Hydrobiologia

, Volume 719, Issue 1, pp 483–507

Ecological status assessment in mediterranean rivers: complexities and challenges in developing tools for assessing ecological status and defining reference conditions

Authors

    • Nelson Mandela Metropolitan University
MEDITERRANEAN CLIMATE STREAMS Review Paper

DOI: 10.1007/s10750-012-1305-8

Cite this article as:
Dallas, H.F. Hydrobiologia (2013) 719: 483. doi:10.1007/s10750-012-1305-8

Abstract

Rivers in mediterranean regions are subject to hydrological extremes. They range from highly stable, perennial ground- or snow-fed systems to highly ephemeral, unpredictable ones in semi-arid environments. Spatial and temporal complexity inherent in these systems presents challenges for ecological status assessment and defining reference conditions, particularly as many areas have been extensively transformed through anthropogenic activities. Temporal variability driven by sequential and predictable, seasonal events of flooding and drying accentuates the need to take season and/or hydrological period into account. Intermittent streams, which are common in mediterranean regions (med-regions) and which have aquatic communities distinct from perennial streams, are often not incorporated in bioassessment and present distinct challenges. Med-regions are also known for their high biodiversity and rates of endemism, as well as large numbers of introduced species. Med-regions are expected to be among the most affected by global climate change and, in these systems, climate change is an additional driver influencing ecosystems that are already stressed. From this review it is evident that an understanding of responses of indices, metrics, and models to climate change in comparison to existing stresses, and the development of thermally specific bioassessment tools are needed for this region.

Keywords

BioassessmentIndicesMetricsPeriphytonInvertebrateFishRiparian

Introduction

Assessment of ecological status involves an evaluation of the effect of anthropogenic activity on a natural system, such as a specific river or stream. Ecological status is determined through bioassessment, using one or more biotic components, together with environmental factors that directly or indirectly affect spatial and temporal variation in the biota (Stoddard et al., 2006). It assumes that responses, condition, and/or community integrity of biota can be used to assess the ecological integrity of an ecosystem (Ollis et al., 2006), and it commonly relies on the concept of comparing current condition to expected condition, where “expected” represents a condition at a pre-defined time, such as the reference or natural condition in the absence of human disturbance or alteration (Stoddard et al., 2006). Critical to the efficacy of ecological status assessment in rivers is the ability to characterize biological assemblages at reference sites. Multivariate (e.g., Wright et al., 1993; Reynoldson et al., 1995; Smith et al., 1999) and multimetric (e.g., Karr, 1981; Rehn et al., 2005; Munné & Prat, 2009) techniques are commonly used to characterize the reference conditions, while variability at reference sites is often modeled (e.g., Wright et al., 1993; Chessman et al., 1999; Simpson & Norris, 2000; Linke et al., 2005; Feio et al., 2007a, b). Such modeling allows for the prediction of benchmarks, with recent research favoring site-specific modeling approaches as a way to improve both accuracy and precision of predictions (Hawkins et al., 2010). The process of estimating and adjusting for differences in reference site quality is also deemed important (Cao & Hawkins, 2010) to facilitate comparison of bioassessment results across regions. Numerous papers have addressed the topic of bioassessment (e.g., Bonada et al., 2006a) and reference conditions (e.g., Stoddard et al., 2006), and a recent review (Hawkins et al., 2010) provides an extensive and insightful overview of the reference conditions for ecological and water-quality assessments.

Streams and rivers in mediterranean regions (med-rivers) are physically, chemically, and biologically controlled by sequential, predictable, seasonal events of flooding and drying over an annual cycle (Gasith & Resh, 1999). Med-regions are environmentally complex landscapes with a range of stressor types (Smith et al., 1999; Boulton et al., 1999; Chessman & Royal, 2004; Figueroa et al., 2007; Stewart, 2011). Derivation of reference conditions in these regions is thus particularly challenging if the intrinsic spatial and temporal heterogeneity are to be taken into account, and selection of reference sites is thus a critical step in successful implementation of bioassessment programs. Many med-regions, particularly lowland areas, have been substantially modified by human activities (Dallas, 2000; Chessman & Royal, 2004; Figueroa et al., 2007; Sánchez-Montoya et al., 2009a; Mazor et al., 2011; Munné & Prat, 2011). The challenge lies in identifying adequate numbers of reference sites within a river type, which ensures the incorporation of intrinsic variability of environmental conditions within the river type. Reference site selection becomes an exercise in balancing Type I error (the risk of keeping disturbed sites in the reference pool) and Type II error (unnecessarily rejecting minimally disturbed sites from under-represented river types) (pers. comm., P. Ode, Aquatic Bioassessment Laboratory, Water Pollution Control Laboratory, California Department of Fish and Game, California).

Several authors have described criteria for selecting reference sites in med-regions (e.g., Dallas, 2000; Stoddard et al., 2005; Chaves et al., 2006; Sánchez-Montoya et al., 2009a; Munné & Prat, 2011; pers. comm., P. Ode) and criteria commonly used include hydromorphological, physico-chemical, and biological factors acting at catchment, reach, and site scales. In their review, Sánchez-Montoya et al. (2009a) proposed 20 Mediterranean Reference Criteria (MRC) that accounted for characteristics within the med-rivers in Spain. These included criteria related to river morphology and habitat, hydrological conditions and regulation, point sources of pollution, diffuse sources of pollution and land use, riparian vegetation zone, and introduced species. Ode (pers. comm.) expanded on these reference screens and thresholds to accommodate ecological stressors known to be important to rivers in California. In their study, each site was characterized with a suite of landuse and landcover metrics for both its natural characteristics and for potential anthropogenic stressors. Sites were then screened with a subset of metrics using thresholds that represented low levels of anthropogenic stress. Fundamental to their approach was to ensure that intrinsic variability was accounted for and that reference sites within a region (or river type) captured important environmental gradients and were thus representative of existing conditions. The objectives of this paper are to provide an overview of the bioassessment tools developed for assessing condition in med-rivers, and to highlight the complexities and challenges of deriving reference conditions for med-rivers. The review has focused on the English-language literature, although other literature, which is directed to the interests and problems of specific audiences (e.g., Prat, 2002), is also available.

Bioassessment in med-regions

This section provides an overview of the biotic components and bioassessment methods currently applied in med-regions, namely in the Mediterranean Basin, United States (California), Chile (central), South Africa (south-western Cape), and Australia (the south-western and south) (Table 1). The body of literature in this field is enormous, and while this overview has attempted to include a range of publications, it is likely that some studies may have been inadvertently omitted or duplicative of information in other references cited.
Table 1

Bioassessment indices [based on four biotic components B: benthic algae/periphyton (including diatoms), M: macroinvertebrates, F: fish and V: vegetation] applied in med-regions [MB: Mediterranean Basin; C: California; Ch: central Chile; SA: South Africa (south-western Cape region) and A: Australia (south and south-western region)]

Biotic component

Med-region

Index

Index type

Reference(s)

B

MB

Diatom-based Eutrophication Pollution Index (EPI-D)

BI

Torrisi & Dell’Uomo (2006), Martín et al. (2010), Dell’Uomo & Torrisi (2011)

B

MB

Nutrients (TI-D), organic matter (SI-D) and mineral salts (HI-D)

BI

Dell’Uomo & Torrisi (2011)

B

MB

Specific Polluosensitivity Index (SPI), standardized Biological Diatom Index (BDI), European Economic Community Index (CEC) and Generic Diatom Index (GDI)

BI

Feio et al. (2009a)

B

MB

MoDi

PM

Feio et al. (2007b, 2009a)

B

C

Periphyton-IBI (Sierra Nevada)

MMI

Herbst & Blinn (2007)

B

C

Nutrient (MVI) and Watershed Disturbance (WD MVI)—(Western United States)

MMI

Stevenson et al. (2008)

B

C

Under development (Indicators based on diatom and soft-bodied algal assemblages, biomass based on chlorophyll-a and ash-free dry mass, and algal percent cover)

BI

Fetscher & McLaughlin (2008)

