«Medium-Range Weather Prediction Austin Woods Medium-Range Weather Prediction The European Approach The story of the European Centre for Medium-Range ...»
Geostationary satellites, at an altitude of about 36,000 km, provide a continuous view of the earth from an apparently stationary position above the equator. Instruments on polar-orbiting satellites, flying at much lower altitude, typically around 800 to 1,200 km, provide more precise details about the atmosphere, including its temperature and moisture profiles, from the surface to the highest levels. A polar satellite’s orbit is fixed relative to a line from the middle of the earth to the sun; the earth is rotating within the orbit. They cover polar regions that cannot be viewed from a geostationary orbit. The lack of in-situ observational coverage in parts of the globe, particularly the Pacific Ocean and the Southern Hemisphere, has led to an increasing role for satellite data. An important programme initiated in the 1990s was the EUMETSAT Polar System, a joint venture with the US agency NOAA. EUMETSAT will assume responsibility for the “morning” — at local time — polar orbit, while the USA will continue with the “afternoon” coverage. EUMETSAT and NOAA instruments will be carried on its Metop satellites, a series of satellites providing service well into the second decade of the 21st century.
EUMETSAT has thus become one of the major partners providing satellite systems for observing our planet, and Europe has taken a leading place in monitoring global weather and climate. Its success ensures the availability of
160 Chapter 13key satellite data for Europe and for many developing countries. We will see in Chapter 14 the approval in February 2005 of a ten-year strategic plan for the Global Earth Observation System of Systems (GEOSS). At first, GEOSS will build on existing satellites and sensors. These will include not only the operational EUMETSAT and NOAA satellites, but also ESA’s Envisat, launched in March 2002, and NASA’s Earth Observing System satellites.
Later launches will be coordinated.
Tony Hollingsworth’s visionary understanding of the importance of investing in the assimilation of satellite data lead to the Centre developing a strong research programme that has exploited a wide variety of satellite data. The data were used not only to improve the analyses and forecasts, but also to verify forecasts. Systematic errors and model biases that otherwise could have gone unnoticed were revealed. Increasingly, work had to be planned well in advance, so that data from a new satellite instrument could be used as soon as possible after launch. Hollingsworth worked hard to ensure good working relations and strong interaction with satellite agencies during his years as the Centre’s Head of Data Division and later as Head of Research.
The Centre used satellite-measured winds generated at the European Space Operations Centre (ESOC) as soon as they became available in the late 1970s. Infrared radiation emitted by the cloud tops to the geostationary satellite could be used to estimate the cloud top temperature. The height of the clouds could then be found by comparing the cloud top temperature with the analysed temperature at different levels of the model atmosphere. Visible clouds were tracked to provide estimates of the wind speed at the height of the clouds, provided of course that clouds anchored in the lee of high ground were ignored! Feedback from monitoring at the Centre helped to improve the estimates. Use of these data improved the small-scale flow in the tropical analyses and close to frontal systems.
In later years, separate estimates of the wind were made from the movement of features detected in high-resolution measurements of the water vapour. Careful quality control was required to produce usable wind fields;
the technique was refined during the years.
The Centre was improving its use of TIROS Operational Vertical Sounder (TOVS) data from the polar-orbiting satellites. Research in 1987–88 concentrated on determining the information content of the temperature and humidity data, and evaluating the techniques used to retrieve temperature and humidity from the radiances measured by the satellite instruments.
The impact of data on the Centre’s forecasts was of course carefully monitored. By 1989, the quality and resolution of the analysis had improved to the extent that the Director, Lennart Bengtsson, reported to Council that Data from on high 161 while the impact of satellite data was positive and large in the Southern Hemisphere, where there were comparatively few other data sources, there was one startling conclusion: “it has been found that the overall impact of satellite temperature soundings has had a minor negative effect on the forecast quality... over the Northern Hemisphere!” Following the significant improvements to the Centre’s system in the preceding two years, the errors in the “background” fields, that is, in the short-range forecasts, were smaller than the errors in the temperatures retrieved from the satellite measurements. Use in the Centre’s system of temperature data calculated from satellite measurements actually degraded the quality of the analyses;
the short-range forecasts had become more accurate than the data! And these were from measurements that had been made at very great expense.
With hindsight, part of the problem lay with the data assimilation system itself. In a well-tuned system, inaccurate observations can be used in such a way that they will do no harm; they will be given a “weighting” corresponding to their accuracy and the usefulness of their data content. For example at the time of writing, the Centre’s background forecasts are often more accurate than the (actually quite accurate) radiosonde data, but these radiosonde data are still used to advantage in the system.
However, in those years, use of satellite data was far from optimum. The instruments measure the radiation upwelling to space from gases in the atmosphere. A complicated retrieval procedure was required to provide estimates of the temperature. Processing the raw radiance data, the actual instrument measurements, was needed in the early years, because the data assimilation systems then in use could not properly handle unprocessed data.
When early satellites were launched, the numerical models needed not what the satellites measured (i.e. infrared, microwave and other radiation coming from gases and clouds in the air, and from the sea, ice and earth below), but the temperature and wind. These were the quantities that had been available from balloons and aircraft, so the assimilation systems had been designed for these kinds of data. However, the act of processing the radiance data to retrieve these numbers introduced errors. As well, even with the most careful processing, spurious signals could be introduced into the data.
One of the benefits of the Centre’s variational assimilation system being developed in co-operation with Météo France was that it could use the raw, unprocessed, radiance data directly. In a sense, instead of taking the satellite measurements, and trying artificially to extract or retrieve quantities that the models required, such as temperature, the variational system was able to tune the model atmosphere so that the radiation that would be emitted from the top of the model atmosphere towards space would correspond to the satellite readings.
