Keywords

1 Introduction

Freshwater ecosystems are severely threatened from human-generated pressures, including water abstraction, pollution, construction of reservoirs, and invasive species. The continuous deleterious effects of human pressures have promoted the need for biological monitoring as well as the development of biological indices [1–3]. Fish are among the taxonomic groups with more longevity in aquatic environments and are excellent ecological indicators for a number of reasons [4]. Fish assemblages have been shown in a number of regions to respond to anthropogenic disturbances including flow regulation (e.g., [5]), habitat fragmentation [6], water pollution [7], land-use change [8], hydrological alteration (e.g., [9]), and acidification [10].

One disadvantage of using fish as ecological indicators is that their population densities are more difficult to estimate accurately and their catchability depends on a number of factors including electrofishing equipment, the characteristics of the river reach [11–13], and species-specific features such as morphology or behavior [14, 15]. The estimation of catchability and intercalibration of data are important to combine data from different fishing teams and to develop protocols for future work or monitoring [12]. Habitat quality is often assessed during fish sampling [16, 17] and inconsistency of habitat assessment among researchers has been also reported by several researchers (e.g., [18–22]).

This chapter summarizes our experience in developing fish-based indices in Catalan Rivers [23, 24] and synthesizes our studies: (1) to estimate the effects of fishing crew and other factors on fish catchability and the resulting fish metrics and on habitat assessments and (2) to attempt to develop type-specific- (i.e., IBICAT2b) and site-specific-based indices (spatially-explicit approach) (i.e., IBICAT2a). We also aim to give a protocol and an Excel for an index (IBICAT2b) that has been validated throughout Catalonia and recently throughout the whole Ebro River basin (Bae et al. unpublished data).

2 Comparison of Electrofishing Crews

Understanding the differences of catchability is particularly important for intercalibration of fish data from various research groups as well as computing fish indices. Several studies have been conducted to balance the compromise between representativeness of fish assemblage in the sampling area and sampling cost (e.g., time, staff, and expenditure), including the comparison of single- vs. multiple-pass electrofishing over various habitats (e.g., [25–30]), and the analysis of electrofishing equipment type (e.g., [31]) and suitable sampling length [30, 32–37]. However, little attention has been paid to assess the differences of catchability among electrofishing crews and equipment and the effects of sampling frequentation in Mediterranean regions.

We compared capture efficiencies based on standard fish descriptors (abundance, observed fish richness and species composition) obtained from four different fishing crews in Mediterranean streams [12]. In eight sites at headwater and middle reaches of a Mediterranean river, we sampled fish in two adjacent stations which had the similar habitat condition at each site using two different methods (single-pass electrofishing without block nets vs. four-pass electrofishing with block nets). During the first fishing day, two different methods were applied, but during the rest of the days only the single pass was applied in order to compare the effects of the consecutive sampling on fish abundance and assemblage structure. We applied a Williams’ crossover design, which is based on a Latin square design and is characterized by that (1) all crews are assigned only once to each sampling site during the four consecutive sampling days; (2) all crews are equally distributed; (3) it allows to test for potential carryover effects. We analyzed the differences in species richness, abundance, and proportional abundances due to the different catchability by the four research teams using generalized linear models (GLMs) with Poisson errors and log link functions (species richness and abundance) or binomial errors and logit link functions (proportional abundance). We also applied the software EstimateS (http://viceroy.eeb.uconn.edu/EstimateS) to estimate richness based on the removal estimates (i.e., four-pass electrofishing) using the second-order jackknife richness estimator (Jack 2; [38]), which is one of the most widely recommended estimators. Furthermore, we estimated population sizes and capture probability for the most abundant species in the four-pass electrofishing using program MARK using four different multinomial models (i.e., a model with constant catchability between different electrofishing passes (P), a model with constant catchability between electrofishing passes (P1), a model with nonconstant catchability between electrofishing passes (P1L), and a model with nonconstant catchability between passes and a quadratic function of fish length (P1L2)). These models were compared using Akaike’s information criterion [39].

