Entropy-Based Analysis of Urban Pollutant-Weather Correlations

Abstract

We employ statistical physics and information-theoretic methods to quantify the dependencies between key atmospheric pollutants and meteorological variables across multiple Indian cities. To capture both linear and nonlinear relationships, we introduce a Composite Correlation Index (CCI) that combines the Pearson correlation coefficient with entropy-based measures, including mutual information and conditional entropy. Based on the CCI values, cities are clustered into distinct groups, uncovering regional similarities in pollutant-meteorology interactions that may reflect shared climatic or environmental conditions. To explore temporal structure and causal dynamics, we analyze the relationship between particulate matter (PM2.5) and relative humidity (RH) using transfer entropy, which reveals a bidirectional flow of information in most locations. Further time-domain analysis via time-delayed mutual information shows that, in many cities, the dependence between PM2.5 and RH peaks at zero lag and decays exponentially thereafter, indicating predominantly contemporaneous interactions with limited memory. This integrative framework provides a robust approach to characterizing atmospheric interaction regimes, bridging statistical physics with environmental complexity and revealing new insights into the pollutant-meteorology dynamics.