«Medium-Range Weather Prediction Austin Woods Medium-Range Weather Prediction The European Approach The story of the European Centre for Medium-Range ...»
While this processing was generally quite effective, at times some data included significant radiances from clouds and precipitation. The Centre’s analysis screening used short-range forecasts to compute clear sky values of the “window channel” radiances, giving better results than the previous processing.
Numerical experiments of these first steps in use of raw satellite data in real time confirmed that useful improvements in the analyses and forecasts had been achieved. Observing system experiments in 1998 confirmed that satellite data had a significant positive impact on both analyses and medium-range forecasts in both Hemispheres.
In 1999, major changes were made to the operational assimilation of the radiance data. In May, after an extensive trial over a four-month period, direct operational assimilation of raw TOVS and ATOVS data began.
Additional levels were introduced in the high atmosphere of the model, and ozone was introduced as another variable in the data assimilation system. An immediate improvement in forecast scores throughout the troposphere and stratosphere was achieved.
Data from on high 165 The total amount of water vapour in a column of the atmosphere could for the first time be measured, in almost all weather conditions, over the oceans with the launch of the first Special Sensor Microwave Imager (SSM/I) instrument, as long ago as June 1987; it was carried on a spacecraft forming part of the US Defence Meteorological Satellite Program. The instrument measured the microwave radiation emitted by water vapour in the atmosphere below. This was useful in principle, for example to diagnose the model’s hydrological cycle. First, the satellite data had to be verified against “ground truth”: the measurements made by radiosonde instruments that happened to coincide with the passage of the satellite. The “ground truth” itself is not always truthful; radiosonde humidity sensors for example are notorious for their errors! Years of research into use of SSM/I data came to fruition in February 1998, when a 1D-Var retrieval of SSM/I data was run as part of the operational suite, giving regular plots of total column water vapour, surface wind speed and cloud liquid water. Also the SSM/I provided wind speed data from over the oceans, but unlike the scatterometer, not wind direction.
The edges of sea-ice fields derived from SSM/I brightness data were up to 300 km better than those used operationally. Tropical precipitation was also estimated from radiance data from the SSM/I instrument. The radiance data being emitted was strongly affected by rain. Using this to modify the initialisation of diabatic heating in the model was first investigated at the end of 1990. It took until 2005 — 15 years later — before research had progressed sufficiently to allow the data from places where it was raining to be assimilated.
In December 1998, the Centre concluded an agreement with the Met Office, under which the Centre participated in a Satellite Applications Facility (SAF) for Numerical Weather Prediction (NWP). The objectives of the SAF were to accelerate the development of techniques for more effective use of satellite data in NWP, and to prepare for effective exploitation of the data coming from satellites planned for launch in the future.
In April 2000, the model-based correction of biases in the TOVS and ATOVS radiance data was applied to the SSM/I radiances. Now there was almost global coverage of wind speed over oceans, and of total column water vapour. The Centre’s development of bias-correction and the improved understanding of the error characteristics of the raw radiances led to a considerable increase in the volume of satellite data assimilated.
The ongoing co-operation between ECMWF staff and those at EUMETSAT and ESA was producing a range of benefits. Operational changes were made in 2000 to the calibration and quality control of Meteosat data by EUMETSAT. In fact, many other users of satellite data were now using the Centre’s statistics as early warnings, or as confirmation of problems.
166 Chapter 13It’s not easy to use radiance data from the channels that measure emissions from the low atmosphere over land, or from cloudy skies. Emissions from the earth’s surface, or from clouds, have to be separated from the radiances from the air. In 2000, an experimental system was developed to analyse the contributions from the surface, and to separate them from the atmospheric data. Adjusting the surface temperature and emissivity within the 4D-Var assimilation system accomplished this. Work continued to optimise the technique, so allowing use of these valuable data over land.