B

C

RIVPACS—Diatoms

PM

Ritz (2010)

B

SA

Under development (testing of several European indices; unique South African index to be developed)*

BI

Taylor et al. (2007a, b, c, d)

B

A

Diatom Species Index for Australian Rivers (DSIAR)*

BI

Chessman et al. (1999, 2007), Chessman & Townsend (2010)

M

MB

Iberian British Monitoring Working Party (IBMWP)

BI

Alba-Tercedor & Sánchez-Ortega (1988), Alba-Tercedor & Prat (1992), Rico et al. (1992), Alba-Tercedor & Pujante (2000)

M

MB

Hellenic Evaluation System

BI

Artemiadou & Lazaridou (2005)

M

MB

ICM-9, ICM-10, ICM11a

MMI

Munné & Prat (2009), Sánchez-Montoya et al. (2010)

M

MB

IM9

MMI

Pinto et al. (2004), Morais et al. (2004)

M

MB

Portuguese: National, North, South

PM

Feio et al. (2007a, b, 2009b)

M

MB

Indice Biotico Esteso (IBE)

BI

Buffagni et al. (2004)

M

MB

Intercalibration Common Metric index (ICMi)

MMI

Buffagni et al. (2004), Buffagni & Furse (2006)

M

MB

MEDiterranean Prediction And Classification System (MEDPACS)

PM

Poquet et al. (2009)

M

C

EMAP-West (per Ecoregion)*

MMI; PM

Stoddard et al. (2005)

M

C

Northern Coastal California Index of Biotic Integrity (NCIBI)

MMI

Rehn et al. (2005)

M

C

Southern Coastal California Index of Biotic Integrity (SCIBI)

MMI; PM

Ode et al. (2005, 2008)

M

C

Urban Index

MMI

Carter et al. (2009)

M

C

Eastern Sierra Macroinvertebrate Index of Biotic Integrity (ESM-IBI)

MMI

Herbst & Silldorff (2009)

M

C

California RIVPACS models

PM

P. Ode. (pers. comm.)

M

Ch

Indice biótico de familias para Chile (ChiBF)

 

Figueroa et al. (2003, 2005, 2007)

M

SA

South African Scoring System—version 5 (SASS5)

BI

Dallas (1997), Chutter (1998), Dickens & Graham (2002)

M

SA

Macroinvertebrate Response Assessment Index (MIRAI)

BI

Thirion (2007)

M

A

Stream Invertebrate Grade Number—Average Level (SIGNAL)

BI

Chessman (1995, 2003a, b)

M

A

Australian River Assessment System (AusRivAS)

PM

Smith et al. (1999), Simpson & Norris (2000)

F

MB

IBI

MMI

Magalhães et al. (2008)

F

MB

IBI-Jucar

MMI

Aparicio et al. (2011)

F

MB

Index of Community Integrity (ICI)

PM

Hermoso et al. (2010, 2011)

F

C

IBI—Sacramento Valley

MMI

May & Brown (2002)

F

C

Watershed IBI (W-IBI)

MMI

Moyle & Randall (1998)

F

C

EMAP-West (per Ecoregion)*

MMI

Stoddard et al. (2005)

F

SA

Fish Response Assessment Index (FRAI)—Northern SA*

MMI

Kleynhans (2007)

F

A

IBI—Southern Australia*

MMI

Harris & Silveira (1999)

F

A

PM—Eastern Australia*

PM

Kennard et al. (2006)

V

MB

Riparian Forest Quality Index (QBR)

MMI

Munné et al. (2003)

V

MB

Iberian Multimetric Plant Index (IMPI)

MMI

Ferreira et al. (2005)

V

MB

Riparian Vegetation Index (RVI)

MMI

Aguiar et al. (2009, 2011)

V

MB

MACrophyte Prediction And Classification System (MACPACS)

PM

Aguiar et al. (2011)

V

MB

Macrophyte Assessment and Classification (MAC)

PM

Aguiar et al. (2011)

V

C

California Rapid Assessment System for Wetlands (CRAM)

MMI

Collin et al. (2008), Stein et al. (2009), Solek et al. (2011)

V

SA

Riparian Vegetation Response Assessment Index (VEGRAI)*

MMI

Kleynhans et al. (2007)

V

A

Index of Stream Condition (ISC)—general*

MMI

Ladson et al. (1999)

V

A

Tropical Rapid Appraisal of Riparian Condition (TRARC)—Northern Australia*

MMI

Dixon et al. (2005)

The index type is given as biotic index (BI), multimetric index (MMI) or predictive model (PM). Relevant references are provided

* Indices or models not developed specifically for med-regions and which have not been tested within them

Benthic algae

Benthic algae (periphyton, including diatoms) is a useful indicator of water quality in streams and rivers and is a reliable tool for the detection and assessment of environmental degradation (Herbst & Blinn, 2007). The biomass and species composition of algae in rivers rapidly reflects stress resulting from both chemical pollutants (toxics, nutrients) and physical habitat disturbance (loss of structural diversity, bank erosion, sedimentation, elevated temperature), providing a biological measure of changing environmental quality (Herbst & Blinn, 2007).

Indices and predictive models based on diatoms have been successfully applied in the Mediterranean Basin (e.g., Feio et al., 2007a, b; Feio et al., 2009a; Martín et al., 2010; Dell’Uomo & Torrisi, 2011), with certain indices more or less accurately reflecting specific anthropogenic impacts. The Eutrophication/Pollution Index has proved to be reliable in med-rivers (Torrisi & Dell’Uomo, 2006; Martín et al., 2010; Dell’Uomo & Torrisi, 2011), as have three partial indices for assessing nutrients, organic matter, and mineral salts. All indices were highly correlated with water quality impairment as well as with other diatoms used widely in other European assessment programs (Dell’Uomo & Torrisi, 2011). Other indices tested (Feio et al., 2009a) include the Specific Polluosensitivity Index, standardized Biological Diatom Index, European Economic Community Index, and Generic Diatom Index, and a predictive model. Overall, the predictive model was considered most appropriate as it expresses both quantitative and qualitative changes in diatom communities and reflects a wider variety of impacts (Feio et al., 2009a). It also conforms to the reference condition approach as prescribed by the Water Framework Directive.

In California, Pan et al. (2006) examined the relationship between environmental variables and benthic diatom assemblages in the heavily impacted California Central Valley ecoregion, and showed that distributional patterns of benthic diatom assemblages were mainly affected by channel morphology, in-stream habitat, and riparian conditions. Herbst & Blinn (2007) developed a preliminary Periphyton-IBI in the Eastern Sierra Nevada, which related clearly to a composite habitat disturbance gradient. At a broader scale, Stevenson et al. (2008) evaluated the efficacy of two multivariate indices (Nutrient and Watershed Disturbance) in response to a stressor gradient in the western United States, where the gradient was defined by increases in conductivity, nutrient concentrations, and % fine sediments as % watershed disturbance. Fetscher & McLaughlin (2008) reported on the incorporation of freshwater algae into California’s bioassessment programs. They suggested that indicators should include diatom and soft-bodied algal assemblages, biomass based on chlorophyll-a and ash-free dry mass, and algal percent cover. They further recommended that field testing was needed to evaluate the need for all three indicators. More recently, Ritz (2010) assessed the ability of predictive models of diatom assemblages to provide an effective method to report on biological degradation in streams along the Central Coast of California. He concluded that the RIVPACS-type predictive model did not perform well and that degraded site identification was not consistent, although it was able to identify likely trends. He suggested that this was partially because of the lack of reference sites in this region.

