[
] 226
The reanalysis of daily
weather observations at ECMWF
Manfred Kloeppel, Adrian Simmons and Sakari Uppala,
European Centre for Medium-Range Weather Forecasts
A
powerful new approach to climate analysis has emerged in
recent years. It applies the tools and techniques of modern
everyday weather forecasting in a process called reanalysis.
The products, or reanalyses, have applicability far beyond that of
traditional climate information.
The Earth’s climate has traditionally been studied by statistical analysis
of observations of particular weather elements such as temperature, wind
and rainfall. Climatological information is often presented in terms of
long-term averages, and sequences of observations are examined for
evidence of warming, increased frequency of severe storms, and so on.
The new approach of reanalysis plays an important role for the
Intergovernmental Panel on Climate Change (IPCC). Three working
groups contributed to the Fourth Assessment Report of the IPCC in 2007,
with Working Group I (‘The Physical Basis of Climate Change’) provid-
ing a comprehensive assessment of the physical science of climate change.
The results presented were based on the extensive scientific literature
that had become available since completion of the IPCC’s Third
Assessment Report in 2001, and had utilized expanded data sets, new
analyses, and more sophisticated climate modelling capabilities. The
Final Report of Working Group I comprises almost 100 references to
reanalysis data sets.
GEO Task CL-06-01 will ‘ensure the development of international
mechanisms to coordinate and maintain sustained climate data repro-
cessing and reanalysis efforts.’ The European Centre for Medium-Range
Weather Forecasts (ECMWF) contributes to this task by its expertise in
reanalysis and provision of reanalysis data sets. A reanalysis workshopwas
held with GEO support at ECMWF in June 2006 to assess and coordi-
nate activities in atmospheric reanalysis.
New approaches to climate analysis
Global weather forecasting has advanced considerably since the
1970s. Forecasts of increasing accuracy have resulted from refining
the numerical model of the atmosphere used to make the forecast,
and from refining the procedure used to determine the initial model
state from which the forecast starts. Both refinements have been made
possible by investment in powerful computer systems, comple-
menting the even larger investment made worldwide in the
meteorological observing system.
In daily forecasting the latest ground- and satellite-based obser-
vations are combined with a short forecast based on earlier
observations to create the initial state for a new forecast. The initial
state describes the elements of weather throughout the atmosphere
and the geophysical properties of land and ocean surfaces. Weather
charts with frontal structures and so forth can be drawn
from this information to help display and diagnose
current weather in the traditional way.
In a reanalysis, the weather observations collected in
past decades are fed into a modern forecasting system
that is much more refined than the systems available
when most of the observations were made. Atmospheric
and surface conditions are reconstructed for each day
of the period over which suitable observations exist.
Reanalysis differs from the traditional climatological
approach in that it processes a wide variety of observa-
tions simultaneously, using the physical laws embodied
in the forecast model and knowledge of the typical errors
of forecasts and observations to interpret conflicting or
indirect observations and fill gaps in observational
coverage.
One of the first reanalyses was started in the mid-1990s
by the National Centers for Environmental Prediction
(NCEP) in the United States, covering the period from
1948 onwards. This reanalysis has been continued in near
real time to the present day. Europe’s first reanalysis, ERA-
15 (1979–1993), was carried out in parallel with the
NCEP reanalysis, until its completion in 1996. Europe
continued as a key player in this activity with ERA-40
(1957-2002), carried out (like ERA-15) by ECMWF with
support provided primarily by Europe’s National Weather
Services and the European Commission. ERA-40 provided
products of hitherto unprecedented scope and quality.
Japan also recently joined the ‘reanalysis club’ and
reanalysed the period 1979–2004. Interaction among
reanalysis producers and with groups involved in obser-
vational data stewardship has eased the task of acquiring
comprehensive past observations in a format that facili-
tates reanalysis and helps more generally in the preparation
of new reanalysis systems.
The capability of reanalysis is illustrated well by one of
the most severe storms of the 20th century over Europe,
which occurred on 31 January/1 February 1953 causing
the greatest surge on record for the North Sea as a whole.
Its amplitude reached 2.74m and 2.97m at Southend and
King’s Lynn in England, and 3.36m in the Netherlands.
Almost 100,000 hectares of eastern England were flooded
and 307 people died. The Netherlands fared much worse
still, as 50 dykes burst and 1,800 people drowned.
S
OCIETAL
B
ENEFIT
A
REAS
– W
EATHER