Toegepast Atmosferisch Onderzoek
NOWCASTING CLOUD COVER
Introduction
Clouds are being observed by satellites already since several decades.
In particular geostationary satellites are very suitable for nowcasting
of clouds because they have a high temporal resolution. Polar satellites
have a much higher spatial resolution, but their temporal resolution
is too low for nowcasting.
Inspite of this long history of observations clouds are still not assimilated in
NWP models. However, since a few years very short range cloud
forecasts using Meteosat imagery as an initialisation are made.
One example is the 'MetCast' model, which is operational at the KNMI since the
summer of 1998, and it forecasts until 12 hours ahead.
This model resembles a complete numerical weather model, though
dynamics is detached from the physics, and radiation computations
are not made. The disconnection of dynamics provides an important advantage:
the MetCast model has no spin up nor spin down of cloud water and
degree of cloud cover.
The MetCast model
In MetCast (Meteosat Cloud Advection SysTem) clouds observed by Meteosat
are advected with winds forecast by the HIRLAM model.
Apart from advection also evaporation and/or condensation of liquid water
is computed. In addition, turbulent diffusion is simulated.
The initial 3-D cloud fields are determined by combining synoptic observations
of cloud base heights and cloud top heights computed by the
MetClock system.
MetClock is a module which automatically computes cloud top temperatures and degree
of total cloud cover from Meteosat images.
Verification of the computed degree of cloud cover by the MetCast model
has revealed that (on the average) the total cloud cover was forecast
more precisely by MetCast, compared to the HIRLAM model; at least until
the first 12 hours during daylight and until the first 5 hours at night.
In particular during the first forecast hours the accuracy of the
predictions was considerably higher than that of traditional weather
models.
The MetCast model makes plus 12 forecasts every 3 hours. On the hours
in between plus 3 forecasts are made.
The horizontal resolution of the model is 30 km, but this can be changed freely.
The vertical resolution is 367 m, this can be changed as well.
For the meteorologist at the KNMI the following 2-D fields are available:
-
degree of cover of low clouds (0 - 2 km)
-
degree of cover of middle clouds (2 - 6 km)
-
degree of cover of high clouds (6 - 11 km)
-
degree of cover of both low and middle clouds (0 - 6 km)
-
degree of cover of all clouds (0 - 11 km)
-
height of highest cloud tops
-
height of lowest cloud base
-
height of 'ceiling'
These 8 parameters are available as time series for the following stations:
- Den Helder
- Schiphol
- Zestienhoven
- Vlissingen
- Beek
- Eelde
- Twente
- Eindhoven
- De Bilt
Also, prognoses of Meteosat images are made for the ir and vis channels.
Applications
The MetCast model can be used in different applications:
- Combining the solar angle and the degree of total cloud cover
can provide predictions of (short wave) global radiation.
- Predictions of clear skies at night can help the forecasting of fog.
- AuToTrend: At present experiments are carried out with a high
resolution MetCast (30 m in the vertical, 20 km in the horizontal)
in aid of aviation. In this high resolution version also several cloud
layers are assimilated, whereas the operational model has only one
cloud layer initially.
- We also have available an experimental version of MetCast in which
a 1-D convection model is embedded. With this model we carry out
nowcasting experiments for electric storms.
Additional information
Starting from 2002 also images from the Meteosat Second Generation satellite
will be available.
These images will provide a higher spatial and temporal resolution
of the cloud top temperature and degree of cloud cover.
At present it is also attempted to apply the 3-D cloud analysis system
(MetClock and the synoptic height of cloud base) to the Hirlam model.
Possibly this application will help to improve predictions of Hirlam
temperatures.
References
-
'Cloud detection using Meteosat imagery and Numerical Weather
Prediction model data', Arnout Feijt, Paul de Valk and Sibbo van
der Veen, Journal of Applied Meteorology, July 2000.
-
'Zeer korte termijn bewolkingsverwachting
met behulp van MetCast',S. H. van der Veen, Technisch Rapprt TR-213,
KNMI, september 1998 (in Dutch).
-
HIRLAM: High
Resolution Limited Area Model. This is an operational numerical
weather prediction model at the KNMI.

Sibbo van der Veen