mindscience/MindEarth

• MindEarth
• • Introduction
  • Latest News
  • Applications
  • • Precipitation Nowcasting
    • Medium-range Forecast
    • Data Preprocessing
  • Installation
  • • Version Dependency
    • Install Dependency
    • Hardware
    • pip install
    • source code install
  • Community
  • • Core Contributor
  • Contribution Guide
  • License

ENGLISH | 简体中文

MindEarth

• Introduction
• Latest News
• Applications
  • Precipitation Nowcasting
  • Medium-range Forecast
  • Data Preprocessing
• Installation
  • Version Dependency
  • Install Dependency
  • Hardware
  • pip install
  • source code install
• Community
  • Core Contributor
• Contribution Guide
• License

Introduction

Weather phenomena are closely related to human production and life, social economy, military activities and other aspects. Accurate weather forecasts can mitigate the impact of disaster weather events, avoid economic losses, and generate continuous fiscal revenue, such as energy, agriculture, transportation and entertainment industries. At present, the weather forecast mainly adopts numerical weather prediction models, which processes the observation data collected by meteorological satellites, observation stations and radars, solves the atmospheric dynamic equations describing weather evolution, and then provides weather and climate prediction information. The prediction process of numerical prediction model involves a lot of computation, which consumes a long time and a large amount of computation resources. Compared with the numerical prediction model, the data-driven deep learning model can effectively reduce the computational cost by several orders of magnitude.

MindEarth is an earth science suite developed based on MindSpore. It supports AI meteorological prediction of short-term, medium-term, and long-term weather and catastrophic weather such as precipitation and typhoon. The aim is to provide efficient and easy-to-use AI meteorological prediction software for industrial scientific research engineers, college teachers and students.

Latest News

• 🔥2023.02.06 MindSpore Helps Ocean Terrain Overscore: Huang Xiaomeng Team of Tsinghua University Releases Global 3 Arc Second (90 m) Sea and Land DEM Data Products.Page.

Applications

Precipitation Nowcasting

Case	Dataset	Network	GPU	NPU
DGMs	Radar dataset	GAN、ConvGRU	✔️	✔️

Medium-range Forecast

Case	Dataset	Network	GPU	NPU
FourCastNet	ERA5 Reanalysis Dataset	AFNO	✔️	✔️
ViT-KNO	ERA5 Reanalysis Dataset	ViT	✔️	✔️
GraphCast	ERA5 Reanalysis Dataset	GNN	✔️	✔️

Data Preprocessing

Case	Dataset	Network	GPU	NPU
DEM Super-resolution	NASADEM、GEBCO_2021	SRGAN	✔️	✔️

Installation

Version Dependency

Because MindEarth is dependent on MindSpore, please click MindSpore Download Page according to the corresponding relationship indicated in the following table. Download and install the corresponding whl package.

MindEarth	Branch	MindSpore	Python
master	master	\	>=3.7
0.1.0	r0.5	>=1.8.1	>=3.7

Install Dependency

pip install -r requirements.txt

Hardware

Hardware	OS	Status
Ascend 910	Ubuntu-x86	✔️
	Ubuntu-aarch64	✔️
	EulerOS-aarch64	✔️
	CentOS-x86	✔️
	CentOS-aarch64	✔️
GPU CUDA 11.1	Ubuntu-x86	✔️

pip install

# gpu and ascend are supported
export DEVICE_NAME=gpu
pip install mindearth_${DEVICE_NAME}

source code install

• Download source code from Gitee.

git clone https://gitee.com/mindspore/mindscience.git
cd {PATH}/mindscience/MindEarth

• Compile in Ascend backend.

bash build.sh -e ascend -j8

• Compile in GPU backend.

export CUDA_PATH={your_cuda_path}
bash build.sh -e gpu -j8

• Install the compiled .whl file.

cd {PATH}/mindscience/MindEarth/output
pip install mindearth_*.whl

Community

Core Contributor

Thanks goes to these wonderful people 🧑‍🤝‍🧑:

yufan, wangzidong, liuhongsheng, zhouhongye, liulei, libokai, chengqiang, dongyonghan, zhouchuansai

Contribution Guide

• Please click here to see how to contribute your code:Contribution Guide

License

Apache License 2.0