«Medium-Range Weather Prediction Austin Woods Medium-Range Weather Prediction The European Approach The story of the European Centre for Medium-Range ...»
In December 1983, the Council endorsed the establishment of a subgroup of the TAC “to follow the work leading to a replacement of the telecommunications sub-system”. The sub-group met in January 1984. In July 1984, an Invitation to Tender for a new telecommunications system was issued.
In May 1985 Council approved a contract with Software Sciences Ltd (SSL) for a new telecommunications system. During 1986 the Regnecentralen telecommunications system was replaced by a system based on a cluster of four Digital Equipment Corporation VAX 11-750 computers.
This, then called the New Telecommunications Computer (NTC), was providing a reliable service by the end of the year, and delays, which had proved a problem with the previous system, had been eliminated. The Regnecentralen was powered off on 20 September 1986, after 7 years and
216 Chapter 1750 days service. By December 1986, all Member States were connected with lines of 2,400 to 9,600 bps, except for the remaining low-speed lines to Italy (100 baud), Yugoslavia and Turkey (50 baud each). In July 1988, all links had become medium-speed, ranging from 2.4 to 14.4 kbps.
In 1986, five Member States co-operated in implementing a National Telecommunications System (NTS), similar to the joint project carried out in 1978-80. The NTS was also based on VAX computers, but now with a modular approach to the software design. DECnet protocols were used as the final “transport layer”. Costs of £50,000 for developing the software for the system were shared between the participating States.
A new dissemination scheme was implemented in June 1986 offering model level data, important for those running their own models, as well as pressure level data, any geographical area, any grid point system or spherical harmonic coefficients and some ad hoc services. From now on Member States could maintain their own product lists using a menu-based interactive system.
From January 1988, ECMWF data on the GTS were in GRIB format at the higher resolution of 2.5° — although the GRID dissemination at 5° resolution was maintained. Line speeds were steadily upgraded. By 1989, sixteen lines were at 9.6 kbps or higher, the remaining two at 4.8 kbps. Five Member States had established NTS connections, five more were in test.
The increased speeds and volumes of transmission led to an acute overload of the system; a VAX 8250, soon upgraded to an 8350, was purchased to alleviate the situation.
In 1992 Météo France started to use the TCP/IP protocol as the “transport layer” for its connection with the Centre. This proved to be an important milestone; since then, TCP/IP has become a standard in computer networking.
In the same year, products to be used as boundary values for limited area models were disseminated for the first time. Links were now increasing to 64 kbps to several Member States, and a link of this speed was set up with the University of London Computer Centre to connect the Centre to JANET, the new “Joint Academic Research and Education Network” in the UK, and through this to the international Internet. Early Internet experience was mixed with some reports of delays and loss of service.
The internal network was expanded by the installation of a fully duplicated network based on fibre optics.
In December 1992, the Council approved the use of the network for electronic traffic routeing, thus enabling Member States to use the connections with the Centre for data exchange between each other.
By early 1993, more than 55,000 products were sent to the Member States each day. Internet use was increasing for acquiring research data.
Communicating the forecasts: mail and 50 baud to RMDCN 217 In December 1993 Council approved 64 kbps as the base speed of the network to be phased in from April 1994. Prior to this, individual States had to pay for the additional cost of lines above 9.8 kbps. France at its own expense upgraded its link to 128 kbps from December 1994. The first link to one of the new Co-operating States from Eastern Europe, Hungary, was ordered in late 1994. As a Co-operating State, Hungary paid the cost of this link. Direct dissemination to Hungary (9.6 kbps) and Iceland (64 kbps) started in 1995.
In April 1994 the last country (Austria) using the NTC protocol suite had started using TCP/IP. Now that all Member States were either using TCP/IP or DECnet as the transport protocol, the VAX 8350 was removed and the VAX 6210/6310 cluster was replaced by two VAX 4100 systems.
