Exploration of dependencies of space weather and geomagnetic proxies with thermospheric density enhancement during geomagnetic storms from 2020 to 2024

• Density peak prominence is the parameter most strongly correlated with solar wind and geomagnetic drivers during CME-driven storms (R > 0.9). • Storm origin plays a major role, and the linear dependence between thermospheric density and space weather proxies differs markedly between CME and HSS events. • For HSS-driven storms, incorporating AE max and dynamic pressure improves the regression model, explaining up to 77% of the variance in density increase. The relationships and linear dependencies between geomagnetic storm drivers and thermospheric density enhancements observed by the Swarm satellites during Solar Cycle 25 (2020–2024) are investigated. A catalog of 125 storm events was compiled using SYM-H criteria and cross-checked against multiple space weather databases to identify their solar origin, including Coronal Mass Ejections (CMEs), solar wind High-Speed Streams (HSS), or a mixed contribution of multiple phenomena. Neutral density variations from Swarm-B and Swarm-C were analyzed and normalized, and features such as peak prominence, relative increase, and integrated orbital decay were extracted for each event. Correlation analyses reveal strong linear dependencies between density enhancements and some geomagnetic indices, solar wind parameters, and coupling potentials (particularly SYM-H*, Akasofu’s potential ∊ ) with correlation coefficients exceeding 0.9 for CME-driven storms. In contrast, HSS-driven storms exhibit moderate dependencies, with the relative density increase correlating most strongly with the southward component of the magnetic field (B Z -) and Akasofu’s potential ∊ . Partial regression analysis demonstrates that combining additional linear predictors, such as AE and dynamic pressure, further improves the characterization of HSS-driven storms, explaining up to approximately 77% of the variance in density increase and significantly reducing prediction error. These findings highlight the relevance of storm origin to linear behaviour, as well as the major contributions of different drivers and energy coupling functions in refining thermospheric density predictions during geomagnetic storms.