No work has yet been undertaken in the med-regions of central Chile, although in South Africa, Taylor et al. (2007a, b, c, d) tested the utility of numerical diatom indices for indicating water quality in river systems in South Africa. These studies concluded that, in general, the European diatom indices could be used with success in South Africa, although the occurrence of endemic species, particularly in natural sites will necessitate the eventual compilation of a diatom index unique to South Africa. This is likely to be particularly important in the naturally acidic med-rivers of the south-western Cape, where there is a high degree of freshwater endemism (Wishart & Day, 2002). Current benthic algal research in the south-western Cape shows distinct patterns in algal communities that generally reflect the outcome of flood disturbance and its interaction with nutrient supply (pers. comm., J. Ewart-Smith, Freshwater Research Unit, University of Cape Town, unpublished data). These communities, which are dominated by acidophilic species, are highly sensitive to changes in both flow and nutrients. The potential exists to utilize shifts in algal divisions and forms that seem to respond to differences in perenniality as a basis for assessing changes in hydrological state. This research further suggests that development of bioassessment tools based on the entire algal community will provide a valuable means of detecting and monitoring both water quality and health of open-canopied med-river ecosystems in the south-western Cape.

In Australia, Chessman et al. (1999) developed and tested (Chessman et al., 2007) the Diatom Species Index for Australian Rivers (DSIAR), with data from diatom sampling of rivers in temperate south-eastern Australia (New South Wales, Queensland, South Australia and Victoria). It was significantly correlated with the river condition, and responded well to the impact from agricultural and urban land use. When tested in a tropical region of northern Australia, DSIAR did not, however, correlate well with an index of catchment condition (Chessman & Townsend, 2010). While the relationships between the index and water chemistry were consistent in both regions (especially pH, salinity, and nutrients), differences in the effects of catchment land-use on water chemistry, resulted in differences in the efficacy of the index in the two regions. No periphyton (or diatom) index has been used or tested for south-western Australia.

Macroinvertebrates

Macroinvertebrates are considered a useful and valuable component for ecological status assessment and have been widely applied in med-regions (e.g., Rosenberg & Resh, 1993; Bonada et al., 2006a; Ollis et al., 2006). In the Mediterranean Basin, the biotic index, IBMWP (or modifications thereof—Hellenic Evaluation System, Artemiadou & Lazaridou, 2005) has been commonly used (Alba-Tercedor & Sánchez-Ortega, 1988; Alba-Tercedor & Prat, 1992; Rico et al., 1992) as have various multimetric indices (Pinto et al., 2004; Morais et al., 2004; Munné & Prat, 2009; Sánchez-Montoya et al., 2010) and predictive models (e.g., Alba-Tercedor & Pujante, 2000; Feio et al., 2007a, b, 2009b; Poquet et al., 2009). The IBMWP index was shown to be comparable with other multimetric indices used in Europe such as ICM-Star (Intercalibration Common Multimetric Index, Munné & Prat, 2009). In evaluating the response of metrics, indices, and multimetric indices to different stressors and a human stressor gradient (Pinto et al., 2004; Munné & Prat, 2009; Sánchez-Montoya et al., 2010), it was shown that ICM-10 (= Iberian Mediterranean Multimetric Index—using quantitative data) and ICM-11a (= Iberian Mediterranean Multimetric Index—using qualitative data) were effective in med-rivers, particularly temporary ones, with ICM-11a showing the highest regression coefficients (Sánchez-Montoya et al., 2010). Multimetric indices had a more linear relationship to the human stressor gradient than IBMWP, and displayed lower variability in the reference values. Nonetheless, IBMWP was recommended as suitable for med-rivers in Spain and ICM-11a for meeting the requirements of Water Framework Directive (Munné & Prat, 2009; Sánchez-Montoya et al., 2010). The application of multimetric IM9 (Pinto et al., 2004), which comprised a summation of Average Score Per Taxon + No. Trichopteran families + percentage Gastropoda, Oligochaeta, and Diptera, was recommended for med-rivers in Portugal with high hydrological variability because it is based on three pure metrics (richness, tolerance, and composition), family-level identification can be used, and it is composed of metrics with different sensitivities.

Multimetric indices based on macroinvertebrates are the most widely applied bioassessment index in the western United States (e.g., Barbour & Hill, 2003; Stoddard et al., 2005) and California (e.g., Ode et al., 2005; Herbst & Silldorff, 2006, 2009; Ode et al., 2008; Ode & Schiff, 2009; Yoder & Plotnikoff, 2009), although predictive models have also received attention (e.g., Hawkins et al., 2000; Stoddard et al., 2005; Herbst & Silldorff, 2006, 2009; Ode et al., 2008; Ritz, 2010; Feio & Poquet, 2011). Ode et al. (2008) evaluated the performance (precision, bias, responsiveness, and sensitivity) of macroinvertebrate O/E indices and multimetric indices developed at a regional scale for California (Ode et al., 2005) to those developed for large-scale condition assessments [US Environmental Protection Agency national Wadeable Streams Assessment—western states only (WSA-West) and the Environmental Monitoring and Assessment Program Western Pilot Study (EMAP-West), Stoddard et al., 2005)]. Their study showed that WSA-West and EMAP-West indices were less precise and accurate in detecting impairment than the CA O/E index, and suggested that calibration to local conditions improves the utility of large-scale models to site or regional assessment. In California, biomonitoring programs use the Southern California Index of Biotic Integrity (SCIBI, Ode et al., 2005) and Northern California Index of Biotic Integrity (NCIBI, Rehn et al., 2005). The SCIBI was developed using reference sites that were predominantly in high-gradient (i.e., >1% slope) streams even though 20–30% of rivers are low-gradient, which prompted the recent study to develop indices and sampling methods for these rivers (Mazor et al., 2010). Herbst & Silldorff (2009) developed an IBI for perennial streams in the eastern Sierra Nevada, while in the urban context, Carter et al. (2009) developed a multimetric index from three ecological categories, including taxonomic composition (% EPT), functional feeding group (shredder richness), and habit (% clingers) based on locally collected data, which allowed for the definition of best available biological conditions, which was defined as the predicted biological potential. Balancing representativeness and biological integrity in the final reference pool and the development of a new multivariate predictive model is currently being focused on in California (pers. comm., P. Ode).

In central Chile, ecological status assessment is in its infancy (Figueroa et al., 2003, 2005, 2007) with focus directed on testing the efficacy of macroinvertebrate indices imported from Europe and elsewhere, including the Family-level Biotic Index (FBI, Hilsenhoff, 1988), Extended Biotic Index (IBE, Ghetti 1986 cited by Figueroa et al., 2007), Biological Monitoring Working Party (BMWP, Armitage et al., 1983), Biotic Index (BI, Chutter, 1972), and Stream Invertebrate Grade Number—Average Level (SIGNAL, Chessman, 1995). The FBI was tested in agricultural catchments and shown to be effective in differentiating between sites of low versus high disturbance, while BI and SIGNAL were more sensitive than IBE and BMWP (Figueroa et al., 2007). Figueroa et al. (2006) also examined colonization patterns of Chilean rivers and found them to be similar to those of other med-rivers in Europe, United States, Australia, and South Africa.

In South Africa, the biotic index, SASS5 (South African Scoring System—Version 5), which was modified from the BMWP system, has been adapted and tested nationally (Dallas, 1997; Chutter, 1998; Dickens & Graham, 2002; Ollis et al., 2006). It has been compared to the IBMWP (Bonada et al., 2006b), where both SASS5 and IBMWP were shown to produce equivalent results and reliably detect water quality impairment. SASS5 forms the backbone of the River Health Programme (RHP, Ollis et al., 2006; DWAF, 2008), and is considered a robust and effective tool for measuring ecological status of South African rivers. It has not been necessary to develop a unique method for the med-region of the south-western Cape, even though several taxa are endemic to this region. Interpretation of data is, however, undertaken regionally (Dallas, 2007b). A second method currently being tested is the Macroinvertebrate Response Assessment Index (MIRAI), which is based on the premise that macroinvertebrate assemblages reflect prevailing flow, water quality, and available habitat at a site (Thirion, 2007). Unlike SASS, which was designed specifically to assess water quality impairment, MIRAI aims to provide a habitat-based cause and effect foundation to interpret the deviation of the macroinvertebrate assemblages from the reference condition. Reference conditions comprise two components: an expected “Reference Score”; and an expected list of “Reference Taxa”, including expected abundances and frequency of occurrence (DWAF, 2008). A regional approach to bioassessment and the derivation of reference conditions has been adopted within the RHP (Dallas, 2002, 2004a; DWAF, 2008), whereby a hierarchical spatial classification system divides the country in a logical and ecologically meaningful way so that variation between rivers (and biotic assemblages) throughout the country is best accounted for. Several aspects related to spatial and temporal heterogeneity in lotic systems and the implications for defining reference conditions for macroinvertebrates have been investigated (Dallas, 2002), including the influence of biotope availability (Dallas, 2007a), and spatial (Dallas, 2004a) and seasonal (Dallas, 2004b) variability in macroinvertebrate assemblages. A biological banding system that uses percentiles and allows intrinsic variability in macroinvertebrate assemblages to be incorporated in data interpretation (Dallas, 2007b; Dallas & Day, 2007) has been developed. Multimetric indices and predictive models have not yet been developed, although preliminary analyses show that further research and testing of these tools would be a valuable contribution to bioassessment of rivers in South Africa (Dallas, 2007c).

