Addressing a grand challenge with a killer App: the progress of AIDA in detecting extreme events in observations and simulations

Giovanni Lapenta and Aida Consortium

Abstract

Space is characterised by some extreme events. We refer here not only to large scale extreme events such as a Carrigton-type CME but also to more common extreme events where local conditions push the physics of our models to their extreme: shocks where the continuum equations fails or reconnection sites where ideal MHD break down. Those are sources of pathological behaviour of continuum equations and become the source of great energy releases, the transformation of magnetic energy into kinetic energy with the formation of a power law tail of intense electric fields and high energy particles.
Detecting these regions is not easy. It is not easy from in-situ observations where the information on large scale system scales and the local measurements are almost never concurrently available.
But it is also di!cult in simulations where the information is fully accessible but overwhelming in size (e.g. one of our particle in cell simulations used for space studies produces about 1TB of data for each time step, and with 20000 cycles it leads to 20 PB, a staggering amount even for modern standard).
AIDA took a new approach to addressing this challenge. It combined the use of di"erent data feeds: 1) time series of the local moments and fields, 2) local point-wise particle velocity and energy distributions, 3) 2D or 3D datasets. For the first two sources the data has been obtained from in situ observations (e.g. MMS, SolO and PSP) or synthetic virtual probes in simulations. For
the third, remote observations (e.g. SDO) or simulations were used.
The novelty is creating the python based AidaPy that gives us the tools to obtain, integrate and analyse all these data sources within a single python-based package. The package is open source and freely available. We invite teh community to use and extend it.
The example provided here will be that of the detection of reconnection and will show the work of AIDA using physics-based, unsupervised and supervised machine learning tools.
[1] Lapenta, G. (2021). Detecting reconnection sites using the Lorentz Transformations for electromagnetic fields. The Astrophysical Journal, 911(2), 147.
[2] Sisti, M., Finelli, F., Pedrazzi, G., Faganello, M., Califano, F., & Ponti, F. D. (2021). Detecting Reconnection Events in Kinetic Vlasov Hybrid Simulations Using Clustering Techniques. The Astrophysical Journal, 908(1), 107.
[3] Hu, A., Sisti, M., Finelli, F., Califano, F., Dargent, J., Faganello, M., ... & Teunissen, J. (2020). Identifying Magnetic Reconnection in 2D Hybrid Vlasov Maxwell Simulations with Convolutional Neural Networks. The Astrophysical Journal, 900(1), 86.

Classification of Solar Energetic Activity using Data Analysis and Clustering of Active Regions

Hanne Baeke, Jorge Amaya, Sara Jamal and Giovanni Lapenta