A Vision for Numerical Weather Prediction in Europe





Many parts of the world have recently experienced high impact meteorological events which have disrupted national services, damaged or destroyed infrastructure and resources, and have resulted in the loss of human life.

·         In 2001 over 1000 people died in Algeria as a result of mudslides and flooding caused by torrential downpours. Over 100mm of rain fell in a few hours over central Algiers, the equivalent to what is normally received in a month.

  • The flooding in Prague (Czech Republic, 2002) led to 200,000 people being evacuated from their homes, with 90 people killed across central and eastern Europe.
  • In Arles (France, 2003) 30,000 people were evacuated as a result of severe flooding, seven people loosing their lives.
  • The summer of 2003 was the warmest on record in many parts of Western Europe. An estimated 20,000 people including at least 10,000 in France, and some 2,000 each in Italy and the UK died as a result of physiological stress brought on by an extreme heat wave during the hottest part of the summer.
  • 140 million people were affected by flooding when over 60% of Bangladesh was flooded in 2004. 600 people died with over 1 million acres of crops destroyed.


Changes in the future climate as a consequence of anthropogenic emission of greenhouse gases are predicted to lead to a warmer world. Not only will there be significant impacts to our environment due to sustained higher temperatures, but extreme heat waves such as that of 2003 may become more commonplace, and a more vigorous hydrological cycle is likely to result in an increased number of extreme, high impact phenomena such as droughts, fluvial flooding, and severe storm events. The impact on society could be great with the disruption of energy production and communication networks weakening national security; destruction of crops could lead to famine and warfare; interruption to water supplies could lead to deterioration in human health; overall the loss of human life will increase.


Today’s Numerical Weather Prediction (NWP) systems are capable of predicting many of these events but not with the required accuracy or at sufficient lead-time to enable all possible mitigating action to be taken to protect life and property; nor are the civil decision making tools in place to reduce the impact of these events.


There is an expectation of further significant improvements in NWP systems over the next decade. Higher resolution models will provide more detail for small scale severe weather events. Continued improvements to the observing network including new satellite instruments, more data from aircraft, improved radar systems and the use of targeted observations will provide better initial conditions which are critical for short and medium range forecasting.  However, because of the inevitable uncertainty in initial conditions and also the uncertainty in forecast models, the further deployment and use of ensembles is essential for the early warning and quantification of the risk of severe weather.


There is a growing consensus that the needs of the global community can best be met through the provision of a global ensemble forecasting system, together with re-locatable regional limited area ensemble systems, supported by targeted observations.


These needs have been recognised through the establishment of THORPEX, a ten-year long international research programme to accelerate improvements in the accuracy of 1 to 14-day weather forecasts for the benefit of society and the economy, as a part of the WMO World Weather Research Programme. Such a system would allow governments and trans-national organisations such as the UN to take intelligent decisions based on the minimisation of societal exposure to weather-related risk, generate optimal mitigation strategies, and maximise the efficiency of the humanitarian effort.




Today, Europe enjoys a diverse NWP capability covering a very broad spectrum. Different model formulations, temporal and spatial ranges and resolutions are deployed to meet a range of customer requirements and funding bases. Mechanisms for research, development and operational implementation also vary. To name but a few, there are modelling communities such as ALADIN, COSMO and HIRLAM. There are Centres with a particular WMO or customer focus, such as the Regional Specialised Meteorological Centres at Toulouse, Offenbach, Rome, Exeter and Reading covering geographical specialisation, atmospheric transport and global medium-range products, and the London World Area Forecast Centre providing global aviation products.


European scientists in many of the groups and centres have been at the forefront of this international effort in increasing the quality of short-range regional predictions and of global predictions on the short-, medium- and longer ranges. As a result, there are Centres in Europe which have developed scientific reputations that are second to none and whose forecast accuracies are recognised as the best in the world.