The NASA AQUA spacecraft was launched in May 2002, carrying the high-resolution Atmospheric InfraRed Sounder (AIRS) instrument. AIRS was the first ‘hyperspectral’ sounder, making measurements in 2,378 spectral channels. Information on profiles of temperature and humidity was provided, at enhanced vertical resolution compared to the previous generation of operational satellite sounders. AIRS was a research forerunner for instruments with similar performance on operational satellites later in the decade.
A subset of radiance data from AIRS was made available to ECMWF in near real time from the end of October. Before this date, significant technical development was made using simulated AIRS data sets provided by NOAA/NESDIS. With this intensive preparation, experiments in cloudscreening, monitoring and assimilation impact could begin almost immediately following the arrival of the real AIRS data. Tony McNally and his colleagues carried out a 100-day trial of the use of AIRS data. They showed that the assimilation of AIRS data had reduced errors in both shortrange and medium-range forecasts, and concluded “that we now have a safe ‘conservative’ assimilation system for AIRS which should be considered for operational implementation”. AIRS data started to be used operationally from October 2003, with small but positive changes to the forecasts. This, the first operational use of advanced infrared sounder data, paved the way for use of data from planned future operational satellites such as Metop, to be launched in 2006, which will carry the Infrared Atmospheric Sounding Interferometer (IASI).
A co-operation agreement was concluded with ESA in May 2005.
As in other areas of its work, we find the Centre starting from small beginnings in its research into, and operational use of, satellite data, and growing as the years passed to provide an impressive body of scientific expertise. Again, we have a flavour of the extensive collaboration between the research teams at the Centre and those outside, at EUMETSAT, in institutions in the Member States, and elsewhere. And again, we see that the groundwork is laid to ensure so far as possible that use of the future global observing system is optimised.
Re-analysis — towards a new ERA
The World Weather Watch is an astonishing technological achievement.
Nations of the world spend billions of Euros each year to measure and probe the atmosphere and oceans of our planet. Many different types of observing
systems are used:
• Satellites passively measure the radiation emitted by the surface of the earth and the sea; from this the temperatures can be deduced. The atmospheric greenhouse gases too are radiating to space; satellites measure this radiation to provide information about the temperature of the air aloft.
• Instruments on satellites emit bursts of high-energy radiation to the sea surface; the reflected radiation measures the waves, and in addition the surface wind speeds can be estimated.
• More than one thousand instrumented balloons drift through the air each day, measuring pressure, temperature and humidity as they rise to 20 km or more. The balloons are tracked by radar, so telling us the wind speed and direction.
• About two thousand buoys have been lowered into the ocean from ships, to sink to a depth of two km, recording salinity and temperature. They drift at this depth for ten days, continuously measuring, before rising to the surface and sending the collected measurements to satellites.
• Hundreds of floating buoys drift on the surface, sending to satellites the wind, and the sea and air temperatures.
• Fleets of commercial aircraft measure wind and temperature every ten minutes high over the earth’s surface.
The expensive part of meteorology is collecting the data; “more data, more data, right now and not later” isn’t cheap. The World Weather Watch costs the nations of the world some billions of Euros each year; the annual budget of the European meteorological satellite organisation EUMETSAT alone is 167
168 Chapter 14close to 300 million Euros. The ECMWF data assimilation system is probably the most advanced system for analysing the data; the Centre’s annual budget is around 40 million Euros. Cartridges worth a few hundred Euros in the Centre’s archive easily holds a year’s worth of these valuable data.
A ten-year strategic plan for the Global Earth Observation System of Systems (GEOSS) was approved at an Earth Observation Summit in February 2005 in Brussels. Initiated by the United States, and with the Centre participating from the start, GEOSS will evolve slowly from national systems to become a coordinated comprehensive set of observations. The aim is to integrate observational systems around the world to avoid existing massive duplication of efforts and ensure that gaps in coverage are filled.
