How many indicator species are required to assess the ecological status of a river?

Preferences of individual macrophyte species for specific environmental conditions make it possible to determine the degree of flowing water degradation (Szoszkiewicz et al. 2017). In every ecosystem, nutrient concentrations change under the influence of contaminants. Changes may also result from fluctuations in river water flow, seasons, weather conditions and the changing self-purification capacity of rivers (Westlake 1975; Dawson 1988). Water quality assessments based on biological characteristics, physical and chemical parameters are complementary, and reflect the state of an aquatic ecosystem. Aquatic organisms are permanently exposed to environmental pressure. If their sensitivity to a specific contaminant has been determined, then the degree of degradation of an aquatic environment resulting from a given pressure can be determined based on field studies, with macrophytes (similarly as fish) facilitating the identification of changes over a longer timeframe, as opposed to phytoplankton and zooplankton, which tend to respond promptly (Wiegleb 1979; Haslam 1982; Holmes et al. 1999; Ceschin et al. 2010; Szoszkiewicz et al. 2017).

Ecological assessment methods for rivers based on macrophytes are applied in many European countries. In the UK, the Mean Trophic Rank (MTR) is commonly used (Holmes et al. 1999). The research conducted in France resulted in the development of the Indice Biologique Macrophytique En Rivière (IBMR) method (Haury et al. 2006). The Polish macrophyte method for rivers is largely based on the British MTR method and the French IBMR method. Their Polish counterpart, using the Macrophyte Index for Rivers (MIR), was developed in 2006 (Szoszkiewicz et al. 2007; Szoszkiewicz et al. 2010) and implemented in the monitoring of flowing waters.

Regardless of the type of biological characteristic (index), in each case the primary problem relates to accurate estimation of the sample size. Spatial and temporal variability are the most important sources of uncertainty affecting the variance of indicator values and classification results for various groups of organisms (Staniszewski & Szoszkiewicz 2006; Staniszewski et al. 2006; Carvalno et al. 2013). The main sources of variability related to macrophytes are associated with the method of on-site data collection (Dudley et al. 2013; Kolada et al. 2011) and the accuracy of macrophyte identification and classification, particularly the estimation of their abundance and cover. Another important source of errors, so far neglected in the literature on the subject, is the underestimation of the sample size – meaning that not all the taxa present were found. On the other hand, new species records are sometimes discovered during on-site studies, which do not provide relevant information, because they may occur anywhere.

The development of the information theory initiated by Shannon (1948) was of great importance for the progress of the probability theory and mathematical statistics. Furthermore, it was also applied in ecological studies (Kullback 1959; Sherwin et al. 2017). The creator of information theory, Eryomin (1998), proposed new scientific directions of research related to information ecology, such as research into the value of information or identification of quantitative and qualitative criteria of information. The concept of entropy in ecology means the expected value of a discrete variable, which is the sum of products of frequencies of species at surveyed sites and their corresponding information values (or relative total percentage cover of a species over the entire study area and its corresponding information value). The amount of information obtained while finding a species at a site may be considered as one of the aspects of diversity.

The aim of this study was to characterize selected macrophyte species in rivers in terms of their indicator value, related to the information they provide in the assessment of the ecological status of rivers. In practice, this will facilitate the assessment of the site completeness in terms of reported taxa, based on the informative value of indicator species for the Macrophyte Index for Rivers (MIR). On this basis, a decision can be made as to whether studies at a given site should be repeated or whether they may be considered complete.

The paper evaluates macrophytes found in medium-sized lowland rivers in Poland in terms of their informative value reflecting the quality of an aquatic ecosystem. A criterion for the sample completeness required for the determination of MIR is proposed.

A total of 90 indicator species with specific indicator values L_i and weighting factors W_i were identified at 100 sites representing five water quality classes (Fig. 2).

It is estimated that the total aquatic flora in the analyzed types of watercourses (lowland, with sandy bottom substrate) comprises approx. 115 vascular plant species (Bernatowicz et al. 1969; Rutkowski 2008; Jusik 2012).

It should be noted that this study identified 19 strongly stenotopic species that were reported in only one quality class. Six species were reported in quality class I: Conocepohalum conicum, Hippuris vulgaris, Hydrocotyle vulgaris, Lysimachia thyrsiflora, Viola palustris, and Menyanthes trifoliata. Carexrostrata, Peucedanum palustre and Ranunculus lingua were recorded in class II. Among the reported taxa, Calla palustris, Calliergonella cuspidate, Enteromorpha sp., Nasturtium officinale, Potamogeton alpinus, Potamogeton obtusifolius, Ranunculus peltatus and Rhynchostegium riparioides were observed only in quality class III, Glyceria plicata – only in class IV and Stigeoclonium sp. – only in class V. Furthermore, a set of 37 taxa was characterized by a very high tolerance (eurytopicity) and comprised a pool of species common to all quality classes. Of the total number of 90 indicator species recorded, 66 species belonged to class I, 68 species – to class II, 66 species – class III, 53 species – class IV, and 50 species – class V.

