Improving the turbulent flux parameterization in regional and global climate modelling systems by extending the classic Monin-Obukhov Similarity Theory with Artificial Neural Networks

• Applicant: Dr. Marcus Breil
• Subject Area: Atmospheric Science
• Term: since 2020
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 456675194

The turbulent fluxes of heat and momentum between the land surface and the atmosphere are fundamental processes in earth’s energy balance. An accurate representation of the turbulent fluxes is therefore a basic prerequisite for successful weather and climate modelling. In weather and climate models, the turbulent fluxes are generally calculated by using the Monin-Obukhov Similarity Theory (MOST). MOST is a semi-empirical method, which is searching for functional interrelations in turbulent motion. In this context, the general concept of the MOST approach is based on some simplifying assumptions, like for example prevailing steady state conditions over a homogeneous land surface structure. For deviating conditions, the classic MOST approach is consequently not always able to identify such consistent interrelations in turbulent motion. As a consequence, the turbulent fluxes are maybe partially spuriously simulated in weather and climate models, constituting a possible reason for consistent biases in several model intercomparison studies.Therefore, the aim of this project is to reduce these uncertainties in weather and climate models by improving the calculation of the turbulent fluxes. We try to achieve this goal by surmounting certain limitations of the classic MOST approach and extending its applicability to non-stationary boundary conditions over heterogeneous land surface structures. For this purpose, Artificial Neural Networks (ANNs) will be used. ANNs constitute a method which is well suited to identify systematic interrelations between a required target quantity and its influencing factors in large amounts of data, which are not detected by conventional methods. In the framework of this project, these capabilities of ANNs will be exploited to identify such systematic interrelations in turbulent motion, also under atmospheric conditions for which the classical MOST approach is failing. To achieve this, parts of the MOST parameterization will be replaced by the ANN and extended with parameters, which account for the specific constraints of the classic approach. This extended MOST approach will be implemented in the ICON modelling system and applied in global and regional weather and climate simulations. Finally, the capability of this approach to improve the description and simulation of the turbulent fluxes in weather and climate models, will be assessed by comparing the results of the extended MOST approach with ICON standard simulations.

• DFG Programme: Research Grants