Turning Delhi’s Smog into Cigarettes for Delivery Riders: How Reliable Is the Data?
We checked two months of OpenAQ Delhi PM2.5 data on AWS and tested the numbers behind RideClean.
The Problem
Delhi’s delivery riders often spend 8–10 hours a day on the road, exposed to changing levels of air pollution. However, they may not know when and where air quality is at its worst. We built RideClean to translate PM2.5 exposure into an easy-to-understand cigarette-equivalent estimate, helping riders better understand their potential exposure.
My Contribution
I helped review the OpenAQ dataset, examine the summary statistics, and check the calculations used to convert PM2.5 levels into cigarette-equivalent estimates. I also looked at how the app communicates typical pollution levels versus live readings when monitoring data is available.
Where the Data Comes From
We used publicly available OpenAQ data hosted on AWS Open Data (S3), covering October 15 to December 15, 2025. The dataset contained 57 monitors and 75,894 station-hours of observations. The analysis used real monitoring data rather than synthetic or sample records.
What the Numbers Show
- Average PM2.5: Approximately 193 µg/m³, around 12.9 times the WHO guideline used in our analysis.
- High pollution exposure: 96.9% of recorded hours exceeded 60 µg/m³.
- Worst time period: 10 PM–2 AM, averaging approximately 202 µg/m³.
- Relatively cleaner period: 2 PM–6 PM, averaging approximately 105 µg/m³.
- Estimated rider exposure: A typical delivery shift was estimated to correspond to approximately 3.3 cigarettes using our conversion assumptions.
How the Conversion Works
The estimate uses Berkeley Earth's cigarette-equivalence approach, under which a full day of exposure at 22 µg/m³ of PM2.5 is approximately equivalent to one cigarette. We also applied an exposure factor of 1.3 for time spent on the road, based on a Delhi study, and a factor of 0.5 for time spent indoors with air conditioning.
These are simplified estimates intended to communicate pollution exposure, not literal measurements of cigarettes smoked or a precise prediction of individual health outcomes.
Limitations
Only a small number of monitors are currently reporting live data. Therefore, RideClean displays typical pollution levels for a location and provides live readings only when a monitor is actively reporting. We make this distinction clear rather than presenting estimated or historical values as live measurements.
What I Learned
One key lesson was that public data needs careful interpretation before it can be turned into a simple user-facing metric. Monitor coverage, timestamps, missing observations, and the assumptions behind exposure estimates all affect how confidently the results can be presented. Making the numbers easy to understand is important, but communicating their limitations is equally important.
AWS Used
- AWS Open Data (S3): Accessing the public air-quality dataset.
- AWS Lambda: Backend processing.
- AWS Amplify: Hosting the application.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article