Last updated: 11 Mar 2019
Final Report ready, it will go to the Editor before ICCARUS 2019.
Project duration: September 2015 – August 2017 (an extension requested until February 2018)
FTEs (plan/used) 4.03/2.69 in COSMO year 2015-2016
3.33/2.83 (used up to MAM2017 + estimated in JJA 2017) in COSMO year 2016-2017
1.13 planned in project extension request in COSMO year 2017-2018
Total FTEs used: 7.13
WG7 members have identified few main issues on which the development work in the ensemble field is mainly needed, leading to the proposal of a new Priority Project in this field and within WG7.
The starting point was the need of improving the spread-skill relation of the convection-permitting (CP) ensembles for surface weather parameters, which has lead to the idea of a coordinated study on this issue, leading to the identification of measures to address it. This implies also that the problem of the description of the model error should be addressed and model perturbation methodologies should be further tested and developed, together with surface/soil perturbations, which are crucial for a good description of the error affecting near-surface parameters.
The project is dealing also with the development of post-processing methods for the CP ensembles, especially focused on selected phenomena and intense events, and calibration.
Finally, the use of Initial Conditions (ICs) obtained from KENDA analyses to initialize the CP ensemble is also addressed, focusing on some issues like member selection and comparison with cheaper techniques for providing ICs to the ensembles.
Status
A short description of the status of ensemble development and planning in the Consortium at the date of summer 2015 is here provided.
DWD: COSMO-DE-EPS, operational at 2.8km horizontal resolution since May 2012. The system receives initial condition perturbations from the 4 global scale operational analyses of DWD, ECMWF, NCEP and JMA. The four operational global runs of these institutions provide also boundary conditions to 4 COSMO runs at 7km horizontal resolution which drive in turn the COSMO-DE runs. Model physics perturbations are also applied, by changing the value of few parameters of the physics schemes, as well as soil moisture perturbations. Developments to be operational in 2016: COSMO-D2-EPS with 2.2 km mesh size, 65 levels, enlarged model domain, 40 members, use of KENDA ICs, advanced physics perturbations likely combined with a stochastic approach.
MCH: COSMO-E, pre-operational in late 2015, planned to be operational in 2016. Alpine domain, 21 members, 2.2km mesh-size, ICs from KENDA at 2.2 km (same domain) with possible back-up solution during the first year, LBCs from IFS-ENS, model perturbations (SPPT).
Italy (ARPA-SIMC, in collaboration with ARPA Piemonte and CNMCA/CNMCA): COSMO-IT-EPS, pre-operational in 2016, 10 members, 2.8 (2.2) km, ICs and BCs from COSMO-ME-EPS, model perturbations (SPPT), soil moisture perturbations (Loglisci, Bonanno); future: ICs from KENDA.
IMGW: running regularly an ensemble based on COSMO at 2.8 km, soil perturbations: c_soil and (preliminary) surface temperature from initial conditions. BCs and ICs: time-lagged approach from COSMO-PL (7km). Pre-operational planned in late 2015/beginning of 2016.
RHM: planning a 2.2 km ensemble over Moscow region. Model perturbations (SPPT) + variations of parameters of physical schemes. Initial conditions perturbations from NCEP EPS and Hydrometcenter of Russia EPS. Pre-operational in 2017. Additionally some experiments with the same ensemble but relocated to the arctic region are expected (only research).
Ensemble systems at CP scale are now in place in several COSMO countries. This provides a good framework for addressing several open issues.
The Consortium ensemble COSMO-LEPS is currently operational at 7 km of horizontal resolution. Main planned developments are: increase the ensemble size from 16 to 20 members, SPPT physics perturbation (under testing) and run COSMO in single precision.
Needs
During the WG7 meeting held in Offenbach (March 2015), the main need recognised by all the participants was to improve the spread-skill relation of the CP ensembles. In the current implementations this means to increase the spread of the ensembles in terms of near-surface parameters (temperature, wind), in order to better match the forecast errors. It is recognised that the ensemble spread cannot be increased up to reaching the forecast error, because the latter includes also the systematic error, which should be tackled by improving the model, not by increasing the spread. What should be better address by the ensembles is the day-to-day error.
To reach this goal, the general feeling was that we should work on a better representation of the model error:
Additional to improving the model error representation, further studies concerning the use of KENDA for ensemble ICs are needed:
The development of KENDA itself and its “tuning” is addressed in the KENDA-O PP, with which a good exchange is already in place.
Ensemble will be used for several operational duties, more than what is done now, therefore it is needed to continue working about ensemble post-processing and interpretation (and there is the need to coordinate COSMO efforts within the EUMETNET project SRWNP-EPS II):
Last but not least, an important need is to save resources: scientists working in ensemble development are not many, therefore we should use as much as possible what is already available and focus the effort on the most needed developments.
The spread/skill relation of the different ensembles will be studied for the main surface weather parameters, aiming at identify similarities and differences among the ensembles and at highlighting the deficiencies in spread.
