Our challenge, ‘Light the Path’, tasked us with analyzing data from Earth-observing satellites to explore the relationship between COVID-19 and human activity. Our project uses mobility data, COVID-19 data, and two types of light data to determine correlations and relationships between human activity and the spread of the disease. Night time imagery can be misleading with the considerations of weather patterns, so by using mobility data in tandem, it provides a more complete picture of the patterns of humans. Using this information, we determined a set of risk factors that contribute to outbreaks. We then created a model to analyze these risk factors and estimate a given community’s likelihood of an outbreak. This model can be used to combat future pandemics by identifying potential hotspots. This could potentially limit the spread of the disease and even prevent an epicenter from starting.
As current college students, we were all forced to move locations in March due to the COVID-19 pandemic. Although the decisions made by colleges and universities to close campuses and suggest (if not require) students to return home made college towns safer, it also caused mass movements towards other populous areas, which could increase the chances of outbreaks happening there. We were particularly interested in finding correlations between human mobility during the pandemic, as well as using light data from NASA and other agencies to determine risk factors in different areas.
Part of our project included using satellite data to show changes in human activity in a specific area. To this end, we used data from VIIRS/NPP Daily Gridded Day Night Band 500m Linear Lat Lon Grid Night (VNP46A1) and analyzed it using python. Because of the sheer size of each data file (2400x2400 matrix), we did not have the time to do much computational calculations with the data sets, so we took the average temperature brightness of each HDF5 file from each week from January 2020 to May 2020 for a section of the the east coast of the US and plotted them using matplotlib to determine if there were any trends in overall brightness that could be correlated to human mobility and/or increases in COVID-19.
We then used mobility data found on Google Data and Cuebiq to analyze the patterns of human populations and how they changed during the pandemic. We were particularly interested with the locations that people were moving to and from and the percentage of people staying home. We used this data as a basis to understand what areas to focus on when looking at the CDC Mobility data.
By looking at the CDC Mobility Data we could see the mobility trends and how they impacted the amount of cases in a given area. We were mainly interested in seeing the trends so we used the 7 day absolute instances as the measurement of cases. The CDC website overlays the COVID cases with the mobility trends (as seen in the previous resources), so it was a good visual for us to see that when mobility increases, the amount of cases increases as well, suggesting a correlation between the two. Knowing this correlation between mobility and cases existed, we then looked to compare it to real world light data.
We looked at images of Earth at night captured by the VIIRS instrument aboard the Suomi satellite. We used images from before and during the pandemic to observe changes in visible light pollution in various parts of the United States. We developed a MATLAB program that would compare the average brightness of two images and used this to determine whether a certain area had increased or decreased in brightness due to the pandemic.
We sampled various locations that had experienced varying degrees of success responding to COVID. To study urban trends in places where social distancing was enacted and obeyed at varying levels, we looked at New York, Miami, and Atlanta. To study overall movement between urban and rural areas, we looked at Cape May County, New Jersey. Based on our findings in the light data, we were able to draw a few conclusions about how people are moving and how they impact COVID cases.
Our first conclusion is that in areas with lower light (relative to non-pandemic conditions), people are moving less and cases begin to decline. Our second conclusion is that in areas with equal or greater light, people are moving the same amount or more, likely due to poor social distancing practices. In these areas, we observe a spike in cases. Our third conclusion is that generally, people are moving out of urban areas and into more rural ones that may not be properly equipped to handle a rise in cases.
Using this data, we made a rudimentary risk analysis model in MATLAB. Users can input their starting region out of three choices: a city, a suburb, and a beach town. Users can then input whether light pollution and mobility have increased, decreased, or stayed the same, and implement various degrees of social distancing. The program accounts for these choices and returns a level of risk to the immediate region and the likelihood of an outbreak spreading to other regions.
Based on this risk analysis, we created a model in Figma to visualize our findings and show how they could be used in the future. For example, seeing what needs to be changed to take a high risk scenario, like an outbreak in a city, and decrease the risk of spread.
You can test the prototype here.
One of the problems we encountered was finding up-to-date and relevant data from satellites. The data we ended up using -- brightness temperature data from VNP46A1 -- is not the best predictor of human activity, at least for our purposes. Another challenge was our lack of knowledge about how to make sense of this complex set of data. We also hit a limitation in computing power. The python program we wrote to analyze the VNP46A1 data takes a while to compute the brightness temperature since the data given is a 2D array of size 2400x2400, which gives us 5,760,000 data points. To process one data file it would take roughly two minutes. This computing limitation made it infeasible to process a larger amount of data, create a more in-depth analysis, and review proper computation.
Space Data
Mobility
COVID