Ecological theories and breakfast cereal naturally go hand in hand, at least for us Conservation Science folks. We have a tradition of using cereal for scientific experiments (check out our results from 2015 and 2016), testing MacArthur and Wilson’s Island Biogeography Theory, and some of us among the teaching staff have been known to play “Guess the theory” during breakfast – a stimulating start of day putting your ecological knowledge to the test!

Surveying our island populations

Ready to colonise our islands!

Counting our species

This Tuesday morning, though, there was no question as to which theory we were referring to – Island Biogeography. In particular, we tested the ideas that the bigger an island is, the more species it can support, and the more isolated an island is, the fewer species that will have the opportunity to colonise.

The premise of our experiment is simple. While in our usual day to day lives it may be true that no man (or woman) is an island, for the purpose of this experiment, everyone was indeed an island, or at least they were responsible for one. Each student grabbed a container of a particular size and placed it at a random distance from the “mainland”. We then temporally abandoned our islands, returned to the mainland, which happened to be supporting a great abundance of breakfast cereal species. With hands full of cereal and lined up along the mainland, we turned our backs to our islands, and threw the cereal in their direction.

We set out our hypotheses, measured, counted, and then went through a quick coding exercise to unwrap the data presents!

Using linear models

Using linear models after excluding islands with zero species

Using smoothed loess curves

Fig. 1.  Species-area and species-isolation relationships presented using different analytical approaches.

Our data were quite zero-inflated as you can see from the first set of plots above. Overall, there was a trend for more species on bigger islands, and fewer species on more isolated islands. After seeing the plots, we discussed how we can improve our cereal experiment in the future. Perhaps we should have used more cereal and repeated our colonisation processes a few times to increase our sample size. It probably didn’t help that we were colonising our islands as hurricane Ophelia was passing through the UK – the winds could have carried away our cereal species in unpredictable directions. Most of the students chose small containers, so we had few data points for islands with large areas. On the third plot above, which we made by fitting a smooth curve,  you can spot an interesting hump-shaped species-isolation relationship. Well, we think we know why! Our islands may have been a bit too close to the mainland, so when we threw the cereal in the air, it would fly over the closest islands and be more likely to land in the islands at an intermediate distance.

Fig. 2. Species-area (A) and species-isolation (B) relationships for brown and white species.

Our breakfast cereal species came in different colours – brown (chocolate) and white (honey), but colour didn’t seem to affect the species-area and species-isolation relationships.

We wrapped up our morning of island hopping with another visit to the mainland, with our metaphorical mainland being Kluane National Park this time. We then took the roles of conservation scientists, tasked with estimating the size of the local brown bear population. Our resources were limited – a box of cereal, a tupperware container, a marker and our minds. Working in small groups, we designed a mark-recapture experiment. Here are the results:

Fig. 3. Population estimates of the brown bear population in Kluane National Park in Canada. Numbers derived using a mark-recapture technique and cereal as a proxy for bears. Black dots indicate actual population numbers. Error bars show standard deviations for Isla and Gergana’s groups, and standard errors for Mariana’s group.

We discussed experimental design, as well as its implications – for example Pedro and Zac’s groups didn’t calculate a measure of uncertainty around their estimates, and Gergana’s group ate their cereal, so they couldn’t count the actual population. Mariana’s group got the most accurate answer, and the first method Gergana’s group used led to the most precise estimate. The first method Gergana’s group used was to mark only 20 individuals, then for their second method, they marked 40, thinking that as more individuals are marked, the estimates become more precise, not the case this time though!

You can download the data we collected during this week’s ConSci session here – Island Biogeography dataset and Bear population dataset.

You can download our R Script here.

Keen to learn more about coding, models and data visualisation? Here are a few relevant Coding Club tutorials:

• Efficient data formatting & manipulation
• Making beautiful & informative graphs
• Intro to linear models

By Gergana

Imagine. You’re a consulting company hired by Kluane National Park in the Yukon Territory to conduct a survey of the Grizzly Bear population in the park by the Canadian Government. You have been given one summer to conduct the field data collection and a fixed amount of helicopter time – no you cannot just go and count every individual!  You need to provide a population estimate.

Now imagine, that the bears are Pom-Bears (crisps) in a bowl rather than a National Park and the tools you have to hand are a coloured pen and your own mathematical knowhow.  That is the scenario that each of the Edinburgh Conservation Science students were faced with a few weeks back.

Not only did the Cons. Sci. students need to figure out how to sample a population to estimate the total population of individuals. They had to figure out the mathematical formula to make their estimates without the aid of the internet. There were moments of struggle as numbers were jotted down, then crossed out, and sums done by hand, then checked with a calculator and minorly adjusted. But in the end six different tutorial groups came up with six different estimates with error for their populations of Grizzly Bears using a technique known as mark-recapture.

