Targeted observations collectively obtained from mobile observation
platforms have historically formed part of an adaptive observing
network. This network is designed to complement the routine observing
network used by Numerical Weather Prediction (NWP) models. Targeted
observations may comprise additional radiosonde ascents, routed
Aircraft Meteorological Data Reporting (AMDAR) aircraft, Automatic
Shipboard Aerological Program (ASAP) ships and drifting buoys. Each
observation is obtained from, or targeted in, a previously defined
targeting region. It is hypothesised that the assimilation of extra
observations in these regions will lead to a reduction of the forecast
error in NWP models. The process of selecting targeting regions
is termed sensitive area prediction.
The Met Office is developing an observation targeting capability
as part of our THORPEX research programme. This builds on experience
gained during the 2003 Atlantic THORPEX Regional Campaign (ATReC),
also known as the Atlantic-TOST.
In this campaign, targeting guidance was based on the Ensemble Transform
Kalman Filter (ETKF). Research to-date has focused on assessing
the quality of guidance using different configurations of the ETKF.
Targeting guidance utility has been evaluated by running a series
of Observation System Experiments (OSEs) to assess the impact
on forecast error for a range of surface and upper air forecast
fields. Future planned work includes further development of the
ETKF to ensure robust and timely delivery of targeting guidance
using the Met Office's 15-day global ensemble running at ECMWF.
Further evaluation of ETKF-generated guidance in the presence of
high-impact weather events is planned.
Figure 1. Example sensitive area
prediction produced using the ETKF initialised from the ECMWF
ensemble. The targeting guidance predicts that deployment of targeted
observations within the shaded areas will improve two-day (T+48)
forecasts over the verification area (the blue box centred over
the UK).
Previous studies have shown that increasing the observation density
for satellite observations with uncorrelated error improves the
analysis accuracy. On the other hand, for operational NWP centres,
the cost of processing a dense observation network must also be
a consideration. Research is being undertaken at the Met Office
to assess the utility of assimilating adaptively thinned Advanced
Microwave Sounding Unit (AMSU) radiance data so that high density
data is used only in areas where they add the most useful information.
Figure 2. Case study demonstrating the utility of
assimilating adaptive AMSU radiance data from ECMWF-initialised
ETKF guidance targeting T+48 forecast verifying within the blue
box at 12Z 21 Oct 2006: a) Operational baseline AMSU data assimilated
thinned at 154 Km; b) Adaptive AMSU radiance data thinned at
40 Km in sensitive areas and 154Km outside; c) Plots of forecast
error for T+48 mean sea level pressure for operational baseline
(root mean square error = 0.953 hPa); d) Found error for adaptively
thinned AMSU data (root mean square error = 0.926 hPa)
Home 
Figure 1. Example sensitive area
prediction produced using the ETKF initialised from the ECMWF
ensemble. The targeting guidance predicts that deployment of targeted
observations within the shaded areas will improve two-day (T+48)
forecasts over the verification area (the blue box centred over
the UK). 
