Clustering patterns of diatom communities, and response to TP
The PCA analysis showed that our sites were well clustered along a nutrient, and a secondary pH gradient (Fig. S2). The NMDS ordination analysis confirmed that the community structure of diatoms in Swedish streams was clearly related to both TP and pH (Fig. 2). While both factors were important, there was evidently a unique impact of each of them. Regarding TP, there was a gap between two clearly TP defined communities.
With our study sites well spread along a TP gradient, and TP being important for diatom community structure, a meaningful TP index could be developed.
Development of the Swedish phosphorus diatom index (PDISE)
The TP-range in our study was 1-902 (median 25, interquartile range 12–57) µg TP/l with a unimodal distribution (dip statistic D = 0.01, p = 0.69) (Fig. 3), thus sufficient to cover the known response range of Swedish benthic diatoms to TP (Kahlert and Gottschalk, 2014). The modeled taxa-specific TP-optima covered the range from 2.2 to 238 µg TP/l and showed an unimodal distribution (dip statistic D = 0.02, p = 0.13). The interquartile range for the taxa-specific optima was 18–54 µg TP/l with a median of 31 µg TP/l (Fig. 4). In contrast, the site-specific optima showed bimodality (dip statistic D = 0.03, p < 0.01), with two peaks having a minimum of sites in between (Fig. 5). In other words, a site tended to have either a diatom community with a community average of a low TP optimum, or vice versa of a high TP optimum. The number of sites having an average “lukewarm” medium TP optimum were few, showing a clear minimum, indicating a threshold at about 30 µg TP/l. This result was surprising, as both the distribution of the TP-range and the taxa-specific optima were unimodal, but it reflected the observed gap of dissimilarities between diatom communities (Fig. 2). To smoothen this strong diatom response and enable a linear indication of TP, the PDISE was developed in a way to bridge the gap, namely we used a square root transformation of a diatoms relative abundance to relax the gap along the TP gradient.
The final PDISE had a linear correlation with log10TP of r2 = 0.68 (Fig. 6). The outliers at very low and very high TP concentrations indicated that linearity was given for a range of about 4 to 100 µg TP/l. Before and after, the few existing data did not follow the linear model, but rather laid on a plateau. Distribution of PDISE values was still bimodal (dip statistic D = 0.04, p < 0.01) (Fig. 6B) despite our attempts of relaxing the gap between clustered diatom communities along the TP gradient.
The bimodal pattern of response
The bimodal pattern of the benthic diatom response assembling taxa in either a “low phosphorus” or a “high phosphorus” community, i.e. a diatom assemblage with an average of either low or high TP-optimum, has to our knowledge not been shown earlier. This pattern could not be explained by a gap of sites matching the minimum of the observed pattern nor by a matching pattern for the taxon-specific TP-optima. It seems as if diatom assemblages with intermediate, and also with very low or very high average community TP-optima, might be less stable, as there were much less sites with such assemblages. The unstable range in the middle of the studied TP gradient was found to cover the community optimum averages of about 20–40 µg TP/l, which is in agreement with other studies showing sudden changes in primary producer assemblages over the same TP range (Gottschalk, 2014 and references therein). Obviously, there are “lukewarm” diatom taxa preferring intermediate phosphorus concentrations, however there seem to be no stable “lukewarm” diatom communities under natural conditions, at least in Nordic streams.
The analysis of the diatom taxa composition of the sites with an average site-specific low, high, or intermediate TP-optimum showed indeed that a typical intermediate assemblage seemed to be lacking. The SIMPER results confirmed that the taxa contributing most to group differences were associated with either the low TP- or the high TP-group, whereas the group of sites with intermediate average site-specific TP-optima did not have a similar clear setup of distinct taxa, reflecting the gap between sites of similar composition shown earlier (Fig. 2, Supplement Tab. S2). The group of sites with high TP site-specific optima were characterized by a high relative abundance of Achnanthidium minutissimum (Kützing) Czarnecki group 3 (mean valve width > 2.8µm), Cocconeis placentula Ehrenberg incl. varieties, Amphora pediculus (Kützing) Grunow and Eolimna minima (Grunow) Lange-Bertalot in Moser & al. (now Sellaphora nigri (De Not.) C.E. Wetzel et Ector comb. nov. emend.), whereas the low TP sites were characterized by Achnanthidium minutissimum (Kützing) Czarnecki group 2 (mean valve width 2.-2.8µm), Fragilaria gracilis Østrup, Eunotia incisa Gregory, Tabellaria flocculosa (Roth) Kützing, Brachysira neoexilis Lange-Bertalot and Eunotia minor (Kützing) Grunow in Van Heurck. Only the last of the taxa contributing to a cumulative difference of 50% between the low, intermediate, and high TP site groups was typical for the intermediate sites (Staurosira venter (Ehrenberg) Cleve & Moeller). E. minor was also a prominent example of the taxa with a rather “lukewarm” TP-optimum (sensitivity class 3), however it was equally abundant in both intermediate and low TP sites.
