«Medium-Range Weather Prediction Austin Woods Medium-Range Weather Prediction The European Approach The story of the European Centre for Medium-Range ...»
Tim Stockdale, now with Burridge as Director, worked as a consultant on a joint seasonal prediction project between ECMWF, the Max-PlanckInstitut für Meteorologie (MPI) in Hamburg — where Bengtsson now was — and KNMI in the Netherlands. Stockdale spent some months in Hamburg in 1992. The joint project was able to complete the development of the Hamburg Ocean Primitive Equation (HOPE) model and couple it to the ECMWF atmospheric model, thus giving the Centre its first coupled oceanatmosphere model, albeit a model strictly for research.
In May 1992, Burridge reported to Council that ongoing research at the Centre showed that “in the tropics, interannual variations in the sea surface temperature impart a high degree of predictability to the atmospheric fields”. Further “an ocean model developed at MPI over a number of years has been coupled to a T21 version of the Centre’s model”.
A meeting held at the Centre in December 1992 considered a “scientific assessment of the scientific prospects for monthly and seasonal forecasting”. The document, prepared by Tim Palmer, by now Head of the Section, and Prof David Anderson from Oxford University, was published in the Quarterly Journal of the Royal Meteorological Society in 1994. Evidence presented included the theoretical basis for seasonal prediction, a review of the results of experiments of various kinds that had been carried out by groups in Europe and the USA, and the modelling needs including those for assimilation of data. A careful distinction was made between the potential
136 Chapter 11for such prediction in the tropics — where it was expected that useful skill could be achieved — and areas such as Europe, where the potential for seasonal forecasts was limited. The effects of coupling between the tropical oceans and the atmosphere were greatest in the tropics. North and south of the tropics, including over the Atlantic and Europe, there are large-scale energy transformations, for example at frontal zones, which are much less affected by the tropical ocean temperatures — though even here, a strong El Niño can extend its influence.
At the request of the Council, a Workshop on seasonal forecasting chaired by Jean-Claude André of Météo France was held at the Centre in February
1994. Its aim was to prepare a feasibility study, including costing, of an experimental programme of seasonal forecasting, and to analyse the economic benefit of seasonal forecasting with the help of potential users of the forecasts. Council discussed the Report of the Workshop in June, including a proposal for a Plan of Action. There was wide support among delegates for the Centre to have an experimental programme of seasonal prediction, although the UK delegate expressed a preference for operational prediction to be done by a network of National Meteorological Services.
Meanwhile, in Australia, the Bureau of Meteorology had developed a comprehensive, robust ocean data assimilation system based on the ECMWF Optimum Interpolation system used for the atmosphere. The system had been running since 1988. In late 1994, Stockdale visited the Bureau for some months, where he installed the Centre’s coupled system on the Bureau’s computers. He experimented with the ocean data assimilation system. Stockdale took back to the Centre the software for this system, giving the Centre now all the necessary ingredients to carry out coupled seasonal forecasts.
In December 1994, the Council finally approved “an experimental programme of seasonal prediction with a view to improving medium-range forecasts” — exactly ten years after Bengtsson had first raised the issue. The reference to “improving medium-range forecasts” gave the assurance that the programme would lie legitimately within the ECMWF core programme.
David Anderson was recruited from Oxford University in early 1995 to head the four scientists of the Seasonal Forecasting Group at the Centre.
Steady advances were made in the following years with help of funding from the EU. One of the early projects was PROVOST, a European Project on “Prediction of Climate Variations on Seasonal to Interannual Timescales”, run in 1995–98, and coordinated by the Centre. This quantified potential predictability using several atmospheric General Circulation Seasonal prediction 137 Models (GCMs) to represent the response of the atmosphere to anomalies of the Sea Surface Temperatures (SSTs). Observed SSTs were used in the experiments, not those predicted by the coupled ocean-atmosphere model developed by the Seasonal Forecasting Group at the Centre.
