Faculty Sponsor: Pavel Oleinikov
Abstract: The purpose of this research is to explore how well publicly available, and thus lower quality, satellite data can be used to track human activity. We made use of the Covid-19 pandemic and subsequent lockdowns, where activity deceased in certain areas and increased in others. Using heat emissions were a proxy for human activity, we fit a robust pixel-wise regression that predicted surface temperature from the Landsat-8 dataset as a function of various weather and environmental variables over our region of interest, Montreal, before 2020. When compared with post-Covid observations, we expected to see decreased heat emissions and thus decreased activity in areas impacted by Covid lockdowns, and increased emissions in residential areas. Our findings varied in their rigor. Visually, a map of the ratio of post-Covid observed results to predicted results proved promising: We were able to make out numerous public areas like malls and schools that showed up as “blue spots” on the map – spots where the observed values were lower than the predicted values. However, upon exploring the data more robustly using T-Tests on the means of specific regions of interest, our tests did not yield any statistically significant findings. This may be due to the lower quality of the publicly available satellite data, and further research will likely benefit from using better quality private satellite data.
Poster-Final-Version-pdf