The continuing success of the European effort in NWP is based on our ability to increase European capability whilst at the same time maintaining a diversity of modelling approaches. This has been achieved through a long established balance between the differing national requirements, financial contributions and scientific effort. The multi-model seasonal ensemble run at ECMWF is an example of where diversity has led to increased capability to the benefit of Europe as a whole. The friendly competition and collaboration between Met Services and ECMWF has resulted in the proving of ideas in systems that use fundamentally different numerical techniques, thereby demonstrating their scientific validity. These differing approaches have not hindered collaboration, but have stimulated discussion and collaboration through EUMETSAT SAFs, EU funded projects and EUMETNET SRNWP.


There is now the opportunity to build on this diversity, maximising the value of European meteorology with the capability at hand. Concerted cooperation within Europe would provide society with the information it will need to ensure the conservation of human life. This will allow Europe to actively respond to the findings of THORPEX, enabling our expertise to benefit the rest of the world.




The vision is based on the following realities and presumptions:


  1. There is the continued development of a limited number of independently formulated state of the art NWP systems with that number potentially reducing. National Meteorological and Hydrological Services (NMHSs) will still function to provide service delivery to its designated geographical area of responsibility, including the issuing of severe weather warnings to its government and population
  2. The resolution of existing global and regional models will increase over the next twenty years to the extent that global models will supplant regional models and regional models will supplant country scale models for all but the most detailed customer requirements.
  3. Uncertainties in model formulation and the initial state of the atmosphere result in deterministic forecasts that are not always correct.  On both global and regional scales, there is a need to quantify the uncertainty to give a better chance of identifying a high impact, low probability event. Ensemble Prediction Systems (EPS) have been established for this purpose at a number of Centres. A Multi-model EPS could provide significant additional benefits over and above a single EPS. They have been shown to capture uncertainties associated with model formulation and the significant increased size of the super ensemble is more likely to capture low-frequency high-impact events. However, multi-model ensembles are unlikely to represent the full spectrum of model uncertainty and a more generalised framework for the formulation of weather prediction may be required, whereby specific realisations of sub-grid processes are represented by stochastic-dynamic parametrisations. Extensive numerical experimentation, probably in an operational context, will be required to assess the merits of stochastic-dynamic parametrisations within the context of multi-model and other related representations of model uncertainty.
  4. There is a growing European capability through joint activities that is focused on delivering tools to enhance the mitigation of meteorological related disasters (i.e. EURORISK).
  5. This implies the need for a much greater collaborative effort between NHMS’s both globally and within Europe who have the independent capability to maintain, develop and run either global or regional high resolution models supported by EPS and to support the interchange of lateral boundary conditions between the different systems.


A Co-ordinated Europe


In summary the vision is of a cooperative of Met Services who combine their various skills and capabilities to ensure all within Europe are provided with the best environmental forecast possible, allowing impact scenarios to provide optimum planning information for our nations.


This vision relies upon supporting technology to enable multiple, geographically separated centres to generate a timely and reliant ensemble product. In the near future virtual computing across Europe may allow the easy exchange of data and products in a timely and efficient manner. The issues of GRID technology is being investigated through the EU funded projects, which are illustrating the limitations of the technology available to us today. However, data is already being reliably exchanged through this technology.


We would expect prototype systems at today’s resolutions to be in place on the 5 year timeframe. The full vision could be realised within 10 years and be the basis of European and Global Meteorology through until 2025.







Working diagram of vision


Global NWP and Multi-Model Ensemble Prediction System


Europe has several organisations which have a global NWP capability. Such organisations would coordinate their independently formulated NWP systems to generate high resolution global NWP and supporting EPS, for the period 0-14 days. These organisations would be linked together to exchange data over a distributed GRID network to contribute to a single MM-EPS product.


Target resolutions for the initial implementation of 20km for global deterministic NWP and 50km for global EPS is achievable with projected computer systems during the next few years. In time atmosphere-ocean coupled models would be implemented.


