Publication: Interpolation and extrapolation techniques based Neural Network in estimating the missing ionospheric TEC data
No Thumbnail Available
Date
2016
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Institute of Electrical and Electronics Engineers Inc.
Abstract
This paper investigates the capabilities of Neural Network (NN) application in estimating the missing ionospheric total electron content (TEC) data via interpolation and extrapolation methods. TEC is an important parameter that describes the state of the ionosphere. A multilayer feed-forward network with a back propagation algorithm is applied to estimate the TEC over Parit Raja (Lat. 1�52?N, Long. 103�06?E) an equatorial latitude station in Malaysia. The solar and magnetic indices, seasonal variation as well as diurnal variation are used as the input spaces in the NN to estimate the missing GPS TEC. The studies period is based on short term data during the medium solar activity period from 2005 to 2006. Normalized RMSE, RMSE and relative correction are computed for both methods to evaluate the capability of NN to interpolate and extrapolate the missing data. The results show that NN is able to interpolate the missing TEC data within the input range better than extrapolating the missing data outside the input range. The NN2 model finds hard to extrapolate the missing TEC data especially during the night hours, where the GPS TEC values are underestimated. Overall the relative correction of NN1 model is above 90% while for NN2 is below 90% for all ranges of missing rate. � 2016 IEEE.
Description
Keywords
Backpropagation, Backpropagation algorithms, Interpolation, Ionosphere, Solar energy, Diurnal variation, Extrapolation methods, Extrapolation techniques, Ionospheric total electron content, Magnetic indices, Multi-layer feed-forward networks, Neural network (nn), Seasonal variation, Extrapolation