<<

HIRLAM strategy 2006-2015 Draft 20051104

Management summary

This document outlines the strategy for the HIRLAM Programme for the period 2006-2015. The purpose of the HIRLAM scientific cooperation is to provide the member institutes with a state-of-the-art short-range (<48h) numerical weather analysis and prediction (NWP) system, as the best possible means of support for their operational activities. This system should be versatile and flexible enough to cater to the changing needs of exisiting and new users and stakeholders. Awith the achievement of such an NWP system, the HIRLAM member institutes confidently expect to remain at the forefront of operational in Europe.

To achieve these aims, the HIRLAM consortium has chosen the following strategic approach: - For many user groups, weather information with very great spatial detail is of critical importance. In response to these needs, the highest priority in the HIRLAM Programme in the period 2006-2010 is to develop a high-quality operational mesoscale analysis and forecast system (horizontal resolution ~2km), in a strategic alliance with the ALADIN consortium. The availability of such a mesoscale model will improve prediction quality for local (severe) weather, and is expected to open up a whole new range of applications for HIRLAM, such as urban air quality modelling and hydrology - Public demand for information on the uncertainty of atmospheric evolution, required for the assessment of weather- related risks, is expected to continue to rise. HIRLAM will therefore continue to focus on both basic research on the mechanisms underlying predictability in short-range forecasting, and on the development of reliable and practical applications for short-range probabilistic (ensemble) forecasting. - By increasing the capability of the model to assimilate a wide range of earth-system observations, and by extending the model physics to include more aspects of the biosphere, atmospheric chemistry and hydrosphere, the HIRLAM system will be gradually transformed from an into a more integrated earth system model. - Through these increases in both detail and scope, the intention is for HIRLAM to become a much more versatile and relevant tool for a far wider user community, including e.g. air and road transport authorities, air quality modelling, the sustainable energy sector and water management. - The role of HIRLAM as a regional climate modelling tool is to be enhanced by means of intensification of scientific cooperation with the climate research community. - The HIRLAM consortium has the ambition to retain a prominent role in the field of both operational modelling and NWP expertise in Europe. The consortium will therefore strive to establish an acknowledged role of the HIRLAM system in joint European operational NWP activities (particularly in the field of multi-model short-range ), and to obtain recognition for the HIRLAM consortium as a European center of expertise on short-range NWP. The aim is to achieve this through fostering cooperations in NWP research in which HIRLAM can play to its strengths, such as its advanced knowledge in the field of data assimilation.

1

1. Introduction

HIRLAM’s prime long-term goal is to continue to provide the HIRLAM member states with a state-of-the-art operational short and very short range numerical weather prediction system. This system should be at least competitive with alternative forecast models in terms of measurable forecast quality. The main application for the HIRLAM system is the production of general weather forecasts, with particular emphasis on its use for the detection and forecasting of severe weather. The HIRLAM system forms the basis of a very wide range of national operational applications, such as oceanographic and storm surge forecasting, road condition predictions, aviation, hydrological forecasting etc. Other applications involve regional climate modelling and use of the model as a tool in atmospheric research studies.

In accordance with the findings of the 2005 review of HIRLAM, and in preparation of the MoU for the HIRLAM-A programme, it was decided to formulate a scientific strategy for HIRLAM for the next decade. The document below outlines this strategy for the period 2006-2015.

2. General developments foreseen for the period 2006-2015

The role and scope of the HIRLAM system is gradually changing, as it adapts to the changing needs of its end users and stakeholders. In the past, HIRLAM has been primarily an intermediary tool in support of the forecasters involved in operational production; its main customers were essentially the forecasters within the HIRLAM institutes. More and more, the model itself is becoming the main production tool, providing output directly to end users. As the model resolution increases, HIRLAM is moving into the nowcasting range, opening up a whole new range of applications (and user communities) for the model. Also, by extending the model with coupling mechanisms to e.g. the biosphere and chemistry, HIRLAM is gradually transforming from an atmospheric model into a more integrated earth-system model. Through this increase in both detail and scope, the model will become a much more versatile and relevant tool for a far wider user community, including e.g. air and road transport authorities, the energy sector, air quality modelling and hydrology.