The basic characteristics of the MIR index (maximum, minimum, mean and median values) in five quality classes were determined for the obtained database (Table 1) in order to present a general description of the environment.

The largest range (20.3), i.e. the MIR values were the most diverse at sites with waters of the highest eutrophic levels. Furthermore, the range for quality class IV was 17.2, followed by 14.4 for class III and 12.6 for class II. It can be observed that the highest disproportions occurred among the MIR values for the sites in class I. This may be due to the fact that at some sites mainly rare and scarce species of high indicator value were identified, while other sites were dominated by species typical of all water quality classes. Furthermore, it should be noted that the decreasing trend of the MIR mean value in the trophic gradient was maintained. Values of the above-mentioned measures are typical of medium-sized lowland rivers. It can be assumed that the data characterize well ecoregions established by the EU WFD (2000/60/EC): Central Plains (Ecoregion no. 14) and Eastern Plains (Ecoregion no. 16). As a good representation of lowland rivers in Europe, they constitute an adequate group of bioindicators for Europe (Klijn 1989; 1994; Groen et al. 1993; Kondracki 1995).

The L_i and W_i values for indicator species (i=1,…,90) and the obtained characteristics D_i, S_i, ln(1/D_i), ln(1/S_i) are presented in Table 2.

The highest recorded incidence of the ith species (D_i) was 0.0526, which shows that Lemna minor was a species commonly found at sites belonging to all quality classes: at 8 sites in class I, at 13 sites in class II, at 15 sites in class III, and at 19 sites in classes IV and V of the total number of 20 sites. For this reason, based on the relative frequency of occurrence, Lemna minor provided the least information (was the least informative) on the water quality status (ln(1/D_i) = 2.945). Similarly, this species was found to have the highest value of relative percentage cover throughout the study area (S_i), amounting to 0.0750, which confirms its lowest informative value related to its abundance (ln(1/S_i)= 2.590).

Taxa found in only one quality class (19 species) had the lowest D_i values, reaching 0.0007 or 0.0028, and provided the greatest informative value: ln(1/D_i) = 7.2492 or 5.8629. Similarly, these species had the lowest S_i values, ranging from 0.0001 to 0.0036, and were the most informative ones, with ln(1/S_i) ranging from 5.621 to 9.306.

Such results are consistent with the scale of species occurrence and dynamic trends (Rutkowski 2008). The abundance score denotes the abundance of species on a nationwide scale of five classes: 1 – very rare (1–10 sites), rare (10–100 sites), quite frequent (more than 100 sites), frequent in many regions, common throughout (or nearly throughout) the territory of Poland. Macrophytes that are highly ecologically specialized, such as Nasturtium officinale, Potamogeton alpinus, Potamogeton obtusifolius, Hippuris vulgaris, are listed as rare in the Polish Flora (10–100 sites in Poland, abundance score 2/5) or frequent only in some areas (abundance score 3/5). They bring the highest information values ln(1/D_i) and ln(1/S_i) in our research. Even if they are found as single macrophytes, they guarantee sufficient knowledge about a site. On the other hand, taxa that are often frequent in the Polish flora in many regions (abundance score 4/5) and the most common taxa (abundance score 5/5) (Rutkowski 2008), such as Lemna minor, Carex acuta, Carex acutiformis, Glyceria fluitans, Rhizocolonium sp., occur in each of the studied water quality classes and contribute little to the general knowledge about a site, which is also confirmed by their information value ln(1/Di) and ln(1/Si). In the case of species which, according to the frequency scale, can be regarded as frequent in many regions (abundance score 4/5) and can be found in every environment, the information vector remains a valuable indicator of the information at that site.

Entropy for all macrophytes found in all examined river sections was H_(D) = D’ ln(1/D) = 3.948, while entropy for the percentage cover over the entire study area was H_(S) = S’ ln(1/S) = 3.775. Values of the indicators S and D show a similar reaction to the detection of another species at a given site. The maximum value of entropy for a total of 90 identified indicator macrophyte species was 4.5. Furthermore, based on the Hutcheson test comparing the structural indicators, no statistically significant differences were found between the entropies H_(S) for the relative percentage cover over the entire study area (2) and H_(D) for all species found over the entire study area (3) (t=0.03112 and p =0.5).

