TIGGE LAM Plan

GIFS-TIGGE TIGGE LAM Plan TIGGE-LAM 05.06.2018

TIGGE LAM Plan

Contributors: Tiziana Paccagnella Joshua Hacker David Parsons Richard Swinbank Jeanette Onvlee

WORK IN PROGRESS !!! GIFS-TIGGE TIGGE LAM Plan TIGGE-LAM 05.06.2018

Foreword The TIGGE-LAM panel was set up by the GIFS-TIGGE working group to coordinate the contribution from Limited Area Model (LAM) Ensemble Prediction Systems (EPS) to TIGGE (the THORPEX Interactive Grand Global Ensemble) and to the proposed GIFS (Global Interactive Forecast System).

After a couple of years of activity, the Panel was requested by the WWRP Joint Scientific Committee to develop a Strategic Plan outlining the main scientific and development issues on which TIGGE LAM must concentrate to advance LAM EPS and defining specific activities related to these issues. Furthermore, the sensible increase in resolution, both in deterministic and ensemble model applications, is making more and more evident the importance of establishing a broader long-term effort within the WWRP that focuses on improving skill of Mesoscale short-term regional forecasts (XV CAS …). This means that discussions need to take place between the different groups with interests in this subject.

This document has been drawn up in cooperation with the GIFS-TIGGE working group and with contributions from the other working groups in THORPEX and the WWRP with interests and competences related to LAM EPS. These working groups are: the Working Groups on Mesoscale Forecasting Research (WGMFR), the Joint Working Group on Verification Forecasting Research and the working Group on Societal and Economic Research and Applications (SERA).

Why LAM EPS is discussed in section 1 where also TIGGE LAM is introduced. The relationship with the GIFS-TIGGE WG is explained in section 2. Section 3 will describe the main scientific issues related to LAM EPS and Section 4 will present the specific actions to be undertaken.

1. Limited Area Ensemble Prediction and TIGGE-LAM High Impact/ Severe events, hereafter referred as HIW events, have a Mesoscale or convective scale component (e.g., severe convection, Mesoscale convective complexes and role in heavy rainfall, flash floods, polar lows, effects of terrain on winter precipitation) and such events have a large impact on society, the economy and the environment (………………….…..).

Deterministic Limited Area Modelling was introduced to increase atmospheric predictability at “smaller” scales thanks to the higher attainable horizontal resolution. During the last decade, high resolution, non-hydrostatic models have been shown to better capture high impact weather events (e.g., MAP D-PHASE results) even if the deterministic approach is still affected by a large uncertainty especially as regards the event localization in space and time. The loss of predictability of weather phenomena is not increasing only with the time but it is also increasing going down with the space scale. The successful application of the Probabilistic-Ensemble approach at Global scale has been extended also to LAM to complement deterministic products with probability information.

Considering that Global EPS systems are moving to higher resolutions (ECMWF EPS moved to T639 ~30 km at the end of January 2010), the main added value from LAM EPS in the future will be probably in the representation of phenomena at the convective-permitting scale (ref. Met.Office). As LAM goes to finer resolution, many open scientific and technical challenges will remain. Work toward solving general problems with current-generation LAM will benefit future systems at cloud-permitting scales and below. GIFS-TIGGE TIGGE LAM Plan TIGGE-LAM 05.06.2018

It is obvious the importance to avoid a “static vision” based on the present modelling systems set up since models and EPS systems are constantly changing and evolving. Following current plans, the GIFS concept should be shaped and concretized in few years but the GIFS Prototype design must be valid for a much longer period. What is not going to change so fast (neglecting climate change considerations) are the HIW types and TIGGE LAM should contribute to the definition of strategies and methodologies to improve their predictability.

According to the THORPEX core objectives, the GIFS-TIGGE vision is focussed on the optimization of the use of Ensemble Forecasting to maximize the forecast skill. TIGGE LAM is intended to coordinate actions to evaluate which are the events whose predictability can be improved by the use of LAM EPS.

TIGGE LAM should define how best to implement ensemble prediction systems to address regional and local situations and specific high-impact weather types. As a general and natural rule, based on an economy principle, it is important to wring out from global EPS all the available information and to use appropriate LAM EPS systems to better describe and detail what is likely to happen. That is, any progress on global ensemble production and archiving to support TIGGE LAM will facilitate research on the application of dynamic downscaling of forecasts from a variety of global ensemble prediction systems and with the introduction of other perturbations associated to local processes.

TIGGE LAM activity will be developed by: i. providing guidelines ii. coordinating activities iii. fostering research

With the following main Objectives: a. to contribute to the definition of scientific issues to advance LAM EPS b. to reinforce the link and cooperation with the other WWRP working groups with crossing competences. c. to support and foster research on LAM EPS: c.1 by promoting actions to make the access to LAM EPS products easier c.2 by coordinating / stimulating / participating to specific initiatives to address the relevant scientific issues (e.g. Research & Development Projects or Forecast Demonstration Projects, RDPs/FDPs) d. to coordinate the archiving of limited-area ensemble forecasts by providing standards and guidelines and as a complement to the TIGGE archive e. to facilitate the interoperability of the different LAM-EPS f. to facilitate the implementation of new LAM EPS

Thanks to the experience gained during the first period of activity, it is clear that TIGGE LAM cannot coordinate all the aspects of the cooperation in place for global systems (TIGGE). Guidelines, standards, directives, scientific priorities, sharing of tools and methodologies should be valid globally, but specific initiatives and applications must be organized at regional levels between the THORPEX Regional Committees, regional forecast centres and the national hydrometeorological services.

