Determining a Local Quiet Day with a Machine Learning Model Approach

Abstract

Quiet Days (QDs) play a crucial role in modeling diurnal variations and removing their contribution to compute regional geomagnetic indices. This project aims to develop a local QD identification process by adapting van de Kamp's criteria for selecting Local Quiet Days (LQDs). However, empirical testing has revealed that days classified as LQDs are not always truly quiet, as they may lack the magnetic signatures associated with diurnal variation. To address this limitation, we propose an automatic, unsupervised machine learning method that combines Long Short-Term Memory (LSTM) networks with Convolutional Neural Networks (CNNs). This approach seeks to enhance the detection of truly LQDs within specified time windows, ensuring more accurate identification and analysis.