Monitoring the Tuning and Performance of Ionospheric Data Assimilation Through Unusual Storms

Empirical and first principles models of the atmosphere and ionosphere can be improved through the use of data assimilation (DA). However, in order to have a realistic assimilation analysis, some cumbersome estimates of various sources of error like model covariances must be created reasonably accurately. The procedures for estimating such quantities are data intensive and rely on capturing the range of likely states that the model will go through under the forcing from external drivers. Unusual events like the Superstorm in May 2024, may cause excursions in ionosphere density and location that are not covered by the usual covariance estimates based on quieter conditions. There are complicated ways of assessing the overall state of error estimates in a DA system (Desroziers and Ivanov, 2001; Desroziers, et al., 2005; Daley and Barker, 2001). The full techniques involve comparing large matrices which is cumbersome and time consuming. What we desire in an operational system are more simple metrics like calculating the minimum of the cost function using the innovation vectors, which can be calculated during the assimilation at very small computational cost. Here we derive some similar additional metrics based on Desroziers et al., 2005 treatment in observational space to assess the appropriateness of the size of different errors. These metrics work well in systems where the correlation lengths are not much larger than the voxel spacing and the error matrices are sparse. We will show an example of these metrics used for ionospheric data assimilation using the H-IDA4D (High latitude Ionospheric Data Assimilation 4Dvar) model during the Gannon Superstorm. These metrics are computed to provide diagnostic information that that give indications as to whether the model error, observational error, and analysis covariance are globally estimated properly for the DA system. These metrics can then be monitored operationally to ensure that the data assimilation is operating as expected through unusual space weather events or could be used to dynamically change the error estimates.

Desroziers, G. and Ivanov, S., 2001, Diagnosis and Adaptive Tuning of Information Error Parameters in a Variational Assimilation, Q. J., R., Meteorol. Soc. 127, 1433-1452.

Desroziers, G., Berre, L., Chapnik, B., and Poli, P., 2005, Diagnosis of Observation, Background and Analysis-error Statistics in Observation Space, Q. J. R., Meteorol. Soc., 131, 3385-3396.

Daley, R. and Barker, E. 2001, NRL Atmospheric Variational Data Assimilation System, https://apps.dtic.mil/sti/pdfs/ADA396883.pdf