In Australia, macroinvertebrates are most commonly used in bioassessment with a biotic index, SIGNAL most routinely applied (Chessman, 1995, 2003a, b; Growns et al., 1995). A predictive model system, AusRivAS, has been developed and widely tested (Smith et al., 1999; Simpson & Norris, 2000; Halse et al., 2007). AusRivAS is based on the British RIVPACS system, whereby macroinvertebrates from a monitoring site are compared with a database from a large number of reference sites throughout Australia. Both SIGNAL and AusRivAS use family-level taxonomy and biotope-specific sampling, and have been nationally applied (e.g., Chessman et al., 1997, 2006; Smith et al., 1999; Simpson & Norris, 2000; Chessman, 2003a; Metzeling et al., 2003). Smith et al. (1999) developed AusRivAS in Western Australia, including the south-western region, where the wet season channel model proved to be most accurate at distinguishing between reference and disturbed sites.

Fish

The original multimetric IBI was based on fish assemblages and composed of 12 metrics reflecting components of community ecology, including taxonomic richness, habitat and trophic guild composition, individual health, and abundance (Karr, 1981). A common issue related to fish in med-rivers is the low numbers of indigenous (= native) species per site, high number of endemic species, high number of introduced (= alien) species, and the high spatial and temporal variation in species patterns (e.g., Moyle & Randall, 1998). Simple measures of the number and abundance of indigenous fish species are considered useful measure of integrity in these rivers (Moyle & Randall, 1998).

The development of a multimetric index based on IBI in the whole Mediterranean Basin proved to be difficult (Ferreira et al., 2007a, 2007b) as there were considerable differences in fish community patterns between regions (France, Iberia, and Greece). Broadly, species richness per site decreased southwards, while endemicity and proportion of introduced species increased. Sites in the east of the Mediterranean Basin had a higher endemicity, lower site richness, and lower number of introduced species than sites in the west. Evaluation of the responses of ecological traits to anthropogenic disturbance revealed that overall metric response was weaker than in other European regions, with abundance-based metrics performing best (Ferreira et al., 2007a, b). At the catchment scale, Magalhães et al. (2008), Aparicio et al. (2011), and Hermoso et al. (2010) all demonstrated that it is possible to develop indices in species-poor regions. Magalhães et al. (2008) developed an IBI that had scores correlated with composite gradients of human impact and that differed significantly between reference and non-reference sites. The final metrics included proportion of indigenous fish, number of intolerant and intermediate species, number of invertivore indigenous fish, number of phyto-lithophilic and polyphilic species, and catches of introduced species. Aparicio et al. (2011) developed an IBI composed of five metrics related to fish health, age-structure, and abundance and richness of indigenous and introduced species. It correlated with various measures of environmental degradation and with several biotic and habitat indices, including the European Fish Index, despite different methods used for development and contrasting metrics obtained. Hermoso et al. (2010) proposed a combined index, the Index of Community Integrity, which is a site-specific comparison of community composition based on the reference condition approach. The presence–absence of ten indigenous species was modeled and used to evaluate the deviation between the observed and expected community composition. The species-by-species probability measures were summed to give the final score of the index, which was sensitive to both habitat degradation and the degree of dominance of introduced species (or biotic perturbation). Habitat degradation and introduced species have been traditionally overlooked in IBIs, and were found to be unresponsive to natural sources of variation. Given the high within-catchment endemism in med-rivers, they recommended a catchment approach for the development of site-specific predictive models.

Indices based on fish have been developed for parts of California (Moyle & Randall, 1998; May & Brown, 2002). No statewide IBI is currently available (Fetscher & McLaughlin, 2008), although Stoddard et al. (2005) developed a three ecoregion-based multimetric index and predictive model for the Western US as part of the EMAP-West pilot study. Characteristics of fish assemblages in California include low indigenous species diversity, high endemism, barriers to (re)colonization and many introduced/invasive species. The occurrence of many ephemeral streams also creates complexity in index development and testing. The Watershed IBI, used by Moyle & Randall (1998) to evaluate the biological health in the Sierra Nevada region, was based on frogs and fish. May & Brown (2002) investigated fish communities and the efficacy of the IBI within the Sacramento River Basin. Four fish community metrics (percentage of indigenous fish, percentage of intolerant fish, number of tolerant species, and percentage of fish with external anomalies) were responsive to environmental quality, and they concluded that it should be possible to develop an IBI-type index.

Ecological status assessment based on fish has not yet been undertaken in Chile, although endemic species are common (Habit et al., 2006) and introduced fish species widespread (e.g., Soto et al., 2006; southern Chile). In South Africa, the Fish Response Assessment Index is the method incorporated into the RHP. This index is based on the environmental intolerances and preferences of the reference fish assemblage within a river segment, and the response of the constituent species of the fish assemblage to environmental stressors (Kleynhans, 2007). Application of this index, which was developed and tested primarily in the northern regions of South Africa, to the south-western Cape is, however, problematic given the low diversity, low abundance, and generalist species present there. In addition, all the indigenous species are extinct in mainstem rivers and roughly 80% of the river length is dominated by introduced species (pers. comm., B. Paxton, Freshwater Research Unit, University of Cape Town). Nonetheless, a local IBI that incorporates relative abundances, population structure, and individual fish health could be developed for the south-western Cape (pers. comm., B. Paxton).

In Australia, early testing of fish-based multimetric indices in rivers of southern Australia suggested that they were useful for assessment of river health (Harris, 1995; Harris & Silveira, 1999). More recently, Kennard et al. (2006), using a predictive model, demonstrated that fish assemblage composition in south-eastern Queensland rivers can be related to a small set of variables describing catchment scale (elevation, and relative site position within the river network) and local scale environmental features characterizing the reference sites (wetted width, depth, and water velocity). They concluded that the predictive model provided sufficiently accurate and precise predictions of species composition and was sensitive enough to distinguish test sites impacted by several common types of human disturbance, including impacts associated with catchment land use and local riparian, in-stream habitat, and water quality degradation. Further, their model that was developed for sites sampled in one season only (winter), was able to accurately predict fish assemblage composition at sites sampled during other seasons and years, provided that they were not subject to unusually extreme environmental conditions. No bioassessment tools or reference conditions based on fish have been developed for med-rivers in south-western Australia, although a substantial body of literature exists (see Centre for Fish and Fisheries Research, Murdoch University). This region also has low diversity (ten indigenous, ten introduced species), high endemism (80%), and high proportion of introduced species (Morgan et al., 1998).

