Optimizing Preconditioning Length for the TIEGCM to Improve Geomagnetic Storm Prediction Accuracy

Geomagnetic storms can have a range of potential impacts from full power grid outages to fluctuations in aerospace vehicle navigations with enhanced satellite drag. The Thermosphere-Ionosphere-Electrodynamics General Circulation Model (TIEGCM) is a model that simulates and outputs variables from the Thermospheric-Ionospheric system (T-I) to analyze the effects of geomagnetic storms. In this paper, we look at the total electron content (TEC) output of the model. The model runs conducted in this study utilize TIEGCM as a standalone model and TIEGCM as a coupled component of the Multiscale-Atmosphere Geospace Environment (MAGE) model. This study seeks to improve the prediction of storm intensity, and potential risk, by understanding the impacts of “SPINUP” time on TIEGCM. This is an important factor of the preconditioning necessary to set the initial conditions of the model before submitting a run; the default spin-up time is currently 7 days. There has been so far no systematic research in regards to the optimal SPINUP length, and this paper aims to fill this gap by comparing the simulated TEC during a couple of geomagnetic storms (May 2024, November 2021, and March 2013) with different TIEGCM SPINUP lengths to the observational data taken at the same time. In order to directly compare the observational and modeled data, which exist on different grids, have different amounts of available data points and different temporal resolutions, we had to interpolate the model data to the location and time of the observational data. Error metrics such as correlation coefficient and Root Mean Square Error (RMSE) are used to analyze the results. For the November 2021 storm, the modeled RMSE with respect to the observations was lower for longer SPINUP lengths, namely 7-day and 11-day, during the storm main phase, and hardly varied in the onset and recovery stages. In conjunction with this outcome, the correlation coefficient between the averaged observational data and the average modeled data monotonically increased from 1-day to 11-day. This suggests that longer SPINUP length typically leads to better initial conditions. However, although 11-day has a marginally higher correlation, RMSE was the lowest for 7-day SPINUP, and 7-day also requires less computational effort. This implies that the current standard SPINUP length is sufficient, and requires further research to be conducted.