblages for the parameters pH and altitude.
https://static-content.springer.com/image/art%3A10.1007%2Fs00267-010-9521-3/MediaObjects/267_2010_9521_Fig3_HTML.gif
Fig. 3

Decomposition of variance across the biological assemblages in relation to forward selected environmental and pressure variables

The highest levels of variance attributed uniquely to pressure variables were for bird (12.5%) and macrophyte (11.8%) assemblages. Groups of retained pressure variables for these two assemblages were associated with agricultural changes in land use and physical disturbance of the banks (birds). Levels of variance derived purely from single pressure variables for each aquatic faunal assemblage were markedly lower and related to agricultural changes in land use (macroinvertebrates 5.4%) and physical impacts on the river channel (presence of culverts, dams, weirs or sluices, fish 6.5%).

Patterns of concordance for biological assemblages and environmental or pressure data were similar for both Procrustean & RELATE analyses (Table 4). Highly significant levels of concordance occurred between macroinvertebrate assemblages and environmental parameters and bird assemblages and pressure parameters. Different levels of statistical significance of concordance occurred between fish assemblages and environmental parameters, while weak levels of statistical significance were detected between macrophytes and environmental parameters by RELATE but not by Procrustean analyses. No statistically significant levels of concordance were detected between macroinvertebrates, fish and macrophyte assemblages and pressure variables or bird assemblages and environmental variables.
Table 4

Results of concordance analyses between biological assemblages and environmental and pressure data using RELATE (Mantel Test) and Procrustean analyses

 

RELATE

Procrustean

P value

Rho

P value

RSS

M2

Invertebrates × environment

0.558

0.001***

0.8509

0.6699

0.001***

Invertebrates × pressure

0.03

0.354 ns

1.2839

0.8718

0.0791 ns

Fish × environment

0.178

0.018*

1.1308

0.8111

0.005**

Fish × pressure

−0.058

0.746 ns

1.2679

0.866

0.0661 ns

Birds × environment

0.122

0.137 ns

1.2527

0.8604

0.0611 ns

Birds × pressure

0.419

0.001***

0.984

0.7419

0.001***

Macrophytes × environment

0.214

0.018*

1.4416

0.922

0.4354 ns

Macrophytes × pressure

0.084

0.205 ns

1.5142

0.941

0.6256 ns

The RELATE procedure compared similarity matrices while Procrustean analyses compared sample site values from the 1st three axes of Canonical Analyses (biological assemblages) and PCA (environmental and pressure variables)

RSS = residual sum of squares

Discussion

The results of this study clearly highlight the importance of identifying environmental and pressure gradients acting upon river systems that will be subject to restoration, as well as the importance of a priori studies on biological assemblage response patterns and the implementation of typologically appropriate restoration measures.

PCA results showed that higher levels of variance were attributed to environmental parameters that described essentially longitudinal changes in the riverine environment. Benchmark sites in the mid section of the Odelouca, upstream of the cofferdam were characterised by mature riparian galleries and higher habitat diversity. Retained pressure variables indicated physical impacts to the riverbank such resectioning and reprofiling in areas dedicated to agriculture and, to a lesser extent, alterations to the channel itself. These impacts are in accordance with the observations of Hooke (2006) and Aguiar and Ferreira (2005) who state that principal human impacts in Mediterranean systems over the last century are mainly related to changes in land use and alterations in water and channel management. Hooke (2006) mentions the very long history of human impacts on Mediterranean systems. The fact that these systems are largely altered and have been so for such a long period of time may also explain the weak pressure gradient and poor association with biological assemblages that was detected in this study, i.e. most of the system has been disturbed with little variation for an extended period of time (see Harding and others 2006).