Our results indicated that single-pass electrofishing was effective in the study area. It captured a large percentage of abundance (40–60%) as well as species richness (50–100%). Unsurprisingly, electrofishing was more efficient upstream than downstream and all species were generally captured in sampling sites with few species (i.e., headwaters). Furthermore, even though it is more difficult to detect all species in mid-river sections with higher species diversity, single electrofishing showed also high catchability there. Although observed species richness was not significantly influenced by the use of block nets, average CPUE was significantly higher using block nets. In addition, observed species richness was not significantly influenced by the research team, fishing day, or carryover effects. However, total CPUE depended on fishing day, crew, carryover effects, and site. Catchability varied depending on species, size, and removal passes.

In summary, single-pass electrofishing can be adequate to estimate abundance, species composition, and richness in headwaters and middle courses of this Mediterranean region. However, various methodological factors (e.g., reach length, number of passes, fish size, and species) influence electrofishing capture efficiency. Our results also show that the effectiveness of electrofishing depends on fishing crews because of different personal skills and practice. Therefore, electrofishing sampling protocols (e.g., sampling time and effort and equipment type) should be standardized as much as possible to get comparable data [24].

3 Comparison of Habitat Assessments Among Sampling Teams

The assessment of habitat quality is essential in fish studies because each fish species often has specific habitat requirements [40] and altered habitats are considered a major disturbance in aquatic ecosystems [41]. Therefore, habitat assessment has been developed as an integral part of stream biological monitoring [42–45]. However, because habitat assessments are often based mostly on visual observations or a minimal amount of measurement [45], the variability of assessments frequently occurs among researchers (even experienced ones). We compared the differences in scoring the habitat characteristics among four research teams. Each research team conducted the habitat monitoring with the same protocol at each site after finishing the electrofishing described in the previous section. Each team surveyed hydromorphological descriptors, riparian vegetation, aquatic vegetation, refuge type, observed visual impacts, land use, and habitat based on a modified version of the US Rapid Bioassessment Protocol (RBI) [46] for Mediterranean rivers (Table 1), which was used during the sampling of the project to implement the Water Framework Directive (WFD) in Catalonia [23, 24, 47]. Table 1 shows the list of habitat assessment descriptors as well as the significance of the differences among four assessors and a measure of effect size (partial η 2). Of 49 habitat assessment descriptors, 12 were significantly different among the four research teams that assessed them independently (P < 0.05). Percentage of grass in the riparian vegetation showed the highest difference among research groups (Table 1, Fig. 1), and four variables (i.e., degree of clogging, erosion of margins (right and left), and width of riparian vegetation (left margin)) from the Rapid Bioassessment Protocol, which provides a detailed protocol to score these features, were also different among the four assessors. A multivariate test suggested that although overall differences among assessors were not significant (MANOVA Wilks’ λ, F 2, 18.5 = 5.482, P = 0.165), probably due to low power, they were more important (partial η 2 of 0.980 vs. 0.907) than differences among sites, which were significant (F 99, 18.5 = 3.698, P = 0.001) and very clear.

Table 1 Comparison of descriptors of habitat assessment among assessors. Degrees of freedom = 2 and 9
Full size table
Fig. 1
figure 1

Box plots of the scoring of % grass and % riffle among four research groups (see Table 1 for statistical analysis). Each box corresponds to 25th and 75th percentiles; the dark line inside each box represents the median; error bars show the minima and maxima except for outliers (open circles or asterisks, corresponding to values >1.5 box heights from the box)

Full size image

Roper and Scarnecchia [19] reported that although consistency of habitat quality evaluation is improved with uniform training, inconsistency increases among researchers, as the habitat types to be classified become more diverse. Hannaford et al. [45] showed that even if the evaluation of habitat assessment becomes similar among groups after equal training in a certain type of habitats, large differences are still observed in other habitat types. Our results also suggest that the scoring for habitat assessment can be highly inconsistent among different research groups even using the same habitat assessment protocol. Therefore, habitat assessment requires more clear and detailed criteria and more training to make a similar evaluation among groups.

