ecosystems and global change

Talking allochthony in Cambridge and Lille

At the start of December, Andrew organized a three-day workshop in Cambridge, bringing over Brian Kielstra, John Gunn, Nikki Craig and Chris Solomon from Canada, Michael Pace, Grace Wilkinson and Stuart Jones from the USA, Jan Karlsson and Martin Breggen from Sweden, and Jon Grey from the UK. The aim was to see how terrestrially-derived organic matter (tOM) contributes to secondary production in aquatic ecosystems, by synthesizing global data collected from 594 observations on C, N and H, in over 10 zooplankton groups from 158 lakes in the northern hemisphere. Ultimately, the group would like to build a model for each consumer by lake. The ten limnologists worked all day (and all night!) but still had time to experience Cambridge, with time spent at the legendary Eagle pub (where great minds meet) and an exquisite dinner at Peterhouse, the University’s oldest college.

group

A few days later, while Jon, Martin, Jan, Nikki and Stuart traveled home, the rest of the group and I traveled to London to catch the Eurostar to Lille, where we attended the British Ecological Society (BES) and Société Française d’Ecologie (SFE) joint meeting. The meeting was the first of its kind and brought about 1,200 ecologists to the Lille Grand Palais convention center – mainly French and British, but also a lot of attendees from other European countries and the rest of the world.

Continue reading

Lakes turning to jelly

Our latest paper in the Proceedings of the Royal Society on the “jellification” of temperate lakes has gotten an impressive amount of on-line attention.  At the time of writing this blog, Altmetric scores it as the 22nd highest ranking paper ever published in the journal.  You can read summaries from the Washington Post, New York Times, Daily Mail, CBC, CBC Radio, and Yahoo, among others.  I’ve also done four separate interviews this week with BBC radio stations (BBC Radio 5, BBC Wales, BBC Cambridgeshire).  You can catch the latest, with the BBC World Service from the 26th of Nov, below:

 

The main finding of the paper is that a small planktonic animal named Holopedium glacialis has been dramatically increasing in two very different lake regions of Canada as the keystone grazer in these lakes, the water flea (Daphnia spp.), has been disappearing.  Our results show that this is mainly driven by declines in lake water [Ca].  Daphnia need large amounts of Ca to build their body shell, while Holopedium surround themselves in a gelatinous polysaccharide “bubble”:

An individual Holopedium with the jelly capsule clearly visible. (Photo: Ian Gardiner / E-Fauna BC)

This jelly also protects Holopedium from predators.  By contrast, Daphnia are increasingly susceptible to predators at low [Ca] because their ability to induce evolved defences is also impaired. Our analyses show how vanishing Daphnia have now left more algae uneaten for their competitors to exploit, allowing them to multiply in number.  Many media reports have picked up on this as Holopedium liking ‘pollution’, with low [Ca] somehow being the result of this.  But it is more in fact a legacy of pollution.  While we have curbed industrial emissions and reduced acid rain, the historical depletion of base cations from the thin soils of the boreal shield, have left behind much lower [Ca] than present prior to industrial activity.  Ca concentrations have consequently been falling across much of North America and Europe.

Continue reading

Fretting over R2 and outliers

Two things I’ve been thinking about lately.

First, outliers. A classical example of these arises in what is known as Anscombe’s quartet, four datasets with almost nearly identical properties (mean, variance, x-y correlation) yet very different when plotted:

Variation in Anscombe’s quartet despite nearly identical summary statistics.

John Kruschke at “Doing Bayesian Data Analysis” has a terrific example of how to fit robust regressions through these lines using BUGS, along with links to additional code from Rasmus Bååth.  It is definitely worth taking a look at these.  One of the key points is that for small sample sizes, where the population SD is really unknown, it is worth modelling data from a Student’s t distribution as opposed to a normal distribution, as it allows for more flexibility in outliers (such as in y3-y4 above).  These are easy to implement in BUGS and Stan.

 

Second, how worried should we be about R2 values?  I’ve been thinking briefly about this over the past week as I’m not getting my usual impressive 0.9 values.  Does this matter though?  What does a R2 really tell us?  Well, consider the example below:

x1 <- rnorm(10000)
y1 = 5 + 1*x1 + 1*rnorm(10000)
y2 = 5 + 1*x1 + 3*rnorm(10000)

summary(lm(y1 ~ x1))
summary(lm(y2 ~ x1))

If you run the code in R you will see that both models recover the estimated effects of x1 on y1 but that y2 has 3X larger error estimates around this effect.  The resulting R2 is ca. 1/5 that of the y1 model.  Does this mean the model is a poor fit?  Not really…  It speaks much more to the fact that the predictive power of the y2 model is diminished and there are wider prediction intervals, rather than there being anything wrong with the mean estimate of x1 per se.  I’d be curious to hear what others make.

