Accounting for representation uncertainties in data assimilation of Earth radiation belts satellite observations to improve Space Weather forecast

The Earth’s radiation belts are a complex system that endangers satellites instruments and especially electronics onboard spacecraft. Its dynamic may quickly change over several orders of magnitude. Forecasting and reanalyzing this environment is of prime importance for Space Weather and Space Climate assets. Salammbô is a model of the radiation belts dynamics that has been developed at ONERA. In order to correct Salammbô uncertainties, satellite observations coming from ONERA large database of in-situ measurements (IPODE, Ionising Particle Onera DatabasE) are assimilated using an Ensemble Kalman Filter (EnKF). For the EnKF to be optimal, it is important to quantify model and observation errors. As part of the Radiation Belt Forecast and Nowcast activity (RBFAN), this data assimilation tool is used as a prototype of Space Weather service and is available on the ESA SWE Network Portal since July 2023.

This study focuses on the estimation of observations pre-processing errors and their impact on data assimilation, which is a topic not currently covered in our field of research. One of the major sources of uncertainty is related to observation’s locations. Indeed, it is necessary to rely on magnetic field models to convert geographic locations to magnetic coordinates which are used in typical radiation belts codes. In IPODE database, the computation of observation’s magnetic coordinates is done using the Olson-Pfitzer Quiet magnetospheric model (OPQUIET), following recommendations from COSPAR/PRBEM guidelines. OPQUIET has the advantage to be fast to compute. However, it is a static model that does not consider the magnetospheric dynamic. Therefore, OPQUIET makes an error on the coordinates computation which then impacts Salammbô results. This contribution focuses on (1) the observations representation error induced by the use of OPQUIET in comparing its L* computations with the ones computed with the dynamical magnetic field model Tsyganenko 89 model (T89) along 15 years of THEMIS spacecraft orbit, (2) a simple and analytical model allowing to consider this error in the data assimilation scheme, and (3) the impact of this error on Salammbô-EnKF code. We conclude that this error can reach three orders of magnitude and consequently has to be carefully taken into account in the assimilative process.