July 2012 - This project aims to produce more accurate and timely forecasts of hazardous weather such as thunderstorms that can lead to flooding, damaging winds and fires.
Forecasting over time periods of just a few hours ahead is commonly referred to as "nowcasting" and whilst the lead time is short, it can nevertheless provide useful information to the emergency response community. Since March 2012, the Met Office has been running a demonstration trial of an hourly-cycling Numerical Weather Prediction (NWP) system for the southern UK. This combines a 1.5 km resolution version of the Unified Model and 3 km resolution 4D-Variational (4D-Var) data assimilation. This system is referred to as the Nowcasting Demonstration Project (NDP) and every hour it produces analyses and forecasts for the period 0 to 12 hours ahead.
The NDP has been enabled by the installation of a new supercomputer allowing faster production of more technically advanced NWP forecasts and data assimilation in addition to the current operational forecast systems and means that the NDP is available to Met Office forecasters over the period of the 2012 London Olympic and Paralympic Games. This period allows us to perform a demonstration and evaluation of its capabilities and potential for future operational use over the whole UK when more computer power is available in the future.
Ultimately it is hoped that an NWP-based system can replace the Met Office's current UKPP nowcasting system of Lagrangian extrapolation blended with NWP. It will combine an accurate depiction of the current weather with improved representation of the evolution and development of new weather and storm systems by using the fluid dynamical, microphysical and thermodynamical equations governing the behaviour of the atmosphere. Good data assimilation and analysis is vital for these short period forecasts of 0-6 hours and frequent updates using more recent observations of the state of the atmosphere are needed.
The NDP is nested in the UKV model which provides it with boundary conditions, but it produces analyses and forecasts every hour rather than the 6 hours of the larger-domain UKV. The domain of NDP is shown in the figure above and covers just Wales, Southern England and the Midlands as this is all that can be run in the available time on our current supercomputer system.
The data assimilation method is called 4D-Var because it uses all observations over a set time window, in this case of 1 hour, i.e. they are available in 4 dimensions - the 3 spatial dimensions and time. During the data assimilation period/time window, observations of the current state of the atmosphere are combined with the latest available forecast for the same period. Using an iterative minimization method called "variational analysis" (essentially least squares fitting allowing for errors in the forecast and the observations), the evolution of the forecast is gradually changed to be closer to that of the atmosphere as measured by the available observations.
The NWP model needs information on temperature, humidity, cloud, wind, pressure and aerosols (for fog and precipitation formation) but no one observing system provides this information over all time and space. Up to now, operational NWP systems have typically used hourly observations which may take 1-2 hours to reach the Met Office. The NDP requires sub-hourly data that need to reach the Met Office within 5-15 minutes of the observation time so very fast processing and communication links are required.
The NDP is able to assimilate:
In addition, radar-derived surface rain rates, available every 15 minutes, are assimilated in the first hour of the forecast, using the technique of latent-heat nudging - adjustments to the temperature and moisture fields aloft in response to the precipitation.
The data assimilation time window runs from 30 minutes before the hour to 30 minutes after. We wait 15 minutes for observations to arrive so start the forecast at 45 minutes after the hour. The observation processing takes about 1 to 2 minutes, the 4D-Var analysis 3 to 10 minutes and the 12 hour forecast about 10 minutes on 6 nodes of an IBM power 7 computer. The forecast is available approximately 1hour after the nominal analysis time.
Observations are not available uniformly in space, nor are all required variables available at all observation locations. Therefore their information needs to be spread smoothly in horizontal and vertical directions in order to adjust the previous forecast fields and also via so-called balance relationships to adjust other fields. The spreading functions applied to observations in the NDP are sharper than those in the standard UKV. For example, in the UKV the humidity spreading function has a horizontal scale of around 90 km whereas in the NDP it can be between 30 and 2 km depending on the vertical mode. This allows the observational network to generate and retain sharper features in the NDP analysis and forecast.
The figure below compares surface rain rate forecasts from the NDP (top left), UKV (bottom right) and current UKPP nowcast system (bottom left) with the radar derived surface rain rate (top right) at 15UTC (16BST) on 28 May 2012. The NDP and UKPP forecasts are 5 hour forecasts from 10UTC and the UKV is a twelve hour forecast from 03UTC which is all that would have been available to the forecasters at the same time as the NDP and UKPP forecasts. The line of thunderstorms was not present at the analysis times for any of these forecasts. The UKPP blended nowcast failed to predict the storms as the extrapolation could not develop them and the operational 4km resolution UK forecast also did not develop them. The UKV 1.5 km resolution forecast for the UK did develop some storms but they are too far east and isolated. The NDP has benefited from a later analysis time, the hourly cycling, more observations and the more advanced data assimilation system of 4D-Var rather than 3D-Var and has a good indication of the line of thunderstorms and their location.
More Information on the NDP system and example forecasts can be found by following links below:
Last updated: 3 October 2016