Using sensor data and artificial intelligence (AI), they found that particulate matter (PM) travels from surrounding areas with less restrictive policies. The study recommended revising regional programs to include socioeconomic factors, like tax revenue, to create more sustainable and effective air quality management.
Background
Urban air pollution, fueled by things like industrial activity and home heating, is still a major public health issue in cities around the world. Places like Kraków have made progress with clean-air efforts, including bans on solid fuels. However, past research has often examined the problem in isolation, treating emission sources, weather patterns, and social or economic factors separately.
This study took a different approach. It combined a dense network of air quality sensors with detailed information on heating sources in nearby towns and local tax data. With the help of advanced machine learning and explainable models, the researchers built a single, clear picture of how pollution moves across areas and how policies can better reflect both environmental and social realities.
Methodology and Analytical Approach
The research relied on a dense, low-cost sensor network, paired with unique data on heating appliances in the surrounding areas and municipal tax revenue for socioeconomic context. Air quality and weather data was collected via APIs, while machine learning (ML) models like Random Forest were used to accurately fill in any missing data.
A key analytical component was unsupervised machine learning.
The researchers used K-means clustering with dynamic time warping (DTW) to group sensors by similar long-term PM2.5 concentration trends, revealing pollution patterns that weren’t necessarily tied to geography. To explain these patterns, they applied a suite of explainable AI (XAI) techniques, including principal component analysis (PCA) and Random Forest models, to identify important predictors.
Further interpretation tools like H-statistics and accumulated local effects (ALE) plots helped uncover how variables such as heater type, weather conditions, and municipal wealth interact in complex, nonlinear ways.
This spatiotemporal framework allowed the team to go beyond simple correlations. Instead, they built a transparent model that shows how emissions from less affluent, less regulated areas affect air quality in Kraków, offering a data-backed foundation for regional, sustainable policymaking.
Findings and Analysis
The study found a clear socioeconomic link. Municipalities with lower tax revenue per capita had a higher reliance on solid fuel heating. This pattern was especially evident in western municipalities like Czernichów and Skawina, which lie directly upwind of Kraków. Because these areas are frequently impacted by westerly winds, their emissions pose a direct threat to the city’s air quality, making them key targets for intervention.
Using explainable AI, the researchers uncovered two dominant pollution patterns. One group, referred to as the “old furnace group,” exhibited PM2.5 levels that spiked in the evenings, coinciding with times when residents manually fed their heaters. The second group’s pollution levels were more affected by meteorological conditions such as humidity, indicating it was likely receiving transported or background pollution.
Clustering analysis confirmed these trends.
Areas with high PM2.5 levels matched closely with places that still use a significant number of solid fuel heaters. While Kraków itself has nearly eliminated these, the city’s air quality remains tied to what’s happening in neighboring towns.
The findings suggest that effective air quality policy must go beyond city limits, focusing on specific high-impact areas and local heating habits. The researchers also advise against continuing to subsidize modernized solid fuel stoves, even those meeting eco-design standards, as they still contribute to harmful emissions. Instead, full replacement with non-emitting systems should be the goal.
A Call for Smarter, Fairer, Region-Wide Solutions
This study provides clear evidence that local actions, however ambitious, are insufficient when air pollution crosses administrative boundaries.
Despite Kraków’s city-level ban on solid fuels, emissions from nearby municipalities continue to shape the city’s air quality. In effect, Kraków’s clean-air efforts are being undermined by a fragmented policy landscape.
What this research highlights isn’t just the physical movement of pollutants, but the structural mismatch between where emissions originate and where policy authority lies. The data-driven models used in the study expose a systemic gap in that emissions don't stop at city borders, but current policies often do.
Addressing this calls for coordinated, region-wide planning that accounts for behavioral patterns, wind dynamics, and economic inequities.
The authors argue for a shift away from one-time subsidy programs toward sustained public investment in full decarbonization of household heating, especially in lower-income municipalities that contribute disproportionately to regional pollution.
While the study stops short of establishing causality and is based on one representative week in March 2022, it sets a precedent for how explainable AI and spatiotemporal data can be applied to policymaking.
Journal Reference
Sabal, M., Danek, T., Mateusz Zareba, & Elzbieta Weglinska. (2025). Explainable AI for effective management of urban heat sources. Scientific Reports, 15(1), 40616–40616. DOI:10.1038/s41598-025-24305-z. https://www.nature.com/articles/s41598-025-24305-z
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.