4 Development and Comparison of Fish Indices: Type- vs. Site-Specific Approaches

In addition to IBICAT2010 (see [4] in this book), whose development was led by Nuno Caiola, two other approaches (i.e., a type-specific and site-specific) were attempted in Catalan rivers [23]. Type-specific fish indices are based on a classification of sites in a region on homogenous types based on environmental or faunistic features and use different metrics and scorings in the different areas. On the other hand, site-specific approaches do not use a classification and instead predict the reference fish metrics from the environmental features of the sites [48, 49].

The WFD requests that various biotic assemblage descriptors (e.g., metrics) should be integrated into a single index to assess ecological status [3, 50]. These indices should represent the status of impairment in a research area [51–54]. Community metrics (e.g., number of intolerant species) and trophic guilds (e.g., percentage of piscivores), which group species sharing a common ecological trait into a single variable, have been commonly applied to develop bioassessment metrics based on fish assemblages [52, 55] (Table 2). It is assumed that these traits respond to anthropogenic disturbances consistently across a wide spatial extent [53, 54]. In addition, unlike species composition, which varies strongly across regions and biogeographical areas [56], patterns from functional traits are mainly determined by environmental filtering (e.g., [55, 57–61]).

Table 2 Features of the freshwater fish fauna from Catalonia used for development and computation of the indices
Full size table

Most predictive models evaluating ecological status start from comparing the biotic condition at current sampling sites with the expected biota without anthropogenic disturbance or in reference conditions [49, 62, 63]. Thus, changes in biotic condition from anthropogenic disturbance can occur only when the range of variation (or response) in reference (natural) conditions is well known [64, 65].

In this section, we summarize the two approaches (i.e., a site-specific one, IBICAT2a, and a type-specific one, IBICAT2b) based on the same guild classification for the fish fauna of Catalonia (Table 2), which was based on a comprehensive literature review. Fish development was based on a database of 364 sites in Catalonia, visited during 2007–2008, of which 8 sites could not be sampled due to the excessive discharge, 45 sites were dry, 76 sites were sampled but no fish was captured in them, and 235 sites were sampled with fish captured. At the 311 sampled sites, the total number of species (NST) ranged from 0 to 13 (median = 2, mean = 2.3), the number of native species (NSN) was from 0 to 8 (median = 1, mean = 1.4), and the number of introduced species (NSI) was from 0 to 10 (median = 0, mean = 0.82).

For selecting candidate metrics, we carefully reviewed the literature including research papers and reports from different countries. In total, for the 311 sites, we computed 199 candidate metrics, which can be classified into four categories as in the original IBI development [51, 66]: species composition and diversity, trophic composition, abundance, and fish condition. All the metrics were in general computed both for native and introduced species separately and for all species together. The native/alien status was considered at the river basin level.

To validate the new indices with gradients of anthropogenic pressure, we used two different anthropogenic disturbance measures. First, we obtained an official statistic of anthropogenic disturbance (the risk of noncompliance measure, RI_AP) from the Catalan Water Agency (document IMPRESS; [67]). It summarizes many different disturbances such as hydromorphological changes, flow regime alterations, changes in land use and the riparian zone, and point and diffuse sources of pollution [47, 67]. Second, a principal component analysis (PCA) was also used to combine this risk of noncompliance with our local measurement at the sampling sites such as the sum of RBI scores, sum of visual impacts, dissolved oxygen concentration, ammonia concentration, and pH. The first PCA axis summarized well a gradient of anthropogenic disturbance (see [47] for details).

The site-specific approach (IBICAT2a) was developed following leading works in Europe [48, 52, 68]. To define the calibration set (low pressure), we followed the usual method (see, e.g., [69, 70]): only sites where none of the pressures (hydrological regime, river connectivity, morphology, toxic acidification, and nutrient organic inputs) was greater than 2, ranging from 1 (no pressure) to 5 (high pressure) were used. Among 369 sites in Catalonia, 49 sites fulfilled all these criteria (of which 34 sites had fish captures). Then, generalized linear models (GLMs), with appropriate error and link functions depending on the types of metrics, were used in the reference condition sites (calibration set) to develop the expected values of fish metrics given numerous natural environmental variables (climatic and topographic) that are not affected by anthropogenic disturbance. A stepwise procedure based on Akaike’s information criterion was used to select parsimonious, adequate GLMs. Then the observed values on the rest of sites are compared to the expected values (see, e.g., [71, 72]) to compute an index that ranges from 0 (worst conditions) to 1 (reference conditions).

