Heliophysics

These examples apply statistical and machine-learning methods to space-physics problems. Gaps, changing cadence, temporal dependence, rare events, physical baselines, and data provenance are treated as part of the modeling problem.

Examples

• Dst Forecasting builds a one-hour-ahead forecast from hourly OMNI data using time-ordered splits, gap-safe windows, training-only scaling, and a true persistence baseline.

• Plasma-Sheet Modeling compares saved chronological temperature predictions and conditional density maps without presenting model training.

• SEP Occurrence Forecasting examines class imbalance, sample-level validation, and interpretation while making the archive’s provenance limits explicit.

• Coronal-Loop Reconstruction uses corrected, non-overlapping partitions to compare profile reconstruction models.

Each example states what its evidence can support and where the available data or validation design limits the conclusion.