PDISE has an improved response to TP compared to TDI
The PDISE’s response to TP was clearly improved compared to TDI currently used as supporting index in the Swedish standard method for diatoms. The PDISE was higher correlated (r2 = 0.68) and had a more linear response than the TDI. Whereas the TDI also showed a clear response to TP with a correlation coefficient of r2 = 0.51, TDI values tended to cluster heavily at both ends of the TP-range, with a gap in the middle with few values (Fig. 6C-D). The main reason for the better correlation of the PDISE is the improved classification of the sensitivity value of the diatom taxa, adapted to Swedish conditions. The majority of the taxa changed class (Fig. 7), with a tendency to be categorized in a higher class, i.e. indicating higher TP concentrations than in the TDI (Fig. 7). Among the ten most frequent taxa which changed class, we found E. minor and E. bilunaris which both moved two classes up, from indicating low TP concentrations in the TDI to indicating intermediate TP concentrations in the PDISE. On the other hand, T. flocculosa moved one class down, now indicating low TP concentrations in the PDISE. We can only speculate why a taxon might have changed class, i.e. showed a different realized niche than modeled in the UK for the original TDI development. One possibility might be that the somewhat different geochemical background and climate might have impacted the response of a diatom to TP, another that many of the Swedish taxa might have been too rare in the UK database to model a robust TP optimum. It is also possible that the taxa were occurring in a somewhat different nutrient condition due to interactions, e.g. competition, with other diatom taxa, in constellations not occurring in Sweden. Furthermore, several taxa with the same name might actually be species complexes represented by different cryptic species in the two different countries. These cryptic species then might have different ecological optima in the two countries, as we have shown for example for T.flocculosa (Kahlert et al., 2021).
Comparison of PDISE with IPS and TDI
The PDISE was well correlated to both Swedish standard indices IPS and TDI (Fig. 8A-B), indicating the possibility to implement the new index in the standard method, with the possibility to replace the TDI to guide the interpretation of the IPS as supporting nutrient index. The correlations of PDISE to both IPS and TDI were high on average, however some sites were obviously classified differently with the PDISE compared to IPS and TDI (Fig. 8A,B). Especially the correlation of the PDISE versus TDI and IPS was low for the nutrient poor sites, most probably because the PDISE classified those sites more correct (Fig. 6).
Comparison of the PDISE to other diatom nutrient indices
The PDISE is among the best nutrient indices developed so far, with the correlation to TP being among the highest reported for other diatom nutrient indices globally (Poikane et al., 2021). The correlation was better than for the current Swedish diatom indices IPS and TDI (Kahlert, 2011; Kahlert et al., 2021; Kahlert and Gottschalk, 2014). Thus, the PDISE met the expectations for a robust nutrient index adjusted to Swedish conditions and reflects about the optimum of phosphorus indication by freshwater benthic diatoms from environmental samples for Sweden. With that said, even the PDISE is not perfect, and one of the reasons is probably the bimodal pattern of response to phosphorus, which is a challenge when developing an index.
Further improvements of the PDISE
Still, many of the Swedish diatom taxa are for the moment not included in the calculation of the PDISE because their abundance was too low in the Swedish database to model an optimum. TDI on the other hand includes most of the taxa. The PDISE could potentially be improved by testing to use the existing TDI classification also for calculation of the PDISE, and test if the outcome might get even more robust. The TDI has been developed for a long time, and missing values for Swedish species might be replaced by TDI values in case the data coverage is better for those taxa in the TDI dataset. Another further development of the PDISE would be to calibrate it for the use in lakes. The benthic diatom method has been shown to well indicate the environmental status of lakes, and the diatom method is intercalibrated within Europe (Kahlert and Gottschalk, 2014; Kelly et al., 2014). Furthermore, benthic diatoms respond earlier, and are better correlated to nutrient changes than the planktonic community (Gottschalk, 2014; Rimet et al., 2015). Finally, we found that alpine streams, and eutrophicated streams with low pH, were underrepresented in our database. We recommend to test if the PDISE might need revision for different stream types, or lakes, as soon as sufficient data are available.
