Description
Air pollution, specifically elevated PM2.5 levels, poses serious public health and economic risks, contributing to respiratory illness, reduced labor productivity, and increased healthcare costs. Air quality risk is shaped by both how often pollution exceeds safe thresholds and how severe those conditions become when they do. This dataset consists of human health impact risk scores derived from daily PM2.5 concentrations between 2017 and 2022, combining the frequency of exceedance of the World Health Organization’s hazardous PM2.5 level with the 95th percentile severity at each location. This dual-metric approach supports informed decision-making for public health, environmental risk assessment, and climate resilience planning.
Technical Description
This global air quality risk product is derived from daily surface-level PM2.5 concentration data spanning 2017 to 2022. The core methodology combines two key metrics to assess risk at each location: frequency and severity. Frequency is calculated as the proportion of days per year in which PM2.5 concentrations exceed the World Health Organization's 2021 guideline of 15 ?g/m³, providing a normalized measure of how often air pollution crosses hazardous thresholds. Severity is captured using the 95th percentile of PM2.5 values over the full time period, representing peak pollution intensity. The severity component is normalized using a fixed health-relevant ceiling (150.5 ?g/m³, the “unhealthy” level from the AQI classification scheme) —and then blended using a geometric mean to generate a composite air quality risk index.
To further support interpretability and comparative analysis, the continuous risk index is classified into five risk classes using global percentile thresholds. The index is segmented into very low, low, moderate, high, and extreme risk zones based on the 10th, 30th, 60th, and 85th percentiles of global index values, enabling a data-driven classification system. The resulting outputs include the normalized risk index and a categorical risk class band, all provided at ~1 km resolution globally. These layers offer a standardized, threshold-light baseline for assessing historical air quality risks with both epidemiological relevance and geographic specificity.
This dataset contains the following fields:
- air_quality_risk_index: The raw air quality risk index at a point location. This value is the geometric mean of the bad air quality frequency and the normalized severity factor. These values range from 0.0 - 1.0, where increasing values are associated with increased risk (from both frequency and severity).
- air_quality_risk_class: The air quality risk classification at the point location. Values are categorized into a 1-5 risk classification system, ranging from Very Low to Extreme and are derived from the air_quality_risk_index metric in the following way: Very Low (<10th percentile), Low (10th-30th percentile), Moderate (30th-60th), High (60th-85th) and Extreme (>85th percentile).