Street Air Quality

Street-level air pollution measurements and analysis tools.

Delhi Air Quality (May 2026)

Based on 98 PM2.5 readings collected across 5 days:

Statistic	PM2.5 (μg/m³)	CO₂ (ppm)
Mean	117.6	541
Median	94.0	415
Min	22.0	11
Max	584.0	1497

Threshold Exceedance (EPA 2024 Standards)

Threshold	% Readings
Above Good (>9 μg/m³)	100%
Above Moderate (>35.4 μg/m³)	99%
Unhealthy for Sensitive Groups (>55.4 μg/m³)	99%
Unhealthy (>125.4 μg/m³)	21%

100% of readings exceeded the EPA "Good" air quality threshold. Nearly all readings (99%) were in the "Unhealthy for Sensitive Groups" category or worse.

Installation

pip install streetaqi

# With LLM support (for image annotation)
pip install streetaqi[llm]

# With map support
pip install streetaqi[maps]

# All optional dependencies
pip install streetaqi[all]

CLI Commands

Analyze pollution data

Run statistical analysis with publication-ready outputs:

streetaqi analyze --readings exports/pollution_logs.csv --output output/analysis

Outputs:

• output/analysis/figs/fig1_map.html - Interactive map color-coded by PM2.5 level
• output/analysis/figs/fig2_histogram.pdf - PM2.5 and CO₂ distributions
• output/analysis/figs/fig3_boxplot_by_day.pdf - Daily variation
• output/analysis/figs/fig4_pm_co_scatter.pdf - PM2.5 vs CO₂ correlation
• output/analysis/tabs/*.tex - LaTeX tables for publication

OCR sensor images

Extract PM2.5 and CO₂ readings from air quality sensor photos using Claude or Gemini APIs:

streetaqi annotate \
  --images "exports/images/pollution/**/*.jpg" \
  --model gemini-2.0-flash \
  --manifest exports/manifest.json \
  --output output/annotations/

Options:

• --model: gemini-2.0-flash (default), claude-haiku-4-5
• --batch: Use batch API for 50% cost savings (async)
• --manifest: Include logged values for comparison

QC viewer

Generate HTML viewer for quality control:

streetaqi viewer \
  --readings output/annotations/pollution_readings_*.json \
  --output output/annotations/viewer.html

The viewer shows images with OCR readings, compares to logged values, and allows manual corrections. Export corrected data as JSON or CSV.

Data

Delhi (May 2026)

Metric	Count
PM2.5 readings	98
CO₂ readings	98
Unique locations	~98

Data Schema

Each reading in data/rider/{city}/readings.json:

{
  "id": "47a27458-...",
  "timestamp": "2026-05-16T16:46:33.097000+05:30",
  "timestamp_utc": "2026-05-16T11:16:33.097+00:00",
  "gps": {
    "latitude": 28.643719,
    "longitude": 77.2989862
  },
  "reading": {
    "pm25": 99,
    "co": 414
  },
  "image": {
    "local_path": "images/pollution/day-09/001_itinerary-1-part-2.jpg",
    "original_name": "17789301803242784560681094233323.jpg",
    "remote_url": "https://..."
  },
  "metadata": {
    "day": 9,
    "itinerary": "1-2",
    "title": "Itinerary 1 - Part 2",
    "stop_id": "1.2",
    "is_traffic_stop": false,
    "is_traffic_jam": false,
    "note_raw": "1.2",
    "address": "Karkari Mor Flyover, East Delhi, Delhi, India",
    "road_type": "primary",
    "road_name": "Karkari Mor Flyover"
  }
}

Field Descriptions

Field	Description
reading.pm25	PM2.5 concentration (μg/m³)
reading.co	Carbon dioxide (ppm)
metadata.road_type	OSM highway classification
metadata.is_traffic_stop	Reading taken at traffic signal
metadata.is_traffic_jam	Reading taken in traffic jam
metadata.note_raw	Raw rider note (preserved for reference)

Road Type Classifications

From OpenStreetMap highway tags:

Type	Description
motorway	Expressway/freeway
trunk	Major arterial road
primary	Main road (like national highway)
secondary	State highway level
tertiary	Local connecting road
residential	Neighborhood street

Traffic Classification Logic

The rider enters notes like "1.2" or "2.1 traffic stop":

• is_traffic_stop = True if note contains "traffic stop" (case-insensitive)
• is_traffic_jam = True if note contains "traffic jam" (case-insensitive)
• stop_id is extracted as first token (e.g., "1.2" from "1.2 traffic stop")

Data Collection

Data collected using the rider route tool and processed via soundscape.

# Convert rider export to streetaqi format
cd ../soundscape
uv run soundscape convert-rider-pollution \
    --manifest data/rider/export-2026-05-24/manifest.json \
    --output ../streetaqi/data/rider \
    --city delhi \
    --geocode

License

MIT