Examine the relationship and interaction between street activities and noise levels in two scenarios (inside the campus on pedestrian lanes and outside the campus on the streets)
Yuchen Dai, Eric Xia, Haoxiangyu Zheng, Joie Sun
The link to our group’s high-quality video is here:
The link to our presentation slides is here:
The link to our final report is Here
Executive Summary
Noise issues and the detrimental effects have increasingly become a public concern in large cities. This urban sensing project plans to document noise intensity within two sites, comparing the patterns and characteristics between car-dominated and pedestrian-dominated environments. The team provides multiple solutions facing hardware limitations and potential concerns regarding the experiment. With specific codes that enable the sensors to work remotely and independently, the team implements the experiment in front of Avery hall and on Amsterdam Ave across four consecutive days, adjusting procedures and equipment after receiving feedback from the field results. After integrating, analyzing and visualizing the data collected, the team differentiates the trend of noise patterns of the two sites, and produces discussions on potential noise sources as well as contributing factors of pattern change. Based on multiple literature reviews and case studies performed, the team drafts multiple proposals where the idea of noise sensing can be applied within the larger urban scale.
Introduction, including context and motivations
With huge vehicular flows and logistic movements, daily activity patterns in large cities are often accompanied by different sorts of sounds, and among them lots of noises. In fact, noise has become an increasingly significant public concern in New York City, a city with an average noise level of 85 decibels that is much higher than the suggested WHO standard of 55 decibels (Re, 2022). Researchers and studies have shown that indoor activities can be interrupted by unstable and sharp noises from the outdoor environment, while high noise levels can be detrimental to physical and mental health of unprotected residents and workers in the city (NYC Health, 2022). Nevertheless, the city is currently lacking effective regulation codes and policies on noise issues, compared to established systems on fire prevention or architectural design (City & State New York, 2019). The severe noise situation and lack of related policies, combined with the personal experiences of team members, contribute to the research project that senses noise patterns and intensities across different scales. Multiple projects that apply cutting-edge technology to identify and mitigate urban noise also inspires the team to apply technological skills in tackling the issue.
Local Interactions
The team plans to measure and compare the noise simultaneously on the streets with automobile flows and on pedestrian pavements without the presence of cars. Selected sites near pedestrian paths are approx. 40 square meters, where the sensor is positioned about 6 meters away from the core target. PIRs sensors are planned to be placed in the same place, with a measurement range about 3-7 meters. The measurement range of sensors does not change significantly along roads with car traffic. In general, investigations at local scale will be around 10 meters following the spec limitations of Adurino sensors.
The sound sensors and recording of pedestrian flows do not facilitate interaction with people on the site. The sound sensor will automatically record the nearby sound for documenting the noise level without recording (and memorizing) specific sound content, thus approvals or interactions from pedestrians are not needed, and no privacy is violated. The documentation of entry/exit data will be based on an anonymous process and will not involve the recording of pedestrian characteristics and activities. Although the experiment process does not involve any interactions from the senso/facilitator side, pedestrians may choose to interact with the sensors as the equipments are presence in both of the two types of locations, leaving potential for the result to be altered by some of interruptive interactions (such as deliberate shouting to the sensor). However, as all pedestrian activities are considered part of the street noise pattern, the team is not eliminating extreme behaviors from the recording and documentation process.
Technologies
Sensors involved in the data collection process are based on Adurino kits, the core requirement in the class. Both sound sensors and PIR sensors (as extensions of the kits) are used in the initial design of the plan. However, PIR sensors are removed from the kit after trial results show the inability of basic PIR sensors to accurately detect and report pedestrian presence. In order to be more low-key in the experiment, the team plans to connect the sound sensor to a separate battery for recording with no need for carrying large power supply equipment. When the power is connected to our sensor, the sensor will start to operate, recording data automatically. Recoding will be stopped when the power is disconnected.
In addition to reducing the size of our sensors and minimizing possible interactions, the team plans to make the sensors independent of the computers. The team plans to connect an SD Card to the sound sensor and record all data on the SD Card.
