I wrote an algorithm with two inputs: a sound file and a data file. The data I used measures the decline of biodiversity from the year 1970 and is updated yearly. It is known as the Living Planet Index (LPI) and is made by the World Wildlife Fund (WWF) . The LPI is a high-level overview indicator that gives yearly single values to measure the health of global biodiversity, showing the average relative abundance compared to 1970. This is calculated from thousands of species’ population sizes, densities (population size per unit area), abundance (the number of individuals per sample), or some proxy of abundance, such as the number of nests or breeding pairs. The lower the LPI value, the worse global biodiversity decline has become.
The algorithm splits the sound file into chunks, with the first chunk of the song representing 1970, and the second chunk representing 1971, all the way through to the final chunk of the song representing the most recent year in the data. In each chunk, soundwave values are randomly set to zero to damage the song proportionally to the LPI value for that year. This introduces a tiny gap in the song, which interrupts the sound. The lower the LPI value for that year, the more interruptions in that section of the song. The more damage to the natural world, the more interference in the sonification.
The sound input could be any recording presented in the right format. I was careful to focus on songs, poems, and natural world recordings that I felt would be familiar and emotional to my audience. One song seemed the perfect fit, Sigur Rós’s iconic 2006 song, Hoppípolla, used to promote the award-winning BBC series Planet Earth. The awe-inspiring visuals, the most expensive nature documentary series ever commissioned by the BBC, transported viewers to stunning habits rich in wildlife. But it represented a world under threat and the song is a key part of this story: the perfect song to vandalise.
However, in my first attempt the algorithm made the song sound fuzzy and low-quality all the way through, and it was impossible to tell how the LPI values changed by listening to the sonification. I had no idea how to fix it. My idea was at a dead-end.