From the numerous GLMs, we selected 10 metrics considering their significant correlation with anthropogenic disturbance (pressures), their meaningfulness in ecological terms, their complementarity (e.g., different organization levels), and relatively low collinearity. Although the detailed results and a tentative index (IBICAT2a) are given in Sostoa et al. [23], we considered that this index was not suitable because of a number of reasons: (1) the GLMs could not be cross-validated because of low sample sizes and considerable variability in the reference data and probably also because of the considerable environmental heterogeneity of Catalonia; (2) the metrics based on absolute richness and abundance metrics did not behave well (gave unrealistic expected results) probably due to low numbers of reference condition sites (which were mostly at higher elevations) and therefore the index only included relative metrics (i.e., percentages); and (3) dry and fishless sites were not well predicted by predictive models, suggesting many local pressures that are not well captured by available indicators. Therefore, although this approach has been successfully applied in France [48, 52] and across Europe [54, 52, 71] and could potentially be developed in Catalonia, the low sample size available of fish data precludes its current application.

5 IBICAT2b: Development of a Type-Specific Fish Index for Catalonia and the Ebro River Basin

We also attempted a simpler type-specific approach (IBICAT2b), whose results we consider much more reliable than IBICAT2a and that we have validated (through correlation with environmental pressures) throughout Catalonia [23] and the Ebro River (Bae et al. unpublished data). We recommend IBICAT2b as a regional fish index, until further data become available that allow developing a better index. This index uses the official river types based on environmental data that are also used for macroinvertebrate indices and other purposes in Catalonia (e.g., [67, 73, 74]), the whole Ebro River [75], and Spain in general (http://www.chebro.es/; [76, 77]) (Fig. 2, Table 3).

Fig. 2
figure 2

Official river types in Catalonia and the Ebro River [74, 75]. See Table 3 for the meaning of code abbreviations and further details

Full size image
Table 3 River typology and number of sites in each river type
Full size table

In order to select the metrics that reflected well the gradients of anthropogenic disturbance for each river type, we computed the correlations between PC1 (the anthropogenic disturbance described in the previous section) and all the metrics in each typology separately, which is a classical type-specific approach (see [70]). In this procedure, because the total sampling sites in some of the river types were very low (e.g., EP, GEM, GRPM, MMS, and RMS where the total number of sampling sites were less than 11), we used a coarser statistical criteria (P < 0.1). In RMS type, we could not calculate correlations because only two sampling sites were available (Table 3). To select the final metrics for the index in each typology, we considered its diversity (different organization levels and type of metrics), complementarity (as assessed with a principal component analysis, which showed different groups of metrics based on their correlation), and interpretability of results (a few metrics had relationships with PC1 opposite than expected). The final metrics selected are shown in Table 4.

Table 4 Metrics selected for the 10 river types. Metrics in italics represent metrics that increase with anthropogenic disturbance (and vice versa for metrics not in italics)
Full size table

These different metrics were scored following a number of approaches. The number of native species was scored based on expert criteria and the historical records of fish assemblages in Catalonia. For DELT anomalies, we used the traditional IBI scoring: 0–2%, very good; 2–5%, moderate; and >5%, bad [41]. For NIN_15cmintol, only presence/absence was considered, because densities were very low and often null despite a clear relationship with anthropogenic disturbance. For the calibration of the other metrics (i.e., PSI, PII, PIT, PIT_pisciv, PST_pisciv, PST_lithophil, PIT_intol, PST_SL, PTI_intol, PST_lithophil, PIT_rheophil, and PST_intol) (see abbreviations in Table 4), the same approach as in the site-specific approach (IBICAT2a) was used for the scoring of metrics. Figure 3 shows the relationship between one of these metrics (PIT_pisciv) and the anthropogenic pressure index in one of the river types (MHS). As shown in this figure, a quadratic model was often significantly better than a linear model. Using these models and the classes defined for the risk of noncompliance measure (RI_AP < 0.8, no risk; 0.8–1.2, low risk; 1.2–2, average risk; >2 high risk) in the IMPRESS official document for Catalonia [67], we predicted PIT_pisciv values corresponding to each threshold and thus obtained the scoring of metrics.

