Below you see three types of data visualisation (A,B, and C) - all are comprised of three components:

• The clock (top) - shows real-world time when the data was collected. The three clocks are identical in all three visualisations.
• The heatmap (middle) - shows the distribution of CO2 in the lecture theatre at the time specified by the clock above, and the scatter plot below. You will see a transparent floorplan overlay over each of the heatmaps to show how the data corresponds to the photo of the lecture theatre shown above.
• The scatter plot (bottom) - shows average CO2 value in the lecture theatre. The red line corresponds to real-world time and matches the clock above. The opaque bars correspond to the duration of the lecture (10:00 - 11:00). The three scatter plots are identical in all three visualisations.

• Visualisation A - CO2 accumulation (yellow - high, magenta - low)
• Visualisation B - CO2 accumulation (red - high, blue - low)
• Visualisation C - CO2 rate of change (red - increasing, green - decreasing)

The described project builds on the pre-existing sensor deployment in the William Gates Building (also known as the Computer Laboratory) in Cambridge, UK. The building has become a testbed for large-scale sensor deployment since April 2020, with a proprietary BIM, sensor data-streaming and visualisation platform. The following subsection describes the smart building deployment in the WGB in more detail, including the description of the installed infrastructure that allows for sensors to be deployed.

The basis for this project was the installation of 36 new environmental sensors in LT1 in the WGB. Before the sensors could be deployed, sufficient ground work had to be carried out to ensure that the space was suitable for the project. There were two large lecture theatres in the William Gates Building - LT1, comprising 264 fixed seats, and LT2, comprising 138 fixed seats. Both lecture theatres had seating arranged on a slope. Lecture Theatre 1 (pictured below) was selected to be more suitable due to its size and more common use that would allow for more data to be collected.

For the analysis, we were mostly focusing on temperature and CO2 metrics, since those were the most affected by human presence.

The collected results generally indicate that all of the measured metrics in LT1 are highly dependent on the number of people present in the lecture theatre. This means that people attending lectures have a great deal of influence over temperature and CO2 in the lecture theatre. In this section, we will look at how the measured metrics are affected by the student distribution in the lecture theatre.

After analysing the heatmap data it was found that CO2 generally tended to increase at the front of the lecture theatre where the lecturer was presenting as well as in the back of the lecture theatre where the sensors were not too far from the ceiling. Furthermore, it was found that CO2 levels tended to rise very rapidly from the beginning of lecture, reaching its apex at around 15 minutes into the lecture and then staying there during the course of the lecture. While opening the doors was found to temporarily decrease CO2 levels, as soon as the doors were shut CO2 levels jumped up again rapidly within minutes. The distribution of CO2 at the front of the lecture theatre can be explained by most of the students’ preference to sit at the front (shown in Figure 25), as well as the fact that lecturers presenting the lectures did not wear masks that would dissipate the CO2 in space. The high concentration of CO2 at the back was surprising since CO2 is heavier than air and should not go up, especially given the steep slope in LT1 (Figure 1). However, since exhaled air from the human body is a mixture of gases that are warmer than the surrounding air, it generally tends to go up. The sensors located at the back of the theatre were also the highest-placed sensors due to the LT1’s shape, hence the observed effect of CO2 being highest at the back.

Finally, when compared with the student distributions, CO2 had a smaller correlation with the seating positions than temperature, even when taking into account possible sensor miscalibration and subtracting their readings at midnight to form a better baseline. Overall the results for CO2 and temperature analysis show some correlation between the sitting location and the observed changes in environmental phenomena, however, more data has to collectd for definitive conclusions.

The average temperature can fluctuate greatly in the lecture theatre based on the ambient outside temperatures. This means that the HVAC system in the building is not always able to maintain a consistent temperature as on warmer days the ambient temperature in the lecture theatre is higher by several degrees, when compared to cooler days. This becomes especially apparent when analysing the collected heatmap data on a fixed value scale, as seen in Figure 24. This means that the starting state of the heatmaps were not identical on all days, which is important when doing a visual analysis of the data. Overall, the temperature distribution showed that the lecture theatre heated up considerably over the course of lectures, however, there was also noticeable noise in the data. It was also found that temperature increased consistently throughout the lecture duration and decreased slowly after the lecture concluded.

However, raw data alone does not provide a complete picture. After adjusting for uneven baselines among the sensors, the heatmap clearly reveals a correlation between where people are seated and their emitted body-heat radiation:

In order to assess the effectiveness of the new ROTAS visualisations (and see how they compared to the original), an online user study was conducted. The study was designed to ask the participants a mix of general (would they use a similar visualisation tool) and specific (if they liked the colour schemes) questions regarding the visualisations. The participants had to rate the visualisations by rating the question statement using a 7-point Likert scale, 1 meaning they strongly disagreed with the statement, while 7 meaning they strongly agreed.

Overall the results (n=9) indicate that there is a clear preference for the new visualisations (Figure above). Furthermore, all participants stated that they would like to use such a data visualisation frequently, and that it would impact their sitting preference in the lecture theatre. This directly ties in with the main research question showing how the collected data can be used to affect the building occupants. Therefore, it gives an indication that data visualisations like this can amplify the existing sensor deployment in the building to create deeper, more meaningful interactions that affect how building users interact with the building.