WP-S4: Model optimisation and sustainability

Goals of the project:

Main goal of the subproject is to guarantee efficient usability of the PALM code on future HPC computer architectures. Planned are e.g. code optimisations and measures to improve the operability for scientific users. Furthermore, code extensions are required to ensure scalability for large computer systems with more than 100.000 cores. Also, code deficits which have been identified during the first project phase shall be eliminated.

Tasks of the project:

The sub-project includes four work packages.

WP-S4.1: Implementation on an immersed boundary condition. This will allow better representation of building surfaces that are not aligned along the Cartesian model grid. This especially helps to improve the energy balance for building surfaces.

WP-S4.2: Implementation of a compressible set of model equations. This will eliminate the need to solve the Poisson equation for calculating the perturbation pressure, which becomes a performance bottleneck on computer systems with more than 100.000 cores and prevents the scalability and therefore usability of the code on respective computer architectures.

WP-S4.3: Model optimisation and professional tools for model development and application. Part of the optimisation will be the code vectorisation for the new NEC-Aurora vector engines, as well as optimisations of new model components. Furthermore, existing pre- and post-processing tools as well as existing scripts for model operation shall be unified on a uniform code basis (Python). Configuration and steering files shall be converted to a uniform format (yml), too.

WP-S4.4: Tool for analysing cold air drainage flows. The implementation of an IBC (WP-S4.1) will allow PALM-4U to simulate cold air drainage flows much more precisely than it would be possible with a step-wise Cartesian topography. Tools shall be developed to determine e.g. cold air volume flows to quantify air ventilation in urban areas.

Project structure:

PI: Dr. Matthias Sühring, also project scientist

PI: Prof. Dr. Siegfried Raasch

Furthermore, two small companies (Klaus Ketelsen, IT expert; Meteosolutions) are put in charge of code optimisation and improvement of tools for model operation.

Deliverables:

Immersed boundary condition, compressible equation solver, unified scripts and configuration files for model operation, and an analysis tool for cold air drainage flows.

Progress so far:

The calculation of cold air drainage variables has been implemented in the model.

The implementation of unevenly sized subareas by IT specialist Klaus Ketelsen was completed. In combination with the still ongoing extension of the FFT pressure solver for non-cyclic lateral boundary conditions, this removes a number of restrictions regarding the model setup (area size, number of grid points along x/y/z).

The final implementation of the "immersed boundary condition" (cut-cell topography) is carried out in close cooperation with WP-P2.

References:

none

Contacts:

suehring[at]meteo.uni-hannover.de

raasch[at]meteo.uni-hannover.de