Artificial Neural Networks Ionospheric Model (ANNIM)

June 18, 2020 By Lt. General P.C. Katoch (Retd) Photo(s): By PIB
The Author is Former Director General of Information Systems and A Special Forces Veteran, Indian Army


Figure 1: Architecture of the neural network used in ANNIM development (Gowtam et al., 2019)

Tracking the variability of the Ionosphere is important for communication and navigation. The ionospheric variability is greatly influenced by solar originated processes and the neutral atmosphere origin. The ionosphere extending from 60 km to about 1000 km altitude, plays a major role in aiding long distance HF (3 to 30 MHz) communications as well as in the deteriorating performance of satellite radio systems in the VHF (30 to 300 MHz), UHF (300 to 3000 MHz) and even higher frequency bands. The ionosphere influences radio propagation in which the radio refractive index of ionosphere, interaction of radio waves with ionosphere and reflection of radio waves come into play.

The long, relatively unhindered travel of the GPS signal through the virtual vacuum of space too changes as it passes through the earth's atmosphere. Through both refraction and diffraction, the atmosphere alters the apparent speed and, to a lesser extent, the direction of the signal. The ionosphere has layers sometimes known as the mesosphere and thermosphere but neither of that are themselves composed of D, E, and F regions extending 50-90 km, 90-120 km and 120-1000 km respectively. Neither the boundaries between these regions, nor the upper layer of the ionosphere, can be defined strictly. D region has almost no effect on GPS signals and E region has slight but it can scintillate the GPS signal. However F region affects propagation of electromagnetic signals the most. The ionosphere is also not homogeneous. Its behavior in one region of the earth is liable to be unlike its behavior in another. For example, ionospheric disturbances can be particularly harsh in the Polar Regions. But the highest Total Electron Value (TEC) values and the widest variations in the horizontal gradients occur in the band of about 60° of geomagnetic latitude. That band lies 30° north and 30° south of the earth’s magnetic equator. Severity of the ionosphere’s effect on a GPS signal depends on the amount of time that signal spends traveling through it. A signal originating from a satellite near the observer’s horizon must pass through a larger amount of the ionosphere to reach the receiver than does a signal from a satellite near the observer’s zenith. In other words, the longer the signal is in the ionosphere, the greater the ionosphere’s effect on it.

Figure 2: The ionospheric electron density as a function of altitude over Indian longitude during March equinox at 12:00 Hrs. as simulated by ANNIM.

The good news is that now researchers from Indian Institute of Geomagnetism (IIG), Navi Mumbai have developed a new Artificial Neural Network based long-term Ionospheric Model (ANNIM) using long-term ionospheric observations to predict the ionospheric electron density and the peak parameters. Artificial Neural Networks (ANNs) replicate the human brain or biological neurons to solve problems like pattern recognition, classification, clustering, generalisation, linear and non-linear data fitting, and the like. The researchers developed the ANNIM by using an extensive database consisting of nearly two decades of global Digisonde (an instrument that measures real-time on-site electron density of the ionosphere by sending the radio frequency pulses), Global Navigation Satellite System (GNSS) radio occultation and topside sounders observations. These datasets were processed with various quality control measures to eliminate spurious data points (outliers) and prepared for the training. Day number, Universal Time, latitude, longitude, F10.7 index (responsible for Photo-ionization), Kp (represents the disturbed space weather conditions), magnetic declination, inclination, dip latitude, zonal and meridional neutral winds were taken as inputs in the study. The target (output) of ANNs is the electron density as a function of altitude for any given location and time. The data was trained with the ANNs using high-performance computer at IIG to develop the ANNIM. The ANNIM predictions done by the IIG team matched with the incoherent scatter radar and the satellite in situ electron density observations. Further, the ANNIM successfully reproduced large scale anomalies of the ionosphere. The ANNIM also captured the general morphological features of the ionosphere during the disturbed space weather periods, such as geomagnetic storms which occurs when the magnetic cloud originated from Sun (known as Coronal Mass Ejection (CME)) interacts with the Earth’s magnetosphere.

The ANNIM developed by IIG researchers can be utilised as a reference model in ionospheric predictions and has potential applications in calculating the GNSS positioning error. It would have a wider impact, ranging from accurate aiming on enemy targets to scientific research in space weather. According to Dr S. Tulasiram from IIG, one of the two primary researchers, “This development is crucial for armed forces as, over the past few years, Indian Armed Forces have, like all militaries, sought pinpoint accuracy to minimise collateral damage in targeting enemy positions. Although several fighter jet and other missile systems have been fitted with highly-accurate navigation systems, few weapons in service can yet take advantage of this model”, adding, “Better ionospheric modeling will yield better accuracy of positioning in satellite based navigation. The current ionospheric model (AMMIM) accuracy is significantly better compared to other ionospheric models used in single frequency GPS users.” ANNIM is another feather in India’s cap.