Variations in the solar, magnetospheric, and ionospheric environment can affect a variety of ground-based and space-borne technological systems (e.g. high frequency (HF) communications, the Global Positioning System (GPS), ultra high frequency (UHF) satellite communications, as well as spacecrafts, pipelines and cable networks, etc.) These may have important socio-economical impacts, and hence the ability to forecast disturbances in the solar-terrestrial environment is valuable.

Traditional linear modeling techniques often fail to improve on the reference recurrence and persistence models due to the incomplete understanding of the solar-magnetospheric-ionospheric interactions and the typically noisy and non-stationary nature of solar-geophysical data sets. The use of knowledge-independent (time series) analysis techniques is therefore an attractive alternative approach.

The Ionospheric Forecasting Demonstrator (IFD) developed by QinetiQ utilises nonlinear modeling techniques to make real-time forecasts of ionospheric parameters (e.g. foF2).

Click on images to enlarge

Measured and predicted hourly foF2 time series during a storm event in February 2000.

Ionospheric forecasting

Solar-geophysical data sets are typically noisy and can contain many data dropouts. This poses a significant difficulty for prospective nonlinear forecasting schemes, which generally require continuous data. To rectify this problem, missing data points are interpolated with 24-hour recurrence values. The normalized root-mean-squared error of 0.18 to 0.25 in the operational nonlinear model within the IFD corresponds to improvements over the reference persistence model of approximately 50 to 30%.

Nonlinear Modeling

The nonlinear predictions within the IFD are based on Radial Basis Function Neural Network (RBF-NN) models. These have the advantage of being able to find a globally optimum solution to a time series prediction problem in a single pass training process. In addition, the use of Principle Component Analysis (PCA) removes the noise subspace and reduces the effects of over-fitting.

Online real-time forecasting

The IFD is a schedule-based event-driven modular system that can be run in either automatic or interactive mode. It controls data download and processing (including optical character recognition) together with the prediction algorithms and the graphical output of results. Hourly forecasts can be posted via WWW, fax or email. The schedules are fully customizable and the model framework allows the incorporation of additional models, data sources, algorithms and output formats.

An example of six-hourly forecasts for foF2 in the European region, using input time series from the Chilton, UK ionosonde station, can be seen here.

Potential applications

• Frequency management on communications links
• Improvement of propagation prediction codes
• Generation of contour maps for ionospheric parameters
• Ionospheric corrections to satellite navigation systems
• Risk assessment for spacecraft operation
• Prediction of power line and pipeline charging

Disclaimer
The ionospheric data may not be used for commercial gain and QinetiQ accepts no responsibility for their use.