Sonification of Tree Ring Widths

This week in Music 220a - Fundamentals of Computer Generated Sound - we had a homework assignment involving the use of time series data sets to manipulate control parameters for synthesized sounds.  The general term for this process is sonification, which essentially boils down to developing an auditory depiction of a bunch of data.

Sonification is an growing research area, particularly because our auditory system is highly  capable of processing multiple streams of information.  By developing audio representations of data, it is often possible to obtain additional insight into the complex interactions between measured quantities - insights that would be difficult or impossible to obtain from visual inspection of streams of numbers or graphs. 

For the class assignment, I chose to sonify multiple sets of tree ring width data, measured (by others) for 6 trees in California. The size of each data set varies depending on the number of years recorded, with the oldest tree in the group having data from 6000 BC to 1979 AD.  

 

The sound of each tree is represented by a slowed down sample of Max Mathews speaking a single word from the corollary to his famous theorem from the 1950s (any perceivable sound can be created using a computer.)  The corollary states that "Most of the sounds that one can make randomly with the computer are either uninteresting, horrible, or downright dangerous to one's hearing."  Words used in this sonification exercise (in order of appearance) include "sound," "randomly," "horrible," "make," "downright," and "uninteresting."

The successive tree ring width data points modify the playback rate of the corresponding sample, slightly speeding it up or slowing it down.

This recording starts with the oldest tree and jumps ahead 200 years with each iteration, skipping 200 data points.  The data is compressed in this manner in order to ensure that the final file is a reasonable length (and not 30 or 40 minutes...)  When the starting year of a new tree is passed, the tree joins the chorus. The data sampling interval is decreased at the end of the recording to allow the last several trees to "perform."  These trees have much shorter timelines and would not be heard for more than a few seconds at the initial sampling rate.

One note - this file sounds best when listened to on headphones.  It is a binaural recording, which allows the simulation of directional sounds (left and right, front and back).  It should sound as if there are four sound sources - two in the front, two in the back.  It is possible to listen on speakers, but various cancellations will occur and the much of the directionality of each sound will be lost.  The recording also sounds best on headphones with a decent bass response since the samples consist mostly of low frequencies.

  Concentric - (Binaural. Listen on headphones) by cloudveins

 
Normalized Tree Ring Width Data Sets (in order of appearance) from the Time Series Database:

• CA535.dat: Methuseleh Walk - GT Basin. -6000 to 1979
• CA506.dat: White Mountains - Bristlecone Pine. -5431 to 1963
• CA051.dat: Limber Pine. -42 to 1970
• CA507.dat: White Mountains - Bristlecone Pine. 800 to 1954
• CA087.dat. Kaiser Pass - Western Juniper. 1140 to 1981
• CA531.dat: Onion Valley - Foxtail Pine. 1027 to 1987