Research continued at a rapid pace. The coupled system was assembled using the “Ocean Atmosphere Sea Ice Soil” (OASIS) coupler from the Centre Européen de Recherche et de Formation Avancée en Calcul Scientifique, CERFACS, in France. Using the coupler facilitated modelling the exchanges of momentum, heat and freshwater fluxes — precipitation minus evaporation — between atmosphere and ocean. It is these exchanges that drive the ocean circulation. The model ocean passed the changed SSTs back to the atmosphere; thus the model was now predicting the SSTs.
The seasonal forecasts were run to 200 days ahead three days per week, in delayed mode until early 1997, in real time thereafter. By early 1997 a significant El Niño was being predicted by the ECMWF system. The forecast was for a strong El Niño in mid-year — see the figures. The Centre’s team felt nervous; other models were not showing this. In the coming weeks, observations were showing signs of significant warming of the oceans. Was this indeed the beginnings of a major El Niño? — some still had doubts.
In fact the ECMWF model was making an accurate prediction of a major El Niño — that of 1997/98. These were still clearly research forecasts.
However Council for humanitarian reasons agreed to make them available to the world meteorological community through the World Wide Web.
National Meteorological Services in Africa, Asia and South America were being called on to provide their best information on the likely effects of the El Niño in their countries. The Council decision was made in early December 1997. The Centre’s team was proud to have been able to complete the difficult technical steps required, so that the forecast products were on the ECMWF website before Christmas.
By 1998, the team was confident that the model was now in overall good shape. It had systematic model errors, but these were generally known. The team was able to hand over the now (almost) robust seasonal prediction system to the Operations Department. The first quasi-operational experiences were good. The model gave a good prediction of significantly above-average rainfall for the 1998 winter and spring in Australia. The termination of the 1998/99 La Niña event by a rapid warming of the Pacific ocean surface was predicted better by the ECMWF system than by others.
138 Chapter 11
The black line shows the observed evolution of the sea surface temperature anomaly in the Niño-3 area starting in October 1997. The coloured lines show the ensemble forecast to six-months ahead, starting at three-month intervals, made with the Centre’s first real-time seasonal forecast system. The plot was produced by CLIVAR based on data from ECMWF.
Another Workshop on seasonal forecasting was held in early 1999. Agrometeorology, insurance, medicine and weather-derivatives financial sector were all represented, reflecting the increasing worldwide interest in (and marketing of!) seasonal forecasts. There was an increasing demand that ECMWF forecasts be made more widely available, not only for research, but for humanitarian and commercial interests as well. In June 1999, the Council agreed to continue to make a selection of the seasonal forecasts freely available on the ECMWF website, and asked its Policy Advisory Committee to look into commercialisation issues. In November 2000, the Council agreed to make these forecasts available commercially.
By early 2000, prediction of the number of hurricanes in the Atlantic and tropical cyclones in the Pacific, and forecasts of the year-to-year displacement of the cyclone genesis region in the Pacific, were showing promise.
Further, work began on making predictions to a month ahead, intermediate between the medium-range and seasonal time-scales.
European interest in the scientific and technical challenge of seasonal prediction and coupled ocean-atmosphere modelling and analysis was not confined to the Centre. Such models were being developed at the UK Met Office and Météo France. Other groups involved in research in the field included those at Electricité de France, at KNMI in the Netherlands and at
140 Chapter 11CERFACS. As usual there was excellent collaboration between the research staff at the Centre and those in the national services and the institutes.
By end 2001, a European multi-model “ensemble” of seven coupled models of the EU-funded Project entitled “Development of a European Multi-model Ensemble System for Seasonal to Interannual Prediction”, the DEMETER Project, was entering the production phase, with six of the models installed at the Centre. Named after Demeter, the goddess of fertility in ancient Greece, the object of the Project was to develop a well-validated European coupled multi-model ensemble forecast system for reliable seasonal to interannual prediction, including establishing its practical utility, particularly to the agriculture and health sectors. The Centre coordinated the project.
