Adaptive mesh generation, data assimilation and storm prediction
The Met Office Unified Model is a computational model used for both weather predictions, occurring over short time, and climate predictions over longer time periods.
An ongoing challenge in computationally modelling weather and climate is how to mesh the globe. Currently, the Met Office makes use of lat-lon grids where the sphere is divided into grids along lines of latitude and longitude. Unfortunately, these have an undesirable property of clustered grid points over the poles which manifests through data‐communication problems on massively parallel computational architecture. This has led to developments in the mechanisms used to mesh. The next generation Unified Model will make use of a cubed-sphere mesh.
Our next generation atmospheric model will use non-conformal cubed-sphere mesh to overcome scalability limitations of lat-lon grid used in our current model. #GungHo https://t.co/vBTAD32jWf pic.twitter.com/91iZrInG95
— Met Office Science (@MetOffice_Sci) May 24, 2019
This project builds on a long history of collaboration between IMI and the Met Office and stems from funding by NERC. Rather than solving the Unified Model over a grid that is fixed in time, be it lat-lon, or cubed-sphere, one can make use of various adaptive mesh strategies.
This project concerns moving mesh methods, which have a fixed number of nodes but relocate grid points, and adaptive mesh refinement strategies, that locally refine and coarsen the grid. Both share the goal of ensuring resolution at particular points of interest according to some appropriate metric. Each have their own advantages and disadvantages in terms of computational cost, guaranteed run-time, massive parallelisation and conservativity properties.
The aim of the project is to determine the applicability and mechanism of combining these approaches with a particular focus on how adaptive methods can also be used to provide high resolution in the important area of forecasting extreme events.
A particular application that will be further explored is in data assimilation, in which data is continuously integrated into the Met Office forecast. Historically, adaptive methods have proven crucial in removing spurious correlations and significantly improved the accuracy of the near ground forecasts. Future work aims to further improve this.
IMI Researchers Chris Budd and Tristan Pryer intend to leverage existing work through IMI’s follow-on-funding scheme to seek further research funding whilst simultaneously developing industrial links with IMI partners and bringing new industrial connections to the IMI.