«Medium-Range Weather Prediction Austin Woods Medium-Range Weather Prediction The European Approach The story of the European Centre for Medium-Range ...»
During the second quarter of 1996 the first pass through the full 15 years was completed. Monitoring enabled many errors arising during production to be located and rectified. Nevertheless two lengthy periods needed to be re-run. A bug, present also in the operational system, was undetected until the re-analysis had completed up to August 1980. The bug significantly affected humidity at upper levels. Secondly, much cloud track wind data were accidentally excluded from June 1990 to October 1992, due to a change in their format affecting the data in the archive. A re-run of the first period was particularly desirable, as it presented an opportunity to run the FGGE year with the same observations and forcing fields as the National Centers for Environmental Prediction (NCEP). Both re-runs were completed in September 1996.
By November 1996, Burridge was able to report to Council that “the ERA project has completed its production phase with the creation of a new, validated 15-year data set of assimilated data for the period 1979 to 1993”. A Re-analysis Workshop held in July 1996 had almost 100 participants, an indication of the now high level of research interest in the project.
The ERA-15 data set contained global analyses and short-range forecasts of all relevant meteorological parameters, beginning with 1979, the year of the FGGE, and running to 1993. All analyses and forecasts were generated by a modern, consistent data assimilation system. The system included better “first guess” preliminary fields and a more efficient dynamical balancing for the assimilation of observed data. The new FGGE analyses were compared with those of other institutions such as NCEP, and the original GFDL analyses.
“Madden-Julian” oscillations are events that are associated with enhanced deep thunderstorm activity moving eastward from the Indian Ocean into Indonesia, and then into the Western Tropical Pacific. These oscillations give rise to “Oceanic Kelvin Waves” below the ocean surface, which propagate eastward along the equator carrying abnormally warm sub-surface water toward, and eventually to, the South American Coast. An Oceanic Kelvin Wave reaching the coast of South America is a signal that El Niño is coming. The capability of representing “Madden-Julian” oscillations in the re-analysis and in the ECMWF and old GFDL analysis was investigated by comparing with satellite observations. The oscillations were successfully reproduced by the new analysis. However agreement with the satellite data
176 Chapter 14was not quite satisfactory. It was found that the use of satellite-observed wind and aircraft data in the data assimilation needed particular care.
The ERA-15 project was a global effort. It received funding and assistance from many quarters, including:
• ECMWF Council,
• European Union,
• University of California Program for Climate Model Diagnosis and Intercomparison (PCMDI),
• Japan Meteorological Agency (JMA),
• World Climate Research Programme (WCRP) of the World Meteorological Organization (WMO),
• Center for Ocean-Land-Atmosphere Studies (COLA),
• National Center for Atmospheric Research (NCAR),
• National Centers for Environmental Prediction (NCEP), and
• Cray Research Incorporated.
ERA-15 data were made available to ECMWF Member States through the MARS archive, to university users within the UK through the British Atmospheric Data Centre (BADC), to University users in Germany through the Max-Planck-Institut für Meteorologie (MPI), and to the UCAR community in the United States through NCAR.
Close co-operation was also established between the ERA team and the teams responsible both for the NCEP re-analysis, and for the re-analysis performed by the NASA Data Assimilation Office.
Production continued until September 1996. The team was desperately running ERA-15 up to the minute the last of the CRAY systems was powered down on 1 October 1996.
The world moves on! In May 1997, Burridge reported to Council that “an initial assessment has begun into the feasibility of a 40-year re-analysis, making use of the additional observation archives being obtained from NCAR”. This would quickly become another global effort.
Euroclivar was the European component, funded under the Fourth Framework Work Programme, of an international research programme on “Climate Variability and Predictability” (Clivar) addressing many issues of natural climate variability and anthropogenic climate change. The need for a project with the objectives of ERA-40 was recognised by Euroclivar. It “strongly recommended that a new 40-year re-analysis be made in Europe in the next five years”.
