Global Air Quality Analysis: Key Pollution Trends Across Major Cities

By abhishek.verma75000 · July 1, 2026

Our latest analysis explores pollution patterns across major cities worldwide, uncovering trends in particulate matter, AQI classifications, pollutant…

Lahore, Pakistan ranks as the most polluted city for both PM2.5 and PM10 levels. Two bar charts and supporting data tables have been generated showing the top 10 cities for each pollutant. Delhi and Beijing closely follow Lahore in overall particulate pollution.

The AQI Class distribution varies significantly across cities. Two stacked bar charts were generated — one showing raw counts and another showing percentage breakdowns — making it easy to compare air quality profiles. Most cities in Asia are dominated by 'Unhealthy' air quality: Dhaka leads at 73.3%, followed by Delhi at 66%, Lahore at 54.5%, and Beijing at 43%. In contrast, London stands out with 83.2% of records classified as 'Good', and New York also skews positive with 55.6% 'Good'. Jakarta and Sao Paulo fall in the middle, with 'Moderate' being their most common class at 38.7% and 53.9% respectively.

Over the 91-day observation period (November 8, 2025 to February 6, 2026), daily average PM2.5 levels showed a slight upward trend of about 0.115 µg/m³ per day. The chart displays daily fluctuations in light blue alongside a 7-day rolling average in crimson, making it easy to see both short-term variability and the overall direction. The overall daily average was 66.1 µg/m³, with levels starting at 39.4 µg/m³ and ending at 42.7 µg/m³.

The analysis identified distinct peak periods for NO2 and Ozone pollution. NO2 reached its highest weekly average of 45.4 during the week of November 23, 2025, with November 2025 also being the peak month (avg 39.1). Ozone, on the other hand, peaked later — hitting 66.3 during the week of December 7, 2025, with December 2025 as its peak month (avg 57.4). Two charts were generated: a weekly line chart showing trends over time and a monthly bar chart comparing both pollutants side by side.

The analysis reveals that PM2.5 and PM10 are by far the most strongly correlated pollutants, with an almost perfect correlation of r = 0.993. This makes sense as both measure particulate matter of similar origins. CO also shows strong positive ties with both PM2.5 (r = 0.749) and PM10 (r = 0.739), suggesting these pollutants tend to rise and fall together — likely from shared combustion sources. SO2 also correlates moderately with both PM2.5 and PM10 (r = 0.696 each). On the flip side, Ozone shows a slight negative correlation with PM2.5 (r = -0.196), meaning higher particulate levels tend to coincide with slightly lower ozone. Two visualizations were generated: a full correlation heatmap showing all pollutant relationships at a glance, and a bar chart ranking the top correlated pairs by strength.

Across all cities, there were 310 total hourly readings that exceeded the hazardous PM2.5 threshold of 250.4 µg/m³. Beijing, China was the most affected city, recording 157 hazardous hours — more than half of all exceedances. Two charts were generated: one showing monthly exceedance counts per city side-by-side, and another summarizing total hazardous hours by city using a red color scale to highlight severity.

The grouped bar chart and data table break down the average concentrations of NO2, SO2, CO, and Ozone for each AQI Class category in the dataset. The visualization clearly shows how pollutant levels rise as air quality worsens — from Good through Moderate, Unhealthy, and beyond. Each pollutant is represented as a separate bar group, making it easy to compare their relative concentrations within and across AQI classes.

The analysis ranked all cities by their average SO2 and Aerosol Optical Depth (AOD) levels separately, revealing clear pollution hotspots. Data tables were generated showing the top 10 cities for each metric. Beijing, China leads in SO2 with an average of 101.14, while Delhi, India tops the AOD rankings at 0.518. Remarkably, 8 cities appear in the top 10 for BOTH pollutants, indicating compounded industrial pollution signatures across multiple major urban centers worldwide.

Comprehensive summary statistics and visualizations have been generated for all 7 numeric pollutant columns (PM2.5, PM10, NO2, SO2, CO, Ozone, and Aerosol Optical Depth). A styled summary table displays the mean, median, standard deviation, minimum, 25th percentile, 75th percentile, and maximum for each pollutant. A heatmap shows the normalized average levels (relative to each pollutant's maximum), making it easy to compare relative magnitudes across pollutants on a 0–1 scale.

Using the IQR method, outliers were detected in both CO and PM10 columns. CO outliers fall below -1,745.5 or above 3,650.5, with 1,349 total outliers found. PM10 outliers fall below -119.5 or above 238.6, with 436 total outliers found. The most extreme readings for both pollutants were recorded in Beijing, China on December 22, 2025 — CO hit 13,039 and PM10 reached 434. Lahore, Pakistan had the highest outlier percentage at 29.12% of its records. Two side-by-side box plots visualize the outlier distribution for CO and PM10, and a bar chart shows outlier percentages across the top cities.