In 1995, dissemination of individual EPS forecasts started. The data volume increased to 90,000 products, totaling 675 Megabytes. A major achievement in 1996 was the design and implementation of a new distributed Fields Data Base.
Close scientific co-operation with NOAA in the USA continued to be a notable feature of the Centre’s work. EPS products, forecasts to ten days, and rainfall forecasts in support of an intercomparison project were among the products sent to National Centers for Environmental Prediction (NCEP) of NOAA in Washington. The Centre received increasing amounts of satellite data from NOAA.
In May 1994, Regional Association VI of WMO, covering the region of Europe, decided to establish a new telecommunications network, to be called the Regional Meteorological Data Communication Network (RMDCN). This was required to meet the any new GTS requirements for the 49 members of RA VI, about half of whom were ECMWF Member States or Co-operating States. A Steering Group considered the legal and administrative framework, as well as technical and cost/benefit aspects. Not surprisingly, discussions with ECMWF staff showed that the use of a common network for GTS and ECMWF products could be beneficial; a saving of 20% to 40% could be achieved in the overall annual cost of about US$4 million.
The delegate from France to the Council session of July 1996 — who happened to be Dominique Marbouty, later to become Head of the Operations Department and then in 2004 ECMWF Director — submitted a proposal for the integration of the GTS and the ECMWF network. Council requested its Policy Advisory Committee to consider political consequences and its Technical Advisory Committee to consider technical aspects. The Director would liaise with a RA VI Working Group on the matter.
In 1996, the range of ECMWF products on the GTS was extended to seven days in the Northern and Southern Hemispheres, and to five days in the tropics.
218 Chapter 17By mid-1997 145,000 products — a data volume of 1.4 Gbytes - were being disseminated each day. New products and parameters continued to be added, for example temperatures and wetness of the soil at levels below the surface of the earth, and high-resolution wave products. Council agreed that the Centre could be accessed via the Internet, and a secure system protected by a Firewall was soon installed and functioning well. The link to JANET was upgraded to 8 Mbits per second.
A TAC sub-group on the RMDCN met in February and June 1997, and a Workshop was held in March on Managed Network Services. Almost all RA VI Members were now expressing an interest in participating in the project. A network with speeds of 128 kbps was anticipated. A detailed plan was presented to Council in July 1997. In December 1997 Council approved an exchange of letters with WMO detailing the responsibilities of the Centre and of RA VI Members. By mid-1998, the Invitation to Tender for the RMDCN had been completed, and in December 1998 a contract was signed with EQUANT NV for the provision of a Frame Relay network service using TCP/IP as the transport protocol for the RMDCN countries. Work on the new network was soon under way, with a pilot phase involving France and Portugal in early 1999, initial deployment of the network with 31 States participating, in summer 1999, and site acceptance tests running from midOctober. This was not one of the easier projects for the Centre. A multitude of connectivity and throughput problems were experienced during implementation. The initial deployment was complete and the network accepted in March 2000. Soon, most Member States were using the network for most of their meteorological communications.
ECMWF products were being provided increasingly to international scientific activities:
• “Labrador Sea Ocean Convection Experiment”;
• “Fronts and Atlantic Storm-Track Experiment” (FASTEX) which used the products to take decisions on intensive observing periods;
• “North Pacific Experiment” (NORPEX);
• Support operations on the Very Large Telescope at the European Southern Observatory in Chile;
• “Tropospheric Ozone Production about the Spring Equinox” (TOPSE) — this requiring research flights from Denver Colorado to the North Pole;
• “Middle Atmospheric Nitrogen Trend Assessment” (MANTRA) programme run by the University of Toronto, Canada;
• “Aerosol Characterization Experiment — Asia” (ACE-Asia) under the International Global Atmospheric Chemistry Programme;
Communicating the forecasts: mail and 50 baud to RMDCN 219 • “Trace and Chemical Evolution over the Pacific” (TRACE-P) experiment organised by NASA;
• “Dynamics and Chemistry of Marine Stratocumulus” (DYCOMS) project run by UCAR in the USA;
• Mesoscale Alpine Programme.