The correlation coefficient between the information vectors ln(1/D) and ln(1/S) was statistically significant and amounted to 0.917. When such high values of correlation between the information vectors were recorded, without compromising the generalization value of the study, the discussion was limited to the analysis of one of these indices. Since in practice the determination of the percentage cover requires additional skills and, according to the literature, is a major source of error, the value ln(1/D_i) was selected. This confirms the correctness of the statement that further analyses are possible based on one of the selected parameters D.

The next step consisted in the standardization of data from Table 2, which resulted in the so-called species profiles for individual taxa. Standardization involves converting measurements expressed in different units into a scale expressed in the same measurement unit (the variable obtains an average equal to 0 and standard deviation equal to 1). The transformed data are presented in Table 3.

Two synthetic Perkal indices were constructed based on the species profiles (Table 3). The maximum value of the Perkal index Pe_Li,Wi was 4.438, while the minimum and the median were −2.936 and 0.208, respectively. For the information value, Pe_{ln(1/Si),ln(1/Di)} reached a maximum of 3.991, a minimum of −3.569 and a median of −0.009. The Perkal index Pe_Li,Wi assumes 22 different values for 90 recorded macrophytes. The lowest values are ascribed mainly to species found in all five quality classes. Taxa representing one group for the Perkal index based on L_i and W_i (Pe_Li,Wi), with the lowest value of −2.936, include Potamogeton pectinatus, Rhizoclonium sp., Amblystegium riparium and the alga Stigeoclonium sp.The Pe_{ln(1/Si),ln(1/Di)} for those macrophytes was −1.853, −1.281, 0.235 and 3.039, respectively. The latter, clearly different value characterizes Stigeoclonium sp., which was not found in all quality classes like the other species – i.e. it was identified only in quality class V. This may indicate the necessity to verify L_i and W_i values or to use additional informative values. Group IV was composed of the algae Cladophora sp. and Enteromorpha sp. with the identical value of Pe_Li,Wi = −1.418, but with markedly different Pe_{ln(1/Si),ln(1/Di)} values (−2.086 and 2.856, respectively). Similarly, this may confirm the reliability of values for the characteristics L and W or the inclusion of additional information on a given taxa. Group V comprises ten species: Myosotis palustris, Rumex hydrolapathum, Scrophularia umbrosa, Veronica beccabunga, Carex acutiformis, Potamogeton natans, Oedogonium sp./Ulothrix, Polygonum amphibium, Lysimachia vulgaris and Hippuris vulgaris, with Pe_Li,Wi equal to −1.309. The latter species has a much greater Pe_{ln(1/Si),ln(1/Di)} value of 3.991, reflecting its affiliation with quality class I only.

The highest values of the Pe_Li,Wi index (approx. 20% of the values) identify the species serving as the best indicators, because they occur in only one or a maximum of two classes. The decreasing values are closely related to the fact that a given species was reported in an increasing number of sites and had an increasingly large range as regards the river quality classes. The Perkal indices constructed on the information vectors for Pe_{ln(1/Si),ln(1/Di)} additionally distinguish species in terms of their information values ln(1/D_i) and ln(1/S_i). The control of the information values is therefore an additional crucial element in characterizing the investigated species and, in the context of the above analyses, it appears necessary to include, in addition to the standard characteristics of species, the information provided by the new indicator values. Such an approach may result in a reliable identification of biological indicators and thus produce accurate decisions related to environmental permits.

The proposed approach allowed for more precise ecological characteristics of macrophyte species. Species with the same values of L_i and W_i, may be indistinguishable when the MIR is determined, whereas in reality they may differ from one another in their ecological specialization. For example, the taxa Amblystegium riparium, Potamogeton pectinatus, Rhizoclonium sp. and Stigeoclonium sp. are of exactly the same importance for the Polish national macrophyte score of the ecological status (MIR) due to the same and lowest values of L_i and W_i. The use of additional information enables us to observe the fact that the first two species occur in five water quality classes, the third one occurs from class II to class V and the fourth one – only in class V. Similarly, macrophytes with higher values of the indicators L_i and W_i, such as Calla palustris, Caltha palustris, Carex vesicaria, Cicuta virosa, Eleocharis palustris, Equisetum fluviatile, Fontinalis antipyretica, Hydrocharis morsus-ranae, Iris pseudacorus, Nymphaea alba, Ranunculus trichophyllus, do not differ in terms of their indicators L_i and W_i, but they actually differ depending on their occurrence in water quality classes. Calla palustris was found in rivers of quite high ecological quality (class III), Caltha palustris occurred in classes I and II, Carex vesicaria was identified in the first three water quality classes, Eleocharis palustris was found in the first two classes. The species Fontinalis antipyretica, Hydrocharis morsus-ranae, Iris pseudacorus occurred in all five water quality classes. It therefore appears that this seemingly hidden variability should be assumed to determine the ecological status based on the MIR. In the current situation, it seems that these species should not be treated equally.