Following the Arduino code, the team connects the SD Card, the sound sensor, and the Arduino board together, so that they can work simultaneously. The team creates a program code that allows the sensor to get rid of computers when operating and record data independently, storing it in the SD Card instead of simultaneously returning the data to computers. The team only connects the A0, power, and ground of the sound sensor to our Arduino board, analyzing data manually without using the electronic recording port. On the same Arduino board, team members connect the SD Card adapter. It is worth noting that the sound sensor and SD Card adapter are connected through the extended board, the power line and the ground line to realize the connection between the two devices. The Arduino board is specifically set to automatically detect the files in the SD Card through the code. After uploading modified codes and orders to the motherboard, the code in the motherboard will be re-run every time the data is connected, which means that every time the power is turned on, the motherboard will automatically delete the last recorded data and automatically create a new file to record data.
Code Main Steps:
- Import the packages that will be used
- Define integer, output port
- Set pin mode and initialize SD card
- Set the new variable, and convert the received resistance value into decibels
- Save data to SD card with “WriteFile()”
- Set SD card
- open file
- write to file
- close file
The team encounters several problems during the measurement process. As the sensors of the Arduino kit are not accurate, it is impossible to measure the noise level in a very accurate way (as the maximum precision kit sensor could reach was in integers). Meanwhile, the accuracy of data collection is also heavily related to the resistance value which determines feedback value back to the Arduino main board (for data processing and recording), while the resistance in the motherboard is heavily influenced by the wiring, placing (of equipments) as well as the local weather during experiments.
Pilot Sensing Work
The team plans to collect and record noise data at the main entrance of Avery Hall and on Amsterdam Avenue (near the intersection of 118th street and the avenue). The team chooses the site of Avery Hall entrance with familiarity to GSAPP, the facility using the hall, and as the entrance is facing a three way intersection that often brings large flows of students that may add potential changes to the noise pattern. We usually hear a lot of noise when we are in class in Avery Hall and Fayerweather Hall, while the impact of street noise on students is severe, inspiring the team to further investigate the characteristics of the noise, and hoping that our recommendations after the experiment will provide useful suggestions for GSAPP to mitigate the noise issue. The team chooses Amsterdam Ave. as a comparative study case since the public road is the nearest available subject to compare with the Avery site, where the team assumes least impact of other environmental factors on noise levels despite the mobility forms and patterns. The team chooses the two sites also based on minimizing the time and cost traveling to/from research sites, as the two places are near Columbia University.
As mentioned above, the core of the experiment is based on an Arduino kit with sound sensors using written programs that measure the noise level in dB units per approx. 0.25 seconds and is later retrieved through inserted micro SD cards for future analysis. Considering the safety and potential privacy concerns regarding the experiment equipment, the team (2 person per small group) carries the equipment to the field and stays on site until the experiment is completed. On both sites, team members document related factors that may contribute to noise pattern changes, where the Avery group is documenting entering/exiting data into the building, while the Amsterdam group is recording any equipment or vehicles that lead to noise surges or background noise changes.
While the on-site experiment largely follows the team’s research plans, members encountered several problems during the experiment. The team found the sensors (Adurino kits) to be insensitive for precise measurements of noise levels after the first data collection trial, while the functionality of sensors was further interrupted with relatively high temperature (86 degrees) and direct sunshine. The team also finds that wiring of sensors may influence the actual performance of these sensors, while the lack of external protection of the original arduino kit risks the sensors to be damaged by pedestrians and heavy winds. During collection periods later, the team tries to set the wiring of the two sound generators to be as similar as possible, and designs light-colored covering of the sensors to prevent the sensor from being exposed to the sunlight during data collection. To ensure a minimum level of data is collected facing potential cases of sensor failure, team members are also equipped with noise measurement devices to back up existing Arduino kits.
Data Analysis and Visualization
The team collects noise level data in parallel on two sites for four days (Tuesday to Friday) in slightly different 30-minute slots (Around noon from Tuesday to Thursday, afternoon on Friday). For each collection, the sensor measures and reflects around 7,000 data documenting the noise level in dB units in integer. As the interval of data collection is short while the data is precise, the team divided data analysis and visualization into general observations (covering the whole study length) and detailed observations (analyzing a continuous 300 data signals during one observation period). The team applies a random sampling of 30 data points for the general observations and visualizes the vertical and horizontal comparison of observations using the python platform. The team relies on microsoft excels to visualize the trending of 300 data points with calculation of data statistics.
