minseok96/SAR2RDR-Dataset

🌦️ Data-driven Precipitation Nowcasting Using Satellite Imagery

AAAI 2025 (Oral Presentation)

---

🗓️ Update Note (2025-10-10)

All data are currently being uploaded and are expected to be completed by 2025-10-12.

---

📦 Dataset Description

This dataset provides raw radar observations and accompanying processing scripts for reproducible research in precipitation nowcasting using satellite imagery.
It was created as part of the AAAI 2025 oral paper “Data-driven Precipitation Nowcasting Using Satellite Imagery”.

📁 Folder Structure

SAR2RDR-Dataset/
├── RDR/                             # Radar data
│   ├── train/                       # Training radar samples
│   └── test/                        # Test radar samples
│
├── SAT/                             # Satellite data
│   ├── train/
│   │   ├── ir105/                   # IR 10.5 μm imagery
│   │   ├── wv063/                   # Water vapor 6.3 μm imagery
│   │   └── wv073/                   # Water vapor 7.3 μm imagery
│   └── test/
│       ├── ir105/
│       ├── wv063/
│       └── wv073/
│
├── vis_rdr.ipynb                    # Visualization notebook for radar maps
└── processing_rdr.py                # Script for cropping and resizing radar maps

---

⚙️ Usage Notes

We uploaded the raw radar files and processing code for convenience.

• If you want to reproduce our results, please use the provided processing scripts.
• If you prefer to use your own pipeline, you will need to handle data preprocessing independently.

In this study, we cropped specific regions to remove areas that were not observed by the radar.
If you are able to overcome this limitation, you may freely extend or adjust the spatial coverage as needed.
The image below illustrates the cropped region used in our experiments.

📁 Repository Info

• License: Apache 2.0
• Tags: climate, weather, radar, nowcasting, AI4Climate
• Size Category: 100B < n < 1T
• Repo: minseok96/SAR2RDR-Dataset

---

📊 Contents