Adam works on ensemble data assimilation systems, and is currently on secondment to the Korea Meteorological Administration
Areas of expertise:
- Variational data assimilation
- Initialization and gravity-wave control
Adam is a senior scientist on long-term secondment to the Korea Meteorological Administration - one of the Met Office's core collaboration partners for weather and climate research. As well as assisting with the scientific and technical collaboration between the Met Office and KMA, he is also involved in the development of a new global data assimilation system called "4DEnVAR".
One of the key requirements of a data assimilation system is to characterise the errors in the "background" forecasts that are combined with observations to produce updated estimates of the atmosphere's state. Traditionally these errors are specified via rather basic models, whose parameters are obtained from climatological statistics. In reality however we know that the structure of forecast errors is complex and constantly changing, depending on where atmospheric instabilities have occurred in recent days, and how well their effects have been picked up by the observation network. Capturing these complex error structures is the domain of What is an ensemble forecast?, so it is natural to use information from such systems to improve the modelling of forecast errors within the data assimilation system. From 2008 Adam led a project to develop a hybrid 4DVAR system that combines the standard climatological estimates of the background error with data from the Met Office ensemble prediction system The Met Office ensemble system (MOGREPS). The hybrid 4DVAR scheme gives significant improvements in forecast quality and was implemented operationally in the Met Office global forecasting system in July 2011.
One of the main problems with hybrid 4DVAR is its relatively poor computational performance on massively-parallel supercomputers with large numbers of cores, with a trend towards computers with even larger core counts. To prepare for future computer architectures the Met Office is currently developing an alternative to hybrid 4DVAR called "4DEnVAR", which is much more efficient and scalable on massively-parallel computers. Another advantage is its ability to generate initial conditions for ensemble forecasts, allowing a unification between data assimilation and ensemble generation. Currently, Adam's main focus is his contribution to 4DEnVAR development.
Adam joined the Met Office in 1993, and was heavily involved in the development of the Met Office 3DVAR and 4DVAR data assimilation systems, which now form the basis of operational weather forecasting at the Met Office. A particular focus was the development of schemes for controlling the "gravity-wave noise" that results from the approximations typically used within analysis schemes. If left unchecked, these spurious waves contaminate the forecasts, and can lead to problems assimilating later observations. To get around such problems, Adam developed the "Incremental Analysis Update" (IAU) and "Digital Filtering" schemes that are now used within the operational forecasting system. The control of gravity waves was also the subject of Adam's MSc dissertation, completed at the University of Reading with the support of the Met Office. In 2004, Adam moved to the Met Office Hadley Centre to join the UK-Japan Climate Collaboration (UJCC); a project to develop high-resolution climate models on the Earth Simulator supercomputer in Yokohama. On his return to the UK in 2008, he rejoined the Data Assimilation and Ensembles section to lead the development and implementation of the global hybrid 4DVAR system.
His secondment to the Korea Meteorological Administration began in April 2012.
Last updated: Apr 9, 2015 2:31 PM