An instance of the analog weather generator unseen-awg with precomputed similarities and a large set of generated weather data

Project: unseen-awg: spatio-temporal weather generation using analogs and unseen data - Unseen-awg is a method for generating long, multivariate, and spatiotemporal weather data representative of present-day climate by resampling historical weather datasets. It assures temporal consistency in the generated weather by ensuring that successive sampled days have consistent large-scale atmospheric fields. Long datasets of artificial weather data, such as those simulated with unseen-awg, allow anticipating unseen weather and help prepare for possible weather-related hazards. Unseen-awg simulations can be used to drive impact models across sectors influenced by weather such as water, agriculture, and forestry. To this end, the weather generator allows simulating weather over all of Europe.

Here, we provide simulations with the weather generator, an instance with pre-computed components, and the data necessary for using the provided simulations, generating new ones, and generating new instances of unseen-awg.

Project members received financial support by the Federal Ministry of Research, Technology and Space of Germany and by Sächsische Staatsministerium für Wissenschaft, Kultur und Tourismus in the programme Center of Excellence for AI-research "Center for Scalable Data Analytics and Artificial Intelligence Dresden/Leipzig", project identification number: ScaDS.AI.
Summary: This experiment contains an instance of the analog weather generator unseen-awg (Version 1.0) and over 10,000 years of artificial time series generated with unseen-awg for European weather under present-day climate conditions. Long simulations of artificial weather data help studying the weather-related risks across many sectors.

The simulations are composed of 500 21-year-long daily time series and stored as look-up tables. They can be expanded into multivariate spatiotemporal data using the provided reforecast dataset of impact-relevant meteorological variables.

The provided unseen-awg generator instance uses default parameter settings. It can be loaded using the corresponding Python class. The instance includes a large dataset of pre-computed similarities. Compared to using unseen-awg without pre-computation, this enables substantially faster simulations.