Operational meteorology within Europe is increasingly becoming coordinated at a European, rather than a national, level. This trend holds for limited area NWP modelling as well. Coordination at a European level does not necessarily imply centralization; in view of the networking possibilities provided by ICT and the need for a certain level of model diversity (e.g. to cater for the demands of short-range ensemble forecasting), limited area NWP is likely to become the province of a virtual network of operational partners. This entails both increased cooperation and division of labour between NMS’s, and specialization of individual services or consortia, forming centres of expertise. HIRLAM needs to define its future position in this context, and make strategic choices for partnerships accordingly.

The HIRLAM consortium has the ambition to retain a prominent role in the field of both operational modelling and NWP expertise in Europe. Thus, the HIRLAM programme should not just ensure that the HIRLAM system is the best possible tool to meet the operational needs of the participating institutes, and that it is competitive with alternative models, which have been its primary aims so far; the consortium should also strive to obtain an acknowledged role of the HIRLAM system in joint European operational NWP activities, and to achieve recognition for the HIRLAM consortium as a European centre of expertise on short-range NWP. As the 2005 review of HIRLAM has stressed, HIRLAM’s international profile could be significantly enhanced by acting more as a single entity on the operational, as well as the research, level. A joint operational synoptic analysis or forecast production system would not only imply a more efficient use of financial and computer resources, but could also strengthen the visibility of HIRLAM with respect to other LAM consortia, and provide

2 HIRLAM with an operational role, for example within a European cooperation in probabilistic forecasting. In such a way, and by concentrating on its strengths in NWP research, HIRLAM can maintain a definite identity of its own, even when engaging in a close full-code cooperation such as is envisaged with the ALADIN consortium.

2.1 Synoptic scale modelling: Global atmospheric models are steadily increasing in quality and resolution, and they are expected to be able to supplant synoptic-scale limited area models by the end of the period. Synoptic scale limited area forecast systems will be phased out gradually by a new generation of operational mesoscale models for the very short range on one hand, and by the global models on the other. With the expected advances in computational power and model physics, ECMWF estimates to achieve a of ca. 16km horizontal resolution (the present resolution of LAM’s) by 2012. Until that time, limited area models on synoptic scales are still expected to have added value and to be required for the operational purposes of NMS’s. Furthermore, even if and when ECMWF has taken over the role and responsibility for short range synoptic NWP, there will most likely remain a need for running a nesting LAM system to provide optimal boundary conditions to higher resolution local models. This outer model will have a resolution intermediate between the ECMWF model and the local HIRLAM model, and data assimilation and physics as consistent as possible with the local model. In addition, there is an increasing demand for probabilistic synoptic-scale short-range forecast products allowing for the assessment of atmospheric predictability, particularly in situations of high-impact weather. Since the cost of ensemble prediction is much higher than the cost of a single deterministic forecast, the need for synoptic scale LAM modelling for the purpose of ensemble prediction is expected to remain for a long time.

By increasing the capability of the model to assimilate a wide range of high-resolution earth-system observations, and by extending the model physics to include more aspects of e.g. the biosphere and hydrosphere, the model is expected to become more generally useful as an earth-system analysis and forecasting tool. Developments in the Formaterat: Teckenfärg: Blå field of synoptic scale modelling at e.g. ECMWF will be carefully monitored and profited from whereever possible, hereby allowing the HIRLAM researchers to concentrate as much as possible on mesoscale modelling. Intensification of scientific cooperation with the climate research community may enable the role of the HIRLAM system as a regional climate modelling tool to be strengthened.