2. Relationship between the GIFS-TIGGE WG and TIGGE-LAM. The GIFS-TIGGE working group has two main objectives: GIFS-TIGGE TIGGE LAM Plan TIGGE-LAM 05.06.2018

 First, to enhance international cooperation in ensemble prediction, between both operational NWP centres and the academic community. The TIGGE project facilitates research on ensemble prediction methods, especially methods to combine ensembles, correct systematic errors, and aid in decision-making. This is achieved by coordinating the archiving of operational global ensemble forecasts from ten NWP centres, and making them available to the research community via three TIGGE data centres.  Second, to coordinate research and development leading to a future Global Interactive Forecast System (GIFS). The objective of GIFS is the production of improved probabilistic forecasts of high-impact weather, using research & development based on the TIGGE data set and other aspects of the THORPEX research programme. GIFS entails the development of prototype forecast products, initially focused on predictions of tropical cyclones. Subsequent GIFS products will be aimed at improving forecasts of precipitation, and then other high-impact weather.

As noted above, the TIGGE-LAM panel was set up by the GIFS-TIGGE working group to coordinate work on regional ensembles to complement the focus of GIFS-TIGGE on global ensemble forecasts. The GIFS-TIGGE working group agreed to a set of standards for archiving the global ensemble forecasts, and agreed to a standard set of forecast fields to be archived using GRIB2 format. The TIGGE LAM panel has adapted these standards for the archiving of LAM EPS products, as detailed in section 4. It has also been agreed that high priority TIGGE-LAM parameters will be archived at one of the three TIGGE data centres, on a regional basis.

It is envisaged that regional ensembles from TIGGE LAM will contribute to the development of the GIFS, by optimizing the use of the products of the existing systems and via the participation to relevant RDPs and FDPs. A GIFS-RDP will focus on the development of products to improve the prediction of tropical cyclones and high-impact precipitation for particular regions (initially Southern Africa and the SW Pacific), in conjunction with the WMO Severe Weather Forecast Demonstration Project. A North-West Pacific Tropical Cyclone Forecast Project will focus on improving the forecasts of tropical cyclones in that region. Products based on LAM EPS systems, where available, will supplement products available from the global TIGGE data, and demonstrate the additional benefit obtainable from higher resolution ensembles.

There is a close working relationship between GIFS-TIGGE and TIGGE LAM. Some of the members of the GIFS-TIGGE working group are also involved in the TIGGE LAM panel, and the chair of the TIGGE LAM panel is a member of the GIFS-TIGGE working group. GIFS-TIGGE TIGGE LAM Plan TIGGE-LAM 05.06.2018

3. Scientific Issues A primary goal of EPS is to represent and quantify the uncertainty associated with numerical weather prediction. This uncertainty is associated to the many approximations and errors affecting the modelling chain both during the assimilation of meteorological observations and during the model integration. A “good” EPS should provide forecast scenarios statistically indistinguishable from the real atmospheric evolution, quantify the system uncertainty by means of the ensemble spread that is the same amplitude of the forecast error, and should also provide a better deterministic forecast based on the ensemble mean (at least for upper air fields). (ref; ref Montani proceeding Ecmwf 2009.).

The assimilation procedures and the model formulations are by definition complicated scientific and technical problems; errors and approximations are still significant. EPS probabilistic approach represents an honest forecasting approach given our limits, and a logical way to account for these limits by complementing deterministic products with probabilistic information. Since many of the errors and approximations are associated with the assumption of scale separation and Reynolds averaging, LAM EPS is aimed at better representing and accounting for uncertainty in “smaller scale” processes and phenomena. Larger scale uncertainty is assumed to be transmitted through the driving lateral boundary conditions.

Going from global to LAM EPS, the process to represent forecast uncertainty is more complicated since larger scale perturbations interact with perturbations generated to represent smaller scale errors typical of LAM systems. Many possibilities to generate and couple these perturbations are tractable and it is almost impossible to evaluate a-priori the best configurations of such complex systems. These difficulties make necessary the physical development, implementation and testing of LAM EPS prototypes. Much of this work has been completed, and as a community we are poised to leverage this work toward greater gains in understanding and probabilistic forecast skill. This is an important reason to support coordinated scientific actions to facilitate the achievement of conclusive results, the definition of guidelines to support the optimization of existing systems and to facilitate the design of the future LAM EPS.

Although the TIGGE LAM panel is not committed to fund or execute scientific inquiry, it is important to set out the main scientific issues and challenges related to LAM EPS. TIGGE LAM will contribute to research by facilitating the availability of LAM EPS products to the research community and by promoting coordinated research initiatives in cooperation with the GIFS-TIGGE and the other WWRP working groups. TIGGE LAM should also create a natural environment to stimulate the cooperations between LAM EPS community and the other groups with closely related activities. A real advancement on Mesoscale forecasting must be achieved by a strong coordination and cooperation with the WWRP Mesoscale Research working Group and with the WGNE group. Research activities must be defined to allow all the numerical experimentation which is necessary to test and improve LAM also addressing regional peculiarities associates to the specific HIW types. As already well ongoing for TIGGE and the GIFS, it is also essential to have a close link with the Verification WG and with SERA to asses the quality and the added value coming from LAM EPS.

3.1. Mesoscale predictability Mechanisms for nonlinear and chaotic error growth, and the associated loss of predictability, are not well understood at scales characterized by the mesoscale kinetic energy spectrum (k-5/3). Unlike GIFS-TIGGE TIGGE LAM Plan TIGGE-LAM 05.06.2018 large-scale baroclinic flows displaying an up-scale error energy transfer, multiscale interactions involving the mesoscale is still the subject of vigorous debate. The classic view put forth by Lorenz (1969), that smaller scales always lose predictability faster than larger scales (commensurate with faster eddy turnover times), may not always hold. Up-scale error energy transfer may be limited, or mesoscale forcing including for example land-surface fluxes and topography actually slow or limit error growth at those scales. The implication is that study of predictability of specific phenomenon man be required. At present it also means that finding optimal methods for simulating mesoscale error growth may be an empirical endeavour and best studied with ensembles, and we are only now approaching a maturity with mesoscale models that permits robust conclusions about mesoscale predictability.