How close did each group come to the true number of bears in the park?  I will let the numbers speak for themselves!

Lots of samples or more marked individuals increases the accuracy of your estimate, but in the real world tagging extra bears is expensive. Most groups came quite close to estimating their populations, but some were farther off and some (Sandra’s group) forgot to count their bears to see how close they got! What did we learn – that precision and accuracy aren’t necessarily the same thing. Emiel’s group’s estimates are pretty precise, but not super accurate!

In summary, Pom bears are tasty and coming up with accurate population estimates can be challenging, if field sampling of real populations is limited by resources and time.

By Isla

Data for figure:

Code for figure:

library(ggplot2)

data

ggplot() +
geom_point(data=data, aes(x=Tutor, y=Actual, colour=”Actual Pop.”), size=6, pch = 17, show.legend=TRUE) +
geom_errorbar(data=data, aes(x=Tutor, ymin=Estimate-Error, ymax=Estimate+Error), width=.1, colour = “black”, show.legend=FALSE) +
geom_point(data=data, aes(x=Tutor, y=Estimate, colour=”Est. Pop.”, shape = 2), pch = 19, size=6, show.legend=TRUE) +
labs(y = “Population Estimate\n”, x = “\nTutorial Leader”) +
theme_bw() +
theme_classic() +
theme(legend.title = element_blank(), legend.text=element_text(size=25), axis.line.x = element_line(colour = ‘black’), axis.line.y = element_line(colour = ‘black’), axis.title.x = element_text(size=25), axis.text.x = element_text(size=25), axis.title.y = element_text(size=25), axis.text.y = element_text(size=25), axis.ticks.length = unit(0.3, “cm”)) +
scale_colour_manual(values = c(“Est. Pop.” = “black”, “Actual Pop.” = “red”), name = “Legend”) +
guides(colour = guide_legend(override.aes = list(size=6, shape = c(17,19))))

# Code credit to Anne for figuring out the tricky stuff to make the legend work!

Despite the autumn rains settling in, we spent one of our last classes island hopping. From Myasia to McIsland, Spiceland to Big Isi we wondered at the vast variety of life. And yet stopping off at Elbario, Demonstrator Island and Little Joe, we found them barren, desolate, in some cases with no life at all. Why could this be?

As you may have guessed, this week we were looking into a classic theory in ecology: Island Biogeography. Back in the 1960s, heart-throb scientists Robert MacArthur and Edward Owen Wilson were puzzling over the differences in diversity between islands – differences that seemed to show predictable patterns. They discovered that larger islands tended to have a much larger variety of species than smaller ones, but also that isolated islands had fewer species than those close to the mainland. The resulting trade-off has been a cornerstone of ecological theory ever since.

Island biogeography in graphical form.

One of the main challenges to testing this theory is the challenge of conducting experiments. Ethics boards are less than keen on allowing researchers to wipe out all island life nowadays (not that it hasn’t happened in the past – see Simberloff & Wilson, 1969 – https://www.jstor.org/stable/1934856), while waiting for recolonisation is not always within the realms of a typical three year PhD project. Of course, these issues become less problematic if you replace animals with cereal and natural dispersal mechanisms with a class-full of students.

Measuring the distances to our islands in our archipalago.

Using a laser range finder to measure the distance from the dispersal point to each island.

Dispersal in action of chocolate ovoids (Cereala cocoa) and pale corn-moons (Cereala lupina)!

The experiment conducted by our Edinburgh Conservation Science class was remarkably simple. Each student chose a tupperware container (henceforth ‘island’) and placed it in the grassy ocean outside the classroom. From a single point of dispersal on the ‘mainland’ we all threw cereal at the ocean, subjecting its chance of survival to fate and the reasonably stiff October breeze. Everyone then claimed their island, counted the number of colonising ‘species’, and measured both island size and distance from the dispersal point.

As a class we now had data to test the theory of island biogeography. We found that:

1. Species richness decreases with distance from the mainland (linear model slope=-0.77, p<0.05)
2. Species richness increases with larger island area (linear model slope=0.016, p<0.001)

In other words, a success! MacArthur and Wilson can breathe a sigh of relief.

The relationships between ‘island’ species richness, distance from the ‘mainland’ (left), and size of island (right).

We also discussed some of the problems with this dataset. Many of our islands were ambitiously distant in their grassy ocean, resulting in a highly zero-inflated dataset. This might give us the appearance of more certainty in our results that we actually have. In fact, if we remove islands without any species at all, the relationship between richness and distance is no longer significant (p=0.46). Another problem is the shape of the relationship. Just as expected by theory, cereal animal species increased non-linearly with proximity and area; fitting the relationship linearly therefore might not be appropriate. This is extremely important if we want to predict change or extrapolate from our data.