2.2 Mesoscale modelling: The first half of the coming decade will see the coming of age of non-hydrostatic mesoscale models, which are expected to become the norm in operational short range and very short range (<24h) weather forecasting. The high resolutions attainable by these models (order of 1km) will create new possibilities for HIRLAM in the field of applications such as urban air quality modelling and hydrology. Mesoscale models also hold great potential for nowcasting applications which up to now often use a wide variety of both observational and model information. In order to make HIRLAM optimally suitable for such purposes, the specific requirements of nowcasting have to be taken into account, in particular the very strong delivery time restrictions; these may lead to a different approach to model initialisation and data assimilation. In general, the introduction of operational mesoscale models will presumably require the development of new strategies within limited area modelling for: • nesting and the optimal treatment of boundary conditions (should the external model be non-hydrostatic and contain similar physics for optimal results, or not?) • initialisation and data assimilation (as the model resolution becomes higher than that of the observational network, and a high frequency of model updating is required; also the interaction between data assimilation and nesting is likely to become a serious issue) • interpretation, verification and postprocessing (increasingly in probabilistic terms, and adapting the present HIRLAM postprocessing system to allow direct use of the model for typical nowcasting applications)

3 After a mesoscale HIRLAM model has been readied for operational use, the model is expected to be enhanced and extended steadily. Improving the description of surface processes will receive much attention, and introduction of coupling with chemical modules is expected to become more common. By the end of the period, a gradual fusion may have taken place between atmospheric and air quality models. For the next few years, for HIRLAM there will be a real difference between the synoptic and mesoscale models. However, this is expected to disappear: it is envisaged that by 2010 HIRLAM will have a single model code, suitable for use on both synoptic and mesoscales, with switches to apply appropriate physics (and possibly assimilation) options, such as is the case for e.g. the Met Office UM already.

2.3 Probabilistic forecasting: Public demand for information on the uncertainty of atmospheric evolution, required for the assessment of weather-related risks, is expected to continue to rise for all prediction time scales from the very short range to seasonal forecasting. Also, there is an increasing demand for applications aiding users in decision making based on probabilistic information, e.g. in the form of cost-loss analyses. Various techniques for estimating atmospheric predictability in short-range forecasting have been tried out, but so far with limited success. In the next decade, much fundamental research needs to be devoted to understanding the mechanisms underlying predictability on these short time scales, and to the development of feasible methodologies to derive reliable probabilistic information on the evolution of the atmosphere. This may involve a gradual merging of the methods used for data assimilation and probabilistic forecasting. Secondly, practical alternatives to the present standard EPS method are to be explored. In this context, strong coordination and collaboration at a European level appears necessary to avoid duplication of efforts. A third area of research concerns the way in which probabilistic forecasts on synoptic scales can be used for the interpretation of deterministic mesoscale forecasts. Short-range EPS forecasts can and should be exploited in the validation and verification of detailed prediction models. Finally, the implementation of short-range ensemble-type forecasting systems will require the development of flexible tools for post-processing and visualization of this information.

3. HIRLAM goals and strategy

The main ambitions of the HIRLAM consortium members are to optimally serve the changing operational needs of their various user communities, and to remain at the forefront of operational short-and very short range NWP expertise in Europe. In order to fulfil these ambitions, HIRLAM has set itself the following goals for the period 2006-2015:

• Develop and improve the HIRLAM system in such a way that it is the best possible tool available for the operational activities of the HIRLAM institutes: o Make the HIRLAM system into a versatile instrument geared to cater for the needs of a wide range of end users, including new user groups such as the urban air quality and hydrological communities. o Continue to deliver and improve a high-quality synoptic-scale short-range analysis and forecasting system, suitable for the operational activities of the member states, in particular for the detection and forecasting of severe weather. The quality of the system, as measured by routine verification, should be at least as good as from other available global and regional NWP models. o Establish, deliver and improve a high-quality operational mesoscale analysis and forecasting capability: a complete NWP system including data assimilation, initialization, a forecast model based on the ALADIN dynamics and a combination of meso-NH and HIRLAM physics, a postprocessing system capable of producing direct forecasts of a comprehensive range of mesoscale parameters, and a verification system.