Although fundamental knowledge is sparse, empirical evidence suggests that for many phenomena, for example deep convection over flat terrain, does fit the Lorenz paradigm. Adopting a probabilistic approach to prediction is therefore imperative. Successes in global, large-scale, ensemble efforts motivate research and development at finer scales. Cooperation amongst several ongoing LAM efforts could provide the necessary experience and data to address the fundamental predictability problems. Classic measures of predictability, such as the Lyapunov exponent that reveals the typical rate of phase-space trajectory divergence in response to initial perturbations, are difficult to quantify analytically for complex NWP models, but large data sets allow for statistical methods of quantification. Inasmuch as mesoscale predictability may be phenomenon-dependent, large samples of specific event predictions are needed.

3.2. Mesoscale model inadequacy Model errors combine with chaotic error growth from initial-condition errors to limit extrinsic (complete NWP system) predictability and in practice reduce predictive skill. Despite the success of the NWP enterprise, objective methods for identifying and characterizing model inadequacy are not well defined. If we believe that compared to large-scale error, total error at mesoscales has proportionally more model error, model inadequacy is a serious barrier to improving predictions. As yet we do not know how best to simulate error growth due to model inadequacy. The canonical use of multiple models has proven empirically valuable, but is likely not the best long-term solution and in fact may be slowing our efforts to find the best model. This problem does not disappear with the eventuality of global mesoscale ensembles because it is a fundamental challenge for mesoscale modelling.

A cooperative TIGGE-LAM effort could bring much to bear on the model error problem. Multi- model ensembles cannot, by definition, sample from the attractor of the best possible model. Techniques to simulate model errors, such as stochastic perturbations to physics tendencies (e.g. Buizza 1999, Reynolds et al 2009 (??), Berner et al 2009, Bowler et al 2009 – need to check details on these), are promising and more scientifically grounded. Balances and energetics used to formulate those perturbations are better understood at large scales than mesoscales. Production of multi-model ensembles (and ensembles of ensembles) could prove valuable for defining the structures needed for effective mesoscale stochastic perturbations. Much research is needed, and the data sets produced TIGGE-LAM again enables study with large samples. GIFS-TIGGE TIGGE LAM Plan TIGGE-LAM 05.06.2018

3.3. Competition and interaction between larger scale perturbations given by the driving global systems and “local” perturbations generated specifically for the LAM EPS. This is a general issue which encompass many more specific issues listed in this document. It is impossible to work on this topic without considering many aspects at the same time. The problem, extending what stressed by Bowler and Mylne (2009), is essentially the answer to the question: is it possible to generate local perturbations which can improve LAM EPS skill when compared to a pure global EPS downscaling? This question can be linked directly to LAM basics (and hence to the cooperation with the MWFR working group) since it is possible to get out some added value every time the LAM modelling system is able to develop its own dynamical structures and not only to detail and adjust larger scale fields to more detailed surface forcing.

A lot of related research and development is ongoing, and much more is necessary, but it is evident that the answer must be multi-faceted. It is important to carry on scientific investigations to understand which are the main factors affecting this competition and how to use them considering the application of the LAM EPS system. Very high resolution, very short range forecasting (1-3 km up to 24 hours), on limited regions is expected to benefit from perturbations quite different from the perturbations increasing the skill of ensemble forecasting at 10-20 km resolution, continental scale at 72 hours forecast range.

3.3.1. Domain size and “local” perturbations on initial conditions The impact of domain size is more acute for LAM EPS (Errico et al. 1993,…) than LAM in general. It plays a crucial role in the interaction and relative weight of the perturbations generated within the LAM EPS, hereafter referred as “local” perturbations, and the larger scale perturbed forcing which penetrates the integration area through the boundaries. It is reasonable to assume that LAM EPS perturbations on the initial conditions will survive for a limited period of time, which apart from the efficiency and quality of the technique adopted to generate the perturbations themselves, is conditioned by the size of the integration domain. Bigger integration domains will allow local perturbations to grow upscale more and reach greater total amplitude in the LAM domain.

It is again reasonable to assume that a good level of consistency with boundary perturbations will play in favour of a good development of local perturbations (ref…), but methodologies to assure this consistency is still matter to be investigated. One of the most recent and advanced work in this direction has been done recently by the Met. Office (Bowler and Mylne 2009) ) where it is confirmed the limited period of survival of these perturbations and the difficulties in drawing final conclusions in this sense. The optimal way to determine efficient perturbations on the initial conditions is matter of research in many groups, both for global and lam applications (ref, Argence et al 2008, …); TIGGE LAM should stimulate exchanges and cooperations to focus ongoing experimentations on the specific HIW events.

3.3.2. Perturbations associated to soil/surface description Representation of details in lower boundary conditions is increasing steadily with more complex land-surface models and space-borne sensors. Those details can then exert more impact on a short- range mesoscale forecast, Thus we expect this “branch” of perturbations will play a greater role in the future. Not much work on this subject has been done up to now (Sutton et al 2004, 2006) but, also considering the forthcoming increasing in horizontal resolution, methodologies to obtain good perturbations in soil and surface parameters need to be developed and tested. Recent work by GIFS-TIGGE TIGGE LAM Plan TIGGE-LAM 05.06.2018

Hacker (2010) showed that this task may prove formidable because atmospheric preconditioning can result in forecasts sensitive to small-amplitude uncertainty.

Soil moisture is one of the most difficult soil parameter to be estimated. Due the poorness of available observations, and to the strong dependency on local soil properties, in numerical forecasting practice the soil moisture field is often computed as the field which produce the better near surface parameters forecast (Mahfouf 1991, Balsamo et al 2005, Hess at all 2008). It is well known that, especially if large scale forcing is weak, the impact of the soil moisture on local meteorological parameters can be extremely relevant (Cassardo et al 2002) and it is evident the importance of deriving proper methodologies to describe and account for the uncertainty in this field.

Topography has always been a big issue in numerical modelling (Wallace et al 1983) since its effect on the atmospheric flow involves many phenomena (Bougeault, 2003). Going down with the mesh size, its representation can be more realistic but, especially in region with complex and steep orography, it is still necessary to apply some smoothing (Gassman, 2001) to avoid problems like unrealistic precipitation at the grid scale (ref.). Many of the model parameters related to orography and to its subgrid variance are tunable and their setting is almost always based on sensitivity performed on a limited set of cases. Uncertainty related to these effects should be also investigated and represented in high resolution LAM EPS.