 

Finally, I thought we should drop a mention of the student blog PLANeT, run by our Part II undergraduates.  It’s worth checking them out! They regularly post well written and engaging articles on the relationship between plants and society.

Summer of field work

Joanna Wolstenholme, a third year NatSci undergraduate, has just wrapped up seven weeks helping our field campaign in Canada.  She authored this entry, describing her experience.

Sudbury, on first inspection, is a rather spread-out mining town, inhabited by many trucks (most of them blue).  However the more you explore, the more remarkable the town becomes.  It is one of the few areas of the world where remediation has really worked, and the next generation will inherit a greener and cleaner city than the one that their parents inherited.  This remarkable change, from a barren ‘moonscape’ caused by years of acid rain (Sudbury was once the world’s largest point source of sulphur dioxide emissions, thanks to large-scale nickel and copper mining), to an area with burgeoning forest cover and recovering lakes, is a great success story that the area can be immensely proud of.

With this backstory, Sudbury, with its 330+ lakes, makes an ideal experimental location for a group dealing in ecosystems and global change. Our study lake, Daisy Lake, is perfectly set up for studying the effects of terrestrial influences on aquatic ecosystems.  Along its length, the shores and wetlands have recovered to various degrees. One catchment has even been limed – covered with calcium carbonate to neutralise the acidic soils, and so plant growth is relatively lush. Other areas, closer to the smelter at the north end of the lake, are far more barren; bare, stained rock predominates, with a few stunted trees.

One of the streams we study

One of the streams we study

In Daisy, we were studying eight stream deltas, each with very different personalities.  At each site Erik and I measured algae, sediment, and water. This all sounds very easy in theory, but in practice (as with any fieldwork, as I came to learn) things were far harder and more complicated… and often involved some rather novel solutions. If nothing else, this placement has certainly given me plenty of opportunities to stretch my problem solving skills!

My first job was to build algae-collectors, which were plastic tubes with cut up swim floats attached from which 6 microscope-slides dangled from fluorescent string. These floated on the surface, but we also sank clay pot holders as another surface for algae to grow on. We left these in the lake (on a beautiful sunny day) at each of the deltas and then returned to collect them 3 weeks later.  On a more high-tech note, we also made use of two chlorophyll fluorometers to characterise the algal species found in the water column and benthic layer. After several dry runs measuring the amount of algae on Erik’s office floor, we took them out to the lake, and used them at each of the deltas. The unseasonal amount of rain that Sudbury was experiencing, however, complicated things, and meant that in some sites Erik had to swim with the fluorometers, as we couldn’t reach the sediment from the boat.

This slideshow requires JavaScript.

As well as working on Daisy with Erik, I also helped Andrew collect additional data for his survey of terrestrial resource use by aquatic organisms. This meant going out to six other lakes around Sudbury, and six down in the Muskokas, to collect water samples, use fluorometers, and deploy and collect the microscope slide contraptions. Key to the project was collecting clean leaf and algal samples, to go off for stable isotope analysis, to allow Andrew to calculate the influence of the terrestrial systems on the lake ecosystems.

In order to grow clean algal samples without the influence of terrestrial DOM, we collected water from each of the lakes, then filtered it into jars and re-inoculated each jar with a small amount of unfiltered lake water, from which we hoped the algae would regrow. This seemed simple in theory, but involved hours of standing by a vacuum pump watching water drip through a filter. One night, we actually filtered water outside a hotel, so as not to set the fire alarms off! Safe to say we got many odd looks. However, the field trip down to the Muskokas was one of the best perks of the summer. We went down in September, almost at the peak of the colours changing, and had two lovely dry but crisp days. Driving down dirt tracks through beautiful forest, to find beautiful lakes to paddle out into was great fun, and a real adventure! It definitely offset the tedium of filtering.

This slideshow requires JavaScript.

 

The 'Hollywood sign' of Sudbury

The ‘Hollywood sign’ of Sudbury

At the end of my seven weeks here I am very sad to be leaving. It was a great experience, with plenty of messing about on boats, exploring new places, and making new friends. I have learnt a lot about the complications of fieldwork, how to solve problems on the fly with limited supplies, and just what really goes on behind those simple sounding ‘Materials and Methods’.

 

This article is also published on PLANeT.  Joanna’s trip was supported by funding from the Department of Plant Sciences and NERC.

Drooling over plants (literally!)

It is really nice to be able to write about our latest paper, which just came out in Biology Letters to a fair bit of fanfare.  It got picked up by Science, New Scientist – including in their 23 July print issue – and even the Royal Society had a piece about it.  BBC Radio recently had me out to talk to them:

Why so much attention?  Well the paper is one of those quirky but interesting scientific discoveries.  And it is based on drool.  Yup, you read that right, the saliva of large mammals.