Fig. 3
figure 3

Relationship between % piscivorous individuals (PIT_pisciv) and anthropogenic pressure (lg_RI_AP: log-transformed RI_AP) in the MHS river type. Straight line: linear regression model (r 2 = 0.375); dashed line: quadratic regression model (R 2 adj = 0.646). A likelihood ratio test showed that the quadratic model is significantly better than the linear model (P = 0.0003)

Full size image

For all the other metrics, we applied the same procedure as with PIT_pisciv to compute the corresponding thresholds based on RI_AP. Finally, the average of the score for relevant metrics depending on river type was computed to obtain the index and the ecological status.

For large rivers (types EP, GEM, GRPM, and MMEC), we also give a “bad” status, if the study reach is dry or no fish was captured after an adequate sampling. There is published [78] and unpublished (personal observations) evidence that Catalan streams are sometimes dry artificially (due to human water abstraction). Conservatively, we only apply this “bad” status classification to large rivers that should be expected to never run dry or be fishless in natural conditions. For other river types, if the sites are dry or no fish was captured, no status is given, because this might be due to natural causes.

Although both indices (IBICAT2a and IBICAT2b) are very different in terms of the development procedure of indices, both indices showed a similar response to anthropogenic disturbance (i.e., the correlation coefficients were 0.41 for IBICAT2a and PC1, −0.36 for IBICAT2a and lg_RI_AP, 0.40 for IBICAT2b and PC1, and −0.33 for IBICAT2b and lg_RI_AP). There was also high correlation between the two indices (r = 0.71), although the relationship was nonlinear because many metrics in IBICAT2a often had values of 0 or 1, indicating that IBICAT2a should be revised with more reference sites to develop further the predictive models and underlying index. Even though IBICAT2a showed relatively high correlation with anthropogenic disturbances, it has several limitations (see section above) and should not be used. A map with the results of IBICAT2b in Catalonia is given in p. 120 of Sostoa et al. [23].

6 Protocol for the IBICAT2b Multimetric Fish Index

An Excel file is given as an online supplementary material to this book chapter (http://invasiber.org/EGarcia/IBICAT2b.html) for the computation of the IBICAT2b index in Catalonia and the Ebro River. The index should not be used in other regions unless it is validated for them (i.e., correlated with environmental pressures) and it should be first adapted for different fish faunas. The following steps should be followed to compute the index. They are automated if the data are imputed in the Excel file.

  1. 1.

    Obtain the river type of your sampling reach.

    River types for this index are the general ones official for the WFD across Spain: there are 12 different river types in Catalonia (Table 3) and 8 in the whole Ebro River basin (all of them also present in Catalonia). Note, however, that there is a minor difference between Catalan and Spanish types: type 15 corresponds to two different Catalan types. Furthermore, there are some reaches declared as heavily modified water bodies and without any official type. Find the river type of your sampling reach in Fig. 2.

  2. 2.

    If your sampling sites are in EP, GEM, GRPM, or MMEC river types and they were dry or fishless, ecological status is “bad” (IBICAT2b = 1, EQR = 0). If the sites were dry or fishless but belong to other river types, the status cannot be defined with this index. Otherwise, proceed to point 3.

  3. 3.

    Score each metric with the fish data from the study site.

    All metrics should be independently scored from 1 (bad) to 5 (very good) according to the following tables. Metrics 1–4 are common to all river types. The rest of metrics are for some river types only. If some metrics cannot be computed (e.g., metric 2 has not been measured), they can be omitted from the final average.