4 o Extend the deterministic HIRLAM forecast system with an operationally feasible, reliable short- range probabilistic forecast system, which can be used to provide accurate estimates on the level of uncertainty of atmospheric predictions. • Ensure that the HIRLAM system plays a prominent part in operational short-range activities coordinated at a European level (e.g. by becoming one of the elements of an operational European regional multi- model ensemble forecast system, or by playing a role in the European contribution to THORPEX). • Ensure that the HIRLAM consortium is recognized as a centre of excellence on NWP within Europe.

Strategy 2006-2015: The strategy to achieve these goals entails: • Development of a operationally feasible mesoscale analysis and forecast system (horizontal resolution ~2km), based on the AROME model (ALADIN dynamics, Meso-NH or HIRLAM physics) and IFS coding standards; this is to be carried out in collaboration with the ALADIN consortium, in the period 2006-2009. Subsequent improvement of this model, extending it towards a more integrated earth system model (e.g. by coupling with chemical modules), and implementation in applications such as air quality modelling and hydrology. Increase of resolution to ~1km towards the end of the period. • Continued development of the synoptic scale HIRLAM analysis and forecast system (10km horizontal resolution). After the mesoscale model has been set up and validated, it is planned to integrate the synoptic and mesoscale models into a single (IFS-based) model code suitable for use at both scales. This integration should be achieved around 2010 at the latest. Until that moment, maintenance and continued improvement of the present synoptic-scale model code will remain of high operational priority. In order to avoid having to maintain two systems for a long time and to make optimal use of improvements made in the mesoscale and the ECMWF models in the synoptic model, it is advisable to transfer the synoptic scale forecasting system to the IFS as early as possible; however, this should not be begun before the basic mesoscale system has been set up and tested properly. • The synoptical-scale HIRLAM may be surpassed in quality and operational usefulness by ECMWF sometime during the second half of the period, but even so it may still be required to provide boundary conditions for the mesoscale model. In the second half of the period, it will therefore need to be investigated whether and when a transition from HIRLAM to ECMWF as the operational short-range model for synoptic scales should take place, and what the optimal nesting strategy for the mesoscale model should then become. • To ensure the best possible quality assurance for the HIRLAM Reference System, HIRLAM should continue the practice of running of the Reference System operationally in at least one HIRLAM country, and preferably at several institutes. All institutes should strive to have their operational systems remain as close to the Reference System as possible. Continuous quality control and measurable improvement of the model will be achieved through systematic validation and comparative verification against other models. This is a priority task within the programme. • Continued development of 3- and 4D-VAR data assimilation, for a wide range of high-resolution observations, and the application of variational techniques to real-time assessment of observational impact (targeting; contribution to THORPEX). 4D-VAR is expected to become the norm for operational synoptic scale HIRLAM within the first few years. The rapid update cycling necessary for fine-scale models may require the use of 3D-VAR for a period longer than that; ultimately, however, when computer resources allow, 4D-VAR is believed to be the optimal assimilation method for operational mesoscale models as well. Application of HIRLAM for nowcasting purposes, demanding very fast delivery times, may require a simpler, more pragmatic approach. Special attention therefore needs to be paid to fast methods of model initialization. • Continued development of methodologies for probabilistic (ensemble) forecasting suitable for the short range (24-72h) throughout the period. HIRLAM should promote research in this field to be carried out in a broader European context, and should establish a clear role for itself in European operational activities in

