Notes from Wastwater …

My field notes for my most recent visit to Wastwater include a just-decipherable scrawl of “usual Stigonmema/Rivulariaeceae” against one of the samples that I scraped from a rock just below the water level.  These are the dark patches, easily overlooked, found on only the largest, most stable boulders in the littoral zone of the least productive lakes in the country that I’ve written about before (see “Tales from the splash zone …”).   Each time I visit, I diligently scrape a few specks into a sample bottle to check.  And each time, I find much the same genera.  Except last week I found something I had not seen here before.

It was a cyanobacterium consisting of several narrow green filaments (technically “trichomes”) of cells all intertwined within the same sheath.  Watching closely, I could see the individual trichomes gliding back and forth within the sheath.  I recognised the genus as Microcoleus (see “How to make an ecosystem”) but had not found it in freshwaters before.   It is much more common in marine intertidal zones although a few species – including this one, M. lacustris – are found in freshwaters.   

The north-west shore of Wastwater, close to where the sample was collected.  This photograph was taken on a warm day in August 2020 when the shore of the lake was busy with holidaymakers.    The large boulder at the right centre is typical of those on which Microcoleus and other nitrogen-fixing Cyanobacteria are found at or just below the water level.  The photograph at the top of the post shows Wastwater from the south, taken in June 2020.

Microcoleus, like several of the other cyanobacteria I find in these habitats, is capable of fixing nitrogen although, unlike these, this does not take place in specialised cells called “heterocysts”.   You can read more about this in my earlier post on this genus but, for now, we will think about the bigger picture.  I’ve never seen a paper that sets out the role that all these unassuming mats and crusts play in ecosystems such as Wastwater so what follows is largely speculation.   But, as Wastwater is a very nutrient-poor lake, these Cyanobacteria are a potentially useful source of carbon and nitrogen for other organisms.   For various reasons, these organisms inhabit the littoral zone of nutrient-poor lakes, whereas we would expect to find their relatives in the plankton of nutrient-rich lakes (see “Both sides now …”).  There may be some invertebrates that are able to graze on these tough, unappetising patches of Cyanobacteiria; however, I suspect that the main route by which these organisms drive the wider littoral ecosystem in Wastwater is simply by organic compounds leaking out of the cells.   There has been a lot of work on this phenomenon in other ecosystems so it is probably a reasonable extrapolation to assume that this happens in Wastwater too.  This organic carbon will then be absorbed directly by bacteria and some heterotrophic protists and these, in turn, will be eaten by larger organisms.  

Microcoleus lacustris from just below the water line on boulders in the littoral zone of Wastwater, October 2020.  Scale bar: 20 micrometres (= 1/50th of a millimetre).

Quite how important this is in the grand scheme of things is hard to judge.   As soon as there is a sniff of nitrate from other sources – runoff from agriculture for example – these sensitive cyanobacteria will be outcompeted by other algae.   Even when the lake is as pristine as Wastwater or Ennerdale, however, these species only seem to thrive when there are large emergent boulders, and this is not the case for the whole perimeter of either lake.   Hanging on in the splash zone may mean that they can grab nitrogen from the atmosphere, so they are acting as pumps, sucking nitrogen into the lake ecosystems.  It may not be much, but not much added to the very little that is there already could be, proportionately, quite important.   

Some other highlights from this week:

Wrote this whilst listening to: Letter to You, new album by Bruce Springsteen and the E Street Band.  

Cultural highlights:  Enjoyed the gothic romance of Rebecca, just released on Netflix, which included Pentangle’s Let No Man Steal Your Thyme on the soundtrack.  

Currently reading:  The Poisonwood Bible by Barbera Kingsolver. 

Culinary highlight:   Homemade pasties filled with kedgeree, inspired by Great British Bake-off..

Bothersome biomass …

This brief post is really just an addendum to “The wrong kind of green?”, which I wrote earlier this year about the problems of measuring and assessing the quantity of algae present in streams.  Part of this post compared the results of visual estimates of algal cover with biomass, measured using a BenthoTorch and noted that there was often a mismatch.   One of the causes of these mismatches was when algae are missed in a visual assessment and this post describes one other situation that we’ve come across during surveys in west Cumbria.   

The substratum of the River Cocker, just below Crummock Water, is dominated by relatively large but very dark stones (possibly basalt) which give us consistently high signals for Cyanobacteria on the BenthoTorch.   This is the type of habitat where crustose Cyanobacteria such as Chamaesiphon thrive.  They are quite easy to spot when the substratum is light in colour (see “A bigger splash …”) but next to impossible to see against a dark backdrop.

Substrata from the bed of the River Cocker just below Crummock Water.  The photograph at the top of the post shows the River Cocker with its early autumn colours just showing.  

