A benchmark dataset for data-driven medium-range weather forecasting, a topic of high scientific interest for atmospheric and computer scientists alike.

WeatherBench 2 is an update to the global, medium-range (1–14 day) weather forecasting benchmark proposed by rasp_weatherbench_2020, designed with the aim to accelerate progress in data-driven weather modeling. WeatherBench 2 consists of an open-source evaluation framework, publicly available training, ground truth and baseline data as well as a continuously updated website with the latest metrics and state-of-the-art models.

🇩🇪 독일

🇩🇪 독일

This repository contains the ensemble forecasts for 24h precipitation for the manuscript "The extension of the WeatherBench 2 to binary hydroclimatic forecasts" that has been submitted to the Geoscientific Model Development. The data is derived from the forecasts and ground truth data provided in Google's WeatherBench 2 (https://console.cloud.google.com/storage/browser/weatherbench2). For the complete data processing procedures, please refer to Zenodo (https://doi.org/10.5281/zenodo.14691007).

This is an example dataset accompanying the github repository https://github.com/LSX-UniWue/dl-climate-models and contains athmospheric data for the year 2009 which is mostly a subset from WeatherBench at 5.625° horizontal resolution. This dataset is intended to help getting started with our codebase and to perform inference with our provided pre-trained models without requiring to download the complete dataset.

The original and complete data source can be downloaded from

WeatherBench
Stephan Rasp, Peter D. Dueben, Sebastian Scher, Jonathan A. Weyn, Soukayna Mouatadid, and Nils Thuerey, 2020. WeatherBench: A benchmark dataset for data-driven weather forecasting. arXiv: https://arxiv.org/abs/2002.00469

and

solar forcing from the input4MIPs from the CMIP6
Matthes, Katja; Funke, Bernd; Kruschke, Tim; Wahl, Sebastian (2017). input4MIPs.SOLARIS-HEPPA.solar.CMIP.SOLARIS-HEPPA-3-2. Version YYYYMMDD^[1].Earth System Grid Federation

Description: The WEA dataset is derived from the WeatherBench repository and designed for medium-range weather forecasting at five geographically diverse cities: London (UK), New York (US), Hong Kong (China), Cape Town (South Africa), and Singapore. It spans the period from 1979 to 2018, with a temporal resolution of 6 hours and a spatial resolution of 5.625° in both latitude and longitude. Each city is matched to its nearest grid point on the WeatherBench grid using minimal absolute distance in both axes.

Task: Time series forecasting of the 850 hPa temperature (T850) — a widely used mid-tropospheric climate indicator — at each location.

Features:

Target Variable: T850 (850 hPa temperature in Kelvin) at the central grid point.

Exogenous Variables (Total: 44):

Local: Z500 (500 hPa geopotential), t2m (2m temperature), u10 (10m zonal wind), v10 (10m meridional wind).

Spatial context: The same 5 variables (T850 + 4 exogenous) from the surrounding 8 grid points (3×3 window).

Temporal Coverage:

Training Set: From Jan 1, 1980 to the end of the year preceding validation (e.g., up to Dec 31, 2014 for the 2015 validation year).

Validation Set: One year preceding each test year (2015, 2016, 2017).

Test Set: Years 2016, 2017, 2018 — each containing ~1,460 time steps.

Use Cases: Benchmarking time series models for weather forecasting in diverse climatic conditions, studying the impact of spatial and exogenous inputs on model performance, and evaluating generalisation across different latitudes and climate zones.

Source: Based on climate variables from the WeatherBench repository.

ClimateBench is a benchmark dataset for climate model emulation inspired by WeatherBench. It consists of NorESM2 simulation outputs with asociated forcing data processed in to a consistent format from a variety of experiments performed for CMIP6. Multiple ensemble members are included where available.