Sonification of weather data as a non-human-centric artistic approach

Method 1: Listening and data collection

Listening in the context of this paper draws closely on the methods that were promoted in previous deep listening practices.⁶ This involved observing and recording sonic events simultaneously when navigating through the sonic landscapes of different environments. In short, this method of perceiving defined soundwalking practice.

The process of soundwalking first involved observing and listening attentively to a different sound and sonic events that took place in the surrounding environment. This process was facilitated by using a digital audio recorder (Zoom H1n) which enables sound with different properties to be heard and detected as audio signal input. In other words, this facilitation intensified the listening experience, in which sound with lower amplitudes and high frequencies that were less audible, such as forest ambience, bird songs were amplified and heard through the device’s integrated stereo microphones. Thus, based on the properties of sound, the gain of signal inputs was monitored and adjusted accordingly.

Then, soundwalking proceeded with identifying and recording environmental sound with desirable sensible qualities, specifically surrounding ambience that was often perceived as ‘noise’ or ‘insignificant sound’. Based on the understanding that sound possessed distinctive qualities in different spatiotemporal settings, soundwalking was conducted in both natural and urban environments. Decisions such as moving towards or away from the sound source were made on-site based on the changing qualities of the environment. Each recording was recorded in 5-7 minutes, thus durationally these recordings can be perceived as sonic events that documents the multiplicities of sonic components emitted from non-human and human producers.

Soundwalking enables environmental events to be perceived primarily through sonic means, however, these events can also be perceived through other technological means. Hence, we identify weather as events that share similar sensible qualities with ambience recordings, that of being ‘unpredictable’ and ‘indeterminate’. In order to examine the changing process of weather events on a specific timeline, weather data of the soundwalking site was collected through a virtual open-source platform known as Open Weather Map (OWM) that integrates several data sources such as numerical weather prediction (NWP), weather stations and satellite data.⁷ The data was retrieved in numerical format through Application Programming Interface (API) calls and stored externally in comma-separated values (CSV) files, as shown in Figure 1. Hence, weather data collection can be seen as an extension of soundwalking – a method of environmental engagement.

Figure 1. Integrating OWM API in Python environment.

Method 2: Sounding, data analysis and sonification

Recordings of the sonic events were categorized based on the details of recording sites, sonic components and sources. For example, sound recordings of forests were categorized as the natural soundscape that consists of bird song, wind noise and hums, whereas sound recordings of cityscape were categorized as the urban soundscape that consists of noises of transportation and traffic as some of the major sonic components. These details were documented to ensure future navigation.

In order to narrow down the variables of recordings, this paper only focuses on the sound that shares similar sensible qualities with the environment. The sound emitted from other sources such as birds, people and transportation were not sampled or used for analysis. Hence, the sound of the surrounding ambience such as the noise of wind and other geophysical influences were sampled out from the recordings into a shorter duration. These sampled sounds were most often perceived as ‘background noise’ or rumbles, in which the sources of emission were arbitrary or unrecognizable. To examine the sensible qualities of background ambience, the acoustic properties of different samples were compared through audio analysis tools, namely a frequency and amplitude follower.

We selected samples from both urban and natural environments; the comparison is better visualized in the form of spectrograms. The values of amplitude (dB) and frequency (Hz) were annotated on each spectrogram. The analysis was done with Sonic Visualizer. The figures below depict the result of acoustic analysis across a specific duration: Figure 2 shows a higher fluctuation of both amplitude and frequency as compared to Figure 3. The comparison of acoustic properties was summarized in Table 1.

Figure 2. Spectrogram of waveforms, amplitude and frequency of sound recorded in a cityscape.

Figure 3. Spectrogram of waveforms, amplitude and frequency of sound recorded in a forest.

The difference in acoustic properties result in the difference of sensible qualities: samples of cityscape could be perceived as loud, chaotic and noisy, whereas samples of forest could be perceived as soothing, calming and unobtrusive. These identified qualities enable the genesis of idiosyncratic preferences, knowledge and sensitivity on the sensibilities of noise. Thus, it indirectly inspires and influences the decision-making process of creating sound generators in the sonification process.

