344
FLORENCE RABIER
constantly re-adjust the model trajectory to produce a reasonable estimate of
the true atmospheric state. From these analyses of the atmosphere, the model
is run daily up to a few days to produce the forecast products which will
guide the forecasters in their prediction of the weather. At the beginning of
the 80’s, data assimilation was a minor sub-discipline of numerical weather
prediction, where the emphasis was mainly on the forecasting model itself.
Simple correction methods were used to update the forecast such as nudging
and linear optimal interpolation. Over the last two decades or so, this subject
has expanded into quite a mature and motivating area of research and
applications, with in particular the advent of variational methods. Such
major scientific advances, combined with a large increase in available
observations, has brought data assimilation to the forefront of operational
weather forecasting. Its use is also spreading to climate applications through
re-analyses and oceanography/chemistry applications. The experience
gained in data assimilation in meteorology can be shared with scientists
interested in other areas, such as oceanography. This paper will mainly
address the issue of the importance of data in the assimilation process, in the
context of global atmospheric modelling.
Firstly, the impact of observations on the forecast performance will be
illustrated through the 40-year reanalysis performed at ECMWF (European
Centre for Medium-range Weather Forecasts). Secondly, tools will be
described which can help to perform an optimal use of observations, through
data selection and error tuning. Finally, current developments towards an
adaptive system will be described in the context of the THORPEX
programme.
2.
Impact of observations on forecast performance
Operational data used at Numerical Weather Prediction (NWP) centres
are consisting of various data types provided by the global observing system.
The backbone of this system is formed by surface observations from land
and ship stations, and vertical soundings from radiosonde and pilot balloons.
From the 1970s, other data types emerged such as drifting buoys, aircraft
measurements, wind profilers, satellite radiances, satellite cloud-drift winds
and scatterometers. On the one hand, observations such as land stations and
radiosonde observations have been providing a stable source of information
throughout the years, but their horizontal distribution is far from being
homogeneous. On the other hand, satellite observations are blooming and
becoming a major and horizontally homogeneous source of information in
current systems. How did this increase in available observations translate
into analysis and forecast quality? As a partial answer to this question, an
illustration of the impact of observations is now provided in the context of
the ERA-40 project (www.ecmwf.int/research/era). As summarized in
FLORENCE RABIER
constantly re-adjust the model trajectory to produce a reasonable estimate of
the true atmospheric state. From these analyses of the atmosphere, the model
is run daily up to a few days to produce the forecast products which will
guide the forecasters in their prediction of the weather. At the beginning of
the 80’s, data assimilation was a minor sub-discipline of numerical weather
prediction, where the emphasis was mainly on the forecasting model itself.
Simple correction methods were used to update the forecast such as nudging
and linear optimal interpolation. Over the last two decades or so, this subject
has expanded into quite a mature and motivating area of research and
applications, with in particular the advent of variational methods. Such
major scientific advances, combined with a large increase in available
observations, has brought data assimilation to the forefront of operational
weather forecasting. Its use is also spreading to climate applications through
re-analyses and oceanography/chemistry applications. The experience
gained in data assimilation in meteorology can be shared with scientists
interested in other areas, such as oceanography. This paper will mainly
address the issue of the importance of data in the assimilation process, in the
context of global atmospheric modelling.
Firstly, the impact of observations on the forecast performance will be
illustrated through the 40-year reanalysis performed at ECMWF (European
Centre for Medium-range Weather Forecasts). Secondly, tools will be
described which can help to perform an optimal use of observations, through
data selection and error tuning. Finally, current developments towards an
adaptive system will be described in the context of the THORPEX
programme.
2.
Impact of observations on forecast performance
Operational data used at Numerical Weather Prediction (NWP) centres
are consisting of various data types provided by the global observing system.
The backbone of this system is formed by surface observations from land
and ship stations, and vertical soundings from radiosonde and pilot balloons.
From the 1970s, other data types emerged such as drifting buoys, aircraft
measurements, wind profilers, satellite radiances, satellite cloud-drift winds
and scatterometers. On the one hand, observations such as land stations and
radiosonde observations have been providing a stable source of information
throughout the years, but their horizontal distribution is far from being
homogeneous. On the other hand, satellite observations are blooming and
becoming a major and horizontally homogeneous source of information in
current systems. How did this increase in available observations translate
into analysis and forecast quality? As a partial answer to this question, an
illustration of the impact of observations is now provided in the context of
the ERA-40 project (www.ecmwf.int/research/era). As summarized in