Improving Covariance Matrix for Aurora Data Assimilation

Recent advancements in Ionosphere-Thermosphere (IT) system simulations have seen significant improvements by integrating data assimilation methods. However, one limitation identified in existing approaches is the use of a covariance matrix based on a Markov random field that does not represent the real spatial correlations of IT system drivers (i.e., auroral precipitation and electric fields). By employing auroral energy flux data from the Special Sensor Ultraviolet Spectrographic Imagers (SSUSI) observations, we analyzed the correlation of auroral energy flux across different locations from 50º to the pole (geomagnetic latitude). We found that auroral fluxes within the auroral oval are positively correlated with themselves while negatively correlated with those in the polar cap and sub-auroral regions. The positive correlation first decreases from 1 (self-correlation of the reference point itself) with the distance by a characteristic (e-folding) length of ~400-500 km, which corresponds to the region that is adjacent to the reference point at the same side of aurora. There is a secondary peak of the positive correlation at ~4500 km away from the reference point, which corresponds to the opposite side of the aurora oval, suggesting that the processes that drive the aurora temporal variability are connected back in the magnetosphere. We fitted the spatial correlation coefficient as the function of MLAT and MLT and considered the dependence on distance. The fitted function can be projected to all locations, including those where no observations are available. We then create the new covariance matrix based on the fitted function, which better aligns with the spatial correlation observed from the SSUSI data and, at the same time, fills up the data gaps. We further examine the differences in the data assimilation products introduced by the new covariance matrix and how it affects the IT model, such as TIEGCM, to better simulate and understand space weather during geomagnetic storms.