Book of Abstracts

DOI: 10.5676/DWD pub/nwv/iccarus 2021

Max-Planck-Institut für Meteorologie DOI: 10.5676/DWD pub/nwv/iccarus 2021

The CC license “BY-NC-ND” allows others only to download the publication and share it with others as long as they credit the publication, but they can’t change it in any way or use it commercially.

Publisher Editors Deutscher Wetterdienst Daniel Rieger, DWD, Business Area “Research and Development” [email protected] Frankfurter Straße 135 Christian Steger, DWD, 63067 Oﬀenbach [email protected] www.dwd.de Contents

1 Invited & Solicited Talks7 Developing the ICON-A model for direct QBO simulations on GPUs...... 7 ICON-Land/JSBACH: A Framework for the Land Component in the ICON Modeling System...... 7 Advances in forecast quality achieved with ICON-D2, and upcoming further development steps...... 8 Once upon a Time at DWD - The COSMO-Model...... 8 HPC Infrastructure of ICON for high-res simulations...... 9 Waves and drag processes in atmospheric models...... 9 Status and applications of the modelling system ICON-ART...... 9

2 Special Session: ICON-Seamless 11 Roadmap to ICON-Seamless...... 11 The ICON-System at MPI-M and its application for Climate Simula- tions...... 11 Sapphire - towards the next generation of Earth System models...... 11

3 Climate Model Applications 13 Tuning ICON-NWP for Climate applications...... 13 Impact of Urban Canopy Parameters on a Megacity’s Modelled Thermal Environment. 13 COSMO-CLM Performance and Projection of Daily and Hourly Tem- peratures Reaching 50◦C or Higher in Southern Iraq...... 14 COSMO-CLM Russian Arctic hindcast 1980 – 2016: modelling workflow, evaluation techniques and preliminary assessment...... 14 Exploring hail and lightning mechanisms over the Alpine-Adriatic region using HAILCAST and LPI in a convection-resolving model...... 15 Development and Quality Evaluation of an Operational Ensemble-based Regional Reanalysis System...... 16 Using regional reanalysis and satellite data to estimate renewable energy production for applications in the German transport infrastructure...... 16 Current and future large-scale dynamical influences on Lake Victoria’s thunderstorms. 17 Austrian Test Case: CCLM results in different warming scenarios...... 17 Ten years of reanalysis activities at DWD: review and outlook...... 18 Evaluation of Convection-permitting simulations for Germany using gridded observations on high temporal resolution...... 18 Current convection-permitting COSMO-CLM simulations for Germany...... 19 4 Clouds, Chemistry, Aerosol and Radiation 20 Application of Liquid and Ice Clouds Optical Properties Parametriza- tions in Numerical Weather Prediction Models...... 20 Increasing Resolution and Resolving Convection Improve the Simulation of Cloud-Radiative Effects Over the North Atlantic...... 20 Modeling methane from the North Sea region with ICON-ART...... 21 New radiation scheme ecRad and impact of uncertainties in radiation and clouds for ICON...... 21 PerduS and PermaStrom: Daily Mineral dust forecasts using ICON-ART...... 22 Raikoke eruption 2019: Improving ash and SO2 forecasts by coupling ICON-ART to a plume rise model...... 23 The effect of spring 2020 lockdown in Moscow on the cloud characteristics according to COSMO-Ru simulations and measurements...... 24 How did they get there? The Importance of the Particle History for the Ice Habit...... 24 An evaluation of kilometre scale ICON simulations of mixed-phase stratocumulus over the Southern Ocean during CAPRICORN...... 25 Prognostic Ozone for ICON: Enabling UV Forecasts...... 26 Efficient generation of synthetic near-infrared satellite images...... 26 Clouds and precipitation in ICON-LAM from radar and SEVIRI observations...... 27 Improved Ice Aggregation Formulation in the Seifert-Beheng Two- Moment Microphysics Scheme...... 28 Investigation of aircraft icing sensitivities using a parameterization respecting droplet spectra...... 28 Options and extensions for the stochastic shallow convection scheme in ICON...... 29 Quantification of the effects of aerosol deposition on snow by ensemble simulations... 29 Using Microphysical Piggybacking in ICON to explain the sensitivity of simulated convection to the choice of microphysics scheme...... 30

5 Data Assimilation 31 Evaluation of ICON’s model cloud fields using simulated and observed visible satellite images...... 31 Assimilation of remote sensing profiler observations at MeteoSwiss...... 31 EnVAR for ICON-LAM: observations and quality control...... 32 Adaptive non-linear bias correction for visible reflectance data...... 32 Assimilating 3D radar information at convective scales at DWD...... 33 Improving radiation forecasts by assimilating visible satellite images in ICON-D2...... 33

6 Dynamics and Numerics 35 A locally smoothed vertical coordinate to improve fog and low stratus forecasts..... 35 COSMO-EULAG dynamical core with long timestep capability...... 35 ICON model on GPU...... 35 Towards a transient 3D gravity wave parameterization in ICON...... 36

7 Model Infrastructure and Data Processing 37 Zarr formatted Climate Data in DKRZ’s Swift Cloud Object Storage...... 37 The ICON single column mode...... 37