I can add two other situations when accurate estimation of cover of macroalgae (and, by extension, all larger plants) is difficult.  The first is when flow is elevated.   This is, in theory, can be avoided by only surveying when river levels are low.  In practice, organising field work always involves two of us juggling diaries and then monitoring hydrographs in the days immediately before we go out.   We always try to go out when river levels are low but It doesn’t always happen like that.  Earlier this summer, we did fieldwork over two days and a brief shower of light rain that was forecast for the night between turned into a thunderstorm that turned one of our rivers into a torrent where wading would have been dangerous.  Knowing that these were flashy rivers, likely to fall as quickly as they rose, we took the morning off while the level dropped back down.   But we don’t always have this flexibility in our schedules.   

The second is more prosaic: tiredness.   We are searching for organisms that may manifest themselves as small dark dots on submerged stones and I am concerned that, after a long day in the field, it can become harder to concentrate.   I know roughly how long I can keep going for and, slowly, learn that it is a false economy to cram too many site visits into a single day.   It is not an easy calculation: other variables include the distance I have to drive at the beginning and end of the day and the distance between the sites we are visiting.   

The message from the earlier post was that neither measuring biomass nor surveying cover offers a perfect insight into algal diversity and their contribution to primary productivity.  The answer would seem to be to use the two methods in tandem wherever possible.    Neither has a monopoly on portraying reality but, together, they can bring us closer to an understanding of what is really going on at the bottom of our streams and rivers.

Some other highlights from this week:

Wrote this whilst listening to: Wizkid, Lagos-based Afrobeats musician.

Cultural highlights:  Rewatched Parasite, winner of the 2020 Oscar for best picture.

Currently reading:  Boating for Beginners by Jeanette Winterson. 

Culinary highlight:   Nigella Lawson’s take on a Torta di Riso, from last weekend’s Guardian.

Dispatches from Plato’s cave …

Back in March I wrote about the mental processes we go through when identifying organisms, noting that this is not always a perfect process (see “Disagreeable distinctions”).  One of the claims of molecular ecologists is that their new technologies will be able to do this process much more efficiently (and precisely) than human analysts.   Sophisticated computer programs, termed “bioinformatic pipelines” sort the millions of sequences that emerge from metabarcoding analyses, matching as many as possible to sequences from organisms whose names we already know, in order to turn those sequences into usable data.    So long as there is a matching sequence in the reference library, there is a high probability that each sequence will be assigned the right name and, because metabarcoding analyses generally produce 30,000 or more sequences from a sample, compared to the 300-500 individual cells that human analysts count, stochastic variation, too, should be lower.  

That’s the theory.   However, someone has to write the bioinformatics pipeline in the first place, so there is a human element in metabarcoding output too.   When dealing with large numbers of samples (a nationwide assessment program, for example), several people are needed to analyse these by traditional microscopy, each with their own personal strengths and quirks. This means that  the variability between them would have been greater than if only one person – or one bioinformatics pipeline – had made all the decisions that led to the final list of species from a sample.  The question that follows on from this is what happens if, instead of one bioinformatics pipeline, we use several, each encapsulating a slightly different set of decisions about how to identify the target organisms.

When we tested this – using pipelines developed by teams in France, Sweden, UK, Croatia, Switzerland and Germany – we found that there was, indeed, variability, due to differences in the software used and also in the choices made at different steps in the process.   On the whole, however, the differences were minor and the overall outcomes – whether or not we would have classified a stream as being in good or poor health – were generally similar.   You can read more about this study in the paper I’ve referenced at the end of the post.

I was curious to see how the variation we saw between different bioinformatic pipelines compared with the variation we saw between human analysts, so I dug out some data from the UK/Ireland diatom ring test (mentioned in the previous post).   Half a dozen experienced analysts form the “expert group” for these tests, and the mean and standard deviation from their analyses set the targets which other participants have to achieve. We’ve done over 60 ring tests since 2007 so we have a good idea of the variability amongst analysts we should expect.   Maria Kahlert lent me the data collected from a similar exercise from Nordic-Baltic countries, and both sets of results are plotted against the results from the comparison of intercalibration pipelines in the graph below.

Results from the comparison of bioinformatic pipelines (Bailet et al., 2020) along with pooled results from two long-running diatom ring-test exercises.  Norbaf = Nordic/Baltic ring-test (standard deviations of two auditors; n = 23); UK/IE = UK/Ireland diatom ring test (standard deviations of 5-6 experienced analysts; n = 60); rbcL = standard deviation of output from six pipelines using the rbcL barcode (n = 29); 18S = standard deviations of output from five pipelines using the 18S barcode (n = 29).   Note that the UK pipeline was only configured for rbcL and UK/IE results have been divided by 5 in order to make valid comparisons with the other metric scales used in the study.