Aquatic and riparian vegetation

Both aquatic macrophytes and riparian vegetation have been included in indices aimed at assessing ecological status. Riparian vegetation is also sometimes combined with indices of general stream conditions such as Index of Stream Condition (ISC, Ladson et al., 1999) or in conjunction with rapid bioassessment protocols (e.g., Barbour et al., 1999). Ferreira & Aguiar (2006), in a review of the ecology of riparian and aquatic vegetation in western Iberia, documented the use of river plants in river assessment. Initial bioassessment methods using river plants in the Mediterranean Basin were largely based on sensitive species (i.e., indicator indices) or on functional groups (i.e., multimetric indices). They concluded these med-rivers have rich and dynamic plant assemblages, which are spatially and temporally controlled by multiple drivers, but that truly aquatic species are few and extensive human disturbance has limited the identification of undisturbed assemblages. Indices include the Riparian Forest Quality Index (QBR, Munné et al., 2003), which was developed for rivers with a forested riparian zone, and is based on four components of riparian habitat, namely total riparian vegetation cover, cover structure, cover quality, and channel alterations. Within the geographical area studied, the QBR index proved to be independent of regional differences in riparian plant community types. Another multimetric index, the Iberian Multimetric Plant Index (IMPI, Ferreira et al., 2002, 2005), which included eight metrics (three representing composition measures, three reflecting direct human disturbances, one trophic measure, and one related with riparian integrity), was developed. IMPI was later applied to nationwide data in order to obtain a spatially upgraded index (Riparian Vegetation Index, RVI; Aguiar et al., 2009), which is based on structural and functional components of the riparian and aquatic vegetation, and which responded reliably to disturbance. Their study confirmed the hypothesis that plant-based indices of integrity are scale-dependent. More recently, Aguiar et al. (2011) tested the RVI and two predictive models: MACrophyte Prediction And Classification System (MACPACS) and Macrophyte Assessment and Classification (MAC). RVI and MAC, which use quantitative data of taxa abundance, performed reliably in response to human disturbance, in comparison to MACPACS, which is derived from taxon occurrence data.

In the United States, rapid bioassessments of rivers include riparian vegetation components (Barbour et al., 1999) and the California Rapid Assessment System for Wetlands (CRAM, Collins et al., 2008; Solek et al., 2011), which is a standardized, cost-effective tool for assessing the health of wetlands and riparian habitats. It assesses wetland condition based on four attributes, namely landscape context, hydrology, physical structure, and biotic structure. Stein et al. (2009) demonstrated that CRAM is an effective tool for assessing general riverine and estuarine wetland condition based on its correspondence with multiple independent assessments of condition. No specific indices have been developed in Chile, while in South Africa only a general index, the Riparian Vegetation Response Assessment Index (VEGRAI), has been developed, although it has yet to undergo rigorous testing (Kleynhans et al., 2007). In Australia, the ISC Australian index (Ladson et al., 1999) and other indices such as Tropical Rapid Appraisal of Riparian Condition (Dixon et al., 2005) have been developed, but no method has yet been developed for the south-western region.

Challenges and complexities in defining reference conditions for med-rivers

The greatest challenge in defining reference conditions for med-rivers relates to hydrological state, with med-rivers characterized by a large hydrological gradient, ranging from stable, perennial rivers to highly ephemeral, unpredictable ones. Spatial heterogeneity and identification of reference sites in lowland areas is an issue in med-regions, although this is also often a problem in non-med-regions. The high degree of endemism in med-regions further complicates the determination of reference conditions, while the predicted consequences of climate change in med-regions, accentuates the urgent need to develop robust tools for bioassessment in these regions. These issues are discussed below.

Temporal variability and hydrological state

Intra- (seasonal) and inter-annual temporal variability in biota is a characteristic of med-regions, where med-rivers have distinct seasonal patterns of precipitation and temperature that result in hot, dry summers and cool, wet winters. Drying over summer varies within regions and determines hydrological state, with both perennial and intermittent rivers common. Uys & O’Keeffe (1997) defined several hydrological states, including Perennial Seasonal Winter (i.e., winter rainfall regions: flow maximized over winter and early spring, and diminishes over summer) and Intermittent Seasonal Winter (predictable floods/recommencement of flow in winter months, surface flow disappears and the channel may dry in parts during the year, certainly during summer months). Both these states apply to med-regions. Inter-annual variability occurs in response to longer cycles of climate variability such as drought cycles. Both intra- and inter-annual variability leads to changes in hydrological state, which result in other effects such as changes in primary productivity, dissolved oxygen, conductivity and habitat (Gasith & Resh, 1999; Bêche et al., 2006).

Studies have shown that there are distinct ‘wet season’ (winter/spring) and ‘dry season’ (summer) taxa, which are adapted to the flow-altered habitat of a particular season (e.g., Bunn et al., 1986; Bonada, 2003; Bêche et al., 2006). Taxa that are not adapted to these seasonal changes may be eliminated and biological factors like predation, parasitism, and competition may be intensified as low summer flows lead to higher densities of individuals, and more frequent opportunities for biotic interactions, such as competition and predation (Power et al., 1988). Seasonal differences in biotic assemblages may manifest as differences in ecological status at the same site assessed in different seasons, even though no change in river condition has occurred. High temporal variability in community structure has the potential to limit the sensitivity of bioassessment approaches by making it difficult to, for example, construct robust predictive models that do not falsely conclude that observed changes in patterns are a consequence of changes in ecological status (Bunn & Davies, 2000).

This concern prompted the initiation of several studies examining the influence of season on metrics, indices, and models used in ecological status assessment. Results are variable and dependent on the metric or index used, with some showing distinct seasonal differences (e.g., Dallas, 2004b; Mazor et al., 2009), while others do not vary seasonally (Zamora-Muñoz et al., 1995; Sánchez-Montoya et al., 2009b). Feio et al. (2006) showed that a model developed for one season (summer) was inaccurate for other seasons, and that models based on lower taxonomic levels (species and genus) were more inaccurate than family or order level models. In perennial med-rivers, it has also been shown that macroinvertebrate assemblages differed more between wet and dry periods within seasons than between seasons, i.e., spring versus summer (Munné & Prat, 2011). Communities are often driven by hydrological disturbance, which is not confined to winter but often extends into spring and autumn, such that spring communities may change considerably within a season and are therefore less predictable (Ratcliffe, 2009; pers. comm., J. Ewart-Smith, Freshwater Research Unit, University of Cape Town). Taxon richness, IBMWP, and EPT values also differed significantly between wet and dry periods, although IASPT did not (Munné & Prat, 2011). For med-rivers that have distinct hydrological periods, it was thus recommended that ecological status assessments take cognizance of this by using reference conditions derived for each hydrological period. Hydrological period can be classified using the Standardized Precipitation Index (McKee et al., 1995), which is based on rainfall in the 3 months prior to sampling.

Inter-annual differences in hydrology, from severe droughts to exceptionally high rainfall, have contributed to high inter-annual variability of macroinvertebrate assemblages in med-rivers (Gasith & Resh 1999). The extent to which this variability affects aquatic communities and ecological status assessment have been examined by testing persistence of metrics, indices, and models using long-term datasets (e.g., Bêche & Resh, 2007; Rose et al., 2008; Mazor et al., 2009). Mazor et al. (2009), using a 20-year data set from four sites in two northern Californian streams, showed that both seasonal and inter-annual variability contributed to total variability of bioassessment metrics and indices. In particular, EPT richness and O/E scores responded to seasonal changes, while Coleoptera richness and % non-gastropod scrapers responded to annual variability. They suggest that benthic macroinvertebrates may be well adapted to the large yet predictable changes that occur in each season, but not as well adapted to the unpredictable changes that occur in certain years (Mazor et al., 2009). Their study also showed that multimetric indices and predictive models had lower long-term variability compared to single metrics and simple indices. For this reason they suggested that combining metrics into a multimetric index would decrease overall long-term variability as metrics with lower variability (e.g., EPT richness) dampens the influence of highly variable metrics (e.g., % intolerant). In Australia, Rose et al. (2008) assessed the ability of SIGNAL, EPT, Family Richness and AusRivAS, determined in two biotopes: riffle (areas with rapid and turbulent flow) and edge (areas of little or no flow), to differentiate between the effects of a drought and anthropogenic impacts. Generally macroinvertebrates in riffles were more resistant to drought compared to edge biotopes, largely as a consequence of riffles remaining available and sampling protocols related to riffle extent, but they warned that bioassessments during drought should be used with caution, particularly where indices of condition are developed for large areas (e.g., at the catchment level). Bayesian hierarchical modeling, which assumes dependency amongst sampling sites, was used to analyse temporal trends in stream ecological condition (as measured by SIGNAL) using a long-term data set (Webb & King, 2009). Similarly, there was strong evidence of a widespread decline in SIGNAL scores for edge biotopes but not for riffles. The rate of decline in edge biotopes was positively associated with catchment urbanization.