Generally speaking all the parameters used to describe soil type and properties, including vegetation, should be included as source of errors to be accounted for through suitable perturbations. Despite local perturbations in the initial conditions, these soil/surface perturbations have the “advantage” to be active all during the model integration and to give a valuable and positive contribution to the spread at least for near surface parameters. But as yet we do not know the spatial scales associated with lower-boundary uncertainty, and we also do not know how important that uncertainty is compared to other sources of mesoscale uncertainty.

3.3.3. Cycling short-range uncertainty for intial perturbations Active research on mesoscale ensemble-filters and ensemble-transform techniques for LAM-EPS is ongoing world-wide. Recent papers by Bowler et al (2009), Hacker et al (2010) [OTHERS FROM THE RECENT SRNWP-SREF MEETING] shows that some technical challenges have been overcome but that the value of these techniques for mesoscale initial condition perturbations has yet to be realized. Those studies did not make use of ensemble data assimilation, and rather update only the perturbations about a control state. Ensemble data assimilation is also still a research topic at these scales, and those systems have not been thoroughly evaluated at forecast times beyond a few hours. The simpler method of perturbed observations in individually cycling data assimilation systems is another candidate receiving only sparse attention. Burgers (1998) showed the equivalence with other ensemble data assimilation systems under conditions of large ensembles, linear systems, and Gaussian errors. Limitations and benefits of these approaches for LAM-EPS are still unclear given inevitabile sampling error and model inadequacy, and further research is needed.

3.3.4. Multi Model EPS ………………………….. ………………………………… GIFS-TIGGE TIGGE LAM Plan TIGGE-LAM 05.06.2018

3.4.Quantify the additional benefit of Multi-LAM EPS Where domains overlap a natural avenue for exploring and exploiting “ensembles of ensembles” may be possible for longer periods. This issue can be investigated also in different regions during RDPs if lateral boundary conditions (LBCs) to LAM are available. The availability of LBCs is most important to achieving many of the scientific objectives. …………………………………

3.5. Evaluate the performance of convective permitting EPS systems and their benefit over restricted integration domains versus the combination of lower resolution ensemble systems This issue is particularly interesting in a longer time framework; Global EPS systems are still far from these finer scale configurations. This topic is particularly relevant in Europe where many systems are already running with substantial overlapping of the integration domains. TIGGE LAM should also coordinate actions to provide suitable set of boundary conditions to support the research on convection-resolving LAM EPS systems. …………………………………

3.6. Predictability at the convection-resolving scale to support the design of the future LAM EPS systems. Turbulence at convection-resolving scales, and below, is still subject to research and debate. Consequently, predictability is not well-understood at those scales. Understanding predictability is essential for guidance on small-scale EPS. LAM EPS enables studies that can address this general topic now. …………………………………

3.7.Biases in deterministic models and calibration Currently, highly reliable and sharp probabilistic forecasts with mesoscale information results only after a calibration technique has been applied (cf. Raftery 2006 – what other mesoscale examples from Europe?). In the absence of a nearly perfect mesoscale model, the need for calibration will persist. Because the expense of producing large data sets appropriate for calibration is substantial, most calibration research has been completed with global models from operational centres with longer archives, and in many cases simplified versions of global models. Mesoscale variability introduces further challenges to statistical methods. For example, mesoscale error modes may be less Gaussian and further may not be well defined by any known parametric distribution. Research is needed to find the best methods for mesoscale calibration. Results may inform future decisions about ensembles sizes and model resolution to produce highly reliable and sharp forecasts under calibration.

A TIGGE-LAM project can enable research to find appropriate mesoscale ensemble calibration approaches, and also to quantify the tradeoffs between ensemble size, model diversity, and resolution when a good calibration method is available. Much as TIGGE has spurred research on calibration for large-scale models (refs -???), a large and diverse mesoscale ensemble data archive will lower the technical barrier for researchers to propose and test calibration approaches. GIFS-TIGGE TIGGE LAM Plan TIGGE-LAM 05.06.2018

Research is required to compare the benefits of multi-model ensembles with the calibration of a single model using the reforecast data. Initial data for limited area model reforecasts needs to come from a reanalysis carried out with an up to date NWP system, so this topic is closely related to regional reanalysis efforts.

3.8. Probabilistic forecasts for other modelling applications Models of other processes that use NWP output as input include both technical (diagnostic and dynamic) models, and models of decision processes often called decision aids. Probabilistic mesoscale predictions pose new challenges to those applications because the error structures and variability are different from either deterministic forecasts or large-scale ensembles. If we assume that better ensemble predictions can theoretically lead to better predictions from other models or automated decision support, then research and development is needed to take advantage of mesoscale ensembles.

Data sets such as those that may be generated from a TIGGE-LAM project could help decision-aid developers learn to make use of mesoscale probabilistic forecasts. Many decision aids cannot directly ingest probabilistic information, and simply running many realizations may not be an option because of limited computational capability in rapid decision scenarios. Use of output from multiple models when dynamical variables have different inherent scales of variability, and physical variables may not even have the same meaning among models, is a practical complication.

TIGGE-LAM could also be important for overcoming challenges associated with diagnostic and dynamical models used in secondary predictions. The practical problem of variable meanings also exists here. More fundamentally, if a secondary model does not follow statistical linearity, for example resulting in a linear transformation of a pdf or more trivially predicting a Gaussian pdf from Gaussian input, further challenges arise. Questions regarding whether to run an ensemble of the secondary models or to use the model to map NWP pdfs to pdf in other variables need to be studied.