162 Chapter 13There is another advantage to using raw data. It can be a year or more after the launch of a satellite before processed data can be made available; the raw data are available typically within a month after launch.
Bengtsson highlighted “the necessity to undertake major efforts to develop better methods for the determination and use of satellite observations”.
In May 1989, a Workshop was held jointly with EUMETSAT on “the use of satellite data in operational weather prediction”. Urgent work was identified, and close contact was soon forged between the scientists at the Centre and those at EUMETSAT.
At the end of 1989, Council unanimously approved a proposal to set up a Satellite Data Research Unit at the Centre, with Switzerland noting “the very great cost associated with technical operational satellites”. The Unit was established in February 1990 with responsibility “for developing systems to use operationally available satellite data and to assess the performance of future observing systems”. A first task of the new Unit was “to improve the use of satellite temperature soundings... by direct assimilation of clear radiances... ”.
An initial staff of two under John Eyre soon expanded. Many skilled and experienced scientists and consultants, several funded by EUMETSAT, worked on satellite data at the Centre in the following years. In early 2005 there were 19 scientists working in the Satellite Data Section under Section Head Jean-Noël Thépaut. These included Graeme Kelly from Australia, who had been at the Unit from its inception.
Soon after its establishment, the Unit was comparing the cloud-clearing schemes used by the Laboratoire de Météorologie Dynamique (LMD) France, the UK Met Office, NESDIS in the USA and the University of Naples to produce “clear-column radiances”, that is data from regions not affected by clouds. By 1992, rapid and substantial progress was being made in research into the use of satellite data. Use of one-dimensional variational analysis (1D-Var) for retrievals of temperature data over the Northern Hemisphere was showing improvements in forecast skill; this was employed from June 1992 in the operational system. An improvement in the analysis of the humidity was soon seen. A great deal of work was required and many problems had to be overcome before extension to the rest of the globe could be implemented in December 1994.
Arrangements were made to ensure that the Centre would receive wind and ocean wave data in near real time from the scatterometer and altimeter on board the new Earth Resource Satellite (ERS-1) launched in July 1991, to allow calibration and validation of the data. The ERS-1 included three major radar systems among its many instruments.
Data from on high 163
• A scatterometer sent a beam from two antennae. The returned signal bouncing back from ocean waves about 5 cm high provided wind information; the waves are generated directly by the wind.
• The Synthetic Aperture Radar (SAR) was a quite different instrument with much higher power consumption. Elaborate signal processing and the motion of the instrument meant that the instrument could be turned into the equivalent of a radar with a very long antenna, a “synthetic aperture”. Ocean waves and swell were measured.
• The third radar was an altimeter; it sent a radar chirp 50 times each second. The return signal gave a very accurate estimate of the height of the instrument above the variable ocean surface, actually about 780 km, allowing ocean currents to be measured, since the dynamic height of the ocean surface determines the currents. Wave height was also measured, and used in the analysis of the Centre’s ocean wave model.
Feedback to ESA continued to contribute to trouble-shooting the ERS-1 data. For example, ESA software could not discriminate between “upwind” and “downwind” signals from the scatterometer; the ambiguity led to the possibility of incorrect surface winds being retrieved from the satellite data.
With feedback from the Centre’s team, ESA was soon able to develop corrections to the satellite bias problems in (a) the scatterometer calibration and (b) the statistical model that ESA had been using to relate radar backscatter from the ocean waves to estimate the wind. Methods being used to estimate wave heights were also improved.
The ERS-2 satellite, launched in April 1995, provided much useful data in the following years. Data from the scatterometer instrument gave estimates of the surface wind speed and direction. A comparison of background wave height and altimeter wave height data soon showed that use of these data had a beneficial impact on the surface wind field analysis. These data were used operationally from 1996, and improved the model significantly in the tropics, with smaller effects elsewhere over the globe. The Centre monitored the winds and radar backscatter data; quality control procedures were steadily improved. ERS-2 also carried a new instrument GOME that measured ozone. Ozone data were analysed by the Centre from 2002.
“Future system studies” were underway to specify instruments for planned satellites, including some that would not be launched for a decade or more.
Experiments were made on the impact of satellite winds and aircraft reports on the Centre’s forecasts. The team at the Centre was involved in studies to draw up specifications for the advanced instruments required on the Meteosat Second Generation and for the planned Third Generation, as well as for the ground segment for the planned EUMETSAT Polar System.
164 Chapter 13The team developed a system to simulate global data sets to investigate different scenarios for a satellite Doppler wind lidar instrument; this was needed for an “observing system simulation experiment”. Liaison with ESA continued: at the time of writing the Centre was actively involved in preparing to process data from the ADM-Aeolus mission scheduled for launch in October 2007. This, ESA’s second Earth Explorer Core Mission within its Living Planet Programme, was designed to make direct measurements of global three-dimensional wind-fields. Named after Aeolus, who in mythology was appointed “keeper of the winds” by the Greek Gods, the Aeolus satellite will be the first mission to observe the Earth’s wind patterns from space directly.
The Centre’s assimilation system was modified in the late 1990s to allow use of the raw radiance data from operational NOAA polar-orbiting satellites.
• Information from each of the five TOVS and Advanced TOVS (ATOVS) instruments were treated as independent sources of radiance data that was assimilated in their natural scan geometry, thus avoiding the attempt to combine or map the different readings to a single location.
• The data were assimilated where they were measured, avoiding artificial adjustment of the variation of the radiance when an instrument took measurements away from the vertical.
• Since clouds and precipitation interact with atmospheric radiation, it was much easier to use data from areas with clear skies. A battery of tests searched for the characteristic signals of cloud and precipitation.