4 8 NWP Model Applications and Case Studies 38 Case study of an Arctic atmospheric river with the ICON model...... 38 Impact of different external parameters on Turin UHI with COSMO at 1km...... 38 Towards understanding the role of uncertainty in microphysical processes for warm conveyor belt ascent using microphysical heating rates along online trajectories in ICON...... 39 Lagrangian analysis of an Alpine Foehn event...... 39 High-resolution Simulations of Atmospheric CO2 with ICON/MESSy...... 40 A preliminary evaluation of the near-surface evolution of Foehn events in COSMO-1...... 41 How do mesoscale weather systems interact with off-shore wind farms: A study for the Kattegat (Denmark, Sweden) with the mesoscale model COSMO-CLM and satellite scatterometer data...... 41 Pollen forecasts using ICON-ART in a limited area mode...... 42 Representation of the urban environment in ICON-LAM...... 43 SINFONY - the combination of Nowcasting and Numerical Weather Prediction on the convective scale at DWD...... 43 Potential links between tropospheric and stratospheric extremes during the winter of 2019/20...... 44 COSMO-GHG simulations of 14C to support the design of a radiocarbon measurement network in Europe...... 44 Sensitivity of gravity wave drag in seasonal experiments with ICON: Stratospheric dynamics and pathways...... 45 Postprocessing of COSMO and IFS ensemble predictions for providing seamless forecasts...... 46 Evaluation of ICON-LAM Forecasting of a Strong Rain Eevent at the Coast of Sao Paulo State in Brazil...... 46 Improving forecasts of wind resources by including two-way coupling between atmospheric flow and offshore wind farms...... 47 Interaction between stratospheric equatorial waves and gravity waves and its implication in QBO simulation...... 48 First steps towards 1 km horizontal resolution over Germany...... 48

9 Planetary Boundary Layer 50 A preliminary analysis of the impacts of small-scale orography on the stable atmospheric boundary layer...... 50 A Budget-Based Turbulence Length Scale Diagnostic...... 50 Evaluation of thermally driven local winds in the Swiss Alps simulated by COSMO-1...... 51 Slope winds in the convective boundary layer over mountainous terrain: LES results and parameterization approach...... 51 Simulations of the Arctic atmospheric boundary layer around the MOSAiC drift track using ICON-LEM...... 52 An improved sea ice parameterization and tile approach in CCLM...... 52 The two-energies turbulence scheme...... 53

5 10 Predictability and Ensemble Systems 54 Choosing the Optimal Sub-Ensemble of Boundary Conditions to Drive Convection Permitting Ensemble...... 54 Statistical and machine learning methods for postprocessing ensemble forecasts of wind gusts...... 54 Predictability analysis and veriﬁcation of the Lightning Potential Index (LPI) in the COSMO-D2 high resolution EPS...... 54 An estimation of the intrinsic predictability limit and required improvements to approach it...... 55

11 Soil, Vegetation and Ocean 56 Marine extreme events in high-resolution coupled model simulations...... 56 Convergence of Richards Equation and Implications for Inﬁltration and Water Propagation...... 56 PT-SAINT – Outcomes and outlook for multi-layer snow modelling in COSMO..... 57 Implementing ICON in TSMP – Coupling strategy and applications...... 58 Projected future changes in Vb-cyclone precipitation and moisture source region characteristics...... 58

12 Verification (NWP) and Evaluation (Climate) 60 Preliminary tests with ICON-LAM and comparison with COSMO-LM at high resolution over Italy...... 60 Comparison of calculated by COSMO-SIB and ICON-SIB models temperature profiles in the boundary layer with available observation data for Novosibirsk city...... 60 Evaluation of a high-resolution dynamical downscaling of ERA5 with COSMO-CLM for the North Sea...... 61 Polarimetric radar forward operator for model validation and data assimilation..... 62 Object based verification of radar-reflectivities on the convective scale...... 62 Temperature and radiation biases in COSMO-REA6 and their role for representation of extreme events...... 63 Multilayer cloud conditions in trade wind shallow cumulus – confronting two ICON model derivatives with airborne observations...... 64

6 Invited & Solicited Talks

Developing the ICON-A model for direct QBO simulations on GPUs M. A. Giorgetta(1), P. Adamidis(2), D. Alexeev(3), V. Clément(4), J. F. Engels(2), M. Esch(1), L. Kornblueh(1), X. Lapillonne(5), P. Marti(4), R. Pincus(6), S. Rast(1), D. Reinert(7), R. Schnur(1), W. Sawyer(8), and U. Schulzweida(1) (1) Max-Planck-Institut fürMeteorologie, (2) Deutsches Klimarechenzentrum, (3) NVIDIA, (4) C2SM, (5) MeteoSwiss, (6) University of Colorado Boulder, (7) Deutscher Wetterdienst, (8) Centro Svizzero di Calcolo Scientifico Classical numerical models for the global atmosphere, as used for numerical weather forecasting or climate research, have been developed for conventional central processing unit (CPU) architectures. This hinders now the employment of such models on current top performing supercomputers, which achieve their computing power with hybrid architectures, mostly using graphics processing units (GPUs). Thus also scientific applications of such models are restricted to the lesser computer power of CPUs. Here we present the development of a GPU enabled version of the ICON atmosphere model motivated by a research project on the quasi-biennial oscillation, a global scale wind oscillation in the equatorial stratosphere that depends on a broad spectrum of atmospheric waves, which origins from tropical deep convection. Resolving the relevant scales, from a few km to the size of the globe, is a formidable computational problem, which can only be realized now on top performing supercomputers. This motivated porting the ICON model in the specific configuration needed for the research project to the GPU architecture of the ”Piz Daint” computer at the Swiss National Supercomputing Centre, which makes computing resource available through the Partnership for Advanced Computing in Europe (PRACE). The ported code achieves a single node GPU vs. CPU speed-up factor of 6.3, and now allows global experiments at a horizontal resolution of 5 km on 1024 nodes with a turnover of 30 simulated days in 18 hours. The application shows a reasonable weak and strong scaling such that higher resolutions or higher node numbers could be used in future applications.

ICON-Land/JSBACH: A Framework for the Land Component in the ICON Modeling System R. Schnur Max Planck Institute for Meteorology The ICON-Land framework is the land component of the ICON Earth System Model providing the surface boundary conditions for the atmosphere. As a framework, it was designed specifically for the flexible and user-friendly integration of alternative process and surface descriptions in different experiment configurations, coupled to the atmosphere or offline as Dynamic Global Vegetation Model. JSBACH is one particular implementation containing, in addition to the physical processes, also descriptions of biogeographical and biogeochemical processes.