3.9.Verification methods Issues: Fair intercomparison between Global EPS and KM scale ensemble Poor statistics associated to rare-extreme events …………………………………

3.10. Assess the probability of rare meteorological events The greatest challenge for regional EPS systems is to accurately assess the probability of (relatively rare) extreme events, which often are far more important in terms of societal impact than “normal” weather. How best to do this for LAM EPS systems? Is an alternative approach to e.g. ensemble bias correction required in order to best support the accurate forecasting of extreme events, as opposed to the forecasting of normal weather? GIFS-TIGGE TIGGE LAM Plan TIGGE-LAM 05.06.2018

3.11. LAM EPS and Data Assimilation: which are the key characteristics of a LAM EPS to best represent short-range forecast error co variances for use in DA? ………………………………… GIFS-TIGGE TIGGE LAM Plan TIGGE-LAM 05.06.2018

4. Actions & Activities Numerical weather forecasting nowadays is based on a wide spectrum of modeling systems. Deterministic models with their associated assimilation schemes, probabilistic systems based on ensemble applications, larger scale (synoptic) applications and mesoscale applications down to the cloud resolving scale. The different part of the forecasting systems are changing and evolving, all together, and it is really necessary to be adaptive and to keep a very high level of coordination among all the groups which are interconnected by having common scientific interests.

As already mentioned before, the management of the TIGGE LAM coordination cannot be kept at global level due to the intrinsic regionalism of LAM EPS. The management and coordination cascade must be defined Region by Region and activities must be planned taking the best possible advantages from already existing initiatives.

WWRP, THORPEX and GIFS-TIGGE represent a perfect opportunity to coordinate LAM EPS activity both at scientific and implementation level. WWRP offers the opportunity to have cross- coordination with the closely related working groups. THORPEX provides the organizational basis to LAM EPS to contribute to the improvement of Weather Forecasting with specific reference to Regional HIW. GIFS TIGGE represents the main reference since TIGGE LAM must complement global eps by adding value where and when higher resolution and local optimization can play a substantial role.

Actions should then cover all the initiatives which are necessary to exploit at best the unique potential coming from being part of these Programmes.

After the completion of this plan, and the restructuring of the Panel (following action 0), a TIGGE LAM workshop will be organized to discuss and address each of the following actions.

Each action will be classified as GLOBAL, REGIONAL or both

Action 0: Reorganization of the TIGGE LAM Panel - GLOBAL After the first period of activity, is now necessary to give a new structure to the Panel to facilitate the progress of TIGGE-LAM. The key points are:  To give more emphasis to the Regional component of TIGGE LAM and to facilitate the focus on regional activities.  To have Panel members who are in the position to give direct contributions to the activities.  To involve representatives of other working groups with common interests . Action 1: Set-up of cross-working group discussion between TIGGE-LAM, the MWFR working group and the WGNE. GLOBAL Model systematic errors represent a limit in this sense and it is clear that it is useless to adopt an ensemble approach based on a model which is physiologically unable to reproduce the atmospheric processes leading to such events. The research activities related to this type of model assessment represents the bulk of the cooperation with the Mesoscale Weather Research Forecasting Working Group. GIFS-TIGGE TIGGE LAM Plan TIGGE-LAM 05.06.2018

The discussion among these groups should lead to the identification of specific activities to be coordinated finalized to precise scientific objectives. ……………………………………. Action 2: Set-up of cross-working group discussion with the Verification Working Group and SERA. GLOBAL ……………………….. High Impact Weather events, in most of the cases, are not frequent Severe Weather events. This aspect makes statistical evaluations more difficult due to the need to have available observational data-sets appropriate to analyze the statistical distribution of the events. ………………………….. Action 3: Definition of the key requirements for regional ensemble forecasting. REGIONAL Scientific issues focused on Regional HIW events (described in the Regional THORPEX Plans) will constitute the basis to design the TIGGE LAM contribution to the scientific activities to be coordinated during research projects, RDPs and FDPs. The new structure of the TIGGE LAM Panel should help in this sense. A very good example in this direction comes from the African approach (THORPEX Africa Plan and Trieste workshop). For predictability studies a number of target high impact events that occur recently have been identified A 3-year detailed work plan has been drawn to undertake predictability studies with the few selected case studies on sub-regional basis. 3 phases have been identified: i) analysis and documentation of the selected events, ii) deterministic and global model assessment ; iii) modelling studies and ensemble prediction assessment. The SERA working Group is involved to equip the HIW catalogue with all the information required to evaluate the impact of the events and the reduction of losses due to the implementation of the new modelling systems. The JWGFVR is also involved to provide guidelines for model assessment. ……………………………. Action 4: Identification of possible Funding opportunities to support the development and implementation of the Regional activities. GLOBAL - -REGIONAL ………………………… Action 5: Activate link with major mesoscale applications Linking to hydrology (a TIGGE LAM effort in HYMEX from the early days could be a help). REGIONAL ……………………………….. Action 5:Set up of the TIGGE-LAM Database GLOBAL - -REGIONAL Following what done by TIGGE, a list of TIGGE LAM output parameters was defined and it is reported as ANNEX 1 to this plan. The first step was the selection of a sub-set of parameters, labelled as High Priority (HP). It was decided to make the access to these products as easy as possible in order to stimulate scientific investigations and the work of the users (e.g. hydrologists ) who are not familiar with the tools and the methodologies to manage meteorological fields. This is GIFS-TIGGE TIGGE LAM Plan TIGGE-LAM 05.06.2018 particularly important if dealing with more LAM EPS products at the same time. Standards to code and archive these parameters have been defined. They will be:  interpolated on a standard geographical lat/lon grid at 0,1° resolution  encoded in GRIB2

HP parameters should be archived at the three TIGGE Archiving Centres, NCAR, ECMWF and CMA, following a geographical/Regional competence principle (i.e., data from European LAM EPSs will be archived at ECMWF, data from the Americas at NCAR, Asian data at CMA). As regards the data access, the same policy adopted by TIGGE will be proposed with a reduction of the delay from 48 hours to 24 hours. These guidelines should be taken as a reference to support the organization of the regional archives. At the time of writing only Europe ,after having agreed about the procedures to be adopted, is working to implement the European regional archive .

The full content archive will be planned and implemented in coordination with GIFS-TIGGE working group during the development of the GIFS prototype. Larger dataset will probably be stored at the originating centres and access to these products will be agreed following the standards defined by TIGGE LAM.