5 multi-model short-range ensemble systems. Development of EPS-type techniques for the very short range (<24h) is still very much a research issue, and unlikely to lead to results of operational interest before the end of the decade. • Increase the accuracy and usefulness of the HIRLAM model for a wide range of users, by improving and extending the coupling mechanisms of the model atmosphere to the ocean, biosphere, surface and atmospheric chemistry. • Ensure that the HIRLAM system is optimally suited to meet the operational interests of its end users, by means of a regular consultation of the users by the individual HIRLAM institutes and incorporation of their needs in the HIRLAM research plans. • Set up a task force to investigate the potential and consequences of increased cooperation in certain operational activities (e.g. 4D-VAR, ensemble forecasting) within the first three years. Such a form of operational collaboration would free up both computational and manpower resources; central production of a 4D-VAR analysis, for example, would allow efforts to access and assimilate a wide variety of data types to concentrate on a single system. A joint operational assimilation or forecast system could play a much more visible role as a HIRLAM contribution to European programmes such as SAF’s or THORPEX. Decisions in this area depend on strategic choices to be made in the field of data assimilation and short- range ensemble forecasting, but also on national interests and limitations of the participating countries. • Foster European cooperation in short-range numerical weather prediction, and be open to new partners. The most direct example of this is the full-code cooperation with ALADIN. This collaboration will start as a joint work programme, to be carried out by two separate teams. A closer integration of the two consortia is expected to evolve from this. In the field of short-range ensemble forecasting, it is highly desirable to establish a European cooperation with other NMS’s or consortia, in the form of the production of a joint ensemble and exchange of model data. EUMETNET is the most obvious forum in which to discuss and create such a quasi-operational collaboration. EUMETNET/SRNWP can be used as a platform for exchange of information on research, and for joint development activities focussing on pragmatic experimentation with short-range ensemble construction, validation and product development. Research within HIRLAM itself should focus on more fundamental studies on how to determine uncertainty on short time scales, and on evaluating the potential of alternative methods such as ensemble Kalman filtering or weak- constraint 4D-VAR.

4. Research issues

4.1 Observations and data assimilation In the past few years, much effort has been devoted to the development of variational assimilation techniques within HIRLAM, and to obtaining a huge increase in remote sensing data available for data assimilation. In the first few years of the period, the fruits of these efforts will have to be harvested. The main target initially is to implement and extend the use of 4D-VAR as the operational data assimilation system on synoptic scales. Subsequently, it can be made more sophisticated (improved physics in adjoint model, error structure functions etc.). On the mesoscale, for practical purposes it may be necessary to maintain 3D-VAR for some time as a rapid update cycling system for very short range forecasting purposes; one could also envisage, for example, a combination of 4D-VAR with a last time-critical update from 3D-VAR. 3D-VAR in mesoscale models could be enhanced further with a more advanced use of observations and refinement of the balance constraints for higher resolution. In time, increasing computational resources may permit the use of 4D-VAR in local models with a , and thereby the full exploitation of the advantages that 4D-VAR offers for the assimilation of e.g. radar data. The quality of the model surface analysis is of crucial importance to mesoscale models. A reliable high-resolution surface analysis is critically dependent on the availability of high-quality land-use datasets and the way in which these data are used in the atmospheric model; this is a high-priority research area. For surface data assimilation,

6 advanced new data assimilation methods have been developed within e.g. the EU-project ELDAS. In the first few years of the period, these should be tested and implemented, in order to replace the present OI approach. Use of results from data assimilation of internationally available data sources should be investigated, e.g SST and ice cover from observations and ocean models. Both the atmosphere and surface assimilation schemes will be extended to include more parameters relevant to the biosphere and hydrosphere.

HIRLAM Member states will continue to be actively involved in the preparation of new satellite products for operational use, mainly in the context of EUMETSAT SAF’s. One of the interesting challenges in the assimilation of these remote sensing data will be to derive a beneficial, lasting impact from cloud and moisture data on model forecasts. Many new types of satellite data will become available during the next decade (e.g. METOP/EPS, NPOESS, ATSM, ADM/Aeolus, SMOS, MTG,…). Most of these appear to be useful primarily for improving the description of large- scale structures in the atmosphere, and will be assimilated in global models such as ECMWF. This raises a question on assimilation strategy: is the best approach for HIRLAM to continue to make its own synoptic-scale analysis using as many observations as possible (transmission speeds permitting), or to make use of the availability and high quality of the ECMWF analysis for larger scales in some manner yet to be defined, and supplement this with a HIRLAM analysis of high-resolution fast-delivery “local” data (e.g. radar, profilers, GPS)? This question needs to be addressed in the first years of the period.