    Metric 1: number of native species (NSN)

    River type no.

    Catalan abbreviation

    Very good

    Good

    Moderate

    Poor

    Bad

    27

    MHS

    >1

    		 

    1

    		 

    0

    26

    MHC

    >1

    		 

    1

    		 

    0

    11

    MMS

    >1

    		 

    1

    		 

    0

    12

    MMC

    >1

    		 

    1

    		 

    0

    15

    MMEC

    >2

    2

    1

    		 

    0

    9

    RMCV

    >1

    		 

    1

    		 

    0

    8

    RMS

    >1

    		 

    1

    		 

    0

    10

    ZC

    >1

    		 

    1

    		 

    0

    16

    EP

    >3

    3

    2

    1

    0

    18

    TL

    >1

    		 

    1

    		 

    0

    17

    GEM

    >4

    4

    3

    2

    <2

    15

    GRPM

    >3

    3

    2

    1

    0

    Metric 2: percentage of individuals with deformities, eroded fins, lesions and tumors (DELT) abnormality [41]

     

    Very good

    Good

    Moderate

    Poor

    Bad

    DELT

    0–2%

    		 

    >2–5%

    		 

    >5%

    Metric 3: percentage of introduced individuals (PII)

     

    Very good

    Good

    Moderate

    Poor

    Bad

    PII

    0%

    		 

    0–5%

    5–20%

    >20%

    Metric 4: percentage of introduced species (PSI)

     

    Very good

    Good

    Moderate

    Poor

    Bad

    PSI

    0%

    		 

    0–5%

    5–20%

    >20%

    Other metrics: specific metrics for some river types. See Tables 2 and 3 for further abbreviations.

    Table 9
    Full size table

    Therefore, IBICAT2b includes 4–8 metrics depending on river type. Each metric is scored from 1 to 5 (1 = bad, 2 = poor, 3 = moderate, 4 = good, and 5 = very good).

  4. 4.

    The final index is computed as the average of all available metrics. To obtain the ecological status according to IBICAT2b, the following thresholds are used:

     

    Very good

    Good

    Moderate

    Poor

    Bad

    IBICAT2b

    ≥4.5

    3.5–4.5

    2.5–3.5

    1.5–2.5

    <1.5

    EQR

    ≥0.875

    0.875–0.625

    0.625–0.375

    0.375–0.125

    <0.125

7 Concluding Remarks

Another type-specific index (IBICAT2010), quite different from IBICAT2b, was also described in Sostoa et al. [23] (see also [4]). An adaptation of this index (IBIMED), so far (February 2015) not available in published papers, Internet reports, or software, was intercalibrated with EFI+ and the Portuguese fish index [79]. The differences between IBIMED and IBICAT2010 include the addition of some of the rest of Spanish fish species with their guild classification (to allow the computation in other river basins) [79] and apparently different thresholds for the EQR classes. IBIMED has only been successfully validated with qualitative environmental pressures in Mediterranean rivers and the Duero and not the rest of Spanish rivers and was only intercalibrated for Mediterranean rivers (excluding the Duero) [79]. Recent unpublished work throughout the Ebro River (García-Berthou and Bae, unpublished data) shows that IBICAT2b and EFI+ are more related to quantitative environmental pressures than IBIMED/IBICAT2010, which shows problems mainly in the typology and treatment of fishless or dry sites. However, these three indices are correlated and their values could thus be converted (e.g., IBICAT2010 = 0.2099 + 0.1398 IBICAT2b, IBICAT2b = 1.3849 + 2.941 IBICAT2010, r 2 = 0.411, P < 0.0005; EFI+ = 0.2686 + 0.1279 IBICAT2b, IBICAT2b = 1.8573 + 2.2129 EFI+, r 2 = 0.283, P < 0.0005). Overall, our work suggests that fish indices can be successful in Spain but research is needed to improve them and generalize them. The availability of further fish data, user-friendly software, and extensive validation are essential steps toward the improvement of these fish-based indices.