THE ECMWF PERSPECTIVE
367
storm over the next 12 hours. When the analysis kicks in, with its lower
effective resolution, the storm can get weakened. So although the
observations say deepen this storm significantly, the analyses says fill it in.
This conflict in analysing small-scale, but important, features is generic.
3.
The ocean analysis
3.1 Introduction
Ocean analyses are needed to provide initial conditions for both monthly
and seasonal forecasts. For the monthly and seasonal forecast systems we
use essentially the same ocean analysis systems. They are not exactly the
same, as the monthly system needs a faster analysis than the seasonal and
also has different cut-off times for the receipt of data. The resolution of the
ocean model is 1 degree except near the equator where it is 1/3 degree in the
meridional direction to better resolve the equatorial waves which are
important in processes such as El Nino. In future we will explore the need
for even higher resolution models coupled to the medium range weather
forecast model, with potential impact on tropical cyclone prediction, but that
will not be discussed further here.
3.2 The observing system
Ocean observations are mainly of thermal data and mainly in the upper
500m. Some salinity measurements are now available from ARGO but only
in the last few years. Other measurements of salinity e.g. from CTD are
sparse and have not generally been available in real-time. There are almost
no measurements of velocity, except in the surface layer. Fig 3 shows the
thermal data coverage in a typical 10-day window.
Several features are
worthy of note. First, there is quite a substantial coverage in real-time. In
fact most data are now received within a day or so at operational centres
such as ECMWF through a network called the GTS (Global
Telecommunications System). When data are received, each individual
observation is checked. It is compared against the model first guess and also
with an analysis performed without the datum being checked. The actual
quality control is quite complex and will not be gone into in detail. See
rejected are in grey. These are mainly located near the coast since coastal
points are not well represented in the ocean model. Some profiles are
partially accepted; at some depth the model and data differ by too great an
amount and the datum is rejected but data above and below this level might
be accepted.
Smith et al, 1991 for further discussion. Data profiles on fig 3 which are
367
storm over the next 12 hours. When the analysis kicks in, with its lower
effective resolution, the storm can get weakened. So although the
observations say deepen this storm significantly, the analyses says fill it in.
This conflict in analysing small-scale, but important, features is generic.
3.
The ocean analysis
3.1 Introduction
Ocean analyses are needed to provide initial conditions for both monthly
and seasonal forecasts. For the monthly and seasonal forecast systems we
use essentially the same ocean analysis systems. They are not exactly the
same, as the monthly system needs a faster analysis than the seasonal and
also has different cut-off times for the receipt of data. The resolution of the
ocean model is 1 degree except near the equator where it is 1/3 degree in the
meridional direction to better resolve the equatorial waves which are
important in processes such as El Nino. In future we will explore the need
for even higher resolution models coupled to the medium range weather
forecast model, with potential impact on tropical cyclone prediction, but that
will not be discussed further here.
3.2 The observing system
Ocean observations are mainly of thermal data and mainly in the upper
500m. Some salinity measurements are now available from ARGO but only
in the last few years. Other measurements of salinity e.g. from CTD are
sparse and have not generally been available in real-time. There are almost
no measurements of velocity, except in the surface layer. Fig 3 shows the
thermal data coverage in a typical 10-day window.
Several features are
worthy of note. First, there is quite a substantial coverage in real-time. In
fact most data are now received within a day or so at operational centres
such as ECMWF through a network called the GTS (Global
Telecommunications System). When data are received, each individual
observation is checked. It is compared against the model first guess and also
with an analysis performed without the datum being checked. The actual
quality control is quite complex and will not be gone into in detail. See
rejected are in grey. These are mainly located near the coast since coastal
points are not well represented in the ocean model. Some profiles are
partially accepted; at some depth the model and data differ by too great an
amount and the datum is rejected but data above and below this level might
be accepted.
Smith et al, 1991 for further discussion. Data profiles on fig 3 which are