Frequency-Domain clustering of Non-Attainment cities of India using CEEMDAN decomposition of aerosols and Coupled meteorological drivers

• Applied CEEMDAN decomposition to multi-source EO and reanalysis datasets for 131 NCAP cities. • Constructed a multi-parameter frequency-domain clustering framework for urban atmospheric variability. • Distinguished emission-driven high-frequency modes from synoptic and seasonal variability regimes. • Integrated satellite-derived frequency signatures with actionable airshed-level management insights. • Demonstrated a transferable EO-based framework linking aerosols, meteorology, and regional air quality. Understanding the temporal scales that govern aerosol build-up and persistence is essential for diagnosing urban air-quality variability under complex atmospheric forcing. This study develops a multi-parametric, frequency-domain framework for 131 non-attainment cities across India’s National Clean Air Programme (NCAP) using CEEMDAN decomposition. A combined Earth Observation and reanalysis dataset is used to extract frequency-domain variability, with independent in-situ PM 2 . 5 observations employed for validation. Each parameter is decomposed into Intrinsic Mode Functions (IMFs), and their normalized energies are integrated into multi-parameter spectral signatures to identify coherent regimes of atmospheric variability. Seven characteristic cluster types are identified, reflecting synoptic-dominated regimes, turbulence-driven regimes, hybrid synoptic–turbulent modes, and mixed regimes shaped by persistent thermal or moisture anomalies. These signatures differentiate cities regulated by slow-evolving aerosol–moisture coupling from those driven by rapid boundary-layer or wind-shear fluctuations. Winter clusters capture inversion-linked stagnation, pre-monsoon clusters reflect dust-modulated synoptic forcing, monsoon clusters align with convective–stratiform transitions, and post-monsoon clusters are shaped by biomass burning and shallow boundary layers. The framework translates frequency-domain variability into physically interpretable ventilation regimes, offering the first demonstration of how signal decomposition can quantify intrinsic atmospheric capacity for pollutant accumulation and dispersion at the city scale. These insights enable a process-based classification of NCAP cities, supporting dynamic airshed characterization and season-specific emission-control planning.