Limited Area NWP Multi-model Ensemble Prediction System


The global solution is mirrored on a regional scale. There are several groupings and countries within Europe that have great skill in Limited Area Modelling and have significant computing power and advanced supporting infrastructures. Through coordinating this effort these centres would take products from the global centres and run regional NWP supported by regional EPS. Each organisation would be linked through a grid network to allow MM-EPS regional products to be generated. Target resolutions of 4km for regional NWP and 10km for regional EPS are suggested and again this is achievable with projected computer systems during the next few years. Taking the products from the regional and global NWP and MM-EPS, the NMHS’s would then issue forecasts and warnings to their own geographical area of responsibility.  


Centres of Excellence


Within Europe there are already several centres with a particular WMO responsibility or customer focus, such as the Regional Specialised Meteorological Centres at Toulouse, Offenbach, Rome, Exeter and Reading.  Taking this approach one step further in light of this coordinated approach, Centres of Excellence would be established having a specific role and responsibility. The management of the GRID network; the development of post processing techniques; the standardization of NWP model output; development of visualization systems; development of ensemble tools; the coordination of warnings to Europe’s NMHSs for example on the risk of forest fires, fluvial flooding, avalanches, severe storms, high seas; issuing of air quality forecasts, to mention a few would need to be managed and coordinated.


To achieve the greatest level of skill in forecasting in both the medium range and for complex severe impact systems a competent and flexible observation network is necessary.  In data sparse areas targeted observations will be required. The coordination of this targeting would need to be done globally, although the European component (perhaps covering the Atlantic region) would be managed through a Centre of Excellence.


Reliability and Functionality


With such an interdependent system the robustness of the forecasting ensemble system will need to be such that if one or more of the forecasting organisations were to fail, the integrity of the ensemble product should not adversely suffer. Redundancy within the system would therefore be required. Multiple Centres provides such redundancy. Conversely having multiple sources for the products would increase global and European forecasting resilience, mitigate meteorological risk whilst optimally utilising the global community knowledge base.



Country scale NWP


As a consequence of the above, NMHS would have a vastly superior quality of products at their disposal to aid in the prevention of loss of life and property. Resources within each organisation could be freed to allow the development of downstream products for the user community. It is likely that here will still be a need for high resolution (1-2km) non-hydrostatic NWP which will help predict small scale high impact phenomena.  These models could be run at the NMHS or at a regional centre on behalf of an NMHS.


Europe supporting the Global Community


Europe can, as a whole, realise its full potential through adopting a networked approach to NWP, ensuring diversification, whilst making the most of the technical capability at hand. Accepting the need to collaborate more closely will also enable Europe to contribute both global and regional ensembles in support of the THORPEX programme.


The development of re-locatable, high resolution ensemble models could be used to forecast high impact meteorological events not only within Europe, but wherever there was a need. Some of the severest natural disasters occur in the poorest, least developed parts of the world. Having re-locatable high resolution models together with targeted observations would ensure that all of the globe would benefit from the technological capability available within the most developed nations.


Together we’re stronger


The vision reflects the new possibilities created by advancements in science and technology. It is the duty of the National Meteorological Services to utilize these advancements.


The vision is consistent with the role of the National Meteorological Services and thus, if implemented, will not change this role. The National Meteorological Services have been established because national governments have a duty to provide their population with forecasts of high impact weather. Implementation of the vision will, through international cooperation in NWP, improve the quality of such forecasts.


Implementation of the vision will require substantial international cooperation. It is difficult to imagine that such cooperation could take place between competing firms in the private meteorological sector, but it can be realized between governmental National Meteorological Services.


For years, we have had a deterministic approach to weather forecasting, yet is has been clear to everyone that the forecasts should more suitably have been given in the form of probabilities. In fact this has led to a presentational problem with the general public which will be solved through the realisation of the vision.


Looking at the situation in Europe today, we are already on our way towards this vision, although perhaps without being fully conscious of this. The NWP situation in Europe, with its balance between cooperation and diversity, is an excellent basis for delivering a forecasting system that will benefit the whole of Europe, and the global community.