Table 4 presents basic descriptive statistics for information values of all macrophyte species found in the respective quality classes.

For the selected representative characteristic D_i presented in the table, the mean information value was determined for each of the river quality classes. It amounted to 4.192, 4.149, 3.883, 3.763 and 3.799, respectively, for the river quality classes ranked from the cleanest to the most polluted (Table 5), which on average gives an information value of approx. 4.0 for medium-sized lowland rivers. This value is a criterion used to determine the information threshold required for a given site to be considered sufficiently surveyed. It was assumed that irrespective of the quality class, if the total information value provided by all species found at a given site exceeds the value of 4.0, the site could be considered sufficiently inventoried.

It is assumed that a reliable macrophyte study requires the incidence of at least nine indicator species, but the number of bioindicators may be lower if they are the most sensitive ones, i.e. mostly stenobiotic species (Szoszkiewicz 2013; Budka 2018). The analysis of the number of taxa at the investigated lowland river sites and the mean value of information introduced by those taxa showed considerable variability in individual water quality classes (Fig. 3).

Figure 3

Figure 3 shows the relationship between the number of species and the mean information value at a given site.

Based on the conducted study, the mean information value at thirteen sites in river quality class I exceeded 4.0, with 6 to 28 species recorded at a site. It should be noted that at some sites, despite the large number of species found (e.g. 22 species at site 686), the information value was insufficient. On the other hand, there are sites where 6, 8 or 9 species were reported, but this was sufficient in terms of information to conclude that a given site was complete. Twelve sites in quality class II exceeded the threshold for the mean information required for a complete inventory (including two sites with only 7 and 9 species). The other sites did not have a sufficient number of species or a satisfactory information value. Eight sites in class III reached a sufficient information level, including one site with only 8 species. At the other sites, a sufficient information level was obtained following a considerable effort to complete the study (as many as 23 species recorded at a site). Only five sites in quality class IV were completely inventoried (including 8 species at one site), while the information threshold at the other sites was too low to consider the sites as completely inventoried. Five sites in quality class V achieved a high information value (from 4.2 to 4), while in the other cases the species identification must be considered incomplete. Consequently, these sites are not reliable for further analysis (Table 4).

The slow natural response of biological indicators to changing environmental conditions and the relatively small possibility of increasing the sampling frequency for biological parameters are an issue for all European water monitoring systems. Therefore, an attempt to solve some of the methodological dilemmas related to sampling, necessary to obtain reliable assessment results is very relevant and may be the first step toward solving numerous ecological issues, such as the likelihood of misclassification of biological elements in surface waters (Loga & Wierzchołowska-Dziedzic 2017). In specific cases where the indicator value is close to the threshold value between good and moderate status classes, these analyses are of particular importance as their results are crucial for water management and water protection decisions. It should be noted that statistical uncertainty in the evaluated ecological status of homogenous waters based on biological parameters may be much higher than the uncertainty based on frequently measured physicochemical indicators. The requirements of the Water Framework Directive (WFD) regarding the determination of the status of watercourses based on biological indices (in particular MIR) may contribute to an increasing risk of errors in decision making related to water management based on the WFD. This study addresses an unresolved bioethical problem by attempting to answer the question “how many species are required to obtain a reliable assessment result with a minimum level of uncertainty”. Monitoring data used for this analysis represent a relatively large database of lowland rivers in Poland. Considering the results obtained while determining the criterion for lowland rivers, the proposed method could be easily extended to rivers of other types in Poland and Europe. The analyses could be performed for other indicators, such as the phytoplankton index (IFPL, Błachuta et al. 2012; Mischke et al. 2011), the diatom index (IO, Błachuta et al. 2010; Rimet et al. 2012) and the multimetric benthic invertebrate index (MMI, Bis et al. 2013; Lewin et al. 2013). With regard to these elements, there is no problem of insufficient data collection, as the relevant data have been collected by all EU countries for several decades.

Determining the completeness of a sample is consistent with the necessity to integrate many monitoring actions in order to systematize and make rational use of environmental monitoring data. The general concept of a hierarchical perspective was proposed by Loga (2012) and assumed that elementary measurement errors would be considered via a hierarchical structure of procedures applied to define the water status in compliance with WFD.