Imagine the world as a sumptuous frieze, into which is woven—or embroidered, perhaps, as I’m liable to mix my textile metaphors—all the rich diversity of wildlife. Now imagine time advancing as the frieze is unrolled, from left to right, one year at a time. Each year is represented by a column of 100 stitches. As the biosphere comes under ever more intense attack, so does that tapestry come thinner. For each 1% of the wildlife we started with lost, the year loses a stitch. Until the tapestry is almost bare.
Using data from the Living Planet Index (LPI), which measures the abundance of wild vertebrates, to represent biodiversity, it would look something like this:
•Each column (from left to right) represents one year in the world, from 1970 to 2016
•Each stitch removed is the loss of another 1% of wild vertebrates from 1970 levels
•Data from the Living Planet Index: http://stats.livingplanetindex.org/
In total, from 1970 to 2016, the average populations of wild mammals, birds, fish, amphibians and reptiles—so, most vertebrates—fell by an average of 68%. It’s worth reading more about this terrifying collapse and the factors driving it in the Living Planet Report from which the data is taken[1]. Here (following Professor Miles Richardson – see Background) the aim is to convey its scope visually.
The LPI is a complex measurement, put together by the Zoological Society of London (ZSL) for the World Wide Fund for Nature (WWF) using data from over 20,000 populations of over 4,000 species. That’s the point of an index: it is itself a simplified representation, expressing an immensely complex pile of data in one number, often a weighted average to give an overall impression. That raises a few important caveats when reading our tapestry:
- A weighted average population loss of 68% does not mean 68% of individual animals killed, 68% of the species extinct, etc. It refers only to the overall picture.
- 1970 is the baseline year for comparison (chosen as the year from which enough records were available), not as the year in which biodiversity loss began.
As well as the World LPI, the report provides more specific indicators for different regions, environments and factors. Those it highlighted as the most devasting include:
- The regional LPI for Latin America and the Caribbean, down by 94%.
- The world freshwater LPI, down by 84%.
Here are their tapestries, unravelling faster still:
•Each column (from left to right) represents a year in Latin America and the Caribbean, from 1970 to 2016
•Each stitch removed is the loss of another 1% of wild vertebrates from 1970 levels
•Data from the Living Planet Index: http://stats.livingplanetindex.org/
•Each column (from left to right) represents one year in lakes, rivers and other freshwater environments, from 1970 to 2016
•Each stitch removed is the loss of another 1% of wild vertebrates from 1970 levels
•Data from the Living Planet Index: http://stats.livingplanetindex.org/
If this does convey something about the devasting scale of biodiversity loss to you, please feel free to share it, with credit and link to this post.
Background
Why visualise biodiversity loss in this way? Working on this graph has kept bringing to mind David Quammen’s famous analogy[2]:
Let’s start by imagining a fine Persian carpet and a hunting knife. The carpet is twelve by eighteen, say. That gives us 216 square feet of continuous woven material. Is the knife razor-sharp? If not, we hone it. We set about cutting the carpet into thirty-six equal pieces, each one a rectangle, two feet by three….
When we’re finished cutting, we measure the individual pieces, total them up—and find that, lo, there’s still nearly 216 square feet of recognizable carpet-like stuff. But what does it amount to? Have we got thirty-six nice Persian throw rugs? No. All we’re left with is three dozen ragged fragments, each one worthless and commencing to come apart.
Quammen’s point is more specific to the one (hopefully) made by the visualisation, of course. He’s highlighting habitat fragmentation as important driver of biodiversity loss: a fragmented ecosystem is a degraded ecosystem. As such, the violence his carpet undergoes is of a different nature to that needed to convey change in two dimensions.
But the question that this project has sought to answer comes from Professor Miles Richardson: how can the visualisations used to convey the reality of climate change data be adapted to communicate that other dimension of the ecological crisis that is biodiversity loss. His starting point is the warming stripes created by climate scientist Ed Hawkins (which you may recognise from the background of this website’s header). Instead of showing a line rising as the world gets hotter and falling as it cools, they colour years in shades of red or blue to represent how much hotter or colder they were than the average.
It’s important to acknowledge that the stripes lose something compared to a more familiar line graph in terms of the readability of precise data. For any given year, the line graph’s position will give you the temperature and its slope will give you the rate of warming. Try doing that with coloured stripes, and you’d be better off with the raw data. But in terms of communicating the big picture clearly, unambiguously and powerfully, the coloured stripes have an advantage. The contrast is so stark and the association of blue and red with cold and hot respectively so intuitive, that it’s almost automatic. The stripes clearly situates short-term fluctuations in temperatures within a long-term warming.
Richardson, whose work explores the relationship between human psychology and connection to the environment, created biodiversity stripes along similar lines. These fade from green (or, in the case of the freshwater LPI, blue) to grey, as biodiversity is lost.
