To encompass and study the effect of Covid-19 on the socioeconomics of the Indian subcontinent region, we incorporated nighttime light data. In India, transportation of freight(goods) happens during the night. Thus, the study of night-time lights provides a direct insight into the economic activity of the country. Also, the CO, Nox emissions from the movement of such vehicles(large trucks) are higher than the movement due to private vehicles and thus, reduction in their movement has led to a reduction in CO, Nox levels.

Night light data was acquired from National Polar-orbiting Partnership Visible Infrared Imaging Radiometer Suite (NPP-VIIRS). Before acquiring raw data from VIIRS-SNPP, night light data was visualized onNASA’s Worldview Snapshots andNASA Worldview to do initial comparisons and to select suited data for analysis. Then raw data of VIIRS-SNPP was acquired fromNASA Black Marble Product, which was then processed using theNOAA HEG tool,  ErDAS Imagine and ArcMap 10.5 to acquire night light data of the required region. Comparison of the final images, i.e March and April 2019 vs 2020 showed a decrease in nighttime activity.  NightTime ImagesAnalysis

To further understand the numbers behind the mobility socio-economic factor, we analysed Earth Observation Data of Sentinel-2 satellite to get possible locations of vehicles, and hence, obtain the total vehicle count in a particular region.

We used Sentinel Hub API to access the site data by using the library xcube_sh. This allowed us to download various bands of a particular region by giving boundary information- Band 2 (Blue), Band 3 (Green), Band 4 (Red), Band 8 (NIR) and BAND 12 (SWIR). NDVI, NDWI, NDSI were calculated to obtain a road mask. The road mask along with Band 2, Band 3 and Band 4 data, was used to calculate possible vehicle positions within the mask. This program was executed for March - April (10 Days Interval) of 2019 and 2020. It was found that the drop in vehicles during COVID-19 lockdown was from 30% to 50%.

Vehicle Location and Count

An Area Averaged time series plot of The CO Surface Concentration using MERRA-2 data, and NO2 Tropospheric column using MERRA-2 data in Giovannifrom November 2019 to Apri 2020 clearly showed the reduction in CO and NO2 levels, due to reduced human activity as well as a reduction in movement of vehicles. CO levels using Mopitt Datawere plotted on a basemap using Python for better visualisation.

CO and NO2 Level Graphs + Case-study of Delhi

The reduction of CO,NO2 levels might also be due to the closure of factories, industries, construction work, and so forth. The closure of many industrial sectors, in turn, has affected the employment rate and overall GDP of the country.

According to the data offered by Trading economics, India’s annual GDP rate is at an all low time low of 3.1% and is forecasted to decrease further in the upcoming months. The unemployment rate is currently at 7.8% and is forecasted to increase up to 12% in the near future.

Unemployment Rate and GDP Visualisation 

Thus, gradual relaxation across the country is key to improving the economy. To help the government decide which sector(s) -Retail and Recreation, Grocery and Pharmacy, Parks, Transit Stations, Workplaces, and Residential Areas, to open first,  we gathered the mobility of each of these sectors averaged across the 4 months for every state along with the total number of cases in each of these states. We then calculated mutual information between the mobility of each of these sectors with the number of cases and ranked them from highest to lowest to see which correlated the best with the number of Covid-19 cases. We found that mobility around transit stations is the best indicator for the number of cases and mobility around groceries and pharmacies the worst.

Mobility Statistics 

With more finely grained data we can achieve considerably better results. We believe that this mutual information approach can lead to very valuable insights if we were given more data and that is part of our future work.

The time taken for Image processing was a lot longer than we had anticipated, so, we only had enough time to prepare a few datasets. Likewise, more than 40% of our road images were covered with clouds which didn’t allow us to do any analysis on them. There is also a significant amount of noise in the final images that we could not get rid of due to lack of time.

Future Work:
Due to limited time and processing power, it was a hard task to integrate all the data together and show the cumulative behaviour. With enough time and resources, all the data can be processed and compared together to observe a more profound behaviour of impacts factor with economic conditions. The reliability of impact factors is already established in our findings, and thus these factors will assist in creating accurate predictive models. 

Furthermore, based upon findings, an API can be created to allow users to navigate the data and find the current status of disease vulnerability in their neighbourhood. This API will also provide predictive analytics on vulnerability for the neighbourhood. 

Using Deep Learning, we can also use vehicle count data collected from the Sentinel 2 L2A dataset over the time of lockdown which will give us an abundance of data to predict how each city is effectively handling mobility and transportation. This will give us further understanding and foresight regarding which cities can be opened after lockdown.

Check out how cool our app is going to look!

https://drive.google.com/open?id=1d0eoTtBJYd674rgmU7ZkH3ZWgQuNs2mH

Sources of Data Acquisition and visualization:

EO Browser

QGIS

Sentinel Hub

Bounding Box

Sentinel 2

NASA Worldview

NASA Worldview Snapshots

NASA Black Marble Product

Earth Engine Data Catalog

LANCE: NASA near Real-Time Data and Imagery

Giovanni

Google covid-19 mobility data

Mopitt Data and Information

Trading economics

Sentinel-5p

Euro Data Cube

Softwares:

Python 3.6

AcrMap 10.1

ENVI 4.8

Erdas Imagine

NASA NOAA HEG tool