ERA-40 was expected to make a significant contribution to those objectives of the Fifth European Community “Framework V” Programme Re-analysis — towards a new ERA 177 covering “Research, Technological Development and Demonstration” that related to the World Climate Research Programme. It would, for example, provide data in support of projects such as DEMETER, which would explore the potential for seasonal prediction, and PROMISE, a programme on the “Predictability and Variability of Monsoons, and the Agricultural and Hydrological Impacts of Climate Change”. This would make extensive use of ERA-40 analyses for validating climate and seasonal prediction models and for driving crop models for impact studies. Studies of ozone depletion and other aspects of atmospheric chemistry could also benefit from ERA-40.
Within a year, substantial progress had been made in reception and initial processing of data from NCAR. The complete NCAR archive of TOVS satellite data, beginning as early as 1978, had been received. These data had to be processed to a form suitable for the ECMWF variational data assimilation system. An External Advisory Group for ERA-40 had been formed.
Scientists were being seconded from China, Japan and the USA to work on the project. EUMETSAT had agreed to re-process cloud track winds from the 1980s. A bid for EU funding under the Framework V activities was made.
The Centre’s validation programme was augmented by a variety of external validation projects:
• Koninklijk Nederlands Meteorologisch Instituut — Ocean waves
• Max-Planck-Institut für Meteorologie — Hydrological cycle
• Météo France — Ozone, stratospheric analyses, ocean surface fluxes and Alpine snow
• Met Office — Clear sky radiation simulation
• National Center for Atmospheric Research — Observations and mass, heat, energy and moisture diagnostics
• University of Reading — General circulation, climate variability The Centre, in practise Adrian Simmons, would coordinate the project.
The Centre would produce the analyses.
In the planning phase the partners were represented by Klaus Arpe of MPI, Tony Slingo of the Met Office, Pascal Simon of Météo France, Gerbrand Komen of KNMI, Roy Jenne and Kevin Trenberth of NCAR, and Brian Hoskins and Julia Slingo of the University of Reading. ECMWF contributors included Adrian Simmons, Sakari Uppala, Per Kållberg and Keith Edwards. Rex Gibson was Project Manager for the preparatory phase of ERA-40. Simmons and Gibson wrote the proposal for ERA-40. Ongoing management of the complex project was largely shared between Simmons and Uppala, who became project manager on Gibson’s retirement from ECMWF at the end of August 1999.
178 Chapter 14A variational data assimilation system was planned, to make a new synthesis of the in-situ and remotely sensed measurements made over the period beginning in mid-1957. A major improvement had been made to the atmospheric observing system in preparation for the International Geophysical Year of 1958 which had as its goal: “...to observe geophysical phenomena and to secure data from all parts of the world; to conduct this effort on a coordinated basis by fields, and in a space and time, so that results could be collated in a meaningful manner”.
Thus, starting in 1957, ERA-40 would produce analyses every six hours throughout the 40-year period, extended to 45 years as we shall see, supplemented by intermediate three-hour forecasts. The products would be of high temporal and spatial resolution, with grid spacing close to 125 km in the horizontal and with sixty levels in the vertical, extending from the surface to a height of about 65 km.
The basic analysed variables would include not only the conventional wind, temperature and humidity fields, but also stratospheric ozone and ocean wave and soil conditions. Model snow would fall on the model surface and accumulate; the snow depth was adjusted according to observations when available, otherwise, it was allowed to change slowly to the climatological values. The production of a three-dimensional ozone field consistent both with available ozone observations made by satellite, and with the dynamical state of the atmosphere, was needed for investigations of the composition of the atmosphere. Ozone measurements were preferred over climatology for RTTOV, a radiative transfer model for very rapid computation of radiances at the top of the atmosphere and transmittance profiles for a range of operational space borne radiometers. RRTOV was the result of collaboration between the Centre, the Met Office and Météo France.