Continue reading

Many measurements and few observations

Our University colleague Professor Sir David Spiegelhalter has written a brief opinion piece in the latest issue of Science on the future of probabilistic models, particularly for big datasets (think images or genomes).

Two points jumped out at me:

(1) Statistical problems have shifted from many observations (large n) and few parameters (small p) to small n and large p, creating pitfalls when testing large numbers of hypotheses.  This is because the standard “p-value”, which we’ve griped about in the past here, will declare 1 in 20 non-existent relationships “significant” simply by chance.  So procedures are needed to reduce false discoveries.  The bit that I didn’t really follow was why even bother minimizing false discoveries?  Wouldn’t an interpretation of effect sizes be more meaningful?

(2) Inferring causation from observational data will continue to be a challenge, especially when n gets cheap and p remains large.  Statistical theory to deal with causality will be needed more than ever, and thankfully, it is improving.  This is something we’re quite fond of having thought a fair bit about causality in the context of path analysis, structural equation modelling, and directed acyclic graphs (see our J Appl Ecol paper that just came out).  The problem, however, is that these approaches don’t come easy and I struggle to see how they can be used by non-statisticians (the models in our paper took years of faffing!).  Finding ways to make causal inference more accessible is going to be critical in the future.

Fish are a forest product

Our latest paper has just been made freely available in Nature Communications, showing that freshwater fish, an important source of nutrition for humans, are in part produced by forests.  The study focuses on small boreal lakes, which contain upwards of 60% of the world’s freshwater. It suggests that any reductions in forest cover in the boreal ecoregion, such as from industrial activities, will threaten the production of healthy fish populations.

Forest stream at Daisy Lake

Forest stream at Daisy Lake

Small streams that drain forest floors bring microscopic particles of vegetation and soil into water. These get broken down by bacteria, which are then eaten by small invertebrate animals that are main food source for small fish. The research uses a gradient of forest cover in Canada to show that more of this forest organic matter is brought into lakes as the surrounding landscape is vegetated. This produces more bacteria in the near-shore water, which can support more zooplankton, and thus provide more food to small fish. Young fish survive winters and escape predators better if they are larger, so these effects are predicted to carry forward into larger and older animals.

Trapping zooplankton

Trapping zooplankton

The research also uses natural variation in the molecular mass of primary production from land versus water to estimate the proportional of terrestrial resources used by fish. At least 34% of fish biomass was supported by terrestrial vegetation, increasing to 66% with greater forest cover. This suggests that fish increasingly use forest food subsidies as they become available in the small nutrient poor lakes that are characteristic of the boreal ecoregion.

You can read more about the work at the BBC, Planet Earth, weather.com, Al Jazeera America,  or even watch a video at the Weather Network.  It is great to see the work receive this type of reception!  It took over 2 years of my life to produce in collaboration with a number of colleagues mainly at Laurentian University’s Living with Lakes Centre.

Tree genomics meets biogeography

This slideshow requires JavaScript.

We recently returned from a trip to the impressive Umeå Plant Science Centre, where we learnt more about tree genomics.  UPSC is home to a lot of the cutting-edge work sequencing the genome of the European aspen (Populus tremula), though other species as well, such as Norway spruce (Picea abies).  Although work with P. tremula, particularly genome assembly, has been slowed by the much higher levels of sequence variation than in other tree species, it is producing a number of really novel findings, strengthened by large-scale latitudinal studies.

For example, a recent common garden experiment with genotypes from across Sweden found that foliar herbivores reflected the genetic structure of plant defense genes, with fewer herbivores on trees from the local region than those that originated further away.  During our trip, Benedicte Albrectsen kindly took us to see one of these more recent experiments, where different genotypes were being exposed to simulated nitrogen deposition to test how foliar metabolites associated with plant defenses might change in the future.  There was quite a clear N effect as you can see above!

We’re now hoping to start putting together some attempts at merging some of this growing genomic data with biogeography.  This isn’t a new idea by any means,  but we’re hoping to bring some fresh eyes to the questions.  As always, any thoughts are welcomed below!

 

 

Mixture models for bimodality?

Oikos kindly featured our latest paper.  See below!

Oikos Blog

Bimodality – the characteristic of a continuous variable having two distinct modes – is of widespread interest in data analysis. This is because, in some cases, we can use the presence or absence of bimodality to infer something about the underlying processes generating the distribution of a variable that we are interested in studying. In ecology, tests of bimodality have been used in many different contexts, such as to understand body size distributions, functional traits, and transitions among different ecosystem states. But a lack of evidence for bimodality has been reported in many studies. Our paper “Masting, mixtures and modes: are two models better than one?”, now shows that a widely-used statistical test of bimodality can fail to reject the null hypothesis that focal probability distributions are unimodal. We instead promote the use of mixture models as a theory oriented framework for testing hypotheses of bimodality.

Our interest in…

View original post 428 more words