Forecasting Global Thermospheric Neutral Density through Calibration and Data Assimilation of GRACE Measurements into the NRLMSISE-00 model

The uncertainty in thermospheric neutral density (TND) estimates is one of the largest and persistent sources of uncertainty in orbit determination and prediction (OD/OP) of low Earth orbit space objects. The TNDs required for these applications are typically obtained from corresponding models. However, the simulation and forecasting skills of these models are limited due to the model structures and the calibration period of the model parameters. Here, we present an Ensemble Kalman Filter (EnKF)-based Calibration and Data Assimilation (C/DA) approach that provides the opportunity to update the model's states and simultaneously calibrates the model’ s most sensitive parameters, such as those related to solar radiation and geomagnetic activity as well as those controlling the calculation of exospheric temperature. The advantageof this approachis that the calibrated parameters can be applied to simulate the global map of global TNDs and forecasting them in future. In this study, we investigate the improvement of the NRLMSISE-00 model after implementing the C/DA scheme using TNDsderived from the accelerometer measurements of the Gravity Recovery and Climate Experiment mission (GRACE) during February 2015 with a wide range of solar activity. We demonstrate the forecasting skills of C/DA covering the altitude of 300-600 km, though the GRACE measurements were introduced at the altitude of 410 km during the C/DA period. The calibrated model are validated along the Swarm-A, -B, and -C with mean altitude of 480, 480 and 528 km,respectively. The results indicate that our TND forecasts agree well with the POD-derived densities. After implementing the C/DA, the root-mean-squares of error (RMSE) of TND forecasts has been reduced comparedto the original NRLMSISE-00 densities,i.e., 51, 57 and 54% along the Swarm-A,-B and -C, respectively. The numerical assessment is useful to demonstrate the capability of the C/DA technique in reducing the modelling errors and their value for forecasting TNDs for applications such as collision analysis