A coupled ocean-wave model was introduced. Ocean wave height was based on the use of satellite data from the altimeter onboard the ERS satellite, available from 1991. Before then, the waves were driven by the analysed surface winds. ERS also carried a scatterometer to measure microwaves reflected from the ocean surface.
Additional information would be stored on the quality of the observations used and of the analyses generated.
A sophisticated archival/retrieval system would be used to store the results and make them widely available. Compact sub-sets of the data would be generated for worldwide user on the public data server. Customers and users of the results would gain maximum benefit from the information by being provided with extensive documentation.
Re-analysis — towards a new ERA 179 ERA-40 built on experience gained with ERA-15. It adopted the innovative variational analysis techniques, especially for assimilating satellite data. New types of observation and improved specifications of sea-surface temperatures and sea-ice distributions were used.
The partners in this project — and indeed many others — supported the acquisition and preparation of the necessary observations, the trial production and validation of analyses, the assessment of user requirements and the general planning of the project. Per Kållberg, Sami Saarinen from Finland and Angeles Hernandez from Spain were scientists in the Group. Graeme Kelly and many other Research Department scientists contributed to the work, often in their spare time. Institutions in China, Japan and the USA funded the secondment of staff to work on the project. Scientists from the Member States contributed as well. Several other institutions provided copies of their archives of past observational data.
Fujitsu Ltd provided substantial computing support for the project: they donated the VPP300 system that had been installed at the Centre before the VPP700. EUMETSAT re-derived winds from Meteosat-2 images for the period 1982–1988. In addition the World Climate Research Programme and the Global Climate Observing System provided funds in support of an External Advisory Group for the project.
Re-analysis projects must proceed at sufficient speed for them not to be continually overtaken by developments in data-assimilation technique and large-scale computing. Funding from the EU enabled the basic production of the re-analyses to be completed within the planned period of about two years, and enabled the necessary validation and demonstration studies to be undertaken.
By the end of 1998, work was underway, preparing the assimilation systems for experiments to ensure that the systems to be used would meet the scientific and technical requirements for the project. The External Advisory Group met in March 1999; help was forthcoming to get missing satellite data, and advice given on what should be archived. Work was under way in the Met Office in the UK, NCEP in the USA, and in the Arctic Climate System Study project of WMO, to specify consistent sea temperature and ice fields for the ERA-40 period.
A vast range of satellite data was used: cloud track winds, total column water vapour content, radiances (which indicate temperatures), ozone measurements and more. The need for a smooth transition from satellite to satellite was given special attention, particularly in the stratosphere where little other data were available.
180 Chapter 14Preparatory work was required on many kinds of data. Measurements made by radiosondes from different manufacturers had to be made compatible and biases removed, especially for the earlier data. However, by the end of 1999, problems were being systematically identified and corrected, and 25 years of preliminary test assimilations had been completed.
Even though not all the satellite data were ready, it was felt that the remaining problems were manageable, and production began in 2000 with the period from 1989. To have a spin-up, assimilation started from September 1986. The period 1957 to 1988 was delayed pending further studies of the data. By mid-2001, the re-analysis reached to the end of 1990, and the first year, 1957, had been analysed. Data coverage varied a lot during the period; it was notable that — while of course satellite data increased — the coverage of the valuable data from instrumented balloons over the oceans, and from the land area of the former Soviet Union, in the 1950s was far superior to that available in recent years.
Verification showed that the overall analysis quality was higher than expected. However, the value of external validation was soon evident. MPI identified serious deficiencies in the water cycle, which were traced to a coding error in surface-level data as received. Also NCEP reported that incorrect times had been assigned to radiosonde reports. Monitoring at KNMI revealed assimilation of erroneous ERS-1 altimeter ocean-waveheight data. Unrealistic rainfall in the 1990s over tropical oceans was detected by several validation partners’ monitoring, as well by the Centre.
Assimilation was at times suspended while the problems were addressed.