Prediction Interval of Interface Regions: Machine Learning Nowcasting Approach

Stream interaction region (SIR) is one of the space weather phenomena that accelerates upstream particles interface in interplanetary and causes geomagnetic storms. SIRs are large-scale structures vary temporally spatially, both latitudinal radial directions. Predicting arrival times regions (IRs) crucial to protect our navigation communication systems. In this work, a 1D ensemble system comprised Long-short-term memory (LSTM) model Convolution Neural Network (CNN) model—LCNN introduced classify observed IR time series give prediction interval nowcast its transit observer. The outcomes two models combined way boost accuracy predictor prevent error propagation between them. implemented technique classification on datasets from STEREO A B spacecrafts. LCNN IRs provides advanced Notice Time (NT) [20, 160] minutes with sensitivity around 93% geometric mean score gmean 91.7%, skills decrease increasing time. demonstrates an enhancement respect using only either CNN or LSTM models. predicted probabilities recalibrated so frequency becomes average consistent frequency. Application method useful provide by inputting estimating likelihood occurrence