Apart from operational data assimilation, the use of 3- and 4D-VAR can also be extended to the assessment of the impact of real-time observations (observational targeting) and of new observational systems (OSE’s and OSSE’s). In the first area, HIRLAM should decide which role it wishes to play in the context of the Thorpex programme. The second area could be of great use in the context of SAF developments.

4.2 Modelling The highest priority for the period 2006-2009 is to set up a non-hydrostatic mesoscale model, and make it feasible for operational use. Physics schemes suitable for this non-hydrostatic model are to be further developed and improved. An important problem to be solved will be how to optimally couple this mesoscale model to the synoptic scales: the treatment of boundary conditions and the impact of nesting within a model using essentially different physics will be major research issues. Another topic for investigation is the strategy to be adopted for deriving the model initial state: what is the optimal approach for obtaining an accurate, computationally feasible rapid-update analysis system using local data? Major applications for mesoscale modelling are found in e.g. nowcasting, wind energy prediction, hydrology, urban meteorology and air quality modelling. The requirements from these applications on the forecast model itself, and on its postprocessing products, will have to be taken into account in the mesoscale research objectives, from roughly 2007 onwards.

Important gains in forecast quality are still expected from advances in model dynamics and the treatment of boundaries. In the area of physics, significant improvements are anticipated from advances in the treatment of the surface, the boundary layer physics and cloud and convection parametrizations. Specific areas where progress is sought are the initialisation and prediction of low clouds, the description of the stable boundary layer, the modelling of strong convection, the description of surface conditions, and the impact of vegetation and snow cover. The quality of underlying land-use datasets, and the interpretation of these data in the surface description is probably a topic of critical importance, requiring much effort. In order to spend limited manpower optimally, it is intended to concentrate on physics developments suitable for the mesoscale, and test developments that have proven successful there also on synoptic scales.

The trend towards integrated earth-system modelling is expected to extend from global to regional atmospheric models. In particular, improvement of the description of the interaction between atmosphere, biosphere and

7 hydrosphere is expected to become a research issue of increasing prominence. This will have consequences for the modelling of physical quantities relevant to e.g. soil properties, the coupling to hydrological soil/runoff models, or atmospheric chemistry.

4.3 Predictability The EPS methodology is well established as a suitable approach to derive medium-range forecasts of predictability. The great issue now is how to successfully extend this technique (or similar alternatives, such as Ensemble Kalman filtering or Ensemble Bayesian Model Averaging) towards the short and very short range, and towards extreme (severe, rare) weather events. A variety of techniques has been tried out to model predictability in the short-range: breeding modes, targeted singular vectors, stochastic physical parametrizations, multi-model systems, etc. Although the ensembles generated in these ways show added value over purely deterministic forecasts, the probabilistic distributions derived from them are still far from reliable, certainly for the prediction of severe weather events. Probably a combination is required of a variety of models, with several initial states and different parametrizations, and subsequent statistical postprocessing techniques (e.g. for calibration of ensemble- derived probabilities). HIRLAM should encourage and participate in European initiatives in setting up and evaluating multi-model systems, such as SRNWP/PEPS. Research within HIRLAM should focus on improved ways to construct singular vectors, including aspects of model error, and more extensive experimentation with perturbing physics parametrizations. Efforts are also required in the verification of ensemble characteristics, and in improving the reliability of ensemble forecasts, by means of applying techniques which remove effects of ensemble bias and underspread. Finally, the potential of promising alternative approaches will need to be evaluated.

