Operational Numerical Modelling
|Numerical Models of the Atmosphere
Operationally, the Met Office runs two configuration of its
Unified Model. The global model has a horizontal resolution of
0.8333° longitude (432 columns) and 0.5555° latitude (325 rows) giving an approximate
resolution of 60km in mid-latitudes.
The global model is used to provide boundary conditions to the mesoscale
model which is a regional model centered on the United Kingdom.
This model has a resolution of 0.11° latitude by 0.11°
which is approximately 11km. In the mesoscale model there are 146 columns
and 182 rows. Both the global and mesoscale models have 38 levels in
In the mesoscale model the model North Pole is not located at
the geographical North Pole. This is done in order to obtain a
fairly uniform horizontal resolution over the area of interest,
ie the UK. The mesoscale model has its North Pole situated at
In both models, the variables are arranged in the same fashion.
In the horizontal, an Arakawa C grid is used, the u wind components
are east-west staggered from temperature and the v wind
components are north-south staggered, and in the vertical
Charney-Philips grid staggering is used which means that
(potential) temperature, scalars (moisture variables and tracers)
and vertical velocity are staggered from (Exner) pressure, density
and horizontal wind.
Back to top
|The Numerical Weather Prediction System
The forecast models run within a system known as the operational
suite. Once initiated, the operational suite will perform
all tasks needed to produce the forecast with no further
manual intervention. There are however, facilities to manually
override tasks should any problems occur.
The operational suite embraces all the individual tasks that
are required to produce a forecast. Most of the software
within the operational suite has been written in-house. The
suite itself is controlled by what is known as the suite
control system (SCS). The SCS can be used to select which
tasks are run, how they are run and when they are run.
The first task is the
observation processing to extract all the observations that
have been received, to quality control them and finally reformat
them into a form ready for use by the model.
For certain runs of the model, a reconfiguration then occurs.
This is a procedure to incorporate data fields from external
files into the model. This is required to update fields that
have their own standalone analysis, such as the sea surface
temperature, or to update a climatological field.
The data assimilation scheme
is then run. This adjusts the model background field, which is
a forecast from a previous model run, towards the new data
received from the observations.
The main forecast is then run, the length of which varies according
to the particular run of the model, more details are given below.
The forecast data are written into files known as fieldsfiles.
Using these, various plotted charts and maps are produced which
forecasters then use to produce the weather forecast. It is
important that the charts are available at the earliest possible
time and therefore fieldsfiles are produced that cover a
24 hour period only. This enables charts for say T+24 to be
plotted and made available even though the forecast is
Once the forecast has completed various ancillary tasks are run.
These include archiving data, monitoring the observations and
verifying previous forecasts that were for the current time.
There are user friendly graphical user interfaces to set up
the suite and many of the individual tasks such as
the observation processing, the data assimilation, the verification
and the forecast model itself. This enables an user to set up
and run their own experiment as if it were the operational
Back to top
|Producing Forecasts for our Customers
Running a numerical weather prediction model is only part of
the process in producing a weather forecast. Before a forecast
is issued, the output from the model is studied by a
forecaster. For short ranges, the forecaster is able to compare
a model field against actual observations and identify any
possible errors so that appropriate allowances may be made.
The forecaster may be able to add extra detail to the model
forecast. An example of this is in the forecasting of
summer showers which are often on a scale too small for the
model to resolve adequately. The forecaster is also able
to respond quickly and amend a forecast should the situation
medium range forecasts, the forecaster is able to compare the results
from our model with those from other centres such as ECMWF (European
Centre for Medium-Range Weather Forecasts), NCEP (National Centers
for Environmental Prediction) and DWD (Deutscher Wetterdienst).
If all models are producing approximately the same solution confidence
in the forecast would be high. If however, all were showing different
evolutions the confidence would diminish rapidly. Another factor
that influences confidence is the consistency between model runs.
If the model consistency follows the same evolution confidence may
be high but if it suddenly changes then confidence falls rapidly,
and in these situations the solutions of other models may be crucial
as to the forecast issued. Sometimes, alternative forecasts may
be issued with probabilities assigned.
This human-machine partnership is very important in producing accurate
The main forecasting facility of the Met Office is the National
Meteorological Centre (NMC) based in Exeter, Devon. The Met Office
also has forecasters based in various outstations both in the UK
Back to top
The global model is run twice a day to produce forecasts for
up to 6 days (144 hours) ahead. These main runs are initialised
with data valid at 0000Z and 1200Z and are started at approximately
0305Z and 1505Z respectively. The data assimilation scheme uses
observations that have been made within 3 hours of the initial
data time, i.e. the 1200Z run uses observations made between 0900Z
and 1500Z. It takes approximately 70 minutes from the suite
starting to produce a forecast for T+144, approximately 20 minutes
of that comprises the observation processing and data assimilation
To maintain the assimilation cycle, there are intermediate runs
with data times of 0600Z and 1800Z which are run at approximately
1300Z and 0100Z respectively. These only produce forecasts to
9 hours ahead in readiness for the next main run. It can be seen
that these are run with a delay of 7 hours from the initial data
time whereas the main runs only have a delay of 3 hours. Therefore,
to maintain consistency, forecasts from 0000Z and 1200Z are
repeated at 0715Z and 1820Z respectively. These forecasts are
only to 9 hours ahead and provide the background for the
next intermediate run.
A mesoscale model covering the UK and surrounding areas
is also run. The main runs of the
mesoscale model are at data times of 0000Z, 0600Z, 1200Z and 1800Z
and these are initiated at 0200Z, 0800Z, 1400Z and 2000Z respectively.
The mesoscale model uses observations 90 minutes before and after
the data time and produces forecasts for upto 48 hours ahead. To
maintain the assimilation cycle, intermediate runs are run with
data times of 0300Z, 0900Z, 1500Z and 2100Z at 0545Z, 1215Z, 1910Z
and 0010Z respectively, each to 4 hours ahead.
In addition to the main atmospheric models, various ocean and
wave models are run. The operational suite runs 24 hours a day,
365 days a year.
Back to top
In August 2002, the Met Office introduced a completely new
formulation of its operational models. It was centred
around a completely new formulation of the model's dynamical
core, the fundamental equations, hence the term 'new dynamics'
but there were also many enhancements to the physical
parametrizations. A fuller description of the work
may be found in the
June 2002 edition of NWP Gazette and a
factsheet has also been produced.
The Met Office disseminates various products from the
operational numerical weather prediction models in
GRIB and GRID formats. In this section there is a
description of the products that are available.
More about operational products