Given the intrinsic temporal variability (both seasonal and inter-annual) observed in med-rivers, and given the forecasted global patterns of climate change, it is critical that bioassessment programs include long-term monitoring at both reference and non-reference sites (e.g., Nichols et al., 2010). This will facilitate the identification of drivers of variability and prevent erroneous determinations of impairment. Mazor et al. (2009) recommend that bioassessment programs incorporate temporal variability into index development by using multiple years of data for calibration, perhaps requiring an iterative approach with regular updates to establish new thresholds, particularly when these thresholds form the basis for the establishment of biocriteria for regulatory purposes. Predictive models should be developed for each season and/or hydrological period, depending on the observed variation in biotic assemblages.

Perennial versus intermittent med-rivers

Aquatic communities have been shown to vary considerably from perennial to intermittent med-rivers (e.g., Bonada et al., 2006c; Argyroudi et al., 2008). In perennial rivers, EPT orders are appropriate indicator taxa, while in intermittent ones Odonata, Coleoptera, and Heteroptera (OCH) and some Diptera are the major indicative taxa (Munné & Prat, 2011). The OCH share biological traits that allow them to survive in intermittent rivers (only pools) or migrate when the river totally dries up (e.g., Williams, 1996). Bonada et al. (2006c) identified many of these taxa to be exclusive to disconnected pool habitats, compared with riffle or connected pool macro-habitats. Hydrological history of a river can thus be determined by applying the ratio of EPT/OCH or EPT/EPTOCH (Rieradevall et al., 1999; Bonada et al., 2006c). Inter-annual variation in macroinvertebrate assemblages at reference sites in south-western Australia varied little over a 5-year period in perennial streams compared to intermittent, which showed marked variation over time with no obvious pattern (Bunn & Davies, 2000). Munné & Prat (2011) suggested that the accentuated variability in reference assemblages and indices observed in intermittent rivers over different hydrological periods result from flow regimes that are highly variable over years and between years. Morais et al. (2004) tested the robustness of biotic indices and multimetric indices in intermittent rivers under different hydrological conditions and found that IASPT varied least of the metrics, while overall the multimetric index, IM9, was the most suitable assessment methodology. Argyroudi et al. (2008) showed that biotic metrics and indices differed between intermittent (with pools during low flow periods) and ephemeral streams (with dry beds). They concluded that existing European quality indices do not sufficiently differentiate between intermittent and ephemeral med-rivers, and thus cannot reliably discriminate between natural variability and human stressors in these rivers. Intermittent streams, which are common in med-regions, are at present excluded from standard bioassessment protocols in some med-regions even though they represent more than half the streams in, for example, southern California (Mazor et al., 2009).

Complexity arises when perennial rivers are converted to non-perennial ones, through water diversions such as inter-basin water transfers; and non-perennial rivers to perennial ones through augmented flows and runoff. What reference does one use when a river has been converted from one type to another, particularly if there has been a considerable time period since implementation, and an “alternate stable state”, albeit an unnatural one has been attained?

Spatial heterogeneity

Natural variability of reference assemblages (i.e., variability in the absence of impact) is assumed to be less than the change caused by a disturbance (Mazor et al., 2009). High natural variability in assemblages may result in variability in metrics and indices, with resultant inaccurate assessment of ecological status, if spatial heterogeneity is not taken into account in bioassessment programs. Spatial heterogeneity is hierarchical, ranging from broad regional variability, to within-river variability encompassing longitudinally variability down a river course, and site-level variability related to factors such as substratum, habitat and flow (e.g., Frissell et al., 1986). At the regional scale, for example, macroinvertebrate fauna of south-western Australia is known to be depauperate, both at family and species level, compared with other Australian regions (Bunn & Davies, 1990, 1992; Kay et al., 2001). In particular, some groups are absent or poorly represented (e.g., Plecoptera and many families of algal-grazing insects), such that mean EPT Taxa and SIGNAL indices for reference sites were significantly lower in south-western rivers compared to south-eastern ones. This results in a status close to the ‘doubtful water quality’ status, even though the sites were in undisturbed forest (Bunn & Davies, 2000). Incorporation of spatial variability is possible by using an appropriate spatial framework that differentiates river types or by utilizing predictive models that incorporate intrinsic spatial heterogeneity. Interpretation of data derived from biotic indices such as SASS5 and SIGNAL that are applied at a national scale, which include non-med-regions, should be undertaken within a spatial framework such that regional and longitudinal differences are taken into account (e.g., Dallas, 2007b).

Identification of reference sites, particularly for lowland rivers

Gasith & Resh (1999) proposed that human activities may impact med-rivers more than rivers in more humid or arid regions because of the severe competition for water that occurs in med-regions. This is corroborated by several researchers involved in the development of bioassessment tools and the identification of reference sites in these regions (e.g., Sánchez-Montoya et al., 2009a; Munné & Prat, 2011; Mazor et al., 2011). The difficulty of finding reference sites, particularly minimally disturbed sites, in middle and lower reaches of med-rivers is often very difficult as these areas are generally impacted by humans, and commonly have agricultural activities, urban development and regulation of river flows (e.g., Sánchez-Montoya et al., 2009a; Yoder & Plotnikoff, 2009; Munné & Prat, 2011; pers. comm, P. Ode). Low-gradient rivers also often comprise a large proportion of stream length, for example they form 20–30% of stream length in California (Mazor et al., 2010). In southern California, biological condition of perennial streams, as measured by the SCIBI, indicated that 53% of streams were non-reference condition, and that their distribution was closely associated with land use (Mazor et al., 2011). Recently, Ode (pers. comm.) showed that only 38% (615) of the 1,637 sites in California met reference criteria, and when linked to stream length, large regional variation was revealed. For example, 70% of the stream length in the Central Lahontan and South Coast Mountain regions met reference criteria compared to 2% in the Central Valley and the South Coast Xeric regions. This lack of high quality reference sites was one reason for poor model precision when applied to diatoms, with O/E scores not clearly identifying impacted sites (Ritz, 2010). Similarly, in both the south-western Cape and south-western Australia, a large portion of the landscape has been modified by activities such as land-clearing, dryland agriculture, livestock grazing, and regulation of river flows, resulting in widespread salinization and eutrophication of rivers (McComb & Davis, 1993; DEA&DP, 2011). Few rivers remain that are likely to have near-natural biological assemblages (Boulton et al., 1999; Chessman & Royal, 2004). The scarcity of high quality reference sites in certain areas is acknowledged as a complicating factor for ecological status assessment (Yoder & Plotnikoff, 2009), and in such instances an acceptable level of disturbance needs to be included in decision making, for example, by defining reference conditions based on least-disturbed or best-attainable condition. The potential to use historical condition should also be explored although even historical data may represent impacted conditions as human settlement generally pre-dates the collection of data in med-regions.

