Getting Started

This page describes the prerequisites, installation steps, and initial setup required to run CPFA on a Windows environment.

Prerequisites

To use CPFA, you need:

Windows operating system
Anaconda (Python distribution)
Python 3.9.2
An active internet connection to download ERA5 data

Download the Repository

Open your web browser and navigate to the CPFA GitHub repository.
Download the project folder (for example, as a ZIP file).
Extract or move the downloaded folder to the C: drive.

After extraction, the folder will be similar to:

C:\CPFA\

Install Anaconda

Visit the Anaconda website.
Download the installer for Windows.
Install Anaconda (installation in the C: drive is recommended).

Install Python 3.9.2

Although Anaconda provides its own Python distribution, CPFA is documented and tested with Python 3.9.2.

Visit the official Python website.
Download the installer for Python 3.9.2.
Complete the installation.

Creating a Conda Environment

Open Anaconda Prompt as Administrator.
Move to the C: drive by entering cd.. repeatedly until the prompt shows:
```
C:\>
```
Navigate to the CPFA project folder, for example:
```
cd CPFA
```
Create a new virtual environment with Python 3.9.2:
```
conda create -n cpfa_env python=3.9.2
```
Activate the environment:
```
conda activate cpfa_env
```

Installing Required Libraries

With the environment activated, install the required packages:

pip install numpy pandas matplotlib xarray cartopy
pip install onnx==1.13.1
pip install onnxruntime==1.14.0

After installation, the environment is ready to run CPFA.

Download model

Please download the four pre-trained models (~1.1GB each) from Google drive or Baidu netdisk:

The 1-hour model (pangu_weather_1.onnx): ‘https://drive.google.com/file/d/1fg5jkiN_5dHzKb-5H9Aw4MOmfILmeY-S/view’
The 3-hour model (pangu_weather_3.onnx): ‘https://drive.google.com/file/d/1EdoLlAXqE9iZLt9Ej9i-JW9LTJ9Jtewt/view’
The 6-hour model (pangu_weather_6.onnx): ‘https://drive.google.com/file/d/1a4XTktkZa5GCtjQxDJb_fNaqTAUiEJu4/view’
The 24-hour model (pangu_weather_24.onnx): ‘https://drive.google.com/file/d/1lweQlxcn9fG0zKNW8ne1Khr9ehRTI6HP/view’

Folder Structure

The project files should be organized as follows:

root
├── download_data
│   └── ...
├── input_data
│   └── ...
├── output_data
│   └── ...
├── pangu_weather_1.onnx
├── pangu_weather_3.onnx
├── pangu_weather_6.onnx
├── pangu_weather_24.onnx
├── prediction.py
├── transform_nc_to_npy.py
├── visualization.py
└── evaluation.py

Make sure the ONNX model files and scripts are placed at the project root as shown above.