A realistic simulation framework to evaluate ionospheric tomography

Observation of the 3-dimensional (3-D) electron density of the ionosphere is useful to study large-scale physical processes in space weather events. Ionospheric data assimilation and ionospheric tomography are methods that can create an image of the 3-D electron density distribution. While multiple techniques have been developed over the past 30 years, there are relatively few studies that show the accuracy of the algorithms. This paper outlines a novel simulation approach to test the quality of an ionospheric tomographic inversion. The approach uses observations from incoherent scatter radar (ISR) scans and extrapolates them spatially to create a realistic ionospheric representation. A set of total electron content (TEC) measurements can then be simulated using real geometries from satellites and ground receivers. This data set, for which the ‘truth’ ionosphere is known, is used as input for a tomographic inversion algorithm to estimate the spatial distribution of electron density. The reconstructed ionospheric maps are compared with the truth ionosphere to calculate the difference between the images and the truth. To demonstrate the effectiveness of this simulation framework, an inversion algorithm called MIDAS (Multi-Instrument Data Analysis Software) is evaluated for three geographic regions with differing receiver networks. The results show the importance of the distribution and density of GPS receivers and the use of a realistic prior conditioning of the vertical electron density profile. This paper demonstrates that when these requirements are met, MIDAS can reliably estimate the ionospheric electron density. When the region under study is well covered by GPS receivers, as in mainland Europe or North America, the errors in vertical total electron content (vTEC) are smaller than 1 TECu (2–4%). In regions with fewer and more sparsely distributed receivers, the errors can be as high as 20–40%. This is caused by poor data coverage and poor spatial resolution of the reconstruction, which has an important effect on the calibration process of the algorithm.