Advanced detection methods and machine learning analysis of temporal and spatial patterns of equatorial plasma bubble depth

Abstract

Strong plasma density depletions, particularly across broader longitudinal areas with consecutive depletion trains, are challenging to analyse using simple polynomial fitting or moving average techniques for detrending. This study presents a double threshold approach, combining the “Bubble Index” and the “Rolling Barrel Technique,” to detect the depth of plasma irregularities. The proposed method is promising in EPB characterization and depth estimation. We validated the result from our method by analysing the temporal, seasonal, and longitudinal characteristics of Equatorial Plasma Bubbles (EPBs) using data for solar cycle 24 from the Communication/Navigation Outage Forecasting System (C/NOFS), covering about 8 years (August 2008–November 2015). Our Machine Learning (ML) prediction results show that XGBoost outperforms other approaches (Random Forest and LSTM), achieving skill, association, and root mean square error scores of 0.78, 0.88, and 0.14, respectively. Climatological verifications of our analysis show that most EPB events are concentrated between −70° and −30° longitude in the South Atlantic Anomaly (SAA). The depth magnitude of EPBs is inversely proportional to altitude, with shallower depths observed during solar minimum and deeper depths during periods of moderate solar activity. Postsunset bubbles are quite deeper, especially between 19:00 LT and 21:00 LT, having significantly deeper magnitude during equinoxes and shallowest in summer, particularly over South America. As the satellite's apex lowered, the depths of postsunset and postmidnight EPB became nearly equal, highlighting the importance of observation altitude. Analysing EPB depth variation with location and time will enhance understanding of EPB dynamics by identifying how seasonal and solar activity influences EPB formation patterns. These findings will further aid in developing improved ionospheric models for space weather forecasting.