7 In the first part of the presentation, the processes and configurations that are implemented in JSBACH are presented, as well as current work and plans. Ideas on possible synergies between ICON-Land/JSBACH and TERRA, the second land model in ICON used for Numerical Weather Prediction,are discussed as well as strategies to serve the different communities interested in the land component of ICON for (global) Earth System Modeling, (regional) Numerical Weather Prediction, seasonal prediction, or ecosystem modeling.

In the second part, the design and programming infrastructure of ICON-Land are outlined together with examples of how to use the framework to implement processes or surface descriptions. Finally, the port of ICON-Land/JSBACH to GPUs using OPENACC and the CLAW Domain Speciﬁc Language is described.

Advances in forecast quality achieved with ICON-D2, and upcoming further development steps G. Z¨angl Deutscher Wetterdienst On February 10, 2021, the operational convection-permitting weather forecasting model COSMO- D2 and the related ensemble prediction system was replaced with ICON-D2. Thanks to a large set of improvements in the model physics, the data assimilation system, and the coupling between model and data assimilation, the transition to ICON-D2 comes along with a substantial improvement in forecast quality for nearly all variables monitored by our veriﬁcation system. Particularly large progress is achieved for variables for which COSMO-D2 was known to have weaknesses, e.g. 2-m temperature, 2-m humidity and 10-m wind gusts. For these variables, the monthly averaged RMSE values typically decrease by about 20% with peaks exceeding a factor of 1.5 in some months during afternoon and the evening transition phase. For most other variables, typical error reductions are on the order of 10%.

The next development step of ICON, affecting the global system as well as ICON-D2, will include a major upgrade of the model physics, including a change of the radiation scheme from RRTM to ECRad, the usage of satellite-derived longwave surface emissivities and an energy conservation bug fix in the physics-dynamics coupling of the turbulence scheme that necessitated a significant amount of retuning. This upgrade will lead to a further weak-to-moderate (2-4%) improvement of forecast quality for most quantities. Error reductions exceeding 10% will be achieved for global radiation at the surface, which directly benefits from the more accurate radiation scheme. More details on this upcoming model upgrade will be provided in the presentation.

Once upon a Time at DWD - The COSMO-Model U. Sch¨attler Deutscher Wetterdienst After about 4 years of development, operational runs with the COSMO-Model (called Local Model LM by that time) were started at DWD on 1st of December, 1999, and ended in February 2021. Being used for more than 20 years for oﬃcial duties set a new record at DWD.

In this presentation we will give a short history of the COSMO-Model. We will highlight its

8 strengths but also some issues where it could have done better. We will also present the status of the last oﬃcial model version 6.0 and an outline, how the transition to ICON-LAM will be organized in the near future.

HPC Infrastructure of ICON for high-res simulations P. Adamidis DKRZ The ICON model is increasingly being used in high resolution simulations, in order to resolve small scale physical processes in the atmosphere as well as in the ocean. This type of simulations are computationally very intensive. The focus of the presented work is on the optimization of the ”time-to-solution”, in order to increase the eﬀective throughput of such experiments with ICON on modern Peta-/Exascale supercomputing systems. These systems are massively parallel computer architectures, consisting of complex, heterogeneous hardware with distributed memory, and ICON’s scalability plays an important role.

The basic prerequisite for using such supercomputers eﬃciently is memory scaling. However, the ”time-to-solution” includes not only the model calculations (execution), but also the preparation and output of the simulation results on the ﬁle system. For this, ICON has been expanded to include 2 additional HPC infrastructure components, the communication library YAXT and the parallel I/O library CDI-PIO. This increased both, the functionality of the model and the ability to perform high-resolution model calculations.

Waves and drag processes in atmospheric models S. Derbyshire, A. van Niekerk, and P. Sheridan Met Oﬃce (UK)

This talk will informally review the representation of waves inatmospheric models, especially those relevant to drag processes. We will consider both explicit modelling (and numerical effects which influence the effective resolution) and parametrized wave representations. Current international intercomparison work (including the Met Office UM, ICON and other leading models) is starting to show in more detail the differences in representation of orographic gravity waves between different models, and at different scales. An increasingly important aspect of this exercise is to understand better the behaviour in regimes where the waves are partially but not fully resolved.

Status and applications of the modelling system ICON-ART B. Vogel on behalf of the ICON-ART developers and users KIT Karlsruhe The non-hydrostatic global modelling system ICON is used for numerical weather prediction, climate projections, and for large eddy simulations. The integrated modelling framework ICON- ART (ICOsahedral Nonhydrostatic - Aerosols and Reactive Trace gases) extends the numerical modelling system ICON by modules for gas phase chemistry, aerosol dynamics and related feedback processes. ICON-ART includes a ﬂexible tracer structure which easily allows adding or removing

9 tracers with an xml framework. Based on this tracer structure we are running ICON-ART. The presentation gives an overview on the currently included emission schemes, chemistry, and aerosol types and already realized interactions with radiation and clouds. Examples of research and operational applications showing the capabilities of the model system are presented. Amongst them are operational pollen forecast, quasi-operational forecast of mineral dust, methane emission of oil and gas platforms, vegetation ﬁres, and optical properties of internally mixed aerosols.

