Methodology, Algorithm, Ground Validation, Limitations


A four-step methodology was followed to measure reductions in anthropogenic noises during the COVID-19 lockdown:

1. Gather Raw Waveform Data from Seismic Stations

Seismic stations record continuous time-stamped data of ground movements. I extracted the raw waveform data from 14 seismic stations across 9 Canadian provinces and territories. The data was available in the mini Standard for the Exchange of Earthquake Data (miniSEED) format from the Incorporated Research Institutions for Seismology (IRIS) website.

2. Calculate Power Spectral Densities

Fast Fourier Transformations (FFT) were applied to each 15-minute period of the raw waveform data to extract Power Spectral Densities (PSDs). These provided the intensities of seismic vibration across different frequencies over time.

3. Extract Anthropogenic Noises

Seismic vibrations in the range from 4 Hz and 20 Hz are associated with human and cultural noises such as construction, transportation, sports and music events. I extracted these frequencies and binned them in different ranges (5 – 10 Hz, 10 – 15 Hz, 15 – 20 Hz, 20 – 25 Hz) to get higher resolution data. The change in seismic vibrations over different frequencies are viewable under “Transportation and Cultural Noises” menu of the web app

4. Data Aggregation, Trends and Comparison

To visualize the trends, seismic vibrations were averaged daily and plotted. Trends were compared to different time periods: before the lockdown (February 2020), during the first lockdown (March 2020 – May 2020), and after the first lockdown.

A Web App and an Android App was created to visualize the trends, compare the results, and get updated results.

Python Algorithms

Several custom python functions were written by me to execute every step of the methodology above. Python libraries such as numby and matplotlib were used to carry out statistical analysis and graphing. Details about my algorithms are available from my tutorial page. It allows anyone to replicate my results or to obtain the results of changes in seismic vibrations during the COVID-19 lockdown for their cities.

Validation of Results Using Ground and Space-based Measurements

I built an instrument using four sensors (sound, light, PM2.5 pollution, and temperature), Arduino, and a camera to measure ground-level changes in human movements during the first COVID-19 lockdown period in Toronto. The instrument was fixed in front of my home in downtown Toronto, and it collected data over four weeks during the lockdown.

Home Made Instrument to Validate Changes in Human Movements during the COVID-19 Lockdown

I measured changes in street sound levels, pollution levels, and used machine learning to count the number of vehicles from the live camera feed to assess changes in human movements. The pollution levels dropped by 40%, the sound levels by 30% and the number of cars on road decreased by almost 50%. I also obtained data from the NASA Suomi Satellite data to observe changes in the night lights spread over Toronto.

The results from ground-based measurements validated the results of analysis of seismic data on the reductions in anthropogenic noises during the COVID-19 lockdown.

Using Machine Learning on Live Camera Feed of Traffic in front of my Home to Count Vehicular Traffic during the Lockdown


  • Seismic stations are purposely built outside the urban areas so that they are isolated from city-related noises. As seismic stations chosen for this project were at different distances (5 km – 60 km) from the city centers, they were not equally sensitive to changes in human activities.
  • Lockdowns happened in all the cities included in this project. However, the definition of which services were considered essential and remained open varied.
  • Canadian cities had a hard start date of the first lockdown (around 18 March 2020) but the end date of the first lockdown varied. Some cities extended the lockdown while others opened up with some restrictions.

View the Web App: