Forecasting of Ionospheric real GPS TEC and SAMI3 model output parameters using the LSTM deep recurrent neural network

Abstract:

In this meeting, we present data-driven forecasting of ionospheric Total Electron Content (TEC) using the Long-Short Term Memory (LSTM) deep recurrent neural network method. The random forest machine learning method was used to perform a regression analysis and estimate the variable importance of the input parameters. The top five parameters were then selected and used to forecast real GPS TEC up to 5 hours ahead, with 30 minute cadence. The results indicate that very good forecasts with low RMS error (high correlation) can be made in the near future and RMS error increase as we forecast further into the future. Most of the training data input to the random forest method is from NASA’s Space Physics Data Facility (SPDF) OMNI (Operating Missions as a Node on the Internet) characterizing the solar-terrestrial environment. We will also present results of forecasting SAMI3 synthetic TEC, the height of peak electron density at F2-layer (HMF2) the peak electron density at the F2-layer (NMF2) using the LSTM recurrent neural network. In the later case, forecasts have been made 72 hours ahead every hour using SAMI3 synthetic data and other solar and geomagnetic indices as input into the LSTM method.