After this diagnostic work (and partly parallel to it) the problem of the description of the model error will be addressed by testing and developing model perturbation methodologies. Among these, SPPT and Parameter Perturbations (PP) will be further investigated, also in combination, as well as the new scheme developed by DWD. Possible developments will be identified and tested.
Surface/soil perturbation will also be further studied and tested, and the schemes developed in the COTEKINO PP will be tested in a “complete ensemble” framework.
An action in collaboration with WG4 will also be initiated, aiming at developing post-processing methods for the CP ensembles, especially focused on selected phenomena and intense events. Calibration of CP ensembles will also be carried out.
Finally, the use of ICs obtained from KENDA analyses to initialize the CP ensemble will also be tested, focusing on some issues like member selection and comparison with cheaper techniques for providing ICs to the ensembles.
Short report showing the results of the study of the spread/skill relation, showing the spread/skill behaviour of the different CP ensembles and of COSMO-LEPS, and proposing solutions to the spread/skill mismatch.
Short report about the impact of the different model perturbations on the ensemble, with a plan for the development work still needed on model perturbations.
New perturbation techniques for surface/soil of the COSMO model. Short report on the performance of the perturbation methods. The code/algorithm is made available to the Consortium.
Method for ensemble calibration for temperature and methods/algorithms for preparation of products from the ensembles for the prediction of thunderstorm and fog, in form of description and code.
Method/algorithm for member selection from an ensemble. Report on the impact of IC from KENDA and from simpler methods.
Task 1 Study of the spread/skill relation in the ensembles
This task deals with studying the spread/skill relation deficiency in the different ensembles, focusing on near surface weather parameters. The analysis will be extended also to the lower part of the troposphere, in order to better understand the mechanism for error propagation, by checking the spread/skill relation on vertical profiles (against radiosondes).
Estimated needed resources: 0.6-0-8 FTE.
The task is subdivided in 3 subtasks:
The aim of this task is to understand which factors determines the lack of spread for near-surface weather parameters and to propose solutions in terms of which elements of the modelling system should be perturbed to account for this uncertainty.
Task 2 Model perturbation
Also depending on the outcome of the previous task, model perturbations will be further developed. The methodologies already available in the Consortium will be further studied and combined and possible developments will be identified.
Estimated needed resources: 0.5 FTE, then possible developments should be planned.
The task is subdivided in 3 subtasks:
The aim of this task is to make more effective use of the model perturbation methods which are already available and identify possible developments of those.
Task 3 Lower boundary perturbation
The perturbation of the model lower boundary (surface/soil) will be further developed. New research will be carried out on new perturbations, and the methodologies developed in the COTEKINO PP will be consolidated and tested in a “complete ensemble” framework, in combination with other perturbations.
Estimated needed resources: 2 FTE.
The task is subdivided in 3 subtasks:
The aim of this task is to provide methods for lower boundary perturbations ready to be put in operations in the CP ensembles.
Task 4 Post-processing and interpretation
In collaboration with WG4, a task on post-processing and interpretation of the CP ensemble is also activated.
Work on calibration of ensemble output, though recognised as needed, is still not planned in the project and FTEs to carry out this work have not been identified.
The work of this task will address mainly the development of methods to make effective use of CP ensemble output, including the development of probabilistic products for selected weather phenomena, in coordination with the SRNWP-EPS II project.
Estimated needed resources: 2-3 FTEs.
This task is divided in two main branches:
The aim of this task is to propose and develop method for effective use of ensemble output at high resolution.
Task 5 Initial Conditions for the CP ensembles
The ICs to be provided to the CP ensemble members will also be addressed, by working on the method to derive them from the KENDA analyses and by studying alternative cheaper options to provide them.
Estimated needed resources: 1.5 FTE.
The task is divided in 3 subtasks:
The aim of this task is to consolidate the method used in the CP ensemble to determine the ICs of the runs.
to be filled
ARPA Piemonte: Nicola Loglisci; action 3 (0.35 FTE per year)
ARPA SIMC: Chiara Marsigli, Andrea Montani, Tommaso Diomede; actions 1 (0.2 FTEs), 2 (0.2 FTEs in 1 year), 4 (0.3 FTEs in 2 years), 5 (0.4 FTEs in 2 year)
CNMCA/CNMCA: Francesca Marcucci and possible fellowship; action 4 (0.5 FTE per year)
DWD: Christoph Gebhardt; actions 1, 2 and 5 (0.1 FTE per year in total)
IMGW: Andrzej Mazur, Grzegorz Duniec, Witold Interewicz; actions 1 (0.15 FTE per year), 3 (1 FTE per year) and 4 (0.3 FTE per year)
MeteoSwiss: André Walser, Kristina Klasa; actions 1 (0.1 FTE in 2 years), 2 and 3 (0.6 FTEs in 18 months), 5 (0.4 FTE in 2016, Master Thesis)
RHM: Elena Astakhova; Dmitry Alferov actions 1 (0.1 FTEs), 2 (0.1 FTEs per year) and 5 (0.15 in two years).
