Starting point / motivation
The exploratory project AIMetSat (Introduction of Artificial Intelligence in Satellite-Related Meteorological Forecasting Procedures) is processed through a collaboration between three young, agile and innovative companies, each of them providing expert knowledge in a particular field of science.
MetGIS GmbH is specialized in ultra high resolution meteorological forecasts, Xephor GmbH (like MetGIS from Austria) is one of the world’s leading companies in the field of AGI (Artificial General Intelligence), while the Swiss startup ExoLabs GmbH focuses on the sophisticated processing of EO data.
The principal motivation for this project is to explore the huge potential that the use of AI in combination with EO data provides to improve the quality of meteorological forecasts and the simulation of the snow cover.
Contents and goals
Out of the problems that can be solved with this approach, two are focused upon: One is the general deficiency of numerical weather forecast models to correctly predict the weather conditions in intra-alpine valleys and basins (especially concerning the temperature distribution and the formation of fog and low clouds), and the other is the correct simulation of the snow cover.
To tackle these problems, Exolabs develops novel approaches to use a combination of Sentinel-2, Sentinel-3 and MODIS EO data to distinguish between fog or low stratus and higher clouds and prepares high resolution time series of EO data for the training of the AI of Xephor. The latter is also fed by various datasets composed by MetGIS (observation data from meteorological stations and snow measurement sites; time series of the output of numerical weather prediction models).
MetGIS also prepares AI input datasets derived from high resolution digital elevation models, such as terrain slope, aspect and shading. Use of the latter in operational forecasting environments (which is aimed at for the future) would be an absolute novelty.
The overall goal of the project is to verify in which way the combination of EO data and AI can be used to further increase the accuracy of MetGIS forecasts and of the simulation of snow cover. The AI will be fed and trained with a variety of datasets, differing in terms of spatial coverage and resolution, time resolution, combination of input parameters etc.
At the end of the project weather predictions processed with the current standard forecasting techniques of MetGIS are compared to predictions using the AI approach. A report will be generated giving details of the verification study, and recommendations about the best ways to proceed concerning the introduction of AI in the operational forecasting process of MetGIS.
- ExoLabs Ltd. liab. Co
- Xephor Solutions GmbH
Dr. Gerald Spreitzhofer
Lange Gasse 16/18