Predicting Swarm Equatorial Plasma Bubbles Via Supervised Machine Learning

Equatorial Plasma Bubbles (EPBs) are plumes of low density plasma that rise up from the bottomside of the F layer towards the exosphere. EPBs are known causes of radio wave scintillations which can degrade communications with spacecraft. We build a random forest regressor to predict and forecast the probability of an EPB [0-1] detected by the IBI processor on-board the SWARM spacecraft. We use 8-years of Swarm data from 2014 to 2021 and transform the data from a time series into a 5 dimensional space consisting of latitude, longitude, mlt, year, and day-of-the-year. We also add Kp, F10.7cm and solar wind speed. The observations of EPBs with respect to geolocation, local time, season and solar activity mostly agrees with existing work, whilst the link geomagnetic activity is less clear. The prediction has an accuracy of 88% and performs well across the EPB specific spatiotemporal scales. This proves that the XGBoost method is able to successfully capture the climatological and daily variability of SWARM EPBs. Capturing the daily variance has long evaded researchers because of local and stochastic features within the ionosphere. We take advantage of Shapley Values to explain the model and to gain insight into the physics of EPBs. We find that as the solar wind speed increases the probability of an EPB decreases. We also identify a spike in EPB probability around the Earth-Sun perihelion. Both of these insights were derived directly from the XGBoost and Shapley technique.

A Deep Neural Network for Pixel-Level Electromagnetic Particle Identification in the MicroBooNE Liquid Argon Time Projection Chamber

We have developed a convolutional neural network (CNN) that can make a pixel-level prediction of objects in image data recorded by a liquid argon time projection chamber (LArTPC) for the first time. We describe the network design, training techniques, and software tools developed to train this network. The goal of this work is to develop a complete deep neural network based data reconstruction chain for the MicroBooNE detector. We show the first demonstration of a network's validity on real LArTPC data using MicroBooNE collection plane images. The demonstration is performed for stopping muon and a $\nu_\mu$ charged current neutral pion data samples.