10 Special Session: ICON-Seamless

Roadmap to ICON-Seamless B. Fr¨uh(1), P. Korn (2), W. M¨uller(2), and R. Potthast (1) (1) Deutscher Wetterdienst, (2) Max-Planck Institute for Meteorology

We outline the project for the development of a new Earth system model for weather and climate predictions (in the following ICON-Seamless) developed jointly by the Max Planck Institute for Meteorology (MPI-M), the Karlsruhe Institute of Technology (KIT), German Cli-mate Computing Centre (DKRZ) and Deutscher Wetterdienst DWD. It is based on the com-ponents of the numerical weather prediction model ICON-NWP of DWD and the ocean-sea ice model ICON-O of MPI-M. ICON-Seamless comprises the aerosol and trace gas model ICON-ART of the KIT. ICON-Seamless will enable us to carry out climate forecasts and pro-jections from medium resolution up to ultra-high resolution climate simulations. The role and necessary extension of NWP physics for climate simulations will be examined in detail. We pursue a model development approach that integrates the needs of the NWP and the climate research community for both model and data assimilation. The impact of each development step and component on DWD’s operational tasks in weather prediction and on the quality of the climate simulation will be evaluated.

The ICON-System at MPI-M and its application for Climate Simulations P. Korn(1), and W. Müller(1) Max Planck Institute for Meteorology The talk gives an overview of ICON-related modelling activities at the Max-Planck Institute for Meteorology (MPI-M). Special emphasis is placed on coupled ICON configurations where the atmospheric component ICON-A is coupled to the ocean-sea-ice model ICON-O. Simulations are shown that cover over long-term simulations at coarse resolution up to short-term experiments on ultra high-resolution. We demonstrate that ICON is able to simulate a broad range of scales of the climate system and opens new scientific opportunities. The talk concludes with an outlook on forthcoming modelling initiatives.

Sapphire - towards the next generation of Earth System models D. Klocke(1,2), and Sapphire-team(2,3) (1) HErZ, DWD (2) MPI-Met (3) DKRZ

The Sapphire version of ICON pioneers the representation of the coupled climate system at resolutions that explicitly represent convective storms in the tropics and meso-scale ocean eddies in the extra-tropics. The goal is to unify existing approaches for regional large eddy simulations and for strongly parameterized coupled global simulations. Model development is driven by ambitious scientific experiments that target scientific questions where global km-scale resolutions simulations can contribute to our scientific understandingand technical capabilities of Sapphire are extended. Initial experimental targets include global short-period uncoupled (2.5 km resolution)

11 and coupled (5 km resolution) simulations, as well as longer higher-resolution experiments over selected regions. As computer performance increases in the future, the length of global experiments, or the domain size of regional experiments, or the grid resolution can be increased. The presentation will outline the Sapphire strategy and present results from selected experiments that shape the model development.

12 Climate Model Applications

Tuning ICON-NWP for Climate applications T. Van Pham(1), S. Brienen(1), K. Fr¨ohlich(1), and B. Fr¨uh(1) (1) Deutscher Wetterdienst

At Deutscher Wetterdienst (DWD), the global earth system model MPI-ESM has been used in recent years for climate forecasting at different time scales from seasonal and decadal forecasts to climate projection. However, in recent years, the new numerical weather prediction model ICON- NWP was introduced and used in operational forecast at DWD. That gave us an opportunity to realise our vision of a united forecasting system at DWD in the direction of seamless prediction. In the next years, we want to prepare the ICON-NWP for climate application at global scale. First step is to investigate its suitability for running at climatological space and time scales and to tune this atmospheric model for climate simulations. Second step is to couple ICON-NWP with the global ocean model ICON-O. A model tuning statergy was developed as a guideline for all of the ICON-NWP tuning work at DWD, now and the future. First long climate experiment with ICON-NWP showed large biases in radiation balance at the top of atmosphere. It implies that using the ICON model with NWP physics for climate simulations is technically possible but finding a proper model configuration for our purpose would require serious effort. Model parameter sensitivity was tested to serve the tuning work later on. Results of these sensitivity investigations as well as of the first tuning experiments will be shown.

Impact of Urban Canopy Parameters on a Megacity’s Modelled Thermal Environment M. Varentsov(1,2,3,4), T. Samsonov(1,2,4), and M. Demuzere(5) (1) Research Computing Center/Faculty of Geography, Lomonosov Moscow State University, Russia, (2) Hydrometeorological Research Center of Russian Federation, Russia (3) A.M. Obukhov Institute of Atmospheric Physics, Russia, (4) Moscow Center for Fundamental and Applied Mathematics, Russia, (5) Department of Geography, Ruhr-University Bochum, Germany

Urban canopy parameters (UCPs) are essential in order to accurately model the complex interplay between urban areas and their environment. This study compares three different approaches to define the UCPs for Moscow (Russia), using the COSMO numerical weather prediction and climate model coupled to TERRA URB urban parameterization. In addition to the default urban description based on the global datasets and hard-coded constants (1), we present a protocol to define the requiredUCPs based on Local Climate Zones (LCZs) (2) and further compare it with a reference UCP dataset, assembled from OpenStreetMap data, recent global land cover data and other satellite imagery (3). The test simulations are conducted for contrasting summer and winter conditions and are evaluated against a dense network of in-situ observations. For the summer period, advanced approaches (2) and (3) show almost similar performance and provide noticeable improvements with respect to default urban description (1). Additional improvements are obtained when using spatially varying urban thermal parameters instead of the hard-coded constants. The LCZ-based approach worsens model performance for winter however, due to the underestimation of the anthropogenic heat flux (AHF). These results confirm the potential of LCZs in providing

13 internationally consistent urban data for weather and climate modelling applications, as well as sup- plementing more comprehensive approaches. Yet our results also underline the continued need to improve the description of built-up and impervious areas and the AHF in urban parameterizations.

