Validation of Ionospheric Tomography Using Residual Minimization Training Neural Network

Abstract

SUMMARY A numerical simulation has been performed to evaluate the performance of ionospheric tomography using the residual minimization training neural network (RMTNN) method. The results indicate that reconstruction with high precision is possible when the standard deviation of the noise is about 2.5% or less of the average value of the observed data (slant total electron content [STEC]). Moreover, in the daytime, when the value of STEC becomes large, the signal to noise ratio (SNR) increases and the reconstruction accuracy becomes high; at night when the SNR falls, conversely, it becomes low. The results of detectability tests show that the RMTNN method has a good performance at about F‐layer height with shape and peak intensity reconstruction. In conclusion, the newly developed RMTNN ionospheric tomography is effective in reconstructing the 3D electron density distribution from realistic STEC data in the daytime, and is able to estimate images around the F‐layer.