Action 6: TIGGE LAM Data Policy GLOBAL - -REGIONAL

TIGGE LAM data providers should agree about data access policy following what has already been done by the TIGGE data providers. The proposal for the non-real time data is the same of TIGGE with a delay of 24 hours. The data policy referred to the real-time availability of TIGGE LAM products for the GIFS will be evaluated and decided in a later stage. Real time availability of products during FDPs and RDPs will be asked specifically from time to time.

Action 7: Implementation of regional observational dataset for objective verification of Mesoscale Deterministic and Ensemble forecasting. REGIONAL Observational datasets suitable for model verification are usually available for limited periods of time, and over restricted areas, in correspondence to scientific projects, RDPs or FDPs. It would be very useful to coordinate activities to implement regional observational dataset in regions where observational networks are particularly dense or rich of information which can be used to asses modelling systems skill in forecasting high impact weather events. In regions where national policies put severe restrictions on the availability of data from local networks, alternative solutions should be identified to allow scientific investigations while preserving the commercial value of the data. The production of High-Res analyses could be an option. The availability of suitable observational data and the adoption of common methodologies for Verification (following the guidelines and the methodologies defined by the WG on Verification) would allow to support and speed-up research and to reach conclusive results based on high quality data and tools.

Action 8: Definition of further HIW related specific products. REGIONAL This will be done during RDPs, FDPs and during the development of the GIFS GIFS-TIGGE TIGGE LAM Plan TIGGE-LAM 05.06.2018

Action 8: Set up of specific projects and collaborations addressed to the specific scientific issues. GLOBAL - - REGIONAL This action(s) should complement the activities carried on during RDPs and FDPs. The key point is to take advantages of Regional specificities (Global Models and EPS available, LAM and LAM EPS available, specific observational networks, etc.) to define a precise project linked to specific scientific questions. As an example Europe can rely on two of the best EPS, the Met Office and ECMWF ensembles, which provide products for verification and comparison, and high quality Boundary conditions for running more LAM EPS in a coordinated way to address a particular phenomena as a scientific focal point (flooding, winter storms, cold air damming in basins, etc).

These actions cannot be undertaken without the strong support and coordination by the Regional Committees.

Action 10: Coordinate the participation of LAM EPS groups to Research Projects, RDPs and RDPs. GLOBAL - - REGIONAL Following the GIFS-TIGGE approach, RDPs and FDPs seem to be the best way to allow a coordinated work on LAM EPS on a regional basis and focuse on the specific HIW types.. Research during RDPs should allow to develop and test the different components of the forecasting system: deterministic forecasts and data assimilation, EPS Global and LAM, combined and calibrated products. This approach is for sure a good way to have a wide WWRP cooperation to advance Mesoscale forecasting by exploiting all the best available modelling tools and by assessing the best way to combine them. RDPs are also important to tailor model outputs to user needs. Another relevant aspect of research during RDPs is the possibility to support the development of LAM EPS also based on the relocation of systems already running and tested over different regions with different climatology and HIW phenomena. Participation to FDPs is again important to exploit the real time feasibility and value of LAM EPS products when used in operations and as a complement to the other available numerical products.

Action 11: Define standards to exchange meteorological fields required as Initial and Boundary conditions – Interoperability GLOBAL The interoperability concept covers many aspects related to the ease of use of operational products exchanged among different centres. At the basic level, interoperability means standardisation of field format, coding, transmission ,etc. At the higher level, it includes the possibility of coupling different GLOBAL and LAM systems. This last task is really tough both as regards the development of the required SW interfaces and also the long term sustainability. It implies a very high level of communication and coordinated SW maintenance by the involved centres. Due to the overlapping of interests and objectives, interoperability aspects will be carried on in liaison with the SRNWP Interoperability programme, a three-year EUMETNET programme started in September 2008 and lead by the Met Office. ICs and BCs will be provided on the original computational grid to avoid loss of information and possible detrimental effects due to double interpolation (model levels – pressure levels – model levels). These effects might be considered negligible for some applications, but there are cases where detrimental effects must be avoided as much as possible. One of these cases is when perturbations at boundaries and in the initial conditions are obtained through a multi-model ICs/BCs GIFS-TIGGE TIGGE LAM Plan TIGGE-LAM 05.06.2018 approach and are extracted from the highest resolution deterministic operational suites (the best possible deterministic configurations). In this case, the driving model systems are run in their best configuration and the boundary conditions derived from their output must be “handled with care” to preserve the original quality. This requirement on the boundary conditions is even more crucial for LAM EPS systems run over very restricted domains, which is the case of the new generation cloud- permitting O(2-3km) and the future cloud-resolving systems.

TIGGE LAM strategy is to agree about the format and content of the initial and boundary condition files, and to leave the physical implementation of data exchange to specific initiatives. The latter may include either regional or bilateral agreements. The definition of these standards will make cooperation easier and faster during RDPs, FDPs or other project campaigns and the implementation of new LAM EPS systems.

Action 12: Provide LAM EPS products in format compliant with GIFS-TIGGE directives GLOBAL ………………. ……………….. GIFS-TIGGE TIGGE LAM Plan TIGGE-LAM 05.06.2018

References and reference documents JSC09 (http://www.wmo.int/pages/prog/arep/wwrp/new/documents/JSC_09_report_Decisions_V5.pdf) Brooks, H. E., Tracton, M. S., Stensrud, D. J., DiMego, G. and Toth, Z., 1995: Short-range ensemble forecasting: Report from a workshop, 25-27 July 1994. Bull. Amer. Meteor. Soc., 76, 1617- 1624.

References per subject; preliminary list to be completed and/or modified

Bougeault P. “Sub-grib scale orography parametrizations”; Key issues in the parametrization of subgrid physical processes- ECMWF Seminar 3-7-September 2001

Hess R., M. Lange and W. Wergen. 2008. Evaluation of the variational soil moisture assimilation scheme at Deutscher Wetterdienst Q.J.R.Met.Soc. Volume 134 Issue 635 , Pages 1357 - 1633 (July 2008 Part B) Mahfouf J-F. 1991. Analysis of soil moisture from near-surface parameters: A feasibility study. J. Appl. Meteorol. 30: 1534–1547.