to be filled
------------------------------------------------------
Extension: from Sep 2017 to Feb 2018
------------------------------------------------------
Project Extension: COSMO year 2017-2018
A short prolongation of PP SPRED is proposed in order to complete Tasks of the existing project and provide the related deliverables. The extension of PP SPRED is requested due to a delay in Task 1 “Study of the spread/skill relation in the ensembles”, to a lack of resources which are now available in Task 2 “Model perturbation”, due to a completion of Task 3 “Lower boundary perturbation”. On top, the work carried out in Task 4 “Post-processing and interpretation” is an on-going work both at CNMCA and IMGW. Therefore it is suggested to consider this Task for an extension to February 2018.
The project will be finalized after the successful completion of the Tasks listed below and the final project report will be delivered. The requested extension is until the end of February 2018 . Part of the unused FTEs are moved to the extension period of the project (0.5 FTEs were unused). Other FTEs are newly requested because they were not actually planned in the original PP (even if needed to carry out the work intended to) due to lack of resources.
The total requested resources for the COSMO year 2017-2018 are 0.9 FTE.
Task 1: Study of the spread/skill relation in the ensembles
It was planned to prepare a short report showing the results of the study, with proposals of solutions to the spread/skill mismatch. This was not completed, due to lack of resources. The work will be completed until December 2017.
Goal: write a short report showing the results of the study, with proposals of solutions to the spread/skill mismatch.
Subtasks
1.3: Write a short report.
Contributors:
1.3: C. Marsigli (Arpae), all contributors to Task 1
Start 09.2017 - End 12.2017: 0.03 FTE (CM) + review of all contributors
Task 2: Model perturbation
Based on the results of Task 1 (study of the spread/skill relation), model perturbations could be further developed or new methods could be tested.
For the period September – February 2017 MeteoSwiss has the availability of resources for investigating the impact of the Boundary Layer perturbation method recently introduced in the COSMO model following the work by K. Kober (perturbations proportional to the variance of T, W and QV fluctuations calculated in the turbulence scheme). This investigation was already planned (original project plan) but no resources to carry it out were available before therefore FTEs were not actually planned in the task.
In the period September – February 2017, resources are available at RHM for starting the work on testing the application of the stochastic pattern generator (SPG), developed by M.Tsyrulnikov and D. Gayfulin, for generation of model perturbations. This work was also foreseen in the PP but it was delayed due to lack of resources, same as for the previous point. Work will probably extend after the end of this PP, therefore it will be considered as object of a successive PP.
The evaluation of the EM scheme (model for model error) of DWD continues after the original end of the PP SPRED since more investigations and tuning are needed.
Goal:
- Investigations of convective cases with additional PBL perturbations in COSMO-E.
- test the applicability of the stochastic pattern generator (SPG) for generation of model perturbations.
- continue the evaluation of the EM scheme
Subtasks
2.3: developments of the techniques.
Contributors:
2.3: A. Walser, S. Klaus (MCH)
Start 09.2017 - End 02.2018: 0.4 FTE
2.3: E. Astakhova, D.Gayfulin, M.Tsyrulnikov, D. Alferov (RHM)
Start 09.2017 - End 02.2018: 0.15 FTE
2.3: C. Gebhardt (DWD)
Start 09.2017 - End 02.2018: 0.1 FTE
Task 3: Lower boundary perturbation
IMGW planned the perturbation of the soil temperature. In order to bring this work to completion, it should be evaluated the correlation of amplitude of perturbation with the soil type and/or land cover (using the preceding research).
Goal:
- Complete soil temperature perturbation method.
Subtasks
3.1: Investigate and develop new perturbation methods.
Contributors:
3.1: Grzegorz Duniec, Andrzej Mazur (IMGW)
Start 09.2017 - End 02.2018: 0.15 FTE
Task 4: Post-processing and interpretation
The work of this task addresses the development of methods to make effective use of CP ensemble output, including calibration and development of probabilistic products for selected weather phenomena. The work is on-going at IMGW and CNMCA and it is needed for effective use of ensembles.
Goal:
- Continue the work of the Task.
Subtasks
4.1: calibration of CP ensembles.
4.2: development of probabilistic products for specific tasks.
Contributors:
4.1: Grzegorz Duniec, Witek Interewicz, Andrzej Mazur (IMGW)
Start 09.2017 - End 02.2018: 0.15 FTE
4.2: Mario Papa, Francesca Marcucci (CNMCA)
Start 09.2017 - End 02.2018: 0.15 FTE
Extension of PP SPRED until February 2018 is proposed due to the delay in the completion in some Tasks mainly due to lack of resources and to the availability of resources for this period. Estimated resources (in terms of FTEs): total of 0.9 FTE is required.