Forecasts were provided to many non-Member States for limited periods in support of environmental emergencies, for example to the States of North Africa to help with a locust plague, to Glavgidromet in Tashkent for selected locations in Uzbekistan around the Aral Sea, to the Czech Republic prior to the Co-operation Agreement coming into force at a time of major flooding, and to Pakistan for monitoring heavy monsoon rains. In early 2000, new dissemination streams were introduced in support of scientific field experiments.
As its name suggests, the National Centers for Environmental Prediction (NCEP) is the civilian weather forecast centre for the United States. A fire in the power supply of the CRAY C-90 computer at NCEP in Suitland, Maryland, USA on 27 September 1999 was fought by over-enthusiastic firefighters, who sprayed the insides of the US$30 million machine with an ordinary fire extinguisher. Computer rooms have special carbon dioxidebased fire extinguishers, but firefighters picked up an all-purpose carbonate extinguisher outside the data center. They were very thorough in applying the chemical dry carbonate; the residue of the fire extinguisher left the computer beyond repair. NCEP was now left without a functioning supercomputer. On request, the Centre gave permission to NCEP to use ECMWF ensemble forecasts for real-time medium-range (6–10 day, and “week two”) weather prediction. In addition they were used for a new “Threats Assessment” outlook developed in the USA to give warning of weather and climate events that posed a potential threat to life, property or economic interests.
During 1999, with the turn of the millennium and the Y2K problem approaching, the dissemination software was completely redesigned and rewritten to run on a Hewlett Packard High Availability System HP9000.
This brought to an end the Centre’s use of the system based on VAX computers and DECnet, begun in 1986. All States were now using TCP/IP as the transport protocol for dissemination, communications with the Centre for use of the computer system, and all related use of the network.
The ECMWF web site was being increasingly used. Extensive revisions to the structure, content and style of the site were implemented in December 1999 and again in early 2002, when the public and Member State sites were merged, with controlled access to the parts of the site
220 Chapter 17restricted to authorized users. Researchers were allowed immediate and free access to very extensive ECMWF data bases.
The RMDCN network was being connected to more and more States. The Russian Federation signed its Accession Agreement in 2001. EUMETSAT connected in the same year. At the end of the year, experts from China, Kenya, Russia and France, under a WMO umbrella, were analyzing the possibility of connecting the RMDCN to Nairobi, Dakar, Algiers, Cairo, Jeddah, New Delhi and Beijing. Tokyo was interested also. The system was seen as a significant opportunity to improve the global meteorological telecommunications network. The Centre acted as focal point for technical, financial and administrative matters. WMO regarded the system as “an outstanding success in its functionality and reliability, and also in its cost structure and cost development... it is exemplary”. China became a member of the RMDCN in 2002. Also South America started work on an RMDCN for that region; the Centre was represented at a meeting in November 2001, and helped in the preparation of the Specifications of Requirements and the contract.
By 2002, more than one million products were being disseminated each day, with a volume of 12 Gbytes, almost doubling the number and volume of the previous year.
By 2004, with the exception of the connection with the UK Met Office all connections to the Member States and Co-operating States were via the RMDCN, with speeds ranged from 64 kbps to 1.5 Mbps. The Centre had a highly resilient connection to the RMDCN via two 34 Mbps links. A 2 Mbps private leased circuit was used for the connection to the UK Met Office. The network base speed became 768 kbps for communication between the Centre and the Member States in early 2005.
As of early 2005, 40 countries were connected to the RMDCN, as well as EUMETSAT, and 2,650,000 products were being disseminated daily. The Centre’s Internet connection was a 60 Mbps link to JANET. Its connections to the rest of the Internet now included a link to the Geant network, which provided a high-speed backbone between most research networks within Europe and to the USA Chapter 18