More than 60 nations and 30 international organisations, including EUMETSAT, the European Commission and the European Space Agency, are working to establish the network of Earth observation systems. WMO will host the secretariat. GEOSS will focus on benefiting society. Weather prediction, our understanding of climate variability, agriculture, and human health and well-being will all be beneficiaries.
Truly vast amounts of information for the Global Observation System are stored in the ECMWF archives: observations of weather from all over the globe — temperature, wind, humidity, pressure and more — from the 1950s to the present time. While useful for many applications in its raw form, there are important questions that cannot be answered by the observations without further processing: Has the June temperature at 5,000 m above the North Atlantic changed on the average between the 1960s and the present decade?
Have the wind speeds around the roaring 40s in the Southern Hemisphere increased, decreased or remained unchanged?
Analyses of the global atmosphere have been made from the beginning of the Centre’s work and, like the data, stored in the archives. In principle the analyses can answer questions like these. However the analysis system itself has been steadily developing as the computers became more powerful, as the data sources — especially satellite data — have advanced and as the science progressed. Thus comparison of a temperature analysis made in June 1980 with one made in June 2000 would be misleading.
An analysis, strictly speaking a “re-analysis”, of all the observational data of past years in the database using a single, frozen, modern analysis system has a clear appeal. This difficult and complex project has been accomplished by the “ECMWF Re-Analysis” (ERA) project. We will see that this project exemplifies the truly global co-operative nature of meteorology. As well, it has exposed the Centre to a much wider user community of research scientists worldwide, a critical group who are constantly providing the Centre with feedback on the quality of its output.
Re-analysis — towards a new ERA 169 The start of ERA goes back to the data collected and analysed in real time from the beginning of operational forecasting at the Centre, during the FGGE (First GARP Global Experiment) period of December 1978 to November 1979. In the early years Bengtsson kept in mind the possibility of using the FGGE assimilation system as the Centre’s back-up system in case of delay in implementing the operational system. Sakari Uppala from Finland joined the Centre in June 1978 to work with Per Kållberg from Sweden, who was already at the Centre. Kållberg was appointed as Project Manager in July 1978.
Kållberg and Uppala formed the basis of the Centre’s “FGGE Section”.
Scientists from other interested institutes were seconded to work at the Centre in this effort in the following years: from Norway — Knut Bjorheim, the USA — Paul Julian and Steve Tracton, Japan — Masao Kanamitsu, and from Australia — Peter Price. And of course Bengtsson had a very keen interest in the everyday progress and decisions in the project. Wiin-Nielsen too kept himself informed.
Some of the raw instrument readings, called “Level I” data, for example radiance data from satellite sensors, had to be converted by the institutes receiving them to provide “Level II” weather parameters such as temperature and wind. Some of these were available within 10 hours of observation time.
These formed “Dataset IIa” and were available for operational analyses.
Others were delayed for up to several months to build the best possible observational dataset. This, called “Dataset IIb”, included all the special observations deployed during FGGE such as drifting buoys, special aircraft data and balloon soundings, some radar data, constant level balloons, and cloud track winds from geostationary satellites. Lots of surface data were received. Archiving capacity was being stretched beyond its limits, and much of these data were not included, with surface data thus making up a small fraction of the total volume. The Level IIa data were collected and managed by a complex WMO data processing and management system before reaching the Centre. The final IIb Datasets were merged at the Space Based and Special Observing System Data Centre in Sweden, but — as noted above — with a delay of several months; this was a complex operation.
Level IIIa analyses were those produced operationally at the National Meteorological Center (NMC) Washington and other institutes from the Level IIa data. Much later the IIIb analyses were produced by the ECMWF FGGE system using the non-real time Level IIb data. Parallel to the work at ECMWF, the Geophysical Fluid Dynamics Laboratory (GFDL) in Princeton in the USA produced its own version of the Level IIIb analyses. The Centre worked closely with GFDL in planning and carrying out the re-analyses.