Does ‘stacking’ work for displaying trends? Our vision on how to plot typical energy data.

BP launched on the 11th of June their 2019 Statistical Energy Review. The data published every year as part of this review is very valuable for anyone working in the energy sector, with detailed information about world energy consumption (and production), CO2 emissions by source, per geographical area, per country, etc. Along with this information, BP releases a report, where the data is visualised through different graphical displays. Looking at the 2018 edition, we have seen that they use a very typical graph to report energy trends: the stacked area display, which is also used by other very well known institutions like the International Energy Agency (IEA). Nevertheless, staked area graphs –actually stacked graphs in general– are quite ineffective. Let’s explore why.

We will discuss the particular graph below, which shows the World energy consumption by source, from 1992 until 2017. Each source is represented by an area, and each are is stacked one on top of each other until reaching the total consumption.

What are the main problems of such a graph? Namely:

  • A colourblind person will not be able to read it.
  • It’s very complicated to ‘isolate’ the evolution of one particular energy source.
  • Specific events for a certain energy source cannot be distinguished.

The colours and design are not suited for a colourblind person

The display uses a legend to identify the energy sources, which forces reading the graph in two steps. Furthermore, the chosen palette of colours is unfortunate for a colourblind person. Here’s how a protanope (red-color blind) would see it:

Oil (originally in green), nuclear (originally in orange) and renewables (originally in light orange) cannot be well distinguished by a person with colour deficiency. As we discussed in this post, colourblindness is much more common than we thing. By using legends and not paying attention to the selected colours, BP’s graph is unreadable for a great part of the population.

The evolution of a specific energy source can’t be easily isolated

Would you be able to say whether nuclear consumption changed in the last years or whether it remain constant? What about hydroelectricity, how much did it increase? How does coal consumption relate to oil consumption? Answering to these questions is quite tricky with a stacked area graph.

We can’t figure out how specific events affected a particular energy source

Let’s take the example of renewables, which come after oil, natural gas, nuclear and hydroelectricity in the stacking. It is obvious that the financial crisis in 2009 influenced energy consumption, but how were renewables in particular affected? Were they actually affected at all? It is impossible to figure this out in the stacked area graph.

What would be an alternative to such a graph?

It all comes down to the message that we want to communicate. Ideally, the caption of a figure should give us a hint. Let’s look at what BP writes in their report:

World primary energy consumption grew by 2.2% in 2017, up from 1.2% in 2016 and the highest since 2013. Growth was below average in Asia Pacific, the Middle East and S. & Cent. America but above average in other regions. All fuels except coal and hydroelectricity grew at above-average rates. Natural gas provided the largest increment to energy consumption at 83 million tonnes of oil equivalent (mtoe), followed by renewable power (69 mtoe) and oil (65 mtoe).

None of these messages can be actually withdrawn from the graph… As an alternative to the stacked area graph, we have produced the displays below in order to unveil the real value of the data.

A line graph that shows the evolution of each energy source

Which clearly shows, among other messages, that:

  • As opposed to oil, coal and natural gas, renewables consumption did not decrease in 2009 due to the financial crisis.
  • Nuclear consumption has remained constant, and even decreased, in the last years.
  • Oil energy consumption is larger than coal; however, not as much as in the nineties since coal consumption started to substantially grow around 2002.

Small multiples to display and highlight the growth of each energy source

In order to display the growth rates of total world energy consumption and of consumption per energy source, we have utilised small multiples. The same plot is repeated seven times, each one of them highlighting one line. Total consumption is placed first; then, each energy source appears in descending order according to the growth rate in the last year. From this graph, it is clear that renewables are growing at substantially higher rates than the rest of energy sources. Although fossil fuels appear like increasing fast in the previous graph, their growth rates are not as high.

A line graph to compare energy consumption by geographical area

This line graph highlights the impressive growth on energy consumption in Asia Pacific since 2000, probably due to industrialisation in China. Its consumption almost doubles the one in North America. On the other hand, Europe has remained constant at around 2000 mote for the last 25 years.

A bar chart comparing growth per region in the last year

A bar chart allows quickly comparing different categories of one variable. If plotted horizontally, we have enough space for writing down the categories without having to tilt them. Moreover, we’ve shown them in ascending order so that it’s immediately evident which region grew the most, which the least and which ones were below or above average.

When producing this graph, we actually realised an error in BP’s report. They write that “Growth was below average in Asia Pacific, the Middle East and S. & Cent. America but above average in other regions”; however, Asia Pacific, the Middle East, Africa and Europe grew all at above average rates.

Our take-away lesson: focus on the messages, don’t follow the default.