COSMO-CLM Performance and Projection of Daily and Hourly Temperatures Reaching 50◦C or Higher in Southern Iraq Y. Levi(1), and Y. Mann(2) (1) Israel Meteorological Service, (2) Department of Middle Eastern Studies, Bar-Ilan University

Fortunately, extreme temperatures reaching 50 C are not common on our planet. The capability of the consortium for small-scale modelling regional climate model (COSMO-CLM), with 0.44 resolution, to project future trends of an extremely hot environment with direct model output (DMO) is questioned. The temperature distribution of COSMO-CLM output driven by reanalysis and RCP4. 5 scenario in southern Iraq was remarkably good, with a slight temperature overestimation, compared to the overlapping observations from Basra airport. An attempt to enhance the DMO with a statistical downscaling method did not improve the results. The COSMO-CLM projection indicates that a very sharp increase in the number of consecutive hours and days with the temperature reaching 50 C or higher will occur. During 1951-1980, consecutive hours and days reaching 50 C were rare events. By the end of the century, the projected climate in southern Iraq contains up to 13 consecutive hours and 21 consecutive days reaching 50 C or higher. As the average projected temperature will increase by 2 C compared to the recent climate, new records may be expected. However, the major climate change feature is the increase in consecutive hours and days of very high temperatures. These ﬁndings require adaptation measures to support future habitation of the region.

COSMO-CLM Russian Arctic hindcast 1980 – 2016: modelling workflow, evaluation techniques and preliminary assessment V. Platonov(1), and M. Varentsov(2,3,4,5) (1) Department of Meteorology and Climatology, Lomonosov Moscow State University, (2) Research Computing Center, Lomonosov Moscow State University, (3) A.M. Obukhov Institute of Atmospheric Physics, (4) Hydrometeorological Research Center of Russian Federation, (5) Moscow Center for Fundamental and Applied Mathematics Detailed long-term hydrometeorological dataset for Russian Arctic seas was created using hydrodynamic modelling via regional nonhydrostatic atmospheric model COSMO-CLM for 1980 - 2016 period with 12 km grid. Many test experiments with different model options for summertime and wintertime periods were evaluated to determine the best model configuration. Verification has showed that optimal model setup included usage of ERA-Interim reanalysis as forcing data, new model version 5.05 with a so-called ICON-based physics and spectral nudging technique. Final long-term experiments were simulated on the MSU Supercomputer Complex “Lomonosov-2” become more than 120 Tb data volume excluding many side files.Primary evaluation of obtained dataset was done for surface wind and temperature variables. There are some mesoscale details in wind sped climatology reproduced by COSMO-CLM dataset including the Svalbard, Severnaya Zemlya islands, and the western coast of the Novaya Zemlya island. At the same time, high wind

14 speed frequencies based on COSMO-CLM data increased compared to ERA-Interim, especially over Barents Sea, Arctic islands (Novaya Zemlya) and some seacoasts and mainland areas. Comparison of two periods (1980 - 1990 and 2010 - 2016) has shown wind speed frequencies above 20.8 m/s has been decreased in the last decade over the Novaya Zemlya, southwest from Svalbard, middle Siberia inlands, however, it has been increased over Franz Josef Land and Severnaya Zemlya. Large-scale temperature climatology patterns have shown a good accordance between ERA-Interim and COSMO-CLM datasets. Significant regional details in temperature patterns manifested in relief and lakes, e.g., over Scandinavian mountains, Eastern Siberian and Taymyr highlands, Novaya Zemlya ranges. The added value in the 1% temperature percentile patterns is more pronounced, especially in the mountainous Eastern Siberia. 37-year period and a large Arctic region covered by dataset with 0.108 grid spacing required a lot of memory. This would be a certain technical issue to share all these data using any host, HTTP or FTP services. At the first stage, we have prepared a subset that included 7 main surface variables within the entire 37-years period and uploaded it to the https://www.figshare.com service (https://doi.org/10.6084/m9.figshare.c.5186714). We plan to extend the list of accessible variables consistently and hope these data would be useful and appropriate for Arctic climate research. The nearest prospect of the COSMO-CLM Russian Arctic dataset application is its comparison with other appropriate datasets including reanalyses, satellite data, observations, etc. This will provide important and useful information about opportunities and restrictions of this dataset regarding different variables and specific regions, outline the limits of its applicability and get framework of possible tasks.

Exploring hail and lightning mechanisms over the Alpine-Adriatic region using HAILCAST and LPI in a convection-resolving model R. Cui(1), N. Ban(2), M.-E. Demory(1), and C. Schär(1) (1) ETH Zurich, Institute for Atmospheric and Climate Science, (2) University of Innsbruck, Department of Atmospheric and Cryospheric Sciences The north and south of the Alps as well as the eastern shores of the Adriatic Sea are hotspots of frequent hail and lightning. Hail and lightning are associated with severe convective storms that happen under the large-scale forcing of surface fronts, upper-level fronts, convergence zones, or local thermal-topographic forcing. Convection-resolving models are run at km-scale resolution, which improves the representation of topography. Moreover, they can explicitly resolve deep convection, thus reducing the uncertainties related to the use of deep convection parameterization in lower resolution models. Both aspects are beneficial for studying processes that drive severe convective storms over mountainous regions.In this study, we analyze convection-resolving simulations of 8 cases performed with the GPU version of the Consortium for Small-scale Modeling (COSMO- crCLIM) regional climate model at 2.2 km horizontal grid spacing over the Alpine-Adriatic region. The cases are selected according to their impacts (size of hailstones, number of lightning strikes and damages). For analysis of hail and lightning, we use the one-dimensional hail growth model HAILCAST and the lightning potential index LPI implemented into the COSMO model, and compare it with observed hail properties and lightning flashes. In addition, we look into key

15 variables for hail formation, including temperature, humidity, CAPE and CIN, bulk wind shear and updraft helicity. By performing a detailed analysis, we identify several environments that are favorable for strong convection and associated hail and lightning, such as ”loaded gun” sounding, conditionally unstable layer and intrusion of dry air aloft. Comparing model simulations with available observations yields very good performance of the model for the simulation of precipitation, hail and lightning. However, there is lower predictability for localized deep convection events driven by local thermal-topographic forcings.