Endemic and introduced species

High levels of endemism characterize med-regions (e.g., Myers et al., 2000; Wishart & Day, 2002; Marr et al., 2010). For example, aquatic invertebrates and fish, respectively, comprise 65% (Wishart & Day, 2002) and 89% endemic species in the south-western Cape (Marr et al., 2010), while for fish, rates of endemism in other med-regions are: 80% in south-western Australia, 60% in California, 52% in Chile, and 44% in the Iberian Peninsula (Marr et al., 2010). Despite the high rates of endemism, all med-regions currently have more introduced fish species than endemic fish species (Marr et al., 2010). Further, many med-regions even have more introduced fish than indigenous/native fish (pers. comm., R. Mazor, Southern California Coastal Water Research Project Authority, California), with some currently having double the number of introduced compared to indigenous fish species (pers. comm., D. Impson, Cape Nature, South Africa). Med-rivers may become even more vulnerable to invasion with predicted global climate change. Med-regions also have high numbers of introduced plant species (e.g., Aguiar et al., 2007) many of which have persisted and become naturalized populations in many parts of the Mediterranean Basin (e.g., Ferreira et al., 2002). Judgments about whether a species is indigenous or introduced to a certain region can be problematic (e.g., Benejam et al., 2008). The issue of endemism and introduced species in the context of ecological status assessment is: (1) to what extent endemism may cause spatial heterogeneity; and (2) should introduced species be included in assessments. While spatial heterogeneity is easily accounted for by using spatial frameworks, river types, or predictive models, the issue of introduced species is more complex. Kennard et al. (2005) suggest that introduced fish may represent both a symptom and a cause of decline in river health and the integrity of indigenous aquatic communities. Many multimetric indices include metrics based on introduced species such as presence, richness, relative abundance, and/or relative biomass of introduced species (e.g., Karr, 1981; Moyle & Randall, 1998; May & Brown, 2002). Many mainstem rivers, for example in the south-western Cape, are dominated by introduced fish species, while having extremely low diversity and abundance of indigenous fish (Woodford et al., 2005). Application of traditional multimetric indices may not be useful and so the potential for developing indices based on introduced species may be an option. Kennard et al. (2005) confirmed that indeed introduced fish species have potential to be used as an initial basis to determine non-biological (i.e., not related to introduced fish) human disturbance impacts and that some introduced species may represent a reliable initial indicator of river health.

Climate change

Med-regions are expected to be among the most affected by global climate change with med-regions generally becoming hotter and drier (e.g., Midgley et al., 2005; CSIRO, 2007; Giorgi & Lionella, 2008; Cayan et al., 2009). Increasing river water temperatures and declining river flows will lead to greater thermal stress on aquatic organisms. Climate change is one of many drivers acting on freshwater systems and should be considered as an additional, perhaps multiplicative, factor in ecosystems that are already severely stressed. It is important to understand the potential amplification of variability that climate change may bring to bear on existing freshwater resources, such as further deterioration in water quality (Dallas & Rivers-Moore, 2009). Ecosystem resilience needs to be promoted, by habitat restoration (e.g., Davies, 2010) and recognition of the link between catchment condition and river health. Climate change could affect biological indicators, interpretation of metrics and indices, reference site condition, and monitoring. Moreover, biological responses to climate change may influence water-quality standards and biocriteria through shifts in baseline conditions (Barbour et al., 2010). As suggested by Hamilton et al. (2010), many widely used taxonomically based metrics are composed of both cold-water and warm-water preference taxa, and differing responses of these temperature-preference groups to climate-induced changes in stream temperatures could undermine assessment of stream condition. Two issues arise, the first is detecting climate change responses and differentiating them from other anthropogenic impacts, and the second is modifying existing metrics/indices/models or developing new ones that are more sensitive to climate change.

Several recent studies have tried to disentangle climate change effects from other effects (e.g., Chessman, 2009; Hamilton et al., 2010; Lawrence et al., 2010; Nichols et al., 2010; Stamp et al., 2010), although only one of which was for a med-region (California, Lawrence et al., 2010). Nichols et al. (2010) assessed the response of macroinvertebrate indices to climate change by analysing indices derived from reference and non-reference sites over a 15-year period and concluded that they were insensitive to the small to large changes in climate, although they did detect extreme climate-related events such as drought and bushfires. Lawrence et al. (2010) showed that a macroinvertebrate-IBI developed for northern California streams, as well as several other metrics including taxon richness, % EPT and O/E scores, was not influenced by temperature extremes (cool and warm) or precipitation extremes (wet and dry). They developed a local climate-change indicator that is composed of the presence/absence of nine thermally sensitive macroinvertebrate taxa, identified to genus level. Hamilton et al. (2010) found that metrics selected for condition assessments were sensitive to changes in temperature, which prompted them to explore the development of temperature-modified metrics. They generated metrics based on the ratio of cold- or warm-water-preference taxa to total invertebrate taxon richness, and a ratio of cold-water or warm-water preference EPT taxon to total EPT taxa. The response of these metrics varied from one ecoregion to another, and was linked to factors such as elevation, but these types of studies warrant further investigation. Using the same data set, Stamp et al. (2010) used thermal preference metrics to examine bioassessment data for climate change effects. Analysis of long-term data sets will provide valuable input into the derivation of thermal metrics, which, together with laboratory based information on thermal tolerances (e.g., Dallas & Ketley, 2011; Dallas & Rivers-Moore, 2012) will provide insight into the identification of bioindicators of thermal change. Long-term monitoring of both reference and non-reference sites is critical if the effects of climate change on aquatic ecosystems are to be understood and factored into ecological status assessments.

Comparison among med-regions and translation of science to legislation

In comparing ecological status assessment and the derivation of reference conditions amongst med-regions, it is clear that, for those med-regions that already have a functioning bioassessment program, there are many commonalities in the approach towards bioassessment and defining reference conditions (Table 2). Benthic macroinvertebrates are the most routinely used component, and while other components have either been used in the past, or methodology based on them has been developed, there are logistical or other constraints that have limited their use in routine monitoring. Most med-regions selected reference sites in an iterative manner, using GIS, available data, and expert opinion for initial reference site selection, followed by more detailed field-based assessments to validate inclusion or exclusion of a potential reference site within the pool of reference sites. Selection of a number of reference sites within a homogenous group (e.g., river type) is considered important such that intrinsic variability could be incorporated. In some med-regions biological criteria are also employed as a further validation step. A variety of multivariate and multimetric techniques are used to characterize the reference conditions, and modeling is frequently undertaken. Bioassessment data are interpreted using biotic, multimetric, and observed/expected indices, with deviation from natural the method for establishing categories or levels of change, for example, using Ecological Categories. Spatial variability is universally recognized as being important and all med-regions have developed frameworks and methods that ensure that intrinsic spatial variability is taken into account, often through the generation of river types and the use of modeling. Temporal variability is understood to be a potential source of error when interpreting bioassessment data and med-regions often constrain the sampling to particular time periods, or develop models that are seasonally specific. Non-perennial streams that flow for approximately 3 months each year are currently included in the Mediterranean Basin and Australia, although ephemeral streams are excluded.
Table 2

Synthesis of ecological status assessments and approaches currently applied in four med-regions

Question

Mediterranean Basin

California

South Africa

Australia

What biotic assemblages are used in routine monitoring?

Benthic macroinvertebrates

Benthic diatoms, benthic soft algae (including cyanobacteria), benthic macroinvertebrates, and riparian vegetation. Vertebrate assemblages were used in routine monitoring through 2000 to 2005, but no longer (because of logistical challenges, limited resources, the absence of fish from many streams, and difficulty in interpreting the data in regions where fish stocking is prevalent)

Five biological response indices are used, including benthic algae (primarily diatoms), benthic macroinvertebrates, fish, riparian vegetation, and habitat integrity. Of these macroinvertebrates are most commonly used for routine monitoring. Three driver indices are used, including hydrology, geomorpohology and physico-chemical, for in EcoStatus/EcoClassification

Benthic macroinvertebrates are mostly used; fish often and diatoms rarely

How are reference sites identified?

Reference sites represent “least-disturbed” conditions. Potential reference sites are selected using GIS, available data and expert knowledge. Field-based assessments are then undertaken to ensure that sites meet all 20 Mediterranean Reference Criteria (MRC). Lastly, one or more biological criteria are used for reference site validation

Reference sites represent “least-disturbed” conditions (although alternative approaches in pervasively disturbed regions are the subject of current research). Sites are screened using GIS stressor data (mostly land use) and is supplemented with field-based measures of habitat and water quality. Ground-truthing reference status was a component of nearly all of the programs that originated the data used to create the reference network

Reference sites represent “least-disturbed” conditions. Potential reference sites are selected using GIS, available data and expert knowledge. Field-based assessments encompassing evaluation of land use, hydrology, habitat, and physico-chemistry are then undertaken to further screen reference sites. Analyses of biological data is undertaken for final site selection

Reference sites used in AusRivAS, which represent “least-disturbed” conditions, were chosen by professional judgment. The Sustainable Rivers Audit (SRA) in the Murray Darling Basin does not use reference sites (Davies et al., 2010). For specific assessment of the impact of water abstraction, Brooks et al. (2011) defined reference sites as those with no upstream entitlement for abstraction

How are reference conditions defined?