Balsamo G, Bouyssel F, Noilhan J, Mahfouf J-F, B´elair S, Deblonde G.2005. ‘A simplified variational analysis scheme for soil moisture: Developments at Meteo-France and MSC’. Pp. 79–96 in Proceedings of ECMWF/ELDAS workshop on Land Surface Assimilation, 8–11November 2004. ECMWF: Reading, UK.

Boccanera F., D. Cesari, M. Elementi, C. Marsigli, T. Paccagnella, 2003. High Resolution Simulations of MAP IOP 2b Case with Lokal Modell Proceeding of the ICAM/MAP conference 2003. available at http://www.map.meteoswiss.ch/map-doc/icam2003/ICAM-MAP2003.htm

C. Cassardo, G. Balsamo, C. Cacciamani, D. Cesari, T. Paccagnella and R. Pelosini, 2002: Impact of soil surface moisture initialization on rainfall in a limited area model: a case study of the 1995 South Ticino flash flood. Hydrol. Process., 16, 1301-1317

Gassman A.. Filtering of LM orography. COSMO Newsletter n°1. Available at www.como- model.org. Wallace, J. M., S. Tibaldi and A. Simmons, 1983: Reduction of systematic forecast errors in the ECMWF model through the introduction of an envelope orography, Q. J. R. Meteor.Soc., 109, 683– 717.

Argence S. , D. Lambert, E. Richard, J..P. Chaboureau, N. Söhne. 2008. Impact of initial condition uncertainties on the predictability of heavy rainfall in the Mediterranean: a case study. Q. J. R. Meteorol. Soc. 134: 1775–1788 (2008)

Perturbations on soil and surface description Sutton, C. J., T. M., Hamill, and T. T. Warner, 2004: Impacts of perturbed soil moisture conditions on short-range ensemble variability Preprints, 16th NWP/20th W&F Conference, Seattle, American Meteorological Society. Sutton C, Hamill TM, Warner TT. 2006: Will perturbing soil moisture improve warmseason ensemble forecast? A proof of concept. Monthly Weather Review 134: 3174-3189.

Perturbations representing model uncertainties and surface description Bowler NE, Arribas A, Mylne KR, Robertson KB and Beare SE. 2008. The MOGREPS short-range ensemble prediction system, Quarterly Journal of the Royal Meteorological Society, in press. Bowler NE and K.R. Mylne. 2009. Ensemble transform Kalman filter perturbations for a regional ensemble prediction system Q. J. R. Meteorol. Soc. 135: 757–766 (2009) Bright, D. R. and Mullen, S. L., 2002: Short-range ensemble forecasts of precipitation during the Southwest Monsoon. Wea. Forecasting, 17, 1080-1100. De Elía, R., Laprise, R., and Denis, B., 2002: Forecasting skill limits of nested, limited-area models: A perfect-model approach. Mon. Wea. Rev., 130, 2006-2023. GIFS-TIGGE TIGGE LAM Plan TIGGE-LAM 05.06.2018

Errico R. M., T. Vukićević and K. Raeder 1993: Comparison of initial and lateral boundary condition sensitivity for a limited-area model Tellus A Volume 45 Issue 5, Pages 539 – 557. Magnusson L., Martin Leutbecher, and Erland Källén, 2008: Comparison between Singular Vectors and Breeding Vectors as Initial Perturbations for the ECMWF Ensemble Prediction System. Monthly Weather Review Volume 136, Issue 11 (November 2008) pp. 4092–4104 Nutter, P. A., 2003: Effects of nesting frequency and lateral boundary perturbations on the dispersion of limited-area ensemble forecasts. PhD thesis. University of Oklahoma, School of Meteorology. Norman, OK. Posselt D. J., T. Vukicevic,2010. Robust Characterization of Model Physics Uncertainty for Simulations of Deep Moist Convection., to appear on Monthly Weather Review. Wang, Y. and Kann, A., 2006: Dealing with the uncertainties in the initial conditions in ALADIN- LAEF. 1st MAP D-PHASE scientific meeting. 6-8 November 2006, Vienna. Multi Model EPS Related references: Tracton MS, Du J, Toth Z and Juang H, 1998. Short Range ensemble forecasting (SREF) at NCEP/EMC 12th Conf. on Numerical Weather Prediction, Phoenix, American Meteorological Society, 269-272 García-Moya, J. A., Callado, A., Santos, C, Santos-Muños, D. and Simarro, J., 2009: Predictability of Short-Range Forecasting: A Multi-model Approach. Nota Téchnica 1 del Servicio de Predecibilidad y Predicciones Extendidas. Agencia Estatal de Meteorología. Spain. Importance of Ensemble Size and post-processing methods that account for a small ensemble size, Related references: Schwartz, C. S., Kain, J. S., Weiss, S. J., Xue, M., Bright, D. R., Kong, F., Thomas, K. W., Levit, J. J., Coniglio, M. C. and Wandishin, M. S., 2009: Toward improved convection-allowing ensembles: Model physics sensitivities and optimizing probabilistic guidance with small ensembles membership. Submitted to Weather and Forecasting. Theis S. E., A. Hense, and U. Damrath, 2005: Probabilistic precipitation forecasts from a deterministic model: A pragmatic approach. Meteor. Appl., 12, 257–268. Quantify the additional benefit of Multi-LAM EPS Evaluate the performance of convective permitting EPS systems Related references: Leoncini, G., Plant, R. S., Gray, S. L., and Clark, P. A., 2009: Perturbation growth at the convective scale for CSIP IOP18. In print on Q. J. R. Meteorol. Soc. Kong, F. and co-authors, 2008: Real-time storm-scale ensemble forecast experiment. 9th WRF User’s Workshop, NCAR Center Green Campus, 23-27 June 2008, Paper 7.3. Schwartz, C. S., Kain, J. S., Weiss, S. J., Xue, M., Bright, D. R., Kong, F., Thomas, K. W., Levit, J. J., Coniglio, M. C. and Wandishin, M. S., 2009: Toward improved convection-allowing ensembles: Model physics sensitivities and optimizing probabilistic guidance with small ensembles membership. Submitted to Weather and Forecasting. Gebhardt C, Theis S, Krahe P, and Renner V: 2008, Experimental ensemble forecasts of precipitation based on a convection-resolving model, Atmospheric Science Letters, 9, 67-72.. Gebhardt, C., Theis, S. E., Paulat, M., Ben Bouallègue, Z., 2009: Uncertainties in COSMO-DE precipitation forecasts introduced by model perturbations and variation of lateral boundaries. Submitted to Atmospheric Research. Evaluate the benefit of higher resolution/ convective permitting systems over restricted integration domains versus the combination of lower resolution ensemble systems

