Understanding the long-term variability of solar magnetic fields is central to improving our ability to interpret and predict solar activity. This thesis investigates the evolution of large-scale flows and magnetic structures in the solar atmosphere using automated, multi-wavelength observational analysis combined with data-driven techniques.

Using long-term observations from SDO/AIA, the Nobeyama Radioheliograph, and historical Meudon synoptic maps, this work presents a unified observational picture of how large-scale flows, and filament evolution encode the dynamics of the solar magnetic cycle. A tracer-independent image-correlation framework is utlised to derive differential rotation across multiple atmospheric heights, revealing systematic variation with temperature, height, and solar cycle phase. Complementary radio observations are used to investigate large-scale flows and the impact of solar activity on them, providing new observational support for the “magnetic tree” paradigm linking atmospheric dynamics to deeper magnetic anchoring. Key results of this thesis is the demonstration that polar filaments act as early precursors of solar cycle amplitude, providing predictive capability years before traditional polar-field–based indicators as well as the very first observational signature of meridional flow-like signature in the hotter atmosphere. Building on this foundation, the final part of the work introduces an ongoing machine-learning framework to automatically detect filaments from kodaikanal hand-drawn suncharts in the cycles 19 and 20 which will be further extended to include prominences as well throughout the entire period of data availability. Together, this thesis connects large-scale solar flows, filament evolution, and data-driven magnetic diagnostics, providing new observational constraints on the solar dynamo and future directions for predictive solar physics.