Classifying Urban Regions by Aggregated Pollutant–Weather Correlation Strength: A Spatiotemporal Study

Koyena Ghosh, Suchismita Banerjee, Urna Basu, Banasri Basu · 2025

Read Classifying Urban Regions by Aggregated Pollutant–Weather Correlation Strength: A Spatiotemporal Study by Koyena Ghosh, Suchismita Banerjee, Urna Basu, Banasri Basu (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-22080.txt.

Introduction

Koyena Ghosh, Suchismita Banerjee, Urna Basu, and Banasri Basu ask how tightly air pollution and weather move together in Indian cities. They build a statistical system that measures dependence between pollutants and weather variables in both snapshot and time-based ways. Cities can then be grouped by the overall strength of that coupling, giving a map of distinct interaction regimes.

The method blends linear and nonlinear measures, including Pearson correlation, mutual information, and relative conditional entropy. Principal component analysis folds these into one interpretable score per pollutant and weather pair, then per city. Robustness checks that drop one variable pair at a time change little. Over time, humidity tends to lead pollutant changes, temperature tends to lag, and links fade quickly.

Catalog record

CREATORKoyena Ghosh, Suchismita Banerjee, Urna Basu, Banasri Basu
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE86 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2512-22080

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Classifying Urban Regions by Aggregated Pollutant–Weather Correlation Strength: A Spatiotemporal Study Koyena Ghosh, Suchismita Banerjee, Urna Basu, Banasri Basu Abstract Understanding pollutant–meteorology interactions is essential for environmental risk assessment. This study develops an entropy-based statistical framework to analyze static and temporal dependencies between urban air pollutants and meteorological variables across multiple Indian cities. Dependence is quantified using complementary linear and nonlinear measures, including Pearson correlation, mutual information, and relative conditional entropy. A key methodological contribution is a PCA-based composite indexing framework that integrates these heterogeneous metrics into a unified and interpretable correlation score. For each pollutant–meteorological pair within a city, PCA is used to extract a joint variability index, while spatial variability is assessed by aggregating correlations across cities. These indices are further combined to derive a comprehensive city-level correlation score that represents overall pollutant–meteorology coupling strength and enables classification of cities into distinct interaction regimes. Sensitivity analysis, performed by systematically excluding individual variable pairs, demonstrates the robustness of the framework, with no single pair exerting disproportionate influence. Temporal dependencies are examined using transfer entropy and time-delayed mutual information. Results indicate that relative humidity generally leads changes in pollutant concentrations, whereas ambient temperature tends to lag, highlighting contrasting causal influences. Mutual information peaks at zero lag and decays rapidly, indicating strong short-term interactions with limited persistence. Overall, the proposed framework provides a unified and interpretable approach for assessing complex pollutant–meteorology interactions across diverse locations and time. keywords Pollutant-Weather Correlation, Shannon Entropy, Mutual Information, Principal Component Analysis, Transfer Entropy, Multi-city Comparison † † equal-contributors: These authors contributed equally to this work. † † equal-contributors: These authors contributed equally to this work. 1 Introduction Air pollution remains one of the most pressing environmental health risks worldwide 60 , particularly in urban regions where population density, industrial activity, and vehicular emissions converge to produce complex pollutant mixtures. Understanding how atmospheric conditions influence the formation, transport, and transformation of pollutants is key to characterize the underlying dynamics of air quality. Air pollution dynamics can be viewed as a classic example of an open, driven-dissipative system 14 ; 8 ; 40 ; 34 ; 35 where energy and matter continuously flow through chemical, physical, and weather-related processes, giving rise to persistent fluctuations in concentration levels. These far-from-equilibrium dynamics generate complex spatiotemporal patterns 46 ; 49 ; 26 ; 57 ; 7 , shaped by both natural and anthropogenic factors and often described in terms of probability distributions rather than deterministic trends. From a physics perspective, such systems challenge classical equilibrium models and call for tools from nonlinear dynamics 44 , stochastic processes 61 , and information theory 21 . Pollutants (such as PM 2.5 , PM 10 , SO 2 and NO 2 and many more) are not merely passive tracers; they undergo nonlinear transformations 62 , interact through feedback mechanisms 25 , and are transported via advection-diffusion processes 33 , all modulated by meteorological (such as temperature, precipitation etc.) forcing that introduce stochastic fluctuations. Understanding the interdependence between pollutants and weather variables is essential not only for accurate predictive modeling, environmental monitoring and air quality risk assessment, but also for uncovering emergent behavior in complex environmental systems 17 ; 47 . A substantial body of literature 45 ; 39 ; 36 has examined the relationships between air pollutant concentrations and meteorological variables. However, these investigations 12 ; 42 ; 58 have been overwhelmingly dominated by the use of a single dependence measure , most commonly linear correlation metrics, and are therefore limited in their ability to capture the complex and potentially nonlinear interactions that govern pollutant–meteorology dynamics. Moreover, these approaches overlook both non-stationarity 20 and directional causality. To address these limitations, we adopt an entropy based information-theoretic approach to study the interdependence of pollutants and weather variables. Although a small number of studies have employed nonlinear dependence measures, including mutual information 23 , these measures are typically applied in isolation and are not combined with linear metrics within a unified analytical framework. In parallel, multivariate techniques such as principal component analysis (PCA) and clustering have largely focused on pollutant concentrations or meteorological variables independently 37 ; 41 . However, the lack of a comprehensive measure integrating multiple forms of statistical dependence has constrained robust and comparable assessment of pollutant–meteorology interactions. In this work, we address this issue by adopting an entropy-based information-theoretic framework, rooted in statistical physics and complex systems science, which provides powerful tools to characterize structure and dynamics in multivariate systems with inherent fluctuations 21 ; 6 . Metrics such as Shannon entropy 50 , Mutual Information (MI) 19 ; 51 , and Transfer Entropy (TE) 48 have proven valuable across fields from neuroscience to climate science 32 ; 16 , and here we extend their application to urban air quality dynamics. Using a multi-year dataset covering multiple Indian cities, we analyze interactions between the primary pollutants ( \text{PM}_{2.5} , \text{PM}_{10} , \text{NO}_{2} , and \text{SO}_{2} ) and two of the most important meteorological variables (relative humidity RH and ambient temperature AT). We compute the Pearson correlation coefficient to estimate the linear correlations and use entropy-based measures to quantify uncertainty, dependency, and the directional flow of information 43 ; 9 , thereby capturing both linear and non linear relationships. The resulting correlation data across cities and variable pairs form a high-dimensional dataset that is complex to interpret directly due to the volume of observations and the multiple metrics involved. To systematically extract dominant patterns and reduce dimensional complexity, we amalgamate these diverse metrics into a unified measure of correlation. This is accomplished by employing Principal Component Analysis (PCA) 28 ; 13 to the set of correlation measures, enabling a consistent and comparable representation of pollutant–meteorology relationships across cities. PCA allows us to construct composite indices that summarize the overall strength and structure of pollutant–weather interactions, both within and across cities, by combining information from all three correlation metrics. To the best of our knowledge, this work presents the first implementation of a PCA-based compound indexing methodology for merging multi-metric correlation data between air pollutants and meteorological variables across spatial and variable dimensions. Our results reveal spatial heterogeneity as well as similarities in pollutant–meteorological couplings, suggesting that regional climate and geography play a fundamental role in shaping these dynamics. We find cities cluster into distinct groups with varying levels of interdependence. Additionally, the transfer entropy (TE) analysis provides clear evidence of bidirectional information flow across most cities. In particular, relative humidity (RH) generally exhibits a leading influence on the pollutants, whereas ambient temperature (AT) tends to display the opposite, lagging behavior, with respect to most pollutants. The time delayed mutual information (TDMI) analysis further reveals that, for the majority of cities, the dependence between particulate matter (PM 2.5 and PM 10 ) and RH peaks at zero lag and decays approximately exponentially with increasing lag. This behavior reflects a short memory and limited temporal persistence in the underlying stochastic dynamics. Only in a subset of cities, similar trend is also observed for the gaseous pollutants (NO 2 and SO 2 ) with RH. The sequential key steps of our multi-stage analysis are summarized in the following technical roadmap, providing a structured overview that guides the detailed methodology in the subsequent sections: • Data

Read the complete text