Finally, not all creatures are created equal. Some can swim, some can fly, some leave all the travelling to their offspring. While our cereal showed little of this diversity, we still had a go at seeing if there were differences between chocolate ovoids (Cereala cocoa) and pale corn-moons (Cereala lupina). There’s really too few data for anything to be statistically significant (and they’re in the same genus anyway so probably have similar dispersal mechanisms), but lupina seems to do better on large islands and have a greater dispersal capability. But we’ll be making no conservation management decisions based on this fact alone.

Difference in dispersal and colonisation ability for the two cereal ‘species’.

So all and all, what did we find? Whether it is breakfast cereal and tupperware or real species and real islands, the theory of Island Biogeography holds up!

By Haydn

Data: islandbiogeog2016

Code: island_bio

Biodiversity is a grand and wonderful thing, but it is also a notoriously hard thing to measure. What makes a habitat or landscape “diverse”? This was the topic of last week’s Cons. Sci. lecture, which touched upon patterns of biodiversity and the different metrics we use to quantify biological diversity in the world around us.

To make the lecture more interactive, we thought we would calculate diversity metrics live in class (*code and files below*), as we talked about them, on a dataset collected by the students. However, sending the students out “in the field” would have been impractical in the time frame we had – and frankly, Scotland in October is not the best time to wander around with a butterfly net. Instead, phone in hand, we sent them in the wondrous virtual world of Pokémon Go to see what they could find.

Our Pokébiologists had to record all the Pokémon species found in their vicinity every day for a week.

Students (and devoted tutors) had a week to make a daily screenshot of the Pokémon’s found around them, and were then asked to compile the species observed along with the time and location of “sampling”. Our research question was: “Does the diversity of Pokémon species vary across campuses of the University of Edinburgh?”

Abundance of Pokémon species recorded on Central and Kings Buildings campuses.

From a quick glance at our data above, we see that many more species were sighted at King’s Buildings, the southern campus, compared to the Main campus. However, this seemed mostly due to the fact that observers were mostly based at KB, which biased our sampling effort. We also noticed that some species were seen more often than others. Those Pidgeys, Weedles and Rattatas are everywhere!

We first calculated the alpha diversity in each building where observations were made, and although we did not follow up with a statistical test it seems that the Main Library was less Pokémon-rich than other places.

Alpha diversity of Pokémon species in buildings across the King’s Buildings (turquoise) and the Central (red) campuses. “n” represents the number of “plots” (i.e. screenshots) at each location. The Main Library appears less species-rich than the other buildings.

We then calculated the Shannon-Wiener index, which considers evenness (i.e. whether we have many species with similar abundances or just a few dominant species making most of the individuals recorded). The Shannon-Wiener index was 0.88 for both campuses, which means that the assemblages in both cases are quite diversified and even.

Finally, we wanted to have an idea of how these assemblages varied spatially: did buildings that are close together hold more similar communities than those further apart? We calculated pairwise Jaccard’s distances, which indicate dissimilarity, and confirmed our hypothesis that buildings from different campuses were indeed more dissimilar.

The Jaccard’s distance shows that buildings that are on different campuses (e.g. Drummond St vs KB Centre, or Main Library vs KB Centre) are dissimilar, that is, they do not share many Pokémon species. Buildings that are very close together (e.g. Murray Library and KB Centre) have more similar communities.

Although simple, this exercise allowed us to put into practice some diversity metrics, test hypotheses, learn bits of R code and reflect on sampling bias and the challenges of monitoring biodiversity.  What do these same sampling biases mean for biodiversity data collection in the real world?  – Our in class discussions taught us that this is a big issue facing biodiversity science, if we want a global monitoring programme to capture how biodiversity is changing on planet earth.

We have since then found out that another course had designed a – rather more ambitious – protocol to explore community ecology in New York City, also using Pokémon Go. We can’t wait to see their results!

By Sandra

*If you want to have a closer look at the code, you can download the R script and data files here

In the activity in today’s Biodiversity Change session, each of our five tutorial groups were given the same dataset from the Niwot Ridge Long-term Ecological Research site and asked to summarize the biodiversity trends and potential drivers of change.

We wanted to know if you give five different collaborative groups of scientists the same data will they come up with the same findings and interpretations.  And the answer was… we spotted many of the same trends, but our interpretations were sometimes quite different (see summaries below)!  Sure this was a speed science activity, as we had to do our analyses in 45 min or less, but I think this might have held true even if we had been given more time.

There are many different papers coming out of the Niwot LTER site that attribute the different trends in vegetation change being observed there to a variety of factors, so even the experts have different interpretations of pretty much the same data!