Deliberately echoing the warming stripes, these stripes similarly have a psychological reality to them that a line graph lacks, and can clearly be a powerful way to communicate the scale and speed of biodiversity loss. Despite the criticisms that follow, I was certainly moved by it and inspired. At the same time, comparing Hawkins’ warming stripes and Richardson’s biodiversity stripes seemed to reveal interesting differences between the respective phenomena, data and colours that could point toward other, perhaps more fitting, representations.
The LPI doesn’t fluctuate as temperatures do, with few upward blips on its downward trajectory. As such, Richardson notes, the result of mapping this directly onto colour would be a smooth gradient, not stripy at all. To force it into stripes, he introduces random variation within the LPI confidence interval. This works, but it also suggests that the stripes format doesn’t fit biodiversity as well as it fits temperature.
As for colour, green is certainly associated with life—though mainly plants, which the LPI doesn’t include—fading to grey is a thing. Yet biodiversity loss is often associated with the expansion or intensification of agriculture; the cattle ranches and palm oil plantations that displace rainforests are green enough, as for that matter is England’s green and pesticide-soaked land. Perhaps more importantly, the eye doesn’t differentiate these colours as clearly as blue and red[3].
Richardson’s graph left me wanting to find an alternative to stripes. This meant finding a contrast abundance and diversity with its absence, and to visually delineate years from each other—in other words, to mark time in a discrete, rather than continuous, way. I decided to try representing biodiversity as a rainbow of multiple colours or patterns, with a single colour representing its loss, and the discontinuity from one year to the next through the appearance of new gaps in the pattern. I found myself picturing a mosaic losing its tiles, an array of lights blinking out and, finally, a tapestry.
Evaluation
In tinkering around with this idea[4], a problem quickly became clear. The use of randomness creates more, irrelevant, information, that several of the people I showed it to described as getting confused or distracted by. Why is a given “stitch” red rather than green, why is the gap at the top, not the bottom, why read vertically, not horizontally? To some extent these problems may be fundamental, but I took steps to minimise them: minimising the colour differences between stitches, heightening that between stitches and void, and having a large number of stitches to reduce the probability of persistent horizontal arrangements bridging columns and misdirecting the eye.
Another problem is that, while it’s easy for our minds to allow for the possibility of a stripe being even redder or even bluer than the ones we see, you can’t have a row any fuller than full. As such this could be read as making the baseline year seem like the year that biodiversity loss began, rather than simply the point we started measuring. Yet any time-travelling zoologist from an earlier decade would have had a much bleaker assessment of 1970 than this suggests. The only way of mitigating this effect that I could find was to make the horizontal lines between the stitches the same colour as the void, as a gesture to the life lost unmeasured prior to 1970.
You can be the judge of how far this succeeds. Any attempt to read precise information from the tapestries will no doubt offer a reminder of the value of straightforward graphs, and it certainly can’t compare to the elegant simplicity of the warming stripes. Yet I hope it still brings home the devastating overall trend of biodiversity loss in a clear and poignant way—or failing that, that it inspires you to try and find a better way.
One interesting possibility to explore might be to evenly, rather than randomly, space out the gaps in a given column, in a sort of colour gradient made discrete. And I’d love to see where someone who knew anything about real weaving—who knew how to meaningfully use words like “warp and woof” instead of “row and column”—could take the tapestry metaphor, or an embroiderer that of stitching. If you are working on something similar, or would be interested in collaborating, or simply have feedback, I’d love to see and hear about it in the comments below or by email at colourlessgrn@googlemail.com.
[1] Explore the LPI website https://livingplanet.panda.org/en-gb/ or download the report from https://f.hubspotusercontent20.net/hubfs/4783129/LPR/PDFs/ENGLISH-FULL.pdf
[2] from Quammen’s book Song of the Dodo. I haven’t read the book, but have encountered that citation in at least two other environment books: Silent Earth by Dave Goulson and Rebirding by Benedict Macdonald
[3] Almost all forms of colour blindness still allow for the differentiation of dark and light tones, and most can contrast cold blue-ish hues with warmer orange-ish ones, so pale blue to dark red is fairly universal. Green, however, is one of the most common colours to be indistinct for colour blind people, grey, by its nature, doesn’t contrast that much with anything. Source: https://www.tableau.com/en-gb/about/blog/examining-data-viz-rules-dont-use-red-green-together
[4] Following Richardson, I used Excel. Surprisingly, perhaps the most awkward part was generating appropriate arrays of random numbers. Once that was done, conditional formatting took care of the rest. Each column contains the integers 1-100 in random order. These are compared with the LPI for that year, and voided out if they exceed it. The rest are assigned a colour using the mod function. I experimented with different colour schemes, dimensions, and whether to reshuffle the gaps every year or have them persist in the same place from when they appear.
Green is a colour 
Leave a comment