Use of the PDISE in environmental assessment
The PDISE is no index to assess a general degradation of a stream or lake. Instead, it is solely developed to reflect TP concentrations. Compared to a modeled reference concentration for phosphorus, the PDISE could however be used to assess a deviation from the reference status and indicate eutrophication. It could also be used to follow up countermeasures to minimize eutrophication. The PDISE will be especially valuable to indicate the TP condition upstream a lake to assess the risk for lake eutrophication, as the diatom method has the advantage over other biological methods that it can be used in all Swedish stream types.
Today, the Swedish benthic diatom standard is dedicated to assess ecological status as defined for the WFD (Havs- och vattenmyndigheten, 2018), with a focus to indicate a general degradation of a water body, and to cover especially the good/moderate boundary where measures have to be taken to improve the ecological status class. The current diatom method is not well adapted to indicate TP concentrations, or eutrophication alone, even if both IPS and TDI are correlated to TP. With the aim to cover the entire degradation range from slight increases of nutrients to heavy organic pollution and oxygen depletion, this being harmonized for all European countries, some of which having large problems with organic pollution, the current diatom indices are not harmonized with e.g. the plankton indices. Most of them have been explicitly developed to indicate eutrophication, being the main cause for cyanobacterial blooms, the main problem in many European lakes. With the PDISE, there now is a benthic diatom index which can be used to detect eutrophication in streams similarly to the planktonic indices in lakes. Furthermore, an adapted PDISE could be used in lakes as well for the assessment of TP. In this way, there now is the possibility to develop a better linkage between the recently updated chemical targets for environmental assessment with biological responses, and also to harmonize the diatom method to the method for plankton, avoiding sudden changes in ecological status class assessments just because of a change of the used biological quality element.
We have calculated the TP values for the different ecological status classes as a first attempt to provide more robust TP thresholds for the WFD (Table 1). We found the threshold of change from the low-TP optima diatom assemblage to the high-TP optima assemblage was at about 30 µg TP/l. Indeed, almost exactly this threshold has been found before as important change point where the diatom species thriving at low nutrient conditions were replaced by species preferring high nutrient conditions (Gottschalk, 2014; Smucker et al., 2013). Benthic diatom communities show additionally other thresholds of change as has been demonstrated in both studies. For example, a crossover of diatom guilds was found at 18 µg TP/l in Swedish lakes (Gottschalk, 2014), and a change from a reference community to more tolerant communities in Swedish streams at about 15 µg/l (Kahlert, 2014). The next major change point of diatom assemblage composition seems to be the crossover where general sensitive species are replaced by species defined as tolerant to a general degradation of the habitat. This point expressed in TP concentration was found for streams at about 51 µg TP/l by Smucker et al. (2013), who was also able to couple a change point for a clear increase of the benthic chlorophyll concentration to about the same TP concentration (58 µg/l TP). This crossover of these sensitive species to tolerant ones has also been found in Sweden at about the same TP concentration (56 µg/l, Table 1). This crossover point has then been used in the European intercalibration work for the WFD to mark the important boundary between good and moderate ecological status (Kelly et al., 2014; Kelly et al., 2009).
Table 1
Current Swedish standard IPS class boundaries, corresponding TP concentrations, corresponding PDISE values and from this derived corresponding TP concentrations
Ecological class boundaries using the biological quality element freshwater benthic diatoms
|
EQR
|
IPS
|
TP-range [µg/l]1
|
TP [µg/l] corresponding to boundary (derived from IPS)2
|
PDISE3
|
TP [µg/l]
corresponding to boundary
(derived from PDISE/IPS)4
|
Reference value
|
1
|
19.6
|
High: 10
(7.5–16)
|
|
2.37
|
9
|
High/good
|
0.89
|
17.5
|
Good: 31
(19–50)
|
~ 17.5
|
2.98
|
18
|
Good/moderate
|
0.74
|
14.5
|
Moderate: 76 (63–102)
|
~ 56.5
|
3.85
|
54
|
Moderate/poor
|
0.56
|
11
|
Poor: 102 (84–111)
|
No response to TP for > 100 µg/l
|
4.86
|
No response to TP for > 100 µg/l
|
Poor/bad
|
0.41
|
8
|
Bad:
no data
|
na
|
5.73
|
na
|
1Median & interquartiles of TP [µg/l] for the five ecological status classes of the WFD (implementation of Swedish diatom method (Kahlert et al. 2007) |
2derived from the interquartiles (Kahlert et al. 2007) |
3this study: PDISE = -0.29 * IPS + 8.0536 |
4this study: PDISE = 1.84322log10TP + 0.6493 |