However, the apparently strong environmental gradient is in fact a direct result of the innate covariance that exists between environmental and pressure variables (Allan 2004) that operate in European Mediterranean river systems (Díaz and others 2008; Hooke 2006). The longitudinal environmental gradient is intimately linked with basin orography and relief which in turn influence will patterns of human access and subsequent intervention. Large areas dedicated to agriculture occurred in the lower reaches of the Odelouca where the flood plain was considerably flatter and wider. Agricultural and urban areas also occurred at some sites upstream of the narrower, less accessible mid section where habitat disturbance was less pronounced and more natural conditions prevailed. The recorded instream habitat diversity is a direct result of the cyclic natural disturbance pattern of seasonal flood and drought typical of Mediterranean systems (Gasith and Resh1999) to which many Mediterranean biological assemblages are adapted (Bonada and others 2007; Magalhães and others 2002; Díaz and others 2008), a fundamental concept of the Habitat Templet Theory (Southwood 1977; Southwood 1988) and river templet theory (Townsend and Hildrew 1994). Bêche and others 2009 have shown that both invertebrate assemblage composition and native fish abundance levels are strongly affected by prolonged drought which influence flow in Mediterranean streams of California. These conditions also facilitated the establishment of an invasive fish species. The natural Mediterranean flow related disturbance pattern has been drastically altered at sites downstream of the cofferdam, which also lie in the wider areas of the catchment where levels of human intervention were found to be much higher. The cofferdam is also a physical barrier to fish migration. Thus, the effects of long term human intervention and the natural pattern of flow related disturbance characteristic of Mediterranean systems, result in complex gradients that are not as distinct as would initially appear. The “ideal” biological response to pressures should be type specific and distinct from responses to environmental gradients (Johnson and others 2006a). Our results show that this is not the case in selecting indicator species for typologically relevant restoration measures.

Multivariate RDA results indicated significant relationships between the species data and the explanatory variables for all biological assemblages and allowed the identification of key indicator species. However pRDA results revealed low levels of explained variance across all assemblages, indicating that variables that could have better explained biological distribution patterns, in particular pressures, were not included in this study. Also, the separation of variables into groups of either environmental variables or pressure variables may also have affected results. For example, although classified as an environmental variable, pH is in fact a highly complex amalgam of changes in underlying natural biotic and abiotic gradients, changes in land use and enrichment (Hughes and others 2009; Townsend and others 1989). The longitudinal pH gradient detected in the Odelouca comprised a complex combination of background pH levels (catchment geology, biotic processes) and the result of changes in land use in the lower reaches due to human intervention. Other “environmental” variables such as the presence of bars (BARS) and riparian width (WDTRIP) can also be affected by changes in land use and water abstraction (i.e., very long side bars, formed by the reduced flow regime, were recorded downstream of the cofferdam). The lack of a measure of longitudinal connectivity, such as an estimate of distance from the cofferdam, may have improved the amount of variance described by pressures and the response of the biological assemblages, in particular the fish assemblage.

Results indicated that the variables we classified as “environmental” explained higher levels of variance than pressure variables across all groups, but that the associations were relatively weak, reflected in pRDA, RELATE and Procrustean results. The results of the aforementioned analyses were slightly stronger for macroinvertebrate and fish assemblages, giving information on important selected environmental descriptors of habitat quality that must be taken into account for typologically suitable restoration efforts, such as water velocity (macroinvertebrates), the presence of deposition bars and riparian galleries (fish). Further, Hughes and others (2008) found that the Habitat Quality Assessment Index (HQA—a measure of the structural diversity of natural features of wildlife interest along the river corridor, derived from RHS data), numbers of macroinvertebrates shredders and the EPT metric (Ephemeroptera Plecopter Trichoptera metric) were all significantly higher at benchmark sites, compared to sites below the cofferdam. Other studies on macroinvertebrate assemblages in Mediterranean systems clearly indicated the importance of flow patterns on macroinvertebrate traits and distribution patterns (Bêche and Resh 2007a; Bonada and others 2007). Microhabitat studies of S. aradensis and I. almacai revealed that each species occupied distinct microhabitats that varied with season and age (Santos and Ferreira 2008). I. almacai preferred sheltered habitats with fine sediments in the autumn while younger nase occupied more exposed areas with coarser substrata than adult nase. S. aradensis preferred faster flowing habitats with coarser substrata in the spring compared to the rest of the year whereas younger chub tended to occupy slower flowing areas with finer sediments. These distribution patterns are intimately linked with seasonal flow patterns and associated processes of erosion and deposition.

