Fördergeber: ACRP-Projekt, Klima- und Energiefonds
Titel: High-resolution machine learning for the climate community in Austria (HighResLearn)
Stichwörter: Klimawandel, km-skalige globale Klimasimulationen, Maschinelles Lernen, Data Science, Kommunikation
Leitung: Universität Wien
Kontakt: Maximilian Meindl (maximilian.meindl@univie.ac.at)
Projektpartner: Universität Innsbruck, GeoSphere Austria, CCCA, Klimadashboard
Projektdauer: 01.07.2024-30.06.2027
Das HighResLearn Projekt entwickelt Methoden um die neuesten, hochaufgelösten globalen Klimasimulationen mit km-skaliger Auflösung bestmöglich nutzbar zu machen.
Die Erreichung des Projektziels gliedert sich in zwei Hauptteile:
Zusätzlich dazu wird HighResLearn eng mit der nationalen und internationalen Forscher:innengemeinschaft zusammenarbeiten, um Brücken zwischen der globalen und regionalen Klimamodellierung zu bauen. Ein besonderer Fokus liegt auch auf der Aufbereitung der Ergebnisse für eine breite Öffentlichkeit, zum Beispiel in Zusammenarbeit mit dem Klimadashboard Österreich.
Abstract
High-resolution climate data is crucial for studying regional climate impacts and extremes, especially in topographically complex regions [1]. However, users often face barriers when trying to access and process datasets from multiple sources due to differences in data structure, resolution, grid structure, and naming conventions. ClimXtract is a modular Python toolkit developed to address this challenge. It provides standardized functions for downloading, regridding, and spatially masking multiple climate datasets into a common format compatible with any high-resolution climate dataset for a given regional domain. ClimXtract includes support for variable harmonization (e.g., for temperature and precipitation), interpolation for different grid types, and optional masking to a target domain. It builds upon the libraries xarray [2] and CDO [3], which are widely used in the climate data community, and is designed for domain scientists and non-specialists alike. Together with processed example datasets and Jupyter notebooks, ClimXtract provides the climate community with a reproducible workflow for preparing data for research and downstream applications. While here presented using the ÖKS15 dataset for Austria [4] as an example, ClimXtract can equally be applied to other regions of interest and target formats more generally.
Published in Journal of Open Reseach Software
ClimXtract: A Python Toolkit for Standardizing High-Resolution Climate Datasets on Regional Domains
ClimXtract provides a modular pipeline for preparing high-resolution climate datasets for regional analysis. Its three capabilities are downloading, regridding, and masking, all designed to support interoperability and reproducability. Although the toolkit is designed with Austria and ÖKS15 in mind, all components can be configured to work with any user-defined target grid.
ClimXtract was developed as part of the Austrian Climate Research Programme (ACRP) project HighResLearn. One goal of HighResLearn is to enable the Austrian climate community to effictiently access and process high-resolution global climate model data in conjunction with national-scale reference datasets like ÖKS15. As such, the ClimXtract toolkit lays the foundation for downstream applications, including machine learrning based analysis of climate model performance on regional scales.
Available on https://github.com
