356
FLORENCE RABIER
IASI spectra (8461 radiance data). The procedure is iterative, based on
Rodgers, (1996). At each step, one channel is picked. It is the most
informative channel among those which have not been previously selected.
The analysis error covariance matrix is then updated before proceeding to
the next channel selection step. The choice of channels is based on
information content with respect to the background or current analysis
information. The selection criterion is the entropy reduction ER=-1/2
log 2 det(AB
-1
) or the DFS= Tr (I- AB
-1 ). At each step, one optimises ER=1/2 log 2 det(A i+1 A i
-1
) or the DFS= Tr (I- A i+1 A i
-1
) with A i the analysis error
covariance when using the first i selected channels, and A i+1 the analysis
error covariance when using the first i+1 selected channels. Figure 7 shows
the location of the “optimal” 300 channels selected for the retrieval of
temperature and humidity information for a typical IASI spectrum. This type
of work illustrates the benefit of using information content diagnostics for
the benefit of the optimisation of data assimilation, through data selection.
4.
Towards an adaptive system
Despite the notable increase in forecast skill over the past quartercentury, there is a necessity for further improvements, particularly in highimpact weather defined by their effect on society and the economy. The
international programme THORPEX is a response to the challenge of
improving the skill of high-impact weather forecasts. Its mission Statement
is “Accelerating improvements in the accuracy of high-impact 1-14 day
weather forecasts for the benefit of society and the economy”. Information
on this programme, and in particular the science plan, can be found on the
World Meteorological Organisation web page (www.wmo.int). Research
objectives are developed under four Sub-programmes: Predictability and
Dynamical Processes, Observing Systems,
Data Assimilation and
Observing-Strategies, Societal and Economic Impacts. Among the core
objectives, THORPEX plans to Contribute to the design and demonstration
of interactive forecast systems which include the new concept of “targeted
observations” and to perform THORPEX Observing-System Tests (TOSTs)
and Regional field Campaigns (TReCs) to test and evaluate experimental
remote-sensing and in-situ observing systems, and when feasible,
demonstrate their impact on weather forecasts.
What is Targeting? In the last decade, strategies were developed that
identify locations where additional observations would provide maximal
improvements in the expected skill of forecasts. Targeting strategies are
based on techniques that predict, prior to the actual measurements, the
influence of an observation (or set of observations) on the uncertainty of a
subsequent forecast. Different targeting techniques have been developed:
some involve the adjoint of the linearized version of the forecast model or of
FLORENCE RABIER
IASI spectra (8461 radiance data). The procedure is iterative, based on
Rodgers, (1996). At each step, one channel is picked. It is the most
informative channel among those which have not been previously selected.
The analysis error covariance matrix is then updated before proceeding to
the next channel selection step. The choice of channels is based on
information content with respect to the background or current analysis
information. The selection criterion is the entropy reduction ER=-1/2
log 2 det(AB
-1
) or the DFS= Tr (I- AB
-1 ). At each step, one optimises ER=1/2 log 2 det(A i+1 A i
-1
) or the DFS= Tr (I- A i+1 A i
-1
) with A i the analysis error
covariance when using the first i selected channels, and A i+1 the analysis
error covariance when using the first i+1 selected channels. Figure 7 shows
the location of the “optimal” 300 channels selected for the retrieval of
temperature and humidity information for a typical IASI spectrum. This type
of work illustrates the benefit of using information content diagnostics for
the benefit of the optimisation of data assimilation, through data selection.
4.
Towards an adaptive system
Despite the notable increase in forecast skill over the past quartercentury, there is a necessity for further improvements, particularly in highimpact weather defined by their effect on society and the economy. The
international programme THORPEX is a response to the challenge of
improving the skill of high-impact weather forecasts. Its mission Statement
is “Accelerating improvements in the accuracy of high-impact 1-14 day
weather forecasts for the benefit of society and the economy”. Information
on this programme, and in particular the science plan, can be found on the
World Meteorological Organisation web page (www.wmo.int). Research
objectives are developed under four Sub-programmes: Predictability and
Dynamical Processes, Observing Systems,
Data Assimilation and
Observing-Strategies, Societal and Economic Impacts. Among the core
objectives, THORPEX plans to Contribute to the design and demonstration
of interactive forecast systems which include the new concept of “targeted
observations” and to perform THORPEX Observing-System Tests (TOSTs)
and Regional field Campaigns (TReCs) to test and evaluate experimental
remote-sensing and in-situ observing systems, and when feasible,
demonstrate their impact on weather forecasts.
What is Targeting? In the last decade, strategies were developed that
identify locations where additional observations would provide maximal
improvements in the expected skill of forecasts. Targeting strategies are
based on techniques that predict, prior to the actual measurements, the
influence of an observation (or set of observations) on the uncertainty of a
subsequent forecast. Different targeting techniques have been developed:
some involve the adjoint of the linearized version of the forecast model or of