Rainshift

Important note: The input precipitation (which is in meter) is in different units than the output precipitation (which is in millimeter), so they differ by a factor of 1000.

More infos can be found in our preprint: https://arxiv.org/pdf/2507.04930?

Dataset description

The RainShift dataset builds on three different data products: ERA5 reanalysis, IMERG satellite observations, and static features like land-sea mask and orography.

Input Data

ERA5

ERA5 is the fifth-generation atmospheric reanalysis product of the European Center for Medium-Range Weather Forecasts (ECMWF). Reanalysis data are the result of combining historical observations with Earth system models to obtain global estimates of the observed climate. It provides hourly global data at 0.25degx0.25deg resolution (approximately 25km per pixel in mid-latitudes) on a regular latitude-longitude grid and spans years from 1950 to the present. For compatibility with IMERG, we use data from 2001 onward.

We select nine input variables, including among others different precipitation types, atmospherics water content, and winds, following \citet{Harris2022AGD} and domain-specific knowledge, motivated by the ecPoint model.

• Total precipitation (tp)
• Convective precipitation (cp)
• Convective potential energy (cape)
• Total water content (twc)
• Total liquid water content (tlwc)
• Surface pressure (sp)
• Top-of-the-atmosphere incident solar radiation (tisr)
• eastward component of horizontal wind velocities at 700 hPa (u)
• northward component of horizontal wind velocities at 700 hPa (v)

Geographic Features

To complement ERA5 inputs, we include high-resolution static geographic features at 0.1deg resolution: (i) A land-sea mask indicating the land fraction per pixel, and (ii) an elevation map (geopotential height at surface).

Target Data - IMERG

The Integrated Multi-satellite Retrievals for GPM (IMERG) is a product of NASA’s Global Precipitation Measurement (GPM) mission. IMERG provides precipitation estimates based on the GPM satellite constellation and additional observations like gauge data. IMERG has full coverage between 60deg N and 60deg S at 0.1deg resolution (about 10km per pixel), like ERA5 on a regular latitude-longitude grid. We use the IMERG V07 Final Run product , averaging its half-hourly data to hourly to match ERA5’s temporal resolution. Data was accessed via NASA Goddard’s GES DISC.

Leveraging global satellite data as a target provides ground truth in the Global South, enabling model evaluation in regions with limited observational data. By using a globally consistent product, RainShift isolates the challenge of regional generalization from potential discrepancies between data products, offering a clean and controlled test case.

Dataset instructions

Below you will find instructions for the rainshift_dataset_utils.py script.

It has been build to help setup the dataset.

Requirements

In the environment of your choice, install the script requirements using the following command:

pip install -r requirements.txt

Script use

The script has been build using typer.

It contains 3 commands:

• download: To download the dataset to a folder of your choice
• validate: To validate the zip files after they've been downloaded
• unzip: To unzip the files in a way to preserve dataset folder structure

Use the following command to interact with the script:

# General script help
python3 rainshift_dataset_utils.py --help

# download command help
python3 rainshift_dataset_utils.py download --help

# validate command help
python3 rainshift_dataset_utils.py validate --help

# unzip command help
python3 rainshift_dataset_utils.py unzip --help

The use of the unzip command is necessary, as the zarr arrays are zipped for storage on HuggingFace.

Download

Before downloading, be sure to create the directory where you want to download the dataset.

The download command is just a utility wrapper around the HuggingFace python library.

You can also just use this library directly, like so, to download the full dataset :

from huggingface_hub import HfApi

api = HfApi()
api.snapshot_download(repo_id="RainShift/rainshift", repo_type="dataset", local_dir="<PATH_TO_DIRECTORY>")