Numerical modelling of clouds and atmospheric flows
Posted: 11 Dec 2015 | 14:39
The Met Office/NERC Cloud model (MONC) has been developed in a collaboration between EPCC and the Met Office. MONC delivers a highly scalable and flexible Large Eddy Simulation (LES) model capable of simulating clouds and other turbulent flows at resolutions of tens of metres on very large domains.
Dr Ben Shipway, Met Office Scientist, writes about the project.
Numerical Weather Prediction (NWP) and climate models have developed considerably over the past few decades along with the computational power to drive them.
Thirty years ago, resolvable length scales of atmospheric flows were on the order of 100km in operational models, where now they are on the order of 10km for global operational models and 1km for regional models. With an increase in resolution comes increased accuracy, but even at these higher resolutions the fundamental fluid motions of clouds and turbulent flows remain at the subgrid scale.
In order for models to represent and account for the interaction of these small-scale flows with the largerscale meteorology, physicallybased parametrizations are developed. A key tool in understanding the fundamental physics of these flows and thus development of the parametrizations is Large Eddy Simulation (LES).
A highly scalable LES
The Met Office Large Eddy Model (LEM) has been the workhorse of cloud process modelling in the Met Office and many UK universities. Originally developed in the 1990s and written predominantly in Fortran 77 with severe limitations in its MPI decomposition, it has failed to keep pace with modern HPC architectures and now lacks the scalability enjoyed by its contemporary operational models.
Despite the shortcomings of the software implementation of the LEM, the scientific foundation is well regarded and well tested and so the MONC project sought to produce a new LES model that is built on the science of the LEM, but with modern software design that is capable of running on tens of thousands of cores and enables high resolution, large domain simulations.
A flexible approach
MONC will be used to simulate a wide variety of atmospheric flows, such as dry boundary layers, fog, stratocumulus or deep moist convection. Each requires its own particular scientific configuration using varying levels of complexity or different numerical implementation.
In recognition of the wide variety of choices a scientist may want to have, and with an eye on future development of new or more efficient codes, a flexible plug ‘n’ play approach was adopted when building the new model. This involved a software infrastructure whereby different components of the model can be chosen and configured at run time. A centrally held and updated model state which can be passed to each component allows for a very simple interface and so allows rapid development of new components.
A scalable and flexible model is all very well, but the key thing a scientist requires is the output of the diagnostic state of the system.
The complex nature of many of the simulations which will be carried out with MONC leads to a wealth of interesting and insightful diagnostics which could potentially be used to understand the underlying physics of the behaviour of clouds. To get at this information, the scalable core of the model needs to be extended to a scalable and efficient diagnostics system.
Another innovation in MONC is the diagnostic server. This allows diagnostics to be calculated and gathered on demand and then farms them off to the dedicated processes which can then asynchronously process statistics and write out to netcdf files.
Just the beginning
The original MONC project is now coming to an end and has undoubtedly been a great success. A beta release of the code is in preparation and the code is now hosted on the Met Office Shared Repository, where the community can start to engage. EPCC involvement will continue with follow-on funding to make further optimisations to the dynamical solver and the microphysics components. It is only a matter of time before the fruits of our labour pull through to fundamental understanding of clouds and vital improvements in climate prediction and weather forecasting.
“EPCC has been doing the code development of MONC, and using our experience of HPC we have produced a model which has already been run on over 32,000 cores and over 2 billion grid points. This scale is far beyond what the community can currently work at and we see no reason why MONC can not be scaled beyond this to 100,000 cores in the future. Our modular design of MONC means that it is trivial to plug in additional functionality as future scientists require.”
Nick Brown, EPCC
These images show the same simulation of a bubble of warm air rising through the atmosphere, with different filters applied. To begin with the bubble was warmest in the middle. The simulation has run for 230 seconds and in that time the warmest, less dense, air has risen up faster (higher) than the surrounding cooler, more dense, air.