Skip to content

README.md

Latest commit

 

History

History
92 lines (74 loc) · 2.63 KB

README.md

File metadata and controls

92 lines (74 loc) · 2.63 KB
Raw

Solar Wind Forecasting Dataset Generation

Description

This repository provides scripts to download data and create sampling splits for generating the Solar Wind Forecasting Dataset.
The dataset is derived from NASA OMNI solar wind and IMF data and curated by removing Interplanetary Coronal Mass Ejections (ICMEs) to retain only background solar wind.

  • Source Data: NASA OMNI dataset (CDAWeb)
  • Coverage: 2010-05-13 to 2024-12-31
  • Cadence: Hourly measurements at L1
  • Variables: Solar wind speed (V), IMF components (Bx, By, Bz), proton density (N)
  • Purpose: Training and benchmarking solar wind forecasting models

Repository Structure

solarwind/
│
├── data_process/
│ ├── download_sw_data.py # Download OMNI solar wind data
│ ├── remove_icme_omni.py # Remove ICME intervals from data
│ ├── split_omni_icme.py # Split data into train/val/test sets
│ ├── plot_omni_csvs.py # Plot dataset for inspection
│ └── download_solar_wind_data.sh # Automated setup + data processing script
│
├── csv_files/ # Processed dataset (CSV format, train/val/test splits)
│
├── plots/ # Visualizations of data (time-series plots)
│
├── models/ # Example model implementations (baseline forecasting)
│
└── README.md # Project documentation

Features

  • Hourly-resolution solar wind and IMF dataset with ICME events removed
  • Predefined splits: train, validation, leaky validation, test
  • Variables included:
    • Solar wind speed (V, km/s)
    • IMF components: Bx (GSE), By (GSM), Bz (GSM) (nT)
    • Proton number density (N, cm⁻³)
  • Dataset size: 119,478 samples across all splits
  • CSV format, compatible with pandas, NumPy, and ML frameworks

Usage

1. Download dataset from HuggingFace

huggingface-cli download nasa-impact/Surya-bench-solarwind --repo-type dataset --local-dir csv_files

1.a Process data locally (alternative)

cd data_process
bash download_solar_wind_data.sh

2 Load dataset in Python

import pandas as pd

# Load training set
df_train = pd.read_csv("csv_files/train.csv")
print(df_train.head())

# Example: plot solar wind speed
import matplotlib.pyplot as plt
plt.plot(pd.to_datetime(df_train["Epoch"]), df_train["V"])
plt.title("Solar Wind Speed (Train Set)")
plt.xlabel("Time")
plt.ylabel("Speed (km/s)")
plt.show()

Requirements

  • pandas
  • numpy
  • matplotlib
  • requests

Install dependencies:

pip install pandas numpy matplotlib requests

Contact

Vishal Upendran, [vupendran@seti.org]