An effort to map Air pollution through Satellite Data and bring more realistic measurement which could be used for city planning and pollution monitoring
1. Deep Lens for Air Pollution
Estimation and forecasting of atmospheric pollution from satellite imagery
Sanjeeth Veigas
Shashi Bhushan Singh
Suyash Mishra
Vinay Sudhakaran
2. Urban air pollution is increasing at an
alarming rate!
● In May, WHO declared New Delhi as the most polluted city in the world
● India’s National Health Profile reported 3.5M cases of acute respiratory
infection in 2015
○ 140,000 increase from previous year
○ 30% increase since 2010
● Bengaluru has ~26L children and more than half of them suffer from Asthma
● Increase in air pollution is higher than that of water pollution and that caused
by improper disposal of garbage!
Source: World Health Organization, 2015
5. AOD from MODIS SATELLITE for Bengaluru
Bengaluru Region
https://giovanni.gsfc.nasa.gov/giovanni/
6. PM VS DUST vs AOD
Correlation plot for AOD vs DUST vs PM
7. Ground Sensor : The Troubling Truth
● India uses machines called high-volume samplers to
measure PM 2.5
○ PM 2.5 - Tiny and dangerous airborne particles that are less than 2.5
microns in size and are fine enough to enter deep into the lungs and
bloodstream.
● Most states choose the cheapest manufacturers
○ The use of poor quality materials render the readings untrustworthy!
○ CPCB compared international samplers with Indian ones and found
massive inconsistencies in the readings
● Ground sensors are very expensive and hard to maintain
Source:
http://economictimes.indiatimes.com/industry/auto/news/industry/why-indias-numbers-on-air-quality-cant-be-trusted/articleshow/44808946.cms
8. Deep Lens
Predict pollution in different parts of Bengaluru through satellite
imagery instead of deploying expensive sensors
Goal 1: Establish correlation between ground sensor and satellite imagery at
locations where both the data are available
Goal 2: Use satellite imagery where ground sensor data doesn’t exist to
estimate pollution
9. AOD and PM2.5 are different
AOD from Satellite Data PM2.5 from High-value Samplers
10. Remote Sensing from Orbit
● Earth observations are made at wavelengths
from 260 nm in the ultraviolet (UV) through
to radar wavelengths (0.1–10 cm)
○ The ability to see through clouds exists only at radar
wavelengths
● Atmospheric transparency is needed to be
able to probe down to the surface
○ Cloud cover is a significant limitation of satellite
observations for air quality
I0 = incoming solar radiation. I1 = scattered radiation from
gases and aerosols in the atmosphere. I2 = scattered
radiation from gases and aerosols scattered into the field
of view of the satellite. I3 = transmitted radiation through
the atmosphere to the surface. I4 = reflected radiation
from the Earth’s surface
11. A Brief Survey: Predicting Pollution from
Satellite Imagery
● Many organizations responsible for monitoring air quality are not taking full
advantage of the satellite data for air quality applications
○ Difficulties associated with accessing, processing, and properly interpreting observational data
○ Degree of technical skill required, which is often problematic for air quality agencies with
limited resources
● Limited studies that seek to establish relationship between satellite data and
ground-based measurements
○ Wang and Christopher showed that under certain conditions, PM2.5 mass measured at the
surface and the 550 nm Aerosol Optical Thickness (AOT) from the Moderate Resolution
Imaging Spectro Radiometer (MODIS) are well correlated (R > 0.7)
○ Detailed survey presented in the project report (year 2000 - 2015)
13. Aerosol Data from
Satellite
● Aerosols are one of the greatest sources of uncertainty in climate modeling
○ Vary in time and space leading to variations in cloud microphysics
○ Retrieving meaningful information from satellite data is not trivial task
● MODIS Aerosol Product monitors the ambient Aerosol Optical Thickness
(AOT) over the oceans globally and over a portion of the continents
○ Daily Level 2 data are produced at the spatial resolution of a 10x10 1-km (at nadir)-pixel
array
○ MOD04_L2, containing data collected from the Terra platform; and MYD04_L2, containing
data collected from the Aqua platform
15. Data Collection and Preparation
We have downloaded Modis L2 data for 66 days (1/3/2017 to 6/5/2017) and we
get 2 images per day:
· MODIs Terra
· MODIs Aqua
16. CPCB Ground Sensor Data
Sulfur Dioxide (SO2)
Carbon Monoxide (CO)
Nitrogen Dioxide (NO2)
Nitric Oxide (NO)
Oxides of Nitrogen (NOx)
Ammonia (NH3)
Methane (Methane)
Non Methane Hydrocarbon (NMHC)
Total Hydrocarbon (THC)
Temperature (Temp)
Relative Humidity (RH)
Wind Speed (WS)
Wind Direction (WD)
Vertical Wind Speed (Vertical Wind S)
Solar Radiation (S.Rad)
Bar Pressure (Bar Pressure)
Benzene (Benzene)
Toluene (Toluene)
Ethyl Benzene (Ethyl Benzene)
Xylene (M & P Xylene)
Xylene (O_Xylene)
Ozone (O3)
PM 2.5 (PM2.5)