Decomposing Climate Signals: Applying Wavelet Resolutions for Source-Type Identification in Non-Stationary Atmospheric Pollutant Datasets

Understanding the complex dynamics of modern climate change requires distinguishing between anthropogenic forcing and natural low-frequency atmospheric adaptability. Traditional linear time-series analyses often fall short when evaluating non-stationary climate datasets, where the frequency and amplitude of signals shift over time. This research proposes an advanced signal processing framework that utilizes varying wavelet resolutions to decompose complex atmospheric time-series data and isolate specific pollutant source types. By applying continuous wavelet transforms (CWTs) to the NASA Modern-Era Retrospective Analysis for Research and Applications Version 2 (MERRA-2) aerosol and atmospheric composition of time series, the temporal evolution of the distinct frequency bands can be mapped. The methodology focuses on optimizing wavelet resolutions to effectively separate high-frequency, localized pollution events from long-term, multi-decadal trends. Multi-resolution analysis allows for the exact identification of underlying variance drivers, differentiating the persistent signatures of industrial aerosol emissions from periodic natural drivers such as seasonal biomass burning or cyclic dust transport. Preliminary framework testing demonstrates that selecting the appropriate wavelet resolution is critical for minimizing signal escape and accurately categorizing variance source types. This application is a mathematical approach that provides a more robust diagnostic tool for atmospheric dynamics, offering deeper insights into how intersecting climatic drivers interact across different time scales. Ultimately, bridging high-resolution signal processing with the MERRA-2 meteorological datasets enhances the predictive modelling of future atmospheric composition and air quality scenarios.