WSA+ Model: Training and Plotting Pipeline

This repository contains scripts for training, evaluating, and visualizing the WSA+ model — a neural enhancement of the traditional Wang–Sheeley–Arge (WSA) solar wind relation.
The workflow is divided into two main stages:

1. Optimization — per Carrington Rotation (CR) fitting of WSA empirical parameters using in-situ solar wind observations.
2. Generalization — training a neural network to predict optimized WSA speed maps directly from magnetogram-derived features.

---

Data & Model Checkpoint (Zenodo)

The training dataset and the WSA+ model checkpoint (wsaplus.pt) are hosted on Zenodo:

DOI: https://doi.org/10.5281/zenodo.16883042

---

End-to-End Workflow

flowchart TD
  A["Input Synoptic Magnetograms"] --> B["Optimization: fit_wsa_params.py"]
  B --> C["Parameter Plots: plot_fitted_params.py"]
  C --> D["Generalization Training: train.py"]
  D --> E["Loss Curves: loss_plot.py"]
  D --> F["CR-wise Visuals: post_training_CR_plots.py"]
  D --> G["All CRs Metrics: post_training_CR_metrices.py"]
  D --> H["2D Speed Maps: 3_2D_map_plot.py"]
  D --> I["In-situ Panels: in-situ_maps.py"]
  D --> J["Dataset Comparison: dataset_comparision.py"]

Loading