The distinct distribution patterns across the fish assemblage highlighted the impact of the cofferdam on longitudinal connectivity, a factor that must be taken into account for effective restoration measures. S. aradensis and I. almacai were confined to the reaches of greater habitat quality, the eel A. anguilla was more abundant below the cofferdam and alien species occurred in high numbers at degraded sites affected by the structure of the cofferdam. These findings strongly contrast with surveys carried out in the Odelouca prior to cofferdam construction where S. aradensis and I. almacai were the most abundant and widespread species and the presence of alien species was negligible (Pires and others 2004). The predominance of alien species in the immediate vicinity of dams has also been recorded in “Mediterranean-type” river systems in California (Kondolf 1998) and in rivers suffering major flow alterations in the southwestern United States (Hughes and others 2005). Clearly the lack of a pressure variable quantifying disturbance in longitudinal connectivity affected results. According to Hooke (2006), connectivity within the different parts of Mediterranean river systems is a major driver of the extent and transmission of longitudinal changes.

Both the pRDA and RDA results clearly indicated a strong association between biological distribution patterns and the longitudinal pH gradient. Studies on environmental variables and invertebrates, fish and bryophyte assemblages in Finnish boreal streams (Paavola and others 2006) and physicochemical determinants of macroinvertebrate distribution in UK (Townsend and others 1989) streams have also identified the important role of pH. Studies tend to regard pH as a “natural” environmental variable when in fact it is an amalgam of both natural and anthropogenic factors. Background pH levels are attributable to catchment geology and biotic processes, but true pH levels are affected by multiple processes such as industrial airborne emissions, changes in land use and enrichment from agriculture and urbanization. This reveals not only the extremely complex nature of the longitudinal pH gradient in the lotic environment, but the inherent source of covariance contained within this parameter. Extremely dense stands of aquatic macrophytes and filamentous algae occurred at many of the lower lying sites in the Odelouca due to changes in land use (agriculture and associated bankside riparian clearance), reduced flow and increased sedimentation (reduced connectivity) and the natural longitudinal gradient (natural accretion, increased insolation due to the increased bankfull width). The higher pH levels recorded at these sites resulted from considerable photosynthetic activity as a result of the complex interplay of these factors.

Bird assemblages were most readily associated with the physical impacts related to land use changes. Similar responses have also been recorded in the bird assemblages on the Sado, a Mediterranean river system situated approximately 140 km north of the Odelouca (Moreira and others 1997). The results for the avifauna clearly highlight their increasingly recognized potential for assessing impacts across the wider river environment (Bryce and others 2002; Jansen and Robertson 2001b; Vaughn and others 2007). Explained variance attributed solely to pressure variables was highest in the bird assemblages, followed by the macrophytes. However, RELATE and Procrustean results were highly significant for the former assemblage but not at all significant for the latter. The results obtained for plant assemblages may be due again to the unintentional omission of descriptors of longitudinal connectivity that may have influenced distribution patterns.

The fundamental, multifunctional role of riparian vegetation in river systems is well known, ranging from flood prevention and sediment retention to seed recruitment, provision of shelter, habitat heterogeneity and allochthonous input. The maintenance of riparian vegetation is recognized as vital to the integrity of river ecosystems, although these ecotones are highly sensitive to environmental change (Gregory and others 1991; Naiman and Décamps 1997). Iberian riparian systems have been subject intensive agriculture and forestry, damming, abstraction and urban development, resulting in degraded habitat integrity, reduced longitudinal connectivity and isolated patches of riparian galleries clearly compromising wide scale lotic function (Aguiar and Ferreira 2005; Ferreira and others 2005). Riparian vegetation in semi arid areas tends to exhibit high levels of longitudinal variability in composition, species richness and density due to the environmental conditions (i.e., the strong annual and interannual patterns of flood and drought that characterise Mediterranean rivers) and land use in the immediate flood plain (Aguiar and Ferreira 2005; Raven and others 2009). The RDA and pRDA results reveal that the Odelouca is no different (retained environmental variables were W_VEL and pH, pressure variables were MONO_A and LU250_AG) and restoration efforts must take into account the current situation, namely the isolated but complex riparian stands in the mid section of the river, the influence of large scale environmental and pressure factors and the association of key biological assemblages with this ecotone.