8 Annex 1: Deliverables, time table and resources

A1.1 Deliverables

In the strategy outlined in section 3, the major deliverables of the HIRLAM programme for the next decade have been sketched, together with a rough indication of the time by which they should be achieved. For the HIRLAM-A programme, the main deliverables in the period 2005-2010 can be formulated as follows:

Mesoscale modelling: M1: An operationally feasible basic NH mesoscale forecast model (2km resolution) M2: An operationally feasible basic mesoscale upper air and surface analysis system, capable of assimilating high-resolution remote sensing data. Initially 3D-VAR, later possibly 4D-VAR. M3: Design of a fast, rapid-update analysis sytem suitable for nowcasting purposes M4: Assessment of an optimal nesting/LBC strategy for mesoscale model M5: Assessment of the suitability of, and application of the mesoscale model in, new application areas such as nowcasting, air quality, hydrology.

Synoptic scale modelling: S1: Implementation of 4D-VAR as the Reference assimilation system S2: Inclusion of new types of remote sensing data in the synoptic scale upper air and surface analysis S3: Improved description of clouds/convection and boundary layer/surface processes, as measurable by routine verification, through improved physics parametrizations

EPS: E1: Provision of HIRLAM contribution to European multi-model ensemble forecast system E2: Reliable (calibrated) probabilistic forecasting system suitable for short-range (24-72h) E3: New methodologies suitable for probabilistic forecasting for very short range (<24h)

System: SY1: Reference System run operationally in at least one institute SY2: Routine verification and reporting, also comparative with other models SY3: A verification system suitable for routine validation and verification of mesoscale and probabilistic analyses and forecasts SY4: Setting up and implementation of procedures for user consultation SY5: After development of the mesoscale model: Merging of this system and the present synoptic system into a single code suitable for use at both resolutions. SY6: Completion of the overhaul of code and scripts of the Reference System

These high-level deliverables, sometimes requiring many different tasks and activities, will be further specified and broken down in more specific deliverables and tasks in the work plans of the various HIRLAM projects.

A1.2 Resources

Table 1 lists estimates of the required manpower resources (in man years) for the first half of the period (2005- 2010) for each of the major areas of research within HIRLAM: the synoptic-scale data assimilation system, the synoptic forecast model, the mesoscale data assimilation system and forecast model, short-range ensemble forecasting, and system development and verification. Detailed projections of resources required beyond 2010 are difficult to make, and will not be presented here. For certain activities, such as EPS, it is expected that extra staff will become available through externally funded projects; these additional resources are not indicated in the table.

9

Year Syn. Data Syn model Mesoscale data Mesoscale EPS System/ Total ass. ass. model Verification 2005 6 5 0 5 1 5 22 2006 5 5 2 6 1 5 24 2007 5 5 2 7 1 5 25 2008 4 4 3 9 1 5 26 2009 3 4 4 9 1 5 26 2010 2 3 5 10 1 5 26

Table 1: Estimated staff resources (in man years) for the various elements of HIRLAM research for the period 2005-2010.

The table reflects the efforts that will initially be required for the introduction of 4D-VAR and the set-up of the mesoscale system. Resources will originally be required for both the synoptic and mesoscale forecast model, but the former will decrease in the course of time. In the development of the analysis system, efforts will gradually shift from the synoptic scale to the mesoscale. The quoted resources for short-range EPS are very conservative, as quite a few of the ongoing and planned activities in this area are assumed to take place outside of the HIRLAM programme.

As can be seen from table 1, the scientific and operational challenges facing HIRLAM in the coming decade require manpower and expertise beyond the present (committed) human resources of the HIRLAM-6 Project (roughly 22 fte/year). Throughout the period 2005-2010, the HIRLAM team is faced with the necessity to develop new modelling capabilities for the mesoscale and short-range ensemble forecasting, while retaining the responsibility for maintaining and improving the synoptic scale analysis and forecast system. This increase in required staff resources can be met by either dedicating more staff to the programme, by making more efficient use of present staff e.g. by increased cooperation in operational activities or pre- and post-processing, or by providing more externally funded staff. If the required resources in specific areas cannot be met, the associated deliverables or the time frame in which they are to be produced, may need adjustments.

10