The London Model: forecasting fog at 333m resolution
November 2015 - New research moves Met Office numerical weather prediction into uncharted waters, and already delivers useful improvements to our fog forecasting ability.
Why, in the high-tech dominated world of 21st century air travel, do weather conditions still have such a huge effect on aviation? In particular, low visibility conditions, such as fog, can cause significant disruption when they affect busy airports, particularly if unexpected.
The reason is safety-based and lies in the fact that pilots and air-traffic controllers still rely predominantly on being able to see other aircraft during landing and taxiing on the ground. When this is not possible, the spatial separation between aircraft must be increased. Air-traffic controllers must be able to clearly distinguish the location of different aircraft on their radar screens, and have additional time to convey instructions to the pilots. At busy airports, such as London Heathrow, there is almost no spare capacity to increase this aircraft separation and maintain the flight schedule, leading to possible delays and the potential for cancellation of flights. This can have significant cost implications for the airline operators, create knock-on effects at other airports around the world because planes are no longer where they are expected to be, and cause inconvenience to thousands of passengers.
Therefore, accurate forecasting of low visibility events is crucial for airport operators, as advanced warning and planning for such events can significantly mitigate the cost and passenger inconvenience. With this as our motivation, in 2013 the Met Office started running very high resolution simulations of the Greater London area on a daily basis. Typically, the 1.5km resolution UKV model is used for short-range forecasting over the UK. The "London Model" has a resolution of 333m and is nested inside the UKV, covering a region of 100km by 66km. Previous work has shown added benefits of increased resolution in complex terrain, due to the much better representation of hills and valleys. Although the hills and valleys are less extreme in the London area, they are nevertheless significant (Figure 1).
The London Model runs twice per day. Once at 06 UTC for 36 hours, providing information that could be used on a planning timescale for the following day. A 21 UTC run for 18 hours then provides an overnight update on likely conditions in the early morning, the time most susceptible to fog.
A recently published study has evaluated the performance of the London Model over a 14 month period, demonstrating clear benefits of the enhanced resolution. Variability in the terrain creates near surface turbulence, which can act to break up fog, and the London Model is able to explicitly represent this effect in a way that the UKV is not. For example, Figure 2 shows a case where Heathrow remained fog-free, despite there being patchy fog in nearby locations, including 10km away at RAF Northolt. The London Model represents this very well, whereas the UKV erroneously shows widespread fog across a large area of London, including at Heathrow.
The London Model has therefore provided a useful improvement to our fog forecasting ability over the Greater London area, and when the supercomputer is fully installed, the domain of the London Model will be approximately doubled to include Gatwick, Stansted and Luton airports. However, fog remains a challenging problem to understand and forecast. Because of this, the Met Office is returning to the Shropshire valleys this winter to make intensive observations of fog formation and dissipation, with the aim of further developing both the London Model and UKV Model.
The London Model is by no means a finished product however, and research and development of this model will continue during the coming years. Numerical weather prediction at this scale presents significant challenges, but as these are understood and overcome, the London Model will be useful for far more than just fog forecasting. It will assist with forecasting urban heat-island effects, air quality and localised flash-flooding, help the development of our other global and regional modelling systems, and put the Met Office in a position such that when computer resources allow, we can model the entire of the UK at similar resolution.