Differences in the details of how the two “human” ring-tests are organised may explain the higher standard deviation for the UK/IE exercise relative to the Norbaf exercise.  However, what is striking is that the standard deviations from both of these are much lower than the standard deviations for the comparison of the bioinformatic pipelines.  In other words, whilst any single pipeline gives greater precision than analyses by traditional means, there is lower precision amongst a group of pipelines than amongst a group of human analysts.  

The easy answer is simply to use the same pipeline for all analyses in a project.  If all UK analyses use the UK pipeline, then the original claim that metabaracoding is more precise than traditional approaches still holds.  The question that this exercise raises, however, is not about precision, it is about accuracy.   All bioinformatic pipelines are human constructs; all give slightly different results but is there any way of knowing which of the pipelines gives the most truthful account of the real state of any of the streams included in the study?  

Try this thought experiment: you commission five terrestrial ecologists to survey a number of woodland sites.   All, like our pipelines, send in broadly similar results but, on closer inspection, you find that one is consistently recording more primroses than the others, another does not record ash trees at all, a third overlooks rare sedges entirely, and so on.   No-one would use the same argument that has been used for bioinformatics: that we will get consistent results if we stick to a single surveyor, even though we know that each has a significant systematic error.   

The problem with diatom bioinformatics is that we don’t know the right answer.  Our traditional analytical approaches using light microscopy are also packed full of biases and assumptions (see “As if through a glass darkly …”) so simply striving for a good match between metabarcoding and traditional approaches will not be enough.   But that, too, may offer a crumb of hope: those of us who study diatoms have lived in Plato’s cave for so long that we, of all the biological disciplines, should recognise that analytical approaches can only offer an approximation of reality.  With that as a starting point, we should be able to agree on a common approach to the steps involved in a bioinformatic pipeline.  Then, at least, we’ll all be staring at the same shadow. 


Bailet, B., Apothéoz-Perret-Gentil, L., Baričević, A., Chonova, T., Franc, A., Frigerio, J.-M., Kelly, M., Mora, D., Pfannkuchen, M., Proft, S., Ramon, M., Vasselon, V., Zimmermann, J. & Kahlert, M. (2020).  Diatom DNA metabarcoding for ecological assessment: comparison among bioinformatics pipelines used in six European countries reveals the need for standardization.   Science of the Total Environment 745: 140948.

Some other highlights from this week:

Wrote this whilst listening to:  Jason Isbell and the 400 Unit.   Classic Americana.   

Cultural highlights:  Sweet and Lowdown, directed by Woody Allen, about a Django Reinhardt-obsessed jazz guitarist in the 1930s.   

Currently reading:  The French Lieutenant’s Woman by John Fowles. 

Culinary highlight:  a vegetarian version of the Persian dish Fesenjan – in our case, this was aubergines in a sauce made from crushed walnuts with pomegranate seeds. 

Indistinct distinctions …

My heart always sinks when I see negative word used as a species epithet because it offers a hint that those who originally described the species were struggling to find any particularly distinctive characteristics.   Amphora indistincta is a case in point: The genus Amphora is named due to a supposed resemblance to the clay jugs used to store wine in Classical times but the species name hints at desperation*.   And for good reason.  Until Zlatko Levkov described it in 2009, everyone would have classified individuals from this species as Amphora pediculus.   But Levkov had noticed that there was a subtle difference in the structure of the striae on some populations we were calling A. pediculus and decided that this justified creating a new species.  The problem was that most populations of A. pediculus are only a 100th of a millimetre long and you need extremely good optics in order to see the structure of the closely-packed striae.  We are faced with a similar situation to that which I described in Disagreeable distinctions, where even experts and experienced analysts cannot produce consistent results.  

Amphora pediculus and A. indistincta from Kenfig Pool, South Wales.   The image on the left is A. indistincta(note the uninterrupted striae) whilst the centre image is A. pediculus (there is a clear break in each stria) and the right hand image is probably A. pediculus but details are very difficult to resolve (photographs: Bryan Kennedy).  The image at the top of the post shows Kengig Pool (photo: Graham Rutt, Natural Resources Wales).

A twist to the tale was that Zlatko Levkov’s original description of Amphora indistincta suggested that it preferred less nutrient-rich habitats to A. pediculus.   Because diatoms are most widely-used to evaluate the effect of inorganic nutrients on lakes and streams, this is a potentially important distinction although, crucially, I have never seen any rigorous evaluation of the comparative ecologies of the two species that would support Levkov’s early statement.   I would go a little further: because we think that A. indistincta is associated with low nutrient environments, there is a risk of identification by association: when we are struggling to resolve the striae on a specimen and the rest of the assemblage suggests low nutrient conditions, then we are likely to err towards A. distincta rather than A. pediculus.   

