August 3, 2020 | Written by: Levente Klein and Siyuan Lu
Categorized: COVID-19 | Science
Share this post:
In 1854, physician John Snow produced a famous map showing that cholera deaths were clustered around a pump on Broad Street in London. Snow’s seminal geospatial analysis — conducted with little more than a pen, a map, and his own observations — led him to formulate the theory that germ-contaminated water was the source of cholera. His is recommendation that the pump be taken out of service effectively ended the deadly outbreak.
With similar intent — but using far more sophisticated geospatial technology — we are shedding light on the environmental and societal impacts of the COVID-19 pandemic, which has kept much of the world in various states of quarantine over the past six months.
Geospatial Big Data
The IBM PAIRS Geoscope is a massive data store of aligned, pre-processed geospatial satellite images, weather predictions, socioeconomic data, news information, and more. Leveraging efficient indexing methods to spatially and temporally link data layers, PAIRS is well suited to perform complex analyses and rapid searches that help reveal key interconnections between data sets. A query of PAIRS which has curated six petabytes of data to date, offers a glimpse of the impact of and response to COVID-19.
COVID-19 started to dominate news reports in the U.S. in early March 2020. By March 21, nearly half of all US news coverage was about COVID-19. High levels of public attention coincident with governmental actions like school closures and shelter-in-place mandates were followed by a drastic reduction of activity — so drastic as to be observable from the space. Figure 1 shows a PAIRS query of New York City’s light intensity as observed at night from the VIIRS instrument onboard the NASA Suomi NPP spacecraft. The light intensity dropped by around 20% starting March 15 and remained low from April onwards compared to late February and early March. The dimming of night light indicates reduced business activity and traffic and may offer an early market intelligence signal.
Figure 1. Change of night light intensity of New York City as observed from the NASA Suomi NPP spacecraft.
Greenhouse Gas Emissions
The reduction of activity is also observable from the query result comparing the average atmospheric nitrogen oxides (NOx) measured by ESA’s Sentinel-5P satellite from January to May (Figure 2). In urban areas, road traffic is the dominant source of NOx emission. It is clear that the emission — and thus driving — in major population centers drastically reduced across the world. The steplike reduction in China occurred between January and February while reduction in the U.S. happened between February and March – reflecting the variation in timing of the COVID-19 outbreaks in different parts of the world.
The observation is supported by data from U.S. Department of Transportation as well, which tracks the frequency of trips of different lengths. Transportation CO2 emission can be derived from trip mileage and estimated average emission per vehicle per mile. We estimate that in the U.S. in April 2020 there was approximately 7.3 billion pounds of CO2 emission from car travel, which was 25% less than in January 2020 and 40% less than a year prior in April 2019.
Figure 2. Animation showing atmospheric NOx from January to May 2020 as observed from ESA’s Sentinel-5P satellite.
Conventionally, it would take a data scientist days to obtain the comparisons shown in Figures 1 and 2. Satellites like Sentinel-5P take around 50,000 images of the earth every month — each covering a small region — so a data scientist would have to download numerous photos and use software to stitch them together on a map. With PAIRS, it takes only a simple query and less than 40 lines of code to get the result.
Climate Impact Science
A major focus for PAIRS is climate science and mitigating humanity’s impact on the planet’s climate. Researchers use PAIRS as a foundational technology to accelerate the development of AI methods to drive regional downscaling of climate forecasting, climate impact modeling, greenhouse gas detection, and optimization of supply chain and cloud computing operations to minimize carbon footprint.
Anyone may try the PAIRS instance on the IBM Cloud using this link. Enter PAIRS by registering for a free IBMid. Enterprise clients may deploy PAIRS in a hybrid cloud environment in which the client’s on-prem PAIRS instance containing private data is federated with the IBM instance. Red Hat OpenShift orchestrates many common PAIRS workloads including data ingestion and analytics.