The results of this study reveal a Mediterranean system clearly suffering from the effects of large scale impacts, namely reduced connectivity and habitat fragmentation due to changes in land use and damming. The Odelouca exhibits symptoms typical of river regulation. There is obvious reduced connectivity between upstream and downstream sites. Downstream sites are starved of sediment and variation in flow amplitude, leading to habitat fragmentation, impoverished habitat diversity, and disturbed lateral and vertical connectivity. System function and native biodiversity have been negatively affected and alien species have proliferated in areas of environmental degradation. These impacts are further compounded by changes in land use, resulting in physical alteration of the banks and channels and riparian clearance, further compromising lateral flood plain connectivity. Kondolf (1998) emphasized the importance of the study of both geomorphological and ecological processes during the whole restoration process. RHS and some SIG data provided an overview of geomorphologic processes, while field sampling of both aquatic and terrestrial assemblages prior to the restoration project provided us with information on patterns of changes in relation to environmental quality.

A decade ago Karr (1999) wrote that the “return of fish” was often listed as one of the principal biological goals of restoration. However, the restoration concept embraces a far broader and intricate picture that must take into account the spatial and temporal complexity of river systems and the multidisciplinarity approach necessary for the implementation of successful, typologically appropriate restoration measures. Ormerod (2004) fittingly states that “key species, the communities of which they are a part, and the ecological functions they provide must be the ultimate arbiters of restoration success”. This study has shown, in the case of the Odelouca, that bird species can be used as indicators to detect bankside environmental degradation due to physical impacts and that distinct assemblages (comprising mainly woodland birds) occur at benchmark sites, where riparian integrity is good and instream habitat diversity is greater. Woody riparian communities are vital for increasing habitat diversity and providing refugia for faunal elements. Detailed phytosociological studies of these riparian communities should be made to contribute to appropriate planting initiatives during restoration. Selected benthic macroinvertebrate taxa and endemic fish species are important indicators of instream habitat quality, a direct consequence of less disturbed Mediterranean flow regime. All of these biological assemblages should be monitored during and after restoration and results can be presented in the form of metrics, e.g., number of macroinvertebrate shredder species, invertebrate EPT taxa, number of endemic fish species. Hughes and others (2009) found that metrics describing avian feeding or habitat preferences successfully separated bird species found in benchmark sites from degraded sites.

Palmer and others (2010) point out the importance of large scale factors for ensuring restoration success. The large scale impacts affecting the Odelouca require accordingly large scale but typologically relevant measures to restore connectivity and abiotic and biotic function once the dam is operational. Examples of measures include release of peak flow patterns consistent with Mediterranean flow patterns to re-establish connectivity, guarantee representative baseflow and re-establish typological patterns of erosion and deposition, thereby providing habitat diversity and refugia suitable for native fish species. Waterborne seed recruitment and propagules of riparian plant species (Dixon 2003), which will contribute to natural reestablishment of native riparian species, will also be intimately linked to restored flow patterns. The construction of a fish passage specifically designed to meet the needs of the species in question is fundamental to the success of restoration measures. Longitudinal connectivity is a major restoration goal, but lateral connectivity must also be considered to restore riparian function (Lake and others 2007). Eradication programmes of alien and invasive plant species such as Arundo donax L., which blights many degraded river banks across the Algarve region (Raven and others 2009), will be essential, followed by planting of native riparian species, grown from seed or propagated from the regional species pool. An equally essential part of the restoration process will be raising public and institutional awareness of the benefits of natural restoration measures on biodiversity.

Acknowledgments

This study was carried out as part of a postdoctoral study (FCT grant number SFRH/BPD/26909/2006) funded by the Fundação para a Ciência e a Tecnologia of the Ministerio da Ciência, Tecnologia e Ensino Superior, Portugal. Many thanks to Professor Don Jackson, University of Toronto, Ontario, Canada for advice on statistical procedure and how to use the PROTEST software. Many thanks to Luis Lopes, Rute Caraça and David Santos for carrying out work in the field. Comments from two anonymous referees and the editor greatly helped the revision of the original manuscript. Many thanks also to the editor of Environmental Management for help concerning manuscript submission.

Copyright information

© Springer Science+Business Media, LLC 2010