Research was advancing satisfactorily, making good use of the 40-year Reanalysis data — see Chapter 14.
Another project, “Enhanced Ocean Data Assimilation and Climate Prediction” (ENACT), funded by the European Union in 2002–04, aimed to enhance European capabilities in the fields of global ocean data assimilation and analysis systems associated with climate modelling and prediction.
ENACT emerged from another project for “Developing Use of Altimetry for Climate Studies” (DUACS) — the name is self-explanatory.
The possibility was developing to take advantage of the different models in DEMETER to make real-time operational forecasts, by a so-called “multi-model” approach. The Centre worked with the UK Met Office to install their coupled model on the ECMWF computer system, integrating it with the ECMWF model, with the intention to produce a common set of forecast products — rainfall, temperature and so on.
In May 2002, the Director exchanged letters with Peter Ewins, Chief Executive of the Met Office, formalising the joint research and operational activity of the Centre and the Met Office in this area. All Member States of the Centre were to have full visibility of and full rights of use of the work and its results.
Soon after, the Director exchanged letters with Jean-Pierre Beysson, Director of Météo France, by which Météo France would join in multi-model seasonal forecasting. By late 2004 all three models were running at the Centre.
The data were being archived in the ECMWF MARS archival system.
In 2003 planning for two more EU projects was developing. Both were
funded in 2004:
• ENSEMBLES continued the work of DEMETER to develop multimodel ensemble forecasts and to link climate forecasts to applications in agronomy, health, hydrology, energy and more. In addition, ENSEMBLES would test the skill of multi-model ensembles against other Seasonal prediction 141 techniques for representing model uncertainty, such as stochastic physics and perturbed parameters. Unlike DEMETER, ENSEMBLES integrations would assess the decadal as well as seasonal predictability of climate. The overall aim of ENSEMBLES was to develop a unified European ensemble system for prediction of climate across a range of timescales, from seasons to decades and beyond.
• The “Marine Environment and Security for the European Area” (MERSEA) project aimed to develop a European system for operational monitoring and forecasting on global and regional scales of the ocean physics, biogeochemistry and ecosystems. The prediction timescales of interest extended from days to months. The integrated system would form the ocean component of the future Global Monitoring for Environment and Security (GMES) system.
We have seen that development of the model and data assimilation system for medium-range forecasting complemented the valuable work on short-range prediction in the Member States, with much two-way exchange of ideas, methods and research. In the same way, there was synergy from the joint efforts in seasonal prediction work at the Centre, and the work on climate and climate change, global warming and similar in institutions throughout Europe. Further, there was synergy between the research projects funded by the EU, and the Centre’s requirements for useful operational seasonal prediction. Accelerating research in Europe and indeed throughout the world in these important and related areas was the consequence.
In February 1953 the dikes protecting the Netherlands were breached by the onslaught of hurricane-force northwesterly winds on top of exceptionally high spring tides. The Dutch Surge Warning Service, which had been established after a destructive surge in January 1916, issued forecasts of dangerously high water levels several hours before they occurred. However the floodwaters came in the night, and the warning came too late to allow evacuation by the limited emergency services. The lives of 1,835 people were lost, almost 200,000 hectares of land flooded, 3,000 homes and 300 farms destroyed, and 47,000 heads of cattle drowned; it was the Netherlands’ worst disaster for 300 years. In eastern England, almost 100,000 hectares were flooded and 307 people died in this storm on that terrible night. Flooding caused by storm surges was nothing new to the Netherlands, but this time, the nation and the world were stunned by the extent of the disaster.
The tapestry hanging in the ECMWF Conference Room was a gift from the Netherlands. It shows the 500 hPa circulation on 1 February 1953. The storm over the North Sea was responsible for widespread devastation.The tapestry serves perhaps to remind delegates of the importance of timely and accurate medium-range forecasts.