Using ionospheric imaging combined with feature tracking to automate identification and tracking of polar-cap plasma patches

"Patches" are regions of enhanced electron concentration that propagate through the polar-cap regions of the ionosphere at F-layer altitudes. They correlate well with the appearance of scintillations on trans-ionospheric radio links such as GPS broadcasts and can act as reflecting layers for HF radio links, allowing communication to regions which would be impossible to access in their absence. The aviation industry is particularly affected by both of these phenomena as it is increasingly reliant on GPS based systems for navigation purposes and on HF radio to communicate with aircraft on trans-polar flight paths. With the number of trans-polar commercial flights increasing annually, polar-cap patches present both an increasing problem (loss of accuracy using GPS) and a benefit (unexpected radio links to aircraft). The approach of a solar maximum, when the rate of occurrence of patches increases, exacerbates both of these [Hunsucker, 2003]. It is therefore of great practical as well as scientific interest to understand the formation, propagation, morphology and decay of polar-cap patches as well as the processes that generate scintillation causing electron concentration irregularities within them. Presented here is a technique for imaging the polar-cap ionosphere in terms of electron concentration and subsequently identifying and tracking polar-cap patches within these images in an automated manner. Ionospheric imaging of the polar cap based on GPS ground receiver data has been successfully demonstrated and used in a number of studies. The MIDAS 2.0 ionospheric imaging software uses a Kalman filter approach to weight GPS data assimilated into a background model against a predictive model of the polar convection partern in order to achieve a final solution for each image. When the images, which are fully three dimensional, are plotted as vertical TEC maps, patches can be readily identified by eye. However, automated identification and tracking of these features would allow analysis of patch climatology, statistical analysis of the tracks of patches as they convect and other studies leading to a better understanding of the phenomenon as a whole [R. Burston et al., 2009; Mitchell et al., 2005; Spencer and Mitchell, 2007; Weimer, 2001 ; Yin et al., 2009]. To identify and track the patches a method originally developed for tracking weather systems in model and Numerical Weather Prediction data has been adapted [K.I. Hodges, 1994; 1995; K. I. Hodges, 1999]. This involves taking the vertical TEC data and first removing the large scale background using a discrete cosine transform filter [Denis et al., 2002], spatial scales in the range 200-1000 km are retained. Representative points (feature points) are determined from the filtered TEC fields; this is done by finding the off-grid maxima to represent the feature points. The feature points are linked to form tracks initially using a nearest neighbour search; these are refined using a cost function minimization for track smoothness to produce the smoothest set of tracks. Once tracks have been determined for an extended time period, statistics for the density of tracks and their genesis and lysis are determined. Both the tracking and statistical estimation are carried out directly on the sphere to exclude the types of biases often introduced by using projections. Other properties along the tracks and intensity distributions are also determined. Results to date are presented, including a feasibility case-study based on the geomagnetic storm that occurred during 24-28th July 2004, using vertical TEC data as generated by the MIDAS 2.0 algorithm.