Having encountered these problems when we used Kenfig Pool in South Wales as the subject for one of our periodic ring-tests, I thought that it might be interesting to explore ecological differences between Amphora pediculus and A. indistincta in a little more detail, using some data on Irish lakes lent to me by Bryan Kennedy.   This showed us that the two species often occur together, so a simplistic interpretation of the species having distinct ecological niches must be wrong.   When they do occur together, A. pediculus was usually the more abundant of the two species.   

Relationship between proportion of Amphora pediculus and A. indistincta in samples from Irish lakes where both are present.   The diagonal line indicates slope = 1.   Red = spring samples; blue = summer samples. 

I then divided the data into three groups: those with just Amphora indistincta, those with just A. pediculus and those with both and plotted the range of water chemistry associated with each.   In practice, there were only a couple of samples where only A. indistincta was found so this is really a comparison between samples with both and samples with just A. pediculus.   Not only are there no differences in the ranges of pH, alkalinity and conductivity between the groups, but there is also no difference in the ranges of the two nutrients, total phosphorus (TP) and total oxidized nitrogen, which we might expect if A. indistincta was restricted to low nutrient situations.  One caveat is that, as the graph for TP indicates, most of the lakes in this study would be expected to support at least “good status”, so it is possible that a dataset that spanned a longer nutrient gradient might have shown A. pediculus straddling a wider range than A. indistincta.  On the other hand, it might not.  

Perhaps what this illustrates, above all, is the danger of casual use of ecological descriptors in taxonomic accounts.   Using the OECD boundary values for eutrophication, few of the Irish lakes where Amphora indistincta is present count as “oligotrophic” (no more than 10 micrograms of TP per litre) and most are mesotrophic (10 – 35 micrograms per litre).   I should add that these values were developed primarily for deep lakes, rather than shallow ones and that whilst the UK TP boundaries suggest that A. indistincta is abundant in lakes where nutrient enrichment is not a problem, the Irish boundaries are quite a lot tighter than these.   Looking around Europe, ideas about what represents “oligotrophic”, “mesotrophic” and “eutrophic” vary considerably, meaning that they are not especially helpful terms to use when describing the preferences of diatoms.   And, as I explained in an earlier post (see “The challenging ecology of a freshwater diatom”), the changing state of freshwaters adds yet another layer of complication to the issue.

Range of values for five water chemistry variables measured in lakes where either Amphora indistincta (AIND) is present, A. pediculus is present (APED) or both species are present (BOTH).  “Cond” = conductivity; TP = total phosphorus; TON = total oxidised nitrogen.  Only sites where the sum of Amphora spp is greater than 2% are included.  Horizontal blue and green lines show the position of the high and good status boundaries for very shallow high alkalinity lakes in the UK.

There is a widely-held belief amongst diatomists that the wealth of new understanding about diatom taxonomy converts into a wealth of new ecological insight.  In my experience, this is not necessarily the case and there are certainly very few rigorous studies that support this.  That Amphora indistincta and A. pediculus are two distinct species seems fairly clear (there is now some molecular evidence to support this) but there is no particular reason why they should differ in their response to nutrients.  Maybe there is another factor – resistance to chytrids or to viruses maybe – which would mean that the information that they impart is less about nutrient levels per se than about the resilience of the community (see “baffled by the benthos (2)”).  That would also be useful information to a catchment manager.   Meanwhile, this is another instance of the well-worn truism that we know the shape (or, in this case, stria structure) of everything but the meaning of nothing.


Levkov, Z. (2009). Amphora sensu lato. H. Lange-Bertalot (ed.) Diatoms of Europe: Diatoms of the European Inland Waters and Comparable Habitats. Vol. 5., A.R.G. Gantner Verlag K.G. 916 pp.

Stephanek, J.G. & Kociolek, J.P. (2014).  Molecular phylogeny of Amphora sensu lato (Bacillariophyta): an investigation into the monophyly and classification of the Amphoroid diatoms.  Protist 165: 177-1965.

More about Amphora indistincta and A. pediculus (including Scanning Electron Micrographs) can be found at Diatom Flora of Britain and Ireland: 

* actually, I think the name more likely reflects Zlatko’s dry sense of humour.

Some other highlights from this week:

Wrote this whilst listening to:  Verdi’s La Traviata whose plot, lest we forget, hinges around the implications of a potentially fatal respiratory disease.   

Cultural highlights:  La Traviata, of course.   The music is glorious and, because the words are in Italian, you don’t need to dwell on the subject matter.

Currently reading:  The Walworth Beauty by Michèle Roberts. 

Culinary highlight:  Pasta ai funghi with homemade pappadelle