Development and Quality Evaluation of an Operational Ensemble-based Regional Reanalysis System T. R¨osch(1), M. Borsche(1), F. Kaspar(1), and R. Potthast(1) (1) DWD

In 2011 the DWD began the development of a regional reanalysis together with partners from the Universities of Bonn and Cologne within the HErZ initiative. In the meantime, the operational service has been taken over by the DWD. The regional reanalysis system COSMO-REA6 was used to produce data sets covering 01/1995 - 09/2019. They cover Europe in a spatial resolution of 6 km. The data sets have proven high quality as shown within diﬀerent activities, in particular in the project UERRA or in applications by external users especially from the sector of renewable energies.Despite the good results, COSMO-REA6 represents no longer the state of the art and will be superseded in the next years by an ensemble-based reanalysis system based on ICON and up-to-date data assimilation schemes. Here we present the concept of the new system and the current state of the development.

Using regional reanalysis and satellite data to estimate renewable energy production for applications in the German transport infrastructure J. Osterm¨oller(1),D. Niermann(1), and F. Kaspar(1) (1) Deutscher Wetterdienst

The enhanced development of renewable energy in transport and infrastructure is one topic of the project “Network of Experts” by the Federal Ministry of Transport and Digital Infrastructure (BMVI). The high energy demand of rail, road and waterways needs to be lowered or replaced by renewable energy sources in order to reduce further greenhouse gas emissions and fulfill climate protection objectives.In the first research phase of the topic, relevant knowledge was gained concerning the identification of energy consumers and potentials for savings. For the simulation of electricity generation, different meteorological data sets were examined with regard to solar radiation and wind. It was shown that the regional reanalysis COSMO-REA6 is able to produce realistic mean 10 m winds speeds and wind gusts and is thus suitable for wind energy applications. Other studies have also evaluated the quality at typical hub-heights of wind energy converters ( 100 m). For radiation instead, it was shown that satellite data (SARAH-2) is more appropriate and should be favored for the simulation of electricity generation by photovoltaics (PV). In the study area (Mittellandkanal and Elbe-Seitenkanal), mean capacity factors of PV were calculated and the generated power was estimated depending on the area of building surfaces used.

16 Within the second phase of the project (2020-2025), the ﬁndings will be transferred to the whole German transport infrastructure and new meteorological data sets will be investigated. In particular, a new version of COSMO-REA6 (R6G2) will be considered and applied in these activities.

Current and future large-scale dynamical inﬂuences on Lake Victoria’s thunderstorms J. Van de Walle(1), W. Thiery(2), and N. P.M. van Lipzig(1) (1) KU Leuven, (2) Vrije Universiteit Brussel

Every year, thousands of fishermen lose their lives on Lake Victoria (East Africa), and capsizing accidents with passenger ferries and transport boats are frequently reported. High waves associated with severe thunderstorms are blamed. Therefore, understanding of the regional climate and its future changes is crucial for local communities.Several climate models have been evaluated over this region, with particular high performances shown for high-resolution simulations at convection- permitting scale. These models demonstrated the importance of the lake in determining the timing and intensity of precipitation. In addition, they highlighted the role of easterly trade winds and their interaction with the high mountains surrounding Lake Victoria in explaining rainfall locations. In this study, we set up three 10-year convection-permitting COSMO-CLM simulations in a tropical configuration. The historical simulation is directly downscaled from ERA 5 reanalysis. For two future projection runs, a surrogate global warming approach has been applied. First, the lateral boundary conditions from the ERA 5 are perturbed in accordance with the recent CMIP6 ensemble-mean end-of-century SSP5 8.5 scenario. In this ensemble mean, variations in large-scale atmospheric dynamics are small and the climate change signal is mainly determined by the increased water vapor and the response of the mesoscale circulation to differential lake/land heating. While increased water vapor content tends to increase total over-lake precipitation, weaker circulation results in a reduction instead. Second, instead of the ensemble mean, only one CMIP6 member with substantial large-scale dynamical changes in the region is selected and applied to the surrogate global warming approach, thereby changing both the thermodynamics and the dynamical external fields. The comparison of both simulations enables us to study the effects of changed large-scale dynamics on mean and extreme precipitation in the region, thereby gaining insight in expected future changes of the region’s hazardous thunderstorms.

Austrian Test Case: CCLM results in diﬀerent warming scenarios A. A. N. Mishra(1), D. Maraun(1), and H. Truhetz(1) (1) Wegener Center - University of Graz

June of 2009 witnessed smashing rains for central Europe. Austria amongst other countries suﬀered from ﬂoodings and landslides caused by extreme spells of precipitation. This resulted in monetary losses worth 30 Million Euros. On 24th of June, southeast Styrian district of Feldbach withstood in access of 100mm of rainfall within 24 hours. The landslides caused in this region raise an alarm and thus motivate this study whereby we investigate if the rainfall event did become

17 stronger with time due to Climate Change and would it grow even more intense if it were to occur in a warmer future. Here we have deployed the CCLM in simulating this event in diﬀerent pseudo (surrogate) warming scenarios. The results typify the thermodynamics of extreme rainfall and demonstrate an increase in intensity in a warmer climate, while a marked decrease in the intensity in a cooler climate.