Discussion
Within the level of general observation and analysis using 30-minute data, there is an obvious difference in average noise level between the entrance of Avery and the intersection on Amsterdam streets, where average noise level near Avery (56.2 dB) is significantly lower than Amsterdam (64.9 dB). It is sufficient for the team to argue that difference in mobility patterns (cars v. pedestrians) leads to the difference in background noise levels. While it can be witnessed that the noise level on Amsterdam is fluctuating within the 30 minutes period showing highs and lows after sampling, the noise level in front of Avery is decreasing stably, which potentially can be explained by the fact that all measurements starting slightly after the end of classes that bringing a concentration of flows of students near the site of measurement. In comparison, car flows during the 30 minute period are relatively stable on the streets. While the noise level on the Amsterdam Intersection rises and falls with red v. green signals that determine passing car flows (as Amsterdam Ave has a far larger car flow compared to 118th St.). The trend difference further ascertains the team’s assumptions that automobile movements are the major contributor to a higher background noise level and the fluctuating pattern on Amsterdam St., while pedestrian movements are the major contributor to the change in Avery. While radical fluctuations in Amsterdam data on Tuesday Apr.11 can be factored into multiple coaches and buses that parked near the sensor without stopping the engine, fluctuations in Avery data on Apr.12 and 13 are mostly likely caused by chattings in short distance from the sensor. The general data trend is showing a slightly higher average noise on Amsterdam Ave. on Thursday Apr.13, and a much higher surge on average on Friday Apr.14. While the increase of traffic flow can explain the higher result on Apr.13, the team claims that a much colder research environment on Friday afternoon (while sunlights are completely blocked by university buildings) is altering the electric resistance of the sensor, resulting in a different ground base in measuring. In contrast, the slight fluctuation of Avery data can be attributed to the change of pedestrian flows in different time periods on different days.
The detailed observation and analysis using 90 second data sections demonstrates significant performance differences in the two groups of data. While both two sites show fluctuations in the data, the range of fluctuation on the streets (Amsterdam) is significantly larger with more peaks and troughs, which can be attributed to the passage of high-performance vehicles with large engine noise (observeefd vehicles include trucks, sports cars and modified cars), passage of emergency vehicles with sirens (observed vehicles include ambulances and police cars), and passage of vehicles that press the horns close to the sensor. Fluctuation in the Avery entrance data can be caused by presence of inner campus vehicles (e.g. trolleys or small shuttles) and/or chatting between students near the sensor. The team finds that selected street data on Apr.11 and 12 has witnessed the highest average fluctuation, partially because of the frequent presence of emergency vehicles and construction vehicles at the certain time period. As the sound sensor in Arduino kit is insensitive for short-interval measurements, there may exist significant data errors that can interrupt detailed data analysis.
The research findings have shown very different results for some of the assumptions the team has made prior to the experiment stage. The general trending analysis does not support the team’s guess that these two sites may interact in noise level changes (e.g. Surge in noise on Amsterdam Ave leads to noise increase on Avery site). Nevertheless, the lack of correlation can potentially be explained by the basic functionality of sensors as correlation between surges may not be reflected in intensity (but instead on frequency, pitch, etc.). As the team has assumed that entries and exits in and out of the Avery hall will be the major contributing factors to the noise patterns, the data collected (showing a slightly and stably decreasing trend everyday) does not correlate with documented entry-exit data (more constant with periodical fluctuations). Thus, the team argues that overall pedestrian flows on the passageways and near the lawns in the front of Avery building, rather than the pedestrian flows in/out the building itself, is the major contributor to the noise level pattern in the experiment.
While the team pre-assumes that white noise level in front of Avery may cause detrimental effects for surrounding activities, the actual results from collection signals that the current white noise level is lower than the bar where background noise can be deemed as negative to the living environment. Based on previous literature reviews and case studies, the team claims that white noise surrounding Avery is not the major source of impact that students had complained about. Instead, the team argues that unpredictable surges, both in front of Avery and more in Amsterdam, are more dominant sources of unwanted/interruptive noise to pedestrians and students in Avery and Fayerweather hall, where the classrooms facing the streets will be heavily influenced by the noise surges.
