A storm-time statistical enhancement of NRLMSIS 2.1 thermospheric density during the May 2024 extreme geomagnetic storm

Abstract Thermospheric mass density is a key parameter for low Earth orbit (LEO) satellite drag modeling and space situational awareness, yet extreme geomagnetic storms can trigger rapid and large density enhancements that challenge empirical model accuracy and operational orbit prediction. This study examines the May 2024 extreme geomagnetic storm (Dstmin = −406 nT) and assesses the storm-time performance of the NRLMSIS 2.1 model using thermospheric densities derived from Swarm-C accelerometer observations. We first diagnose the limitations of NRLMSIS 2.1 during this event and show that its storm-time bias exhibits a strong statistical dependence on the solar activity index F10.7 based on logarithmic residuals. To address this deficiency without changing the internal model structure, we develop an event-level statistical enhancement scheme that applies an F10.7-dependent proportional correction to the NRLMSIS 2.1 density output under high solar activity conditions. The storm is divided into initial, main, and recovery phases to evaluate phase-dependent performance, and three additional severe geomagnetic storms are further analyzed to test robustness. The proposed enhancement substantially improves storm-time density modeling, reducing root mean square error and mean absolute error by approximately 30–70 %, with the most pronounced enhancements observed during the initial phase. The method is computationally efficient, physically interpretable, and suitable for rapid thermospheric density correction during extreme space weather events and operational satellite drag applications.