Niwot Ridge in the mountains of Colorado

Summaries of our different findings

Group 1
Species richness: Sites A1+B1 richness is increasing overall
Driver: Temperature and may be suppressed by another factor before the mid-1990s

• Logging?
• Fire?
• Nutrient Availability?

Species richness: D1 richness is decreasing
Driver: Temperature because extinction rates are greater than recruitment rates – species have nowhere to go

Future Research:

• Migration between sites (species composition change with temperature)
• Corridors
• Community Assemblages

Group 2
Species richness: Change is different between subsites
Driver: NO₃ deposition?, No strong evidence for temperature as a driver

Group 3
Species richness: Goes up, down and stays the same at different subsites
Driver: No trend in temperature over the long-term, temperature goes down from 2006 to 2014, Deposition data is variable and seems to track climate, but the change at the alpine site D1 could be due to pocket gopher herbivory and burrowing!

Group 4
Species richness: No clear trends
Driver: Nitrogen deposition – NH₄ is increasing, NO₃ is decreasing, climate?

Group 5
Species richness: Different trends for different plots – increase is greater for B1 > A1 > C1(negative) > D1 (negative)
Driver: General warming trend, increasing NH₄ but decreasing NO₃, glacier melt, tree cover change, and human disturbance might also be factors

By Isla and all five groups

The theme of this weeks class was the quantitative side of conservation science with a lecture on population ecology, the theory of island biogeography and metapopulations.  We talked maths, stats and computers covering the topics of matrix algebra, demography, stochasticity, hierarchical models, programming in R and more. I hope we learned that being quantitative doesn’t have to be scary, as maths aren’t really that hard, and they can even be fun!

Coordinating the data collection from our archipelago of tupperware containers in our island biogeography experiment.

We also got hands on with being quantitative and tested the theory of island biogeography by setting up our own archipelago of tupperware containers in the grass outside of the Crew Annex and pitching handfuls of star-shaped chocolate breakfast cereal our species immigration from a point representing our mainland.  We even used a laser range finder to measure the distances between the “mainland” and our “islands” to calculate our species area and species isolation relationships!  We learned that the theory of island biogeography holds pretty well for our model tupperware-breakfast cereal system with significant relationships for both the species-area and species-isolation curves in our data.  Check out our data, R script (see bottom of post) and super cool figures (see below) that we coded in about 5 minutes or less, the first intro to programming for some members of our class.

Our species-area relationship

Our species-isolation relationship

Then we broke off into tutorial groups to try our hands at mark recapture to estimate the number of grizzly bears (animal-shaped cereal) in populations in Banff National Park (five large tupperware containers) over time.  I asked the students to derive their own equations for the mark recapture experiment and that was somewhat of a challenge given the short amount of time that we had for the activity.  My group had some moments of frustration exclaiming that maths were indeed “hard, too hard” – but they worked through that to get some solid population estimates together. And, by the end of class, all five groups had produced estimates with error of the populations over time (see figure below from the whiteboard).  We learned that population censuses using mark-recapture techniques are a trade off between the number and sizes of the samples you collect and the precision of your results, and that there can be a lot of error when you don’t standardize your methods in advance!

The hand-drawn figure of our mark recapture data – check out those error bars and that outlier measurement in 2005!

So all and all, I hope we learned that conservation science is a quantitative science and that maths and being quantitative shouldn’t be something to be afraid of, but something that we all can embrace as ecologists/conservation scientists in training!

Our very first class R script:

## Island Biogeography
## Conservation Science 20-10-2015

# Set the working directory to the folder on your computer where you saved the data
setwd(“path to the folder here”)

# Load the data – convert from .xlsx to .csv before uploading
data <- read.csv(file=”Island_Biogeo.csv”,head=TRUE,sep=”,”)

# Make sure the dataset has loaded properly. “Island” is the name of your island, “Size” is the area of the bottom of your container (in cm2), “Distance” is the distance from your standing point to your island (in m) and “Immigration” is the number of species that were found in your container
head(data)

# Plot the species-area relationship – pch = 19 makes the points filled circles, col = sets the colour of the points
plot(data$Immigration ~ data$Size, pch = 19, col = “red”, ylab = “Immigration”, xlab = “Size”)

# Is the relationship significant?
lm1 <- lm(Immigration ~ Size, data=data)
summary(lm1)

# Plot the regression line on the graph
abline(lm1)

# Plot the species-isolation relationship – pch = 19 makes the points filled circles, col = sets the colour of the points
plot(data$Immigration ~ data$Distance, pch = 19, col = “blue”, ylab = “Immigration”, xlab = “Distance”)

# Is the relationship significant?
lm2 <- lm(Immigration ~ Distance, data=data)
summary(lm2)

# Plot the regression line on the graph
abline(lm2)

# Further beautification of these figures could occur such as relabling the axes, adding error around the linear model, etc.