Item

Abstract

Within the International GNSS Service (IGS), Ionosphere-Associated Analysis Centers (IAACs) provide global ionospheric maps (GIMs) obtained from the advanced processing of GNSS observations. These products are combined into one consistent product using a weighted average. Such obtained IGS GIMs are widely used in scientific research and as positioning support. Our research aimed to develop a machine-learning model based on TEC values from individual analysis centers, location, time, and information about solar activity. The result of the model is a combined, consistent, and accurate global GIM. This task was accomplished using different machine learning algorithms to detect the relationship between input data and 5-minute vertical TEC derived from altimetry satellites. Our analyses were based on the dataset covering 2005 to 2020. To validate the quality of our model, we used single-frequency code and phase measurements from more than 200 GNSS reference stations evenly distributed over the Earth's surface. We tested the performance of our model for a period of quiet ionospheric activity and days with strong geomagnetic disturbances. We obtained results that, relative to the IGS product, differ up to a few centimeters during both study periods for the northern and eastern components of the topocentric coordinates. For the altitude, we obtained similar results during the calm period, but during the stormy days, we achieved accuracy improvement for more than half of the stations. In some cases, the improvement was even a few decimeters.