Containerized Models to Predict Ionospheric Scintillations

Ensemble's innovative proposal targets ionospheric scintillation characterization, monitoring, and prediction. The technology, designed for North America, aims to deliver precise scintillation indices estimates for amplitude and phase. This proposed effort covers four key tasks: - Data Pipeline Creation: Establishing a robust pipeline to collect and process diverse data relevant to scintillation, ensuring a reliable dataset for ML model training. - Data Transformation and Cleaning: Ensuring temporal and spatial alignment of data sources, addressing gaps, and preparing the dataset for ML model training, including solutions for an imbalanced dataset. - ML Model Development and Validation: Using AWS SageMaker, Ensemble will develop two ML models— convolutional neural network (CNN), and Long Short Term Memory (LSTM). - Model Deployment and Alert System: Deploying the models via AWS SageMaker, generating predictions up to 4 hours into the future at 15-minute intervals. Implementing an alert system for stakeholders when scintillation is predicted. This effort will involve data collection, processing, model development, cloud computing resources, and deployment infrastructure. Target markets include aviation, telecommunications, agriculture, construction, and emergency services, benefiting industries reliant on precise GNSS communication and positioning. Ensemble's methodology bridges critical space weather research gaps, contributing to national security objectives outlined by the SWORM subcommittee. Beyond operational reliability, the proposal offers actionable alerts, improved safety, and enhanced understanding of ionospheric scintillation phenomena.