Predictability at the convection-resolving scale to support the design of the future LAM EPS systems. Related references: Hohenegger C, Schär C: 2007b, Atmospheric predictability at synoptic versus cloud-resolving scales. Bull. Am. Meteorol. Soc. 11: 1783–1793. GIFS-TIGGE TIGGE LAM Plan TIGGE-LAM 05.06.2018

Kong F , Droegemeier K, Hickmon N: 2006, Multiresolution ensemble forecasts of an observed tornadic thunderstorm system. Part I: Comparison of coarse- and fine-grid experiments. Mon. Weather Rev. 134: 807–833. Skamarock W C: 2004, Evaluating mesoscale NWP models using kinetic energy spectra. Mon. Wea. Rev., 132, 3019-3032. How do biases in deterministic models affect EPS skill? Combination and Calibration of LAM EPS products. Related references: Casanova S. and B. Ahrens. 2009. On the Weighting of Multimodel Ensembles in Seasonal and Short-Range. Monthly Weather Review.Volume 137, Issue 11 Di Narzo A. F.and D. Cocchi 2010. A Bayesian hierarchical approach to ensemble weather forecasting. To appear on The Journal of the Royal Statistical Society, Serie C. Diomede T., C. Marsigli, A. Montani, T. Paccagnella, 2009. Test of calibration techniques based on reforecasts for limited-area ensemble precipitation forecasts. EMS Annual Meeting, 28 September - 02 October 2009, Toulouse, France. Presentation available at: http: //www.emetsoc.org/ annual_meetings/documents/2009/NWP2_EMS2009-377.pdf Hamill TM and Whitaker JS: 2006, Probabilistic quantitative precipitation forecasts based on reforecast analogs: theory and application, Monthly Weather Review, 134, 3209-3229. Fundel, F., Walser, A., Liniger, M. A., Frei, C., and Appenzeller, C., 2009: Reliable Precipitation Forecasts for a Limited Area Ensemble Forecast System Using Reforecasts. Mon. Weather Rev., in press. Hagedorn R, Hamill TM and Whitaker JS: 2008, Probabilistic Forecast Calibration Using ECMWF and GFS Ensemble Reforecasts. Part I: 2-meter Temperatures, Monthly Weather Review, in press Hamill TM, Hagedorn R and Whitaker JS: 2008, Probabilistic Forecast Calibration Using ECMWF and GFS Ensemble Reforecasts. Part II: Precipitation, Monthly Weather Review, 136, 2620-2632. Toth Z, Talagrand O, Zhu Y: 2006, The attributes of forecast systems: a general framework for the evaluation and calibration of weather forecasts. Predictability of Weather and Climate. Cambridge University Press, 718 pp. Kann A., C. Wittmann, Y. Wang, and X. Ma. 2009. Calibrating 2-m Temperature of Limited-Area Ensemble Forecasts Using High-Resolution Analysis Monthly Weather ReviewVolume 137, Issue 10 (October 2009)

Verification methods for extreme events Marsigli C, Montani A and Paccagnella T: 2008, A spatial verification method applied to the evaluation of high-resolution ensemble forecasts, Meteorological Applications, 15, 125-143. Coupling with applications: Air Quality, Hydrology LAM EPS and Data Assimilation: which are the key characteristics of a LAM EPS to best represent short-range forecast error co variances for use in DA? Related references: Bonavita M., L. Torrisi and F. Marcucci. 2010. Ensemble data assimilation with the CNMCA regional forecasting system. To appear on the Q. J. R. Meteorol. Soc. (2010) Di Giuseppe F., C. Marsigli and T Paccagnella, 2010: The relevance of background error covariance matrix localization: an application to the variational retrieval of vertical profiles from SEVIRI observations. Accepted for publication on QJRMS Kalnay E, Hunt B, Ott E, Szunyogh I, 2006: Ensemble forecasting and data assimilation: two problems with the same solution? Predictability of Weather and Climate (p. 157-180). Cambridge University Press. Hamill TM, 2006: Ensemble-based atmospheric data assimilation. Predictability of Weather and Climate (p. 124-156). Cambridge University Press. Li H., E. Kalnay, T.Miyoshi, and C.M. Danforth, 2009: Accounting for Model Errors in Ensemble Data Assimilation. Monthly Weather Review, Vol. 137, Iss. 10, pp. 3407–3419. Sakov P., G. Evensen and L. Bertino. 2009. Asynchronous data assimilation with the EnKF Tellus A Volume 62 Issue 1, Pages 24 – 29. GIFS-TIGGE TIGGE LAM Plan TIGGE-LAM 05.06.2018

Peña M., Z. Toth, M. Wei, Controlling noise in ensemble data assimilation schemes. to appear on Monthly Weather Review. Torn R. D., G. J. Hakim and C. Snyder 2006: Boundary Conditions for Limited-Area Ensemble Kalman Filters Mon. Wea. Rev., 134, 2490-2502

Acronyms: EUMETNET NETwork of EUropean METeorological services SRNWP Short Range Numerical Weather Prediction GIFS-TIGGE TIGGE LAM Plan TIGGE-LAM 05.06.2018

List of Annexes: Annex 1. TIGGE LAM output parameter list with GRIB2 coding specifics