Ten years of reanalysis activities at DWD: review and outlook M. Borsche(1), F. Kaspar(1), D. Niermann(1), J. Ostermöller(1),T. Rösch(1), T. Spangehl(1), R. Potthast(1), J. Keller(1,2,3), and S. Wahl(2,3) (1) Deutscher Wetterdienst, Offenbach, Germany, (2) Hans Ertel-Centre for Weather Research, Bonn, Germany, (3) University of Bonn, Bonn, Germany

In 2011, the development of a regional reanalysis based on the COSMO model began within the Hans-Ertel-Centre for Weather Research. The Meteorological Institute of the University of Bonn and the Institute of Geophysics and Meteorology of the University of Cologne focused on the question of whether it is possible to produce a long-term regional reanalysis for applications in the ﬁeld of climate analysis and climate services by using DWD’s weather forecast model.A regional reanalysis for Europe (COSMO-REA6, 6 km spatial resolution, available from 1995) and Central Europe with Germany in the center (COSMO-REA2, 2 km spatial resolution, currently available for 2007-2013) was produced with the COSMO model and the quality has been evaluated with various methods. The data product is widely used by a large and diverse community of researchers, companies, and governmental institutions alike. In this presentation, we summarize the background and current status of DWD’s reanalysis activities and present quality estimates and evaluation results. We provide an overview over recent studies as performed at DWD and at the Hans-Ertel-Centre in combination with REA6. Examples are shown addressing renewable energy applications with a focus on oﬀshore wind speeds as well as for the analysis of weather and climate extremes such as heat waves. Finally, we provide an outlook on our near future plans, which include the preparation of an ICON based reanalysis.

Evaluation of Convection-permitting simulations for Germany using gridded observations on high temporal resolution M. Haller(1), J. Brauch(1), S. Brienen(1), and B. Fr¨uh(1) (1) Deutscher Wetterdienst

The knowledge about frequencies and intensities of extreme events of the local climate is one of the key aspects of climate change studies. For the assessment of the vulnerability and possible adaptation strategies for transport infrastructure in Germany, the project “Network of experts - Adapting transport infrastructure to climate change and extreme weather events” aims at bringing together knowledge in several disciplines, including climatology and hydrology.Global climate models are not able to picture local climate eﬀects as their grid resolution is too coarse and physical parametrizations are often not satisfying. Facing the needs of our project partners for high spatial and temporal resolution of climate model data, we decided to perform regional

18 climate model (RCM) simulations on the convection-permitting scale with partly turning off the convection parametrization. Thus, we may address the small-scale effects of convection and topographic features, which play an important role not only for traffic infrastructure. We performed climate projections with COSMO-CLM on 3 km grid width using the RCP 8.5 scenario, dynamically downscaling two-fold from MIROC5 global model data. The simulations were performed for 30-year time periods: the historical time period goes from 1971 to 2000, while the future time periods include the years 2031-2060 and 2071-2100. The highest temporal resolution is 5 minutes for precipitation and 1 hour or more for all other variables. An additional evaluation run was performed using the reanalyse data ERA-40 and ERA5 as driving data sets. For the analyses and evaluation of our COSMO-CLM simulations, we use gridded observations from HYRAS (daily resolution) and RADKLIM (hourly resolution). We aim at analysing spatial distribution and frequencies of extreme rainfall as well as strong wind events and we examine the added value of high temporal resolution. Results of these analyses will be presented.

Current convection-permitting COSMO-CLM simulations for Germany S. Brienen, M. Haller, and B. Früh Deutscher Wetterdienst Four regional climate simulations at convection-permitting scale have been performed last year at DWD with the COSMO-CLM 5.0 as a contribution to the German research project “Network of Experts – Adapting transport and infrastructure to climate change and extreme weather events“. The focus of this project is to improve the resilience of the national transport infrastructure towards extreme weather and climate change. One major need is the provision of high-resolution climate information. Precipitation is one of the key parameters and the project partners are interested in information on high temporal and spatial scales. These simulations consist of three climate change runs (historical 1971-2000, near future 2031-2060 and far future 2071-2100 with the RCP8.5 scenario) driven by the GCM MIROC5 and one evaluation run driven by the ERA5 reanalysis. For the scenario runs, an intermediate nest on 12 km was used, while the evaluation run was downscaled directly from ERA5 to the final grid spacing of 3 km (0.0275◦). The domain is centered over Germany (approximately the COSMO-DE domain). The overall setup of our simulations was in most parts taken from the configuration for the CLM-community contribution to the CORDEX-FPS convection, which was compiled after a series of tests in the working group “Convection resolving climate simulations”. One specific characteristic of our simulations is a 5-minute output frequency of precipitation, which is of particular interest for hydrologists, which are also participating in the project. The evaluation run serves, in addition to the project aims, also as a reference dataset for other projects at DWD and therefore will be extended continuously near to the present day. Characteristics and some first analysis results of the simulations will be presented on this poster. The simulations show good performance compared to HYRAS observations. In terms of the diurnal cycle of precipitation, we compared our results to hourly radar observations (RADKLIM) and found good agreement between model and observation. Currently, the output data of the simulations is processed and evaluated; the standard variables are then planned to be published on the DWD ESGF node in the next weeks for interested users.

19 Clouds, Chemistry, Aerosol and Radiation

Application of Liquid and Ice Clouds Optical Properties Parametrizations in Numerical Weather Prediction Models H. B. Muskatel(1), U. Blahak(2), P. Khain(1), Y. Levi(1), and Q. Fu(3) (1) IMS, (2) DWD, (3) University of Washington

Parametrization of radiation transfer through clouds is an important factor in the ability of NWP models to correctly describe the weather evolution. Here we present a practical parameterization of both liquid droplets and ice optical properties in the longwave and shortwave radiation. An advanced spectral averaging method is used to calculate the extinction coefficient, single scattering albedo, forward scattered fraction and asymmetry factor taking into account the nonlinear effects of light attenuation in the spectral averaging. An ensemble of 7500 particle size distributions was used for the ice optical properties calculations, which enables the effective size range to be extended up to 570 µm and thus be applicable for larger hydrometeor categories such as snow, graupel, and rain. The new parameterization was applied both in the COSMO limited-area model and in ICON global model and was evaluated by using the COSMO model to simulate stratiform ice and water clouds. Numerical weather prediction models usually determine the asymmetry factor as a function of effective size. For the first time in operational NWP model, the asymmetry factor is parametrized as a function of aspect ratio. The method is available on-line to be applied to any optical properties dataset and spectral intervals of a wide range of radiation transfer models.