Sound generators were created in Pure Data (PD) – an open -source visual programming environment to sonify different parameters of weather data that were collected previously. The sonification process translates numerical data into sonic outputs - in which sonic parameters of generators were modified by incoming data parameters.⁸ Each data point maps to the frequencies and amplitudes as shown in Table 2. Consequently, this mapping allows the changes of data points to be revealed and perceived across time.

To distinguish weather data parameters from one another, the parameters were assigned to three different sound generators which were synthesized to deliver different sensible qualities. In order to represent the distinctive changes of temperature data, a sawtooth wave oscillator with an amplitude envelope of short release time was used to deliver a sharper sonic quality. In contrast, to represent the fluid and irregular physical quality of clouds, a sine wave oscillator and white noise generator were used; enhanced with slight reverberation to ensure a smoother sounding quality. Lastly, rain volume was represented with white noise generators by structures of high-pass and low pass filters to mimic rainy ambience. Figure 4 shows examples of sound generators in the sonification system.

Figure 4. Sound-generators for temperature (left) and cloud coverage (right) in PD.

To afford control in the process of sonification, the system was designed with a few interactive parameters. The front-end of the system was organized into two parts as shown in Figure 5, the top part display control parameters for input weather data, which consist of two options of data sets and controllable data reading intervals (milliseconds), whereas the bottom part display volume control (dB) for sonic output. In consequence, the sonic output varies each time when the control parameters were modified.

Figure 5. Graphic User Interface (GUI) for controlling parameters of data input and output.

For a desirable result, the control parameters were set constant. Daily data was chosen from the dataset and its reading interval was set to 350 milliseconds, which means each data point was read and translated into sound between the given time intervals. Weather data parameters, namely temperature, cloud coverage and rain volume were sonified and exported separately into three (.wav) files with a duration of 3 minutes. Each sonification output was later visualized in spectrogram to draw a relation between weather data and sonic parameters.

Practicalizing sonic methods to perceive and re-enact sensibilities of environmental events

In this paper, we propose sonic methods as practical ways of perceiving and re-enacting sensibilities of environmental events. This proposition presupposes events derived from everyday reality are possible catalysts for artistic creation. Through practicalizing these methods, a relation was drawn between environmental components and intricacies that were perceived, both through sonic and non-sonic means. In the first method, noise, ambience, and weather were identified as specific components that share similar sensible qualities, that of being unpredictable and random in its materialities. That being the case, the noise was highlighted as a prevalent sonic medium that inspires the creation of sound generators which were later used to represent the sensibilities of weather data through sonification.

Environmental perception enables the formation of idiosyncratic experience, ideas, preferences, and affords ways of realizing and re-enacting sensibilities. In the second method, sonification was seen as a method in realizing and re-enacting sonic experiences. This method considers how weather events were experienced, for example, how temperature changes seemingly deliver an alarming effect, or how moving clouds might suggest a sense of wonder. Thus, each data parameter was mapped onto different sound generators to deliver distinctive sensibilities. Sonic output of temperature data was heard as distinctive changes in frequencies; cloud data was heard as subtle changes in frequencies, random changes in amplitude of noise; and rain was heard as ambience based on its occurrences. The outputs suggest an analogical connection between weather data parameters and sonic parameters, this connection is visualized in spectrograms, as shown in Figures 6, 7 and 8.

Figure 6. Comparison of changes in temperature data and changes in frequencies.

Figure 7. Comparison of changes in cloud coverage data and changes in amplitudes.

Figure 8. Comparison of changes in the volume of rain and changes in amplitudes.

The spectrograms suggested the most distinctive changes in the acoustic properties of temperature sonification as compared to others. This method of outputting sonification process enables weather data to be represented as distinctive parameters. However, the output altered if all processes were outputted at once, the combination of different sonic qualities will generate a new sonic event; abstracting the changes in data values, thus blurring the relation between the parameters.

Ultimately, both methods were found to be intricately linked to one another in the processes of creating sonic works. The transition from one method to another is found to be fluid and interchangeable as these methods consist of other undiscussed variables. These uncertainties can be further explored as creative avenues.