Devlog #39

Hello sonification fans! Welcome back to the Loud Numbers development log - a weekly newsletter where we talk about the process of putting together our upcoming data sonification podcast. You’re getting these emails because you signed up at loudnumbers.net, and you can unsubscribe at any time with the link in the footer.

The Ice is Getting Thinner

We didn’t send a devlog last week because it was Easter, but we took the opportunity to make a bunch of progress on the first episode of the podcast - The Natural Lottery. Long-term subscribers will know that this was actually the first sonification we made for the project, but when we came back to it after a year it wasn’t quite up to the standards of the others we made. So we’ve gutted it and rebuilt it almost from scratch with a simpler set of encodings.

For those who are new around here and haven’t caught up with the archive, our first episode tells the story of the Nenana Ice Classic - an annual contest in Nenana, Alaska that involves placing bets on exactly when the frozen Tanana river will melt each spring.

Here’s how it works - a tripod is placed on the frozen ice early in the spring, fixed to a rope attached to a siren on the shore. When the ice floes start to move and break up, the tripod is pulled down the river and the siren sounds to mark the winning date and time. There are big bucks involved - the jackpot regularly tops six figures. Feeling lucky? Tickets are on sale until April 5th, so there’s still time to enter this year’s event.

Here’s the amazing thing, though - the competition started in 1917! That gives us a huge long dataset to work with. On average, the tripod now falls considerably earlier in the spring than it used to because of climate change: 2019’s ice breakup date was the earliest ever. So this is a long-running dataset that shows climate change happening on a local level.

We decided to turn this data into a techno track, because Miriam is very into her techno. But we’ve found it really tricky to sonify, for two reasons.

Firstly, finding the right data-to-audio mapping to communicate something happening earlier each year has been a challenge. We’ve tried many options. In the first version of this track we used rhythm. The earlier the tripod fell, the earlier a sound played in a two-bar group. But this was too subtle to convey the warming trend effectively.

Next, we tried stepped tempo changes, so that the music got faster when the tripod fell earlier. Listen to the example we talked about in Devlog #32. It is pretty disorientating! Then we tried mapping the data to the speed and pitch of a short looping sound, but this didn’t sound quite right either.

So finally, we’ve fallen back on trusty old pitch: the higher the chord in our track, the earlier the tripod fell. It’s not a perfectly intuitive mapping, but it effectively communicates a sense of rising panic and urgency as the chords get higher towards the end of the track (the Logic synth we’re using is even called “Panic Arp”). Plus, you can hear the trend clearly and we think it sounds great. :) Have a listen.

Then the second sonification challenge we’ve had with the Nenana dataset is that, while the overall warming trend over decades is clear, the year-to-year data is very noisy, as you’d expect with any real-world dataset.

We’re using the ten-year average to smooth it out but, even so, there’s not a completely uniform trend in one direction. In the chart above, the long-term trend is easy to see. But it taxes the memory to hear it in a sonification which is several minutes long: you’ve got to remember what the value was at the start of the time series to be able to compare it with the current value. This is difficult! How to make it easier?

One option is to use a reference pitch throughout so that you can compare the current value with it by evaluating the width of the musical interval between them, a bit like visually comparing the height of a bar with an axis starting at zero. Or you can play the whole data range. For example, if you’re mapping data to pitches between C4 to E5 then you can play these notes in the background while the pitches that carry data play at the same time. Then you always have an audible pair of reference points.

Here’s some Sonic Pi code illustrating that, and here’s what it sounds like.

# Axis

use_synth :tri

play :C4, sustain: 5, amp: 0.2

play :E5, sustain: 5, amp: 0.2

# Data

use_synth :tri

play_pattern_timed [:F4, :G4, :B4, :D4, :D5], [1], amp: 0.6

This works well when the range is small (less than a couple of octaves) and the fundamental pitch of each data-carrying sound can be tied to a particular octave (so not multi-octave chords, or Shepard tones).

In a way, though, all music that’s in a key has a built-in reference pitch: it’s called the tonic, or root note. In the Natural Lottery track we’re making use of a tonic reference pitch in this looser way, as an orientation point, without mapping out the scale range precisely. We think it works. We hope you will too, when the episode is released on 5 June 2021.

Elsewhere on the web

• Another long-running local climate dataset: the Economist’s gorgeous chart of cherry blossom peak bloom times in Kyoto since 800 AD (!)

• A research opportunity for sonifying climate data. If you’re a sonification researcher, we’d definitely be up for partnering in some way on this, so get in touch.

• The first note known to have sounded on Earth was an E.

• What a Loud Numbers dinner party would be like.

If you’ve made (or just seen!) any cool sonification things on the web recently then we’d love to hear from you. Send us links - all the links.

xox

-Duncan and Miriam