BP’s data is so rich that has allowed us to produce four different graphs (and we could go on plotting), all transmitting a different key message. By using the ‘default’ stacked area graph, so often seen in the energy scene, BP ended up making a mistake when reporting the growth by area. As we highlight in our graph training, our take-away lesson is: plot the data, focus on the message (adapt the graph to it) and never trust the defaults.

Reporting election results in an effective way

Last 26th of May European elections took place. In Belgium, people also voted for the 150 deputies who are part of the Chamber of Representatives as part of the federal elections. The following day, newspapers, webpages and television news were flooded with graphs showing the results. But are these graphs effective in transmitting the key messages?

How the press reported the results

Let’s take an example. The graph below is extracted from the Belgian newspaper Le Soir and it is similar to the ones that La Libre Belgique and L’Echo reported. It shows the distribution of seats in the parliament, comparing the results between 2014 and 2019.

Results of Federal Belgian elections as reported by Le Soir here.

We believe that this is an ineffective way of reporting and contrasting election results for several reasons. The main problem is the difficulty to get the main messages after a first glance. The results of 2014 are shown in a light blue bar which does not work well when there has been an increase in votes. To get the numbers of 2014 we need to look at the number of seats in 2019 and reduce (if we get a green arrow facing upwards) or increase (if the arrow is red and downwards) the grey number.

In this bar chart it is almost impossible to appreciate (unless we look at it for a while) that the far-left PTB*PVDA party had 0 representatives in 2014 but 12 in 2019. Or that the extremists Vlaams Belang increased from 3 seats in 2014 to 18 seats in 2019. The green parties (Ecolo and Green) are separated by several bars, so we can’t directly appreciate that there’s been a green wave. In essence, bar charts like this one require quite a lot of time to understand and draw conclusions

What would be an effective alternative?

We’ve re-worked the graph of Le Soir into what Edward Tufte calls a table-graphic. The chart, when read vertically, ranks the 13 parties by number of seats in the parliament in 2014 and 2019. The names of the parties are spaced in proportion to the number of seats. Across the two columns, the paired comparisons show how the numbers have changed over the two election years. The slopes are also compared by reading down the collection of lines, and lines of unusual slope stand out from the overall downward pattern.

In this re-worked graph, the messages clearly stand out (as opposed to the bar chart):

  1. All parties, except four, have lost seats in the parliament.
  2. N-VA has lost its privileged position, going down from 33 to 25 seats.
  3. The far-left PTB*PVDA, which had no representation in 2014, has entered the parliament in 2019 with a total of 12 seats.
  4. The support to the Greens (Ecolo and Groen) has raised from 12 to 21 seats.
  5. The right extremists Vlaams Belang have also substantially increased: from 3 seats in 2014 to 18 in 2019.

Interested in data visualisation?

If you’d like to know more about how to get the most out of your data and produce graphs that effectively visualise your messages join our training: Figures of evidence: uncover the real value of your data by making effective graphs. We are offering an independently organised version of this training on the 28th of June in Brussels, register here, there are still places available!

Ready to try Agile for research?
Join our upcoming study

A PhD (or any research project) can feel at times as long and tiring as running a marathon. Yet, as we explained in this post, there are techniques you can use to alleviate the pain and turn your project into a series of dynamic sprints.

Laura Pirro, PhD candidate at Ghent University, realised the potential that Agile project management could have in academia. Agile has revolutionised the software-development sector and it is now used in many other areas (from manufacturing industry to the FBI). Why not using Agile for academic research too? In many aspects, academia is different than industry but Agile philosophy can still be applicable. Laura has worked on adapting Agile to research: you can find the details of the methodology she proposes in this post that she wrote for Nature Careers.

Agile for research has already proven success in Joris Thybaut’s research group at Ghent University (where Laura is pursuing her PhD). We are now looking for enthusiastic researchers and professors who would be happy to test Agile for research during a few months and let us know their results.

How could Agile for research work for you?

If you are a student

As a student, you probably understand that your advisors/supervisors are incredibly busy people with tight timetables. Still, you work hard on your project, and you often need advice on how to proceed with your research. However, setting up a meeting with your supervisor whenever you encounter an issue may feel like a struggle: it sometimes takes so long that by the time you get to meet her/him, you have already figured a way around the problem (although sometimes it might not be the ‘right’ path, so you need to go back and redo some stuff).

Agile for research will allow you to have continuous feedback on your activities in an effective way both for you and your supervisor. First, it will ‘force’ you to organise sprint planning meetings with all the stakeholders involved in your project: supervisor, postdoc, industrial partner. master student… in order to get everyone to agree on the goal and the duration fo the sprint. Thereafter, you will have short (15min) scrums every week with your supervisor so as to answer three questions: what was done the previous week to contribute to the goal? What will be done next week to contribute to the goal? And, are there any impediments? Finally, once the sprint is finished, you will meet again all the stakeholders for the sprint review, retrospective and planning.

