The Terrestrial Systems Modelling Platform (TerrSysMP) is a fully coupled scale-consistent physics-based numerical model system, currently consisting of the COSMO NWP (v4.11) model, the Community Land Model (CLM, v.3,5) and the ParFlow (v3.1) variably saturated surface and subsurface hydrological model, coupled with the external coupler OASIS3/OASIS3(-MCT).

TerrSysMP allows for a physically-based representation of transport processes across scales down to sub-km resolution with explicit feedbacks between the individual compartments, including groundwater dynamics and a full representation of the terrestrial hydrological cycle. The model has been ported and tuned for multiple HPC systems such as highly scalable IBM Blue Gene/Q platform and Linux clusters at Juelich Supercomputing Centre and the Meteorological Institute at University of Bonn. TerrsysMP employs a MPMD paradigm for its execution which require memory and load balancing considerations. Advanced profiling and tracing tools such as Score-p and Scalasca are used for this aspect and the advantages of parallel coupling have been demonstrated in a massively parallel environment.

In its current setups the model can be run from catchment to continental spatial scales. The setups include a high resolution (1km and sub-km) domain set up over Northrhine-Westphalian and a 12 km European domain in context of EURO-CORDEX, which are currently in focus at the TR32 Collaborative Research Centre. Some examples of ongoing developments and extensions, are

  • improvements in the coupling,
  • updates of the component models CLM to v4.5 and COSMO to v5.1,
  • the first steps towards the consideration of bio-geochemical cycles by including CO2 coupling,
  • the implementation of a parallel data assimilation framework in ParFlow+CLM, and
  • steps towards an improved big data readiness in the complete modeling system.

Fully coupled daily experimental forecasts forced with ECMWF and DWD model forecasts, are run and the results are uploaded to a YouTube channel for public dissemination.

 

Slides to this talk