NuLL CaChe - Space Apps Challenge

NuLL CaChe| A One Health Approach

A One Health Approach

Air pollution is a major global environmental health risk, causing an estimated seven million deaths across the globe annually. Your challenge is to take an interdisciplinary approach, using both Earth science and health science, and integrate different types of datasets and applications to study the effects of air pollution.

Air Pollution Hotspots Detection and Integration with Health Science using AI/ML Techniques

Summary

Our solution demonstrates data analysis capabilities with the help of time series analysis and visualization by identifying local sources of pollutants, their concentrations, source sector data and the reason of generation, along with latitude and longitude data, which helps us plot these hotspots on a map. The yearly and monthly average plots help us to study climate trends and seasonal changes. This is then integrated with health science data to identify the health impacts of pollution.

How We Addressed This Challenge

Our planet is plagued by environmental problems and if left unchecked, it can adversely affect livelihoods. Air pollution kills around 1.25 million people in our country (India) and 7 million people worldwide every single year.

Over the years, NASA has launched many Earth-observing satellites. However, selecting, processing, and analyzing satellite data can be a daunting task for many, especially the common masses.

We have developed a user friendly tool that demonstrates the ability of data analysis and visualization with the aid of existing NASA resources and other Space agencies so that we may enhance the public's understanding of what satellite data really means and how can it be leveraged effectively and easily.

People can easily explore our processed data and they’ll also have the option to integrate their findings with our datasets while at the same time be more aware of the extent to which pollution can affect their health.

We enable the user to select an area of interest and filter the types of phenomena that they are interested in monitoring.

We have integrated Earth Science with Health Science to study the effects and impact of pollution on health. Based on our analyses, we will also generate in-depth data reports which we can then provide to Health Officials, Government bodies and Policy Makers to take necessary action.

We aim to raise the general public’s awareness of NASA data around the world.

We hope to make people more aware of how and why air pollution can impact their health, and also increase their awareness of pollution levels in and around them.

We aspire to achieve a National level target of 20-30% reduction of the concentration of pollutants by 2022

How We Developed This Project

Worldview lacks data analysis capabilities because data to generate image pixels are not ready to use. Hence, we have developed a user friendly tool that demonstrates the ability of data analysis and visualization on existing NASA and other Space Agencies' data to discover, visualize and analyze Earth Data.

We created a pipeline to fetch scaled data from ESA for the years 2018, 2019, 2020, and then averaged it. These files were originally in the GEOTIFF format, which we then converted into CSV format. We then processed this data and conducted advanced Data Analysis and Deep Learning techniques to gather information regarding the sources of pollution hotspots. We plotted this data with the help of certain libraries and then identified, analyzed and visualized climate and seasonal trends. We then integrated this data with Health Science data from WHO to study the effects of pollution on health. On the basis of the analyses conducted, we then consolidated our insights and results into elaborate reports which can be used by Pollution Control Boards, Health Officials, Policy Makers and Researchers for taking necessary steps.

Since we also aim to maximize the involvement of the common masses and also find a reliable way to find the sources of anomalous hotspots, we came up with a creative approach that uses Crowdsourcing. Today the world population is about 7- 8 Billion. What better time to leverage the power of collective effort than now! In case of anomalous hotspots, say because of an event like a cricket match or a music concert, people burst fire crackers. The common people can just open up their phone and go to our website and report the event by clicking on the nearest pin to their area which can then be further reviewed.

Technology Stack:

Frontend: React
Backend: Python & Node.js
Deep learning Algorithms: Clustering, Time-Series Analysis & Change Detection
Libraries: Scipy, scikit-learn, netcdf, matplotlib, seaborn, pandas, numpy, folium, plotly, bs4, requests, tqdm

Interesting Facts:

We studied and reported seasonal changes of indoor and outdoor pollutants
We studied the effects of lockdown on pollution levels.
We identified accurate Source Sector Data (Industrial, Residential, Power Stations, Transport, Waste and Mobile) that is responsible for pollution.
We identified the impact that pollution has on the death rates of male and female separately and also studied their effects on different age group.

Link to Our Findings

Future Scope:

During the Hackathon, we started working on European countries and India. We are going to keep updating our solution in the next versions and will gradually cover the entire globe.
Providing an Application Programming Interface so people can directly download the processed satellite data in JSON and CSV.
Demonstration of the ability to obtain image and other atmospheric parameters from existing NASA resources.

How We Used Space Agency Data in This Project

We used the data from ESA’s Sentinel 5 precursor. The portal provided by ESA gives us the near real time and offline data. However, the data is not scaled. We needed to scale the data to 50Km/px because we don’t require the precision it provides. So we look at Earth Engine, which is a Google product which allows one to write scripts to scale the data to the required configuration and download it. So, we took the following datasets from Google Earth Engine:

Sentinel 5p Carbon Monoxide
Sentinel 5p Nitrogen Dioxide
Sentinel 5p Ozone
Sentinel 5p Sulphur Dioxide
Sentinel 5p Methane

We created a pipeline to fetch the scaled data for the years 2018, 2019, 2020 and averaged it. We also downloaded the Monthly data and averaged to get insights into climate trends and seasonal data. All of these files are available in GeoTIFF format which we then converted into CSV for further analysis and JSON for the purpose of plotting these data points.

Sentinel 5p Level-2 products are:

Geolocated total columns of ozone, sulfur dioxide, nitrogen dioxide, carbon monoxide, formaldehyde and methane
Geolocated tropospheric columns of ozone
Geolocated vertical profiles of ozone
Geolocated cloud and aerosol information (e.g. absorbing aerosol index and aerosol layer height)

There are three types of processing required: NRT (near real time), OFFL (Offline), and reprocessing.

1.For NRT processing the availability of products must be within 3 hours after sensing. Total columns of ozone, sulfur dioxide, nitrogen dioxide, carbon monoxide, formaldehyde, vertical profiles of ozone, cloud & aerosol information will be provided in NRT.

2.For OFFL processing the data availability depends on the product. The currently planned NTC delivery times are:

for Level 1B products within about 12 hours after sensing
for methane, tropospheric ozone and corrected total nitrogen dioxide columns within about 5 days after sensing.

3.For reprocessing activities there are no time constraints. Reprocessing of Sentinel-5 Precursor products will be performed when major product upgrades are considered necessary.

Weekly Average Data