If you are a professor/supervisor/student advisor

Following the work of students can be incredibly time consuming and difficult to fit in in your timetable. Furthermore, it is frustrating to dedicate time to students who (sometimes) still seem lost and have an unclear idea of what exactly they should do in their project. These tasks get even harder as the number of supervised students increases.

Agile for research will allow you to structure the way you communicate with your students. In a collaborative effort, the project is divided into layers of activities with an estimated duration of 2-12 weeks. In the sprint planning meeting, the specific goal and duration of the activity is decided with all the stakeholders (student, industrial partner, postdoc, etc.). Thereafter, you will follow the work of your students with short weekly scrums of only 15 minutes. Finally, in the sprint review, retrospective and planning all the stakeholders come back together to remove impediments, adapt to changes and plan the next sprint.

Does this sound appealing? Get in touch with us!

If you believe that Agile for research might be interesting for you, please get in touch with us or with Laura Pirro! We are conducting a study to see how this methodology fits in academia and we would be happy to solve any questions and provide additional information. We look forward to hearing from you!

Anthropomorphism or the art
of humanising nonhuman subjects

Academic writing should be clear and objective. In the pursue of objectivity, some believe that by using the first person and introducing ‘I’ or ‘we’ in their text, the outcome will not sound as rigorous or formal. But attempting to avoid the first person may confuse readers, leaving them wondering ‘who does what?’ as we discussed in our article about the passive voice. Focusing on objectivity may also lead to anthropomorphism.

Continue reading “Anthropomorphism or the art
of humanising nonhuman subjects”

Passive voice in scientific writing: angel or devil?

For years, we were told that in scientific writing we needed to use passive voice to sound formal, neutral and serious. More recently, the contrary philosophy bursted in: suddenly, passive voice had to be by all means avoided as it forces hiding the agent of the sentence and creates confusion. This paradigm shift left many of us in the doubt… is using passive voice in formal, scientific writing right or wrong?

Continue reading “Passive voice in scientific writing: angel or devil?”

Not in the mood to write? Why you should still show up, even if the muse doesn’t

Let’s face it, us, scientists, are passionate about our job. We are usually delighted about carrying out our scientific tasks (experiments, simulations, reviews, etc.). But when it comes to writing our findings, the motivation goes down. We rarely feel we’re ready to write and we rarely feel in the mood to write… the consequence: when we sit down and are supposed to write, we rather start doing other things, we procrastinate. And of course procrastination comes guilt and frustration. Until the deadline dangerously approaches: then, in the last minute, creativity pops up. Well, let us break it for you: that’s not really last minute creativity, that’s stress and adrenaline doing their job.

In our Road to Bootcamp series of posts, we’ve already covered how starting writing your work early enough will let you fully benefit from the ‘magic’ of the writing process; therefore, reducing procrastination. In this post, we’ll focus on how creativity can be boosted—even when you’re convinced that you’re not in the mood to write.

Continue reading “Not in the mood to write? Why you should still show up, even if the muse doesn’t”

Want to procrastinate less and be an effective writer? Start writing your articles early enough

If you ask researchers about their main issues when it comes to writing, procrastination always appears on top of the list. There are several methods that can help you become an effective writer who seldom procrastinates (or who effectively procrastinates—did you know that that’s possible?), so on our Road to the Writing Bootcamp we will be dedicating a series of blog posts to this problem. 

Why do we procrastinate when it comes to writing a scientific document? For multiple reasons, but many of them are related to the fear of the blank page, also known as writer’s block.

Continue reading “Want to procrastinate less and be an effective writer? Start writing your articles early enough”

Succeeding at your scholarship interview:
Advice from Prof. Alessandro Parente

We had the pleasure of interviewing Alessandro Parente, Professor at the Aero-Thermo-Mechanical Department of the Université Libre de Bruxelles (ULB) and frequent member of juries for the FRIA and FNRS fellowships. He talked with us about his experience as a jury member and he gave us some precious tips for students preparing for this type of scholarships.

Continue reading “Succeeding at your scholarship interview:
Advice from Prof. Alessandro Parente”

Are your documents colourblind friendly?

Did you know that one in twelve Caucasian (8%), one in 20 Asian (5%) and one in 25 African (4%) males are colourblind? For the case of women, the probability goes down to one in 200 (0.5%). Still, this means that there are always colourblind people among the readers and the audience of the reports, papers and presentations that you produce. In academia, assuming that your next journal paper is reviewed by three white males (which is rather likely given the population in science nowadays), the probability that at least one of them is colourblind is 22%.

Continue reading “Are your documents colourblind friendly?”