Water is our most precious and most rapidly declining natural resource. We explore pervasive technology as an approach for promoting water conservation in public and private spaces. We hope to motivate immediate reduction in water use as well as higher-order behaviors (seeking new information, etc) through unobtrusive low-cost water flow sensing and several persuasive displays. Early prototypes were installed at public faucets and a private (shared) shower, logging water usage first without and then with ambient displays. This pilot study led to design iterations, culminating in long-term deployment of sensors in four private showers over the course of three weeks. Sensors first logged baseline water usage without visualization. Then, two display styles, ambient and numeric, were deployed in random order, each showing individual and average water consumption. Quantitative data along with participants’ feedback contrast the effectiveness of numeric displays against abstract visualization in this very important domain of water conservation and public health.
We developed an unobtrusive sensor that can be mounted onto existing showers or faucet pipes to measure water usage. The sensor was built using a low-cost, off-the-shelf microphone and an Arduino AVR microcontroller. Audio is sampled at 10kHz, and volume thresholding combined with a sliding sample window are used to detect water flow. False positives such as human voice or ambient sounds are minimized by continuously matching the sampled signal against the waveform uniquely produced by water flow. Water flow is measured in Gallons per Minute (GPM). The flow of water produces a distinct waveform such that the peak amplitude increases with higher flow rates. This relationship is used to approximate water volumes based on two-point calibration: a known volume is filled twice using different flow rates to interpolate a linear relationship between absolute average microphone output (in mV) and time to fill a known volume. A third datapoint ensures that interpolation is within an error tolerance of 10% per gallon. After calibration, each sensor records the volume of water used during each event at the resolution of 0.2 gallons, as well as the time duration.
The faucet display represents individual water use through a ‘traffic-light’ metaphor. The ambient display is green when water is first turned on, turning yellow if water remains on for more than the previously measured average duration, and red once water has been running for longer than one standard deviation above average. The color sequence culminates in a flashing red light when water is left running for significantly long periods of time (more than 2 standard deviations above average). In addition, an LED bar graph below the ambient light represents collective daily use. Each bar represents one tenth of the water consumed during the day when baseline water usage data was logged. When water is turned on, the next consecutive bar begins blinking to signal individual contribution to the total water usage. Thus, the bar graph gradually ‘fills up’ over time and resets every 24 hours.
Two styles of display were developed to visualize water usage in the shower. The numeric display presents current usage to the nearest tenth of a gallon. When the shower is turned on, this number increases based on water flow rate. In addition, the average usage, which was computed during the initial (logging) phase of the study.

The ambient visualization presents this same information as an ambient orb, leveraging a ‘traffic light’ metaphor. First, the orb shows a green light while water usage is below average, fading to yellow and then red as water usage reaches average and 150% of the average amount of water is used, respectively. The light sequence culminates with a flashing red light when usage exceeds 200% of the average.