Urban Interactions – URBAN SCALE
With the construction of prototypes, design of experiments and the support of multiple literature reviews and case studies, the team has found numerous opportunities that are brought by the idea of noise sensing, with a series of implementations that are possible in real world and/or in the future. However, the team has also recognized some of the challenges with the noise sensing idea within a world of smart data.
Equipped with expanded functions that stem from the most basic noise level measurement in the Adurino kit, modified sensors can be used in collection of different types and characteristics of sound data to assist smart decision making within digitized data analytical platforms. With a fuller coverage of noise elements and characteristics, the more comprehensive and intelligent sensor will be granted the capacity to discern between different types and potential sources of sound, and will better inform urban planners and emergency responders in the decision making process. The sensors can also be added with warning functions and ‘sound alternatives’ functions, where the system will send warning signals to the monitoring system when detecting noise surges that are louder above certain bars and/or with special patterns, and can automatically generate more comfortable sound (e.g. Simulating rains, winds, or playing gentle music) with less intruding effects as programmed in the sensing systems. If the coverage of sensor network is extended into the whole urban landscape, the setting of noise warning bar (as well as lots of subsequent standards in construction and transportation planning) can be more flexible and accurate, while a whole monitoring system can be established and fortified to improve the overall livability for pedestrians on the streets.
Major challenges of assembling and introducing noise sensing networks in the urban environment include 1) The detection and discerning of ‘essential surges’, including sirens of emergency vehicles and unavoidable noises from certain construction/maintenance vehicles. While these noises are often detrimental to the hearings and mental health of pedestrians and residents, debates will arise on management methods as many of these noises are necessary in other systems and are hard to be eliminated; 2) Ethical and privacy-related issues may arise facing enhanced functionality of sensors. The team has encountered pedestrian concerns around the potential of sensors to record the actual voices for surveillance purposes (in addition to noise monitoring) during the research stages, while these kinds of concerns may be further enlarged when the range and functionality of sensors are expanded. 3) Costs of the system may surge if an urban-scale network is to be implemented, as sensitive sound sensors are usually expensive and may be energy-consuming.
Despite several challenges ahead, the team can propose some imagined realities of the noise sensing program, including 1) Using the network and data to assist decision makers in siting public facilities where users are sensitive to noises, such as hospitals, public schools, service apartments, and community gathering spaces. 2) Following Tehran’s case where the city improves its infrastructure and protects pedestrians from huge noises on the highways using noise sensors, the team is imaging the design and implementation of flexible noise barriers that separates pedestrian lanes and roads based on real-time data. 3) The noise sensing network can provide information for different levels of sound-proof interventions needed in different places. For instance, two sides of Avery Hall, one facing the streets and one facing the lawns, can be modified differently as noise levels are different. 4) The sensing network can suggest intervention points of green spaces in hotspot regions (of noise) to reduce the extent of noise surge and overall background noise, providing a healthier environment for residents, workers and pedestrians.
Work Cited:
Re, J. (2022, July 22). New Yorkers suffer from excessive noise. Retrieved April 25, 2023, from https://www.ny1.com/nyc/all-boroughs/news/2022/07/22/new-yorkers-suffer-from-excessive-noise
Noise. Noise – NYC Health. (n.d.). Retrieved April 25, 2023, from： https://www.nyc.gov/site/doh/health/health-topics/noise.page
Sax, S. (2019, December 5). New York City needs to better regulate noise. City & State NY. Retrieved April 25, 2023, from https://www.cityandstateny.com/opinion/2019/12/new-york-city-needs-to-better-regulate-noise/176662/
Moudon, A. V. (2009). Real noise from the urban environment. American Journal of Preventive Medicine, 37(2), 167–171. https://doi.org/10.1016/j.amepre.2009.03.019
Söderlund, G. B. W., Sikström, S., Loftesnes, J. M., & Sonuga-Barke, E. J. (2010). The effects of background white noise on memory performance in inattentive school children. Behavioral and Brain Functions, 6(1), 55. https://doi.org/10.1186/1744-9081-6-55
Monazzam, M., Karimi, E., Abbaspour, M., Nassiri, P., & Taghavi, L. (2015). Spatial traffic noise pollution assessment – A case study. International Journal of Occupational Medicine and Environmental Health, 28(3), 625–634. https://doi.org/10.13075/ijomeh.1896.00103