Biological data are used to calculate metrics, specifically IBMWP. Biological quality metrics are calculated from an adequate number of reference sites within each river type. The median value of IBMWP at reference sites within a river type is used as the criterion

Currently available data are used to generate expected conditions using predictive models. Multimetric indices based on observed stressor relationships are also used within certain subregions

The process by which reference conditions are derived varies from one biotic component to another. Currently available data (normally from several reference sites within a river type) are used to generate “expected” or “reference” conditions. Historical data and expert judgement are used when “least impacted” sites are not available. Predictive modeling, while desired, has not yet been incorporated

AusRivAS uses reference data from sampling of reference sites, mostly in the mid-1990s. In the SRA, reference data for fish are based on professional opinion. Reference data for macroinvertebrates are derived from contemporary biological and environmental data by statistical hind-casting with boosted regression trees

How is bioassessment data interpreted?

Five levels are used in the Water Framework Directive (WFD). Thresholds between classes are derived using the 25th percentile of the median of the metric in reference conditions. Assuming that the relationship between pressures and the metric is linear, and taking the reference value as a basis, the other thresholds are fixed at proportional intervals between the reference value and the minimum value of the metric

Bioassessment data are used to compare biological communities to reference condition using Observed/Expected indices or multimetric indices (which are only used within certain regions of the state). Although statistical distributions of indices at reference sites are traditionally used to establish thresholds for identifying non-reference conditions (e.g., two standard deviations below the reference mean), other approaches are under exploration (e.g., benchmarking thresholds with measures of ecosystem function)

Bioassessment data are interpreted using reference conditions where these are available. Comparison includes biotic indices, Observed/Expected indices, and expected taxa. Data are interpreted within the spatial framework and ecological categories are determined for each component. Ecological Categories are used, expressed as A to F, where A represents close to natural (reference) and F critically modified

By comparison with site-specific (and sometimes time-specific) reference data via biotic indices, Observed/Expected indices, and multimetric indices. Index values are used to place sites in bands of degree of inferred human impact

How is spatial variability incorporated?

River types, which attempt to account for intrinsic spatial variability resulting from, for example, differences in climate, geomorphology and geology are defined. Reference conditions are derived for each river type

For reporting and stratification of sampling effort for statewide programs, the state is divided into 6 major regions (three of which are further divided into subregions). Regional divisions are based on hydrologic, ecological, and political boundaries. However, for modeling applications, these boundaries are ignored and spatial gradients are examined either directly (e.g., latitude, longitude, and elevation), or indirectly (through correlated gradients, like climate and geology)

A spatial framework that incorporates ecoregions, which are based on physiography, climate, geology, soils and potential natural vegetation; and longitudinal zones, which reflect broad geomorphological characteristics and distribution patterns of biotic components, is used. Reference conditions are derived for each river type

Site-specific reference data are modeled

How is temporal variability incorporated?

Sampling is always at the same time period each year, normally spring, and where possible under similar hydraulic conditions. This is particularly important in the case of temporary streams

Temporal variability is constrained by sampling within an index period (which varies by region, but is typically two months long). Current research is underway to validate the consistency of bioassessments within and beyond this index period. Sampling never occurs within 4 weeks of a major rain event (i.e., one sufficient to mobilize the streambed). In some probabilistic and reference sampling programs, 10% of sites are replicated within-season (usually within 1 month), and 10% are replicated within years

Sampling is not constrained within any one time period, although winter sampling is generally avoided because of high flows. Sampling period is considered when interpreting results and cognisance is taken of any taxa that are highly seasonal. Hydrological state is noted

AusRivAS has separate predictive models for autumn and spring, but does not adjust for secular variation, e.g., multi-year climatic cycles and trends. SRA does not adjust for temporal variability at all

Are non-perennial streams included?

Yes, but only non-perennial streams that flow for at least 3 months per annum. Hydrological conditions are taken into account, with sampling date set for when the river is flowing and at least one month since the previous flood. Data interpretation for non-perennial rivers is difficult

Currently, no, but long-term intermittent streams will be added within the next few years. Research is underway to incorporate the more ephemeral end of the non-perennial spectrum, (including highly episodic channels, washes, alluvial fans, and other systems that have very short flow durations). Although this later research focuses on California’s deserts, it will apply to the Mediterranean climate regions as well

No, not at present

Non-perennial streams, which flow for a period of time each year, are generally included; but ephemeral streams are not

Has science been translated to legislation?

Yes, in Spain the index defined by Munné & Prat (2009) is used now as an “official” method. In the European Union the methods of each country should be validated (intercalibration exercise) before they can be used for the national administrations

Development of benthic macroinvertebrate index (BMI)-based biological standards (biocriteria) for perennial wadeable streams is underway, and implementation is expected to begin in January 2013. Development of criteria for other indicators and waterbodies is expected thereafter

Biocriteria and thresholds have been developed to a limited extent on an ad hoc basis. A more comprehensive program to define reference conditions and set biocriteria within a spatial framework is required to generate the necessary data on which legislation may be based

The state of Victoria includes biological objectives based on macroinvertebrates in the State Environment Protection Policy (Waters of Victoria). http://www.epa.vic.gov.au/water/epa/wov.asp

Note: Information for the Mediterranean Basin is mainly based on Spain and Chile has been excluded as biomonitoring is still in its infancy

Information for the synthesis was obtained through consultation with regional experts (Narcís Prat: Mediterranean Basin; Raphael Mazor: California; Helen Dallas: South Africa; Bruce Chessman: Australia) and supplemented with information extracted from the literature

The extent to which scientific information has been translated to legislation varies among med-regions, although it is recognized by all med-regions as being of vital importance. Indeed, the translation of scientific work to legislation is a critical stage in the management and protection of aquatic resources. In Spain, the “Instrucción de Planificación Hidrológica” (2008) has based much of its regulations on the Limnetica monograph (e.g., Prat, 2002) and other scientific papers produced by the same group of researchers (GUADALMED project). Biological objectives based on benthic macroinvertebrates have been incorporated in the State Environment Protection Policy in Victoria, Australia, but elsewhere the development of biocriteria and incorporation into legislation is currently underway or yet to begin. The value of focused, output-driven research programs that ultimately lead to development of legislation is therefore of great importance.

Conclusions

Med-regions are environmentally complex areas exposed to multiple anthropogenic stressors. Bioassessment and the derivation of reference conditions for med-rivers requires cognisance to be taken of the intrinsic spatial and temporal variability. Numerous tools developed for bioassessment, and their application and testing in med-regions, have provided insight into the complexities and challenges faced in ecological status assessment of these systems. Med-regions where tools are currently being developed, or have not yet been developed, would benefit greatly through collaboration with regions where the development of such tools is more advanced because many similarities exist in terms of the natural and anthropogenic factors that affect bioassessment and reference conditions, and the solutions are likely to be universal. This is particularly critical as med-regions are expected to be among the most affected by global climate change, which coupled with their well-recognized status as “biodiversity hotspots”, underlines the urgency for directed research aimed at development and testing of bioassessment tools and predictive models in all med-regions. Lastly, the importance of establishing long-term monitoring sites (reference and non-reference) for regular sampling and collection of data has been highlighted by many individuals. Clearly, strategic selection and prioritization of such sites needs to be given precedence.

Acknowledgments

Sincere thanks to Núria Bonada and Vince Resh for inviting me to contribute to this special issue. Thanks to the many authors who provided me with copies of their papers. My colleagues, Nick Rivers-Moore and Justine Ewart-Smith, provided useful comments on the draft manuscript. Thanks to Raphael Mazor for his insightful and useful review of this manuscript, and to a second anonymous reviewer for their valuable comments. Thanks also to Bruce Chessman, Raphael Mazor, and Narcís Prat who provided input to the synthesis (Table 2).

Copyright information

© Springer Science+Business Media B.V. 2012