The Spatiotemporal Dynamics of Air Pollutants and the Universal Thermal Climate Index in 370 Chinese Cities

Kaiqi Huang, Linlin Zhang, Qingyan Meng, Mona Allam, Jing Pan, Shize Chen, X. G. Lei, Mengqi SunAtmosphere 2025 · DOI

TL;DR

This study reveals how air pollutants, beyond just weather, significantly impact urban thermal comfort across China. Using advanced machine learning, researchers found that pollutants like ozone and PM can amplify heat stress in summer and provide insulating warmth in winter, with these effects varying greatly by climate zone and time of day. This work provides crucial insights for urban planning and public health strategies that address both air quality and thermal comfort.

What Problem Does This Paper Address?

Urban livability and public health are increasingly threatened by extreme weather events, making outdoor thermal comfort a critical metric. While meteorological factors like temperature and humidity are known drivers of thermal stress, the influence of air pollution on how we perceive thermal comfort has been largely overlooked. This is a significant gap because cities are experiencing both rising temperatures and worsening air quality, and these two phenomena likely interact in complex ways. Understanding these interactions is vital for developing effective strategies to make urban environments more resilient and healthy, especially in rapidly urbanizing regions like China.

Methods

The researchers employed a multi-pronged approach:

  • Spatiotemporal Analysis: They analyzed data across 370 Chinese cities over a five-year period (2020-2024) to capture variations in thermal comfort and air quality across different locations and times.
  • Machine Learning (XGBoost): To model the complex, non-linear relationships between multiple air pollutants, meteorological variables, and thermal comfort (measured by the Universal Thermal Climate Index - UTCI), they used the eXtreme Gradient Boosting (XGBoost) algorithm. This algorithm is known for its robustness in handling high-dimensional and heterogeneous datasets.
  • SHAP (SHapley Additive exPlanations) for Interpretability: To understand how each pollutant and meteorological factor influences UTCI, they used SHAP values. This technique breaks down the model's predictions to attribute the contribution of each input feature, allowing for both global (overall importance) and local (city-specific) interpretations. This is crucial for understanding the underlying mechanisms.
  • Climate Zone Stratification: The analysis was further refined by stratifying data into four major Köppen climate zones within China, allowing for the examination of region-specific pollutant impacts.

Datasets

  • Universal Thermal Climate Index (UTCI): Gridded UTCI data were obtained from the ERA5 reanalysis dataset provided by the Copernicus Climate Data Store (CDS). This dataset offers global coverage at a 0.25°× 0.25° resolution.
  • Air Quality Data: Hourly air pollution observations for six major pollutants (PM2.5, PM10, SO2, NO2, O3, and CO) were sourced from the publicly available platform maintained by Quotsoft, which compiles data from the China National Environmental Monitoring Center (CNEMC).
  • Meteorological Data: Key meteorological variables (air temperature, downward short-wave and long-wave radiation, wind speed components, and dew point temperature) were extracted from the ERA5 hourly reanalysis dataset.

Key Insights and Ideas

  • Pollutants as "Thermal Regulators": The most striking insight is that air pollutants don't just passively exist; they actively modulate thermal comfort. Ozone and PM can amplify heat stress in summer by absorbing solar radiation, while in winter, PM2.5 and CO can act as "insulators," trapping long-wave radiation and reducing heat loss at night.
  • Diurnal and Seasonal Reversals: The thermal impact of pollutants can flip entirely between day and night. For example, ozone can enhance heat stress during sunny summer days but contribute to cooling at night. This highlights the critical need for diurnal resolution in thermal risk assessments.
  • Climate Zone Matters: The same pollutant can have vastly different effects depending on the local climate. Arid regions might see ozone-driven heat stress exacerbated, while cold regions might experience a slight warming benefit from particulate matter in winter.
  • Meteorology as a Dual Regulator: Weather conditions don't just directly affect thermal comfort; they also influence how pollutants behave and exert their thermal effects. For instance, high humidity can disperse pollutants, reducing their impact.

Key Results

  • Summer Heat Trend: The national average UTCI in China showed a slight but consistent upward trend from 2020-2024, indicating intensifying summer thermal conditions and an increase in severe heat stress events.
  • Winter Stability: The national average winter UTCI remained relatively stable, but extreme cold stress events showed significant spatial heterogeneity, particularly in northern and western regions.
  • Dominant Pollutants: In summer daytime, ozone (O3) and PM10 were the strongest contributors to increased thermal stress. In winter nighttime, PM2.5 and CO showed an insulating effect, moderating cold.
  • City-Specific Variations: SHAP analysis for representative cities across different climate zones revealed distinct pollutant contributions. For instance, in Guangzhou (tropical), pollutants primarily cooled during summer daytime, while in Harbin (cold), they contributed to warming in winter daytime.

Limitations and Future Work

  • ERA5 Resolution: The authors acknowledge that ERA5 data, while consistent, may underrepresent fine-scale urban heterogeneity. Future work could incorporate higher-resolution, ground-based observations.
  • Model Extensions: They suggest extending modeling frameworks to include causal inference or hybrid physics-ML models to enhance robustness.
  • Long-Term Climate Change: The study focused on a five-year period. Future research should investigate these interactions over longer timescales, considering the impacts of climate change.
  • Complex Feedback Loops: While the study captures interactions, explicitly modeling feedback processes like urban heat island effects and anthropogenic heat release could further refine assessments.

Summary updated

Abstract

Outdoor thermal comfort is a critical determinant of urban livability and public health, particularly in the face of the increasing frequency and intensity of extreme weather events. While meteorological variables are well-established drivers of thermal stress, the influence of ambient air pollution on human thermal perception remains poorly understood and largely overlooked in urban climate research. To address this gap, this study investigates the multidimensional effects of six major air pollutants PM2.5, PM10, SO2, NO2, O3, and CO on the Universal Thermal Climate Index (UTCI) across 370 Chinese cities from 2020 to 2024. Using integrated spatiotemporal analysis, we found significant seasonal, diurnal, and climatic heterogeneity in pollutant–UTCI interactions. Our findings reveal that O3 and PM10 amplify thermal stress during summer daytime through photochemical heating and radiative forcing, whereas PM2.5 and CO reduce nocturnal heat loss in winter by trapping long-wave radiation, effectively acting as thermal insulators. These effects are further modulated by local climate: arid regions (e.g., Lanzhou) experience exacerbated O3-driven heat stress, while cold zones (e.g., Harbin) benefit from particulate-induced warming in winter. Meteorological factors serve as dual regulators; temperature and solar radiation directly elevate the UTCI, while wind and humidity govern pollutant dispersion and thus indirectly shape thermal comfort. This study not only advances the scientific understanding of air pollution’s role in urban thermal environments but also provides actionable, data-driven insights for climate-resilient urban planning, public health interventions, and integrated environmental policies that jointly address air quality and thermal comfort in rapidly urbanizing regions.
BibTeX
@inproceedings{huang2025spatiotemporal,
  author = {Kaiqi Huang and Linlin Zhang and Qingyan Meng and Mona Allam and Jing Pan and Shize Chen and X. G. Lei and Mengqi Sun},
  title = {{The Spatiotemporal Dynamics of Air Pollutants and the Universal Thermal Climate Index in 370 Chinese Cities}},
  booktitle = {Atmosphere},
  year = {2025},
  doi = {10.3390/atmos16111263},
}
Figure