Modelling global vertical total electron content with neural networks
Abstract
This study addresses a two-step neural-network model of the vertical total electron content (VTEC), consisting of a temporal and a spatial component. The model is parametrized with geomagnetic and solar wind indices and their time histories combined with geomagnetic and geographic coordinates. The neural network parameters are tuned using five-fold cross validation and the features are chosen using Pearson correlation coefficient, permutation feature importance and mutual information. The performance of the neural networks is tested in extended periods covering a wide range of solar and geomagnetic activity conditions. In addition to increasing the computational efficiency, the proposed approach allows to get physical insights into the dynamics of the ionosphere.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2021
- Bibcode:
- 2021AGUFMNG45B0560K