Increasing Resolution and Resolving Convection Improve the Simulation of Cloud-Radiative Eﬀects Over the North Atlantic F. Senf(1), A. Voigt(2), N. Clerbaux(3), A. H¨unerbein(1), and H. Deneke(1) (1) Leibniz Institute for Tropospheric Research, Leipzig, (2) Karlsruhe Institute of Technology, Karlsruhe, (3) Royal Meteorological Institute of Belgium, Brussels

The impact of parameter choices for simulations of cloud-radiative effects is assessed in the current study. Numerical experiments are carried out using the ICOsahedral Nonhydrostatic (ICON) model with varying grid spacings between 2.5 and 80 km and with different subgrid-scale parameterization approaches. Simulations are performed over the North Atlantic with either one-moment or two-moment microphysics and with convection being parameterized or explicitly resolved by grid-scale dynamics. Simulated cloud-radiative effects are compared to products derived from Meteosat measurements. Furthermore, a sophisticated cloud classification algorithm is applied to understand the differences and dependencies of simulated and observed cloud-radiative effects. The cloud classification algorithm developed for the satellite observations is also applied to the simulation output based on synthetic infrared brightness temperatures, a novel approach that is not impacted by changing insolation and guarantees a consistent and fair comparison. It is found that flux biases originate equally from clear-sky and cloudy parts of the radiation field. Simulated cloud amounts and cloud-radiative effects are dominated by marine, shallow clouds, and their behavior is highly resolution dependent. Bias compensation between shortwave and longwave

20 flux biases, seen in the coarser simulations, is significantly diminished for higher resolutions. Based on the analysis results, it is argued that cloud-microphysical and cloud-radiative properties have to be adjusted to further improve agreement with observed cloud-radiative effects.

Modeling methane from the North Sea region with ICON-ART C. Scharun(1), R. Ruhnke(1), M. Weimer(1), and P. Braesicke(1) (1) Karlsruhe Institute of Technology - Institute of Meteorology and Climate Research

Methane (CH4) is the second most important greenhouse gas after CO2 affecting global warming. Various sources (e.g. fossil fuel production, agriculture and waste, biomass burning and natural wetlands) and sinks (the reaction with the OH-radical as the main sink contributes to tropospheric ozone production) determine the methane budget. Due to its long lifetime in the atmosphere methane can be transported over long distances.Disused and active offshore platforms can emit methane, the amount being difficult to quantify. In addition, explorations of the sea floor in the North Sea showed a release of methane near the boreholes of both, oil and gas producing platforms. The basis of this study is the established emission data base EDGAR (Emission Database for Global Atmospheric Research), an inventory that includes methane emission fluxes in the North Sea region. While methane emission fluxes in the EDGAR inventory and platform locations are matching for most of the oil platforms almost all of the gas platform sources are missing in the database. We develop a method for estimating the missing emission sources based on the EDGAR inventory and the known locations of gas platforms as additional point sources will be inserted in the model. In this study the global model ICON-ART (ICOsahedral Nonhydrostatic model - Aerosols and Reactive Trace gases) is used. ART is an online-coupled model extension for ICON that includes chemical gases and aerosols. One aim of the model is the simulation of interactions between the trace substances and the state of the atmosphere by coupling the spatiotemporal evolution of tracers with atmospheric processes. ICON-ART sensitivity simulations are performed with inserted and adjusted sources to access their influence on the methane and OH-radical distribution on regional (North Sea) and global scales.

New radiation scheme ecRad and impact of uncertainties in radiation and clouds for ICON S. Schäfer(1),M. Köhler(1),R. Hogan(2,3), C. Klinger(4), D. Rieger(1), G. Zängl(1),M. Ahlgrimm(1), and A. de Lozar(1) (1) Deutscher Wetterdienst, (2) European Centre for Medium-Range Weather Forecasts, (3) University of Reading, (4) Ludwig-Maximilians-University Munich

Radiation in the atmosphere provides the energy that drives atmospheric dynamics and physics on all scales, so determining radiative balance correctly is crucial for understanding processes ranging from cloud particle growth to global weather and climate. Radiation schemes in global weather and climate models make assumptions to simplify the complex interaction of radiation with the Earth system, such as treating radiative transfer in only the vertical dimension. Capturing cloud- radiation interactions is particularly challenging since clouds vary strongly on small spatial and temporal scales not resolved in the models, and also interact strongly with radiation. In models,

21 sub-grid atmospheric variables are simplified, describing three-dimensional cloud geometry, cloud particle size and shape and complex scattering functions with a few parameters. Uncertainties in these assumptions contribute to the large lingering uncertainty in the climatic role of clouds.The new modular radiation scheme ecRad provides the opportunity to vary these parametrisations and assumptions individually to determine their impact. Several options are available for the radiation solver, cloud vertical overlap and horizontal inhomogeneity treatment and cloud ice and water optical property parametrisations. The solver SPARTACUS is the only radiation solver in a global model that can treat 3D radiative effects. We use ecRad as the new operational radiation scheme in the DWD global model ICON. We investigate the sensitivity of radiation results to radiation model assumptions and input variables such as cloud particle size and cloud geometry, and the varying role of cloud-radiation interactions in regional cloud regimes. We find that ecRad with an up-to date solar spectrum agrees much better with exact line-by-line radiation calculations than previous radiation models. In ICON, ecRad improves the global radiation balance, model physics and forecast performance as evaluated against observations.

PerduS and PermaStrom: Daily Mineral dust forecasts using ICON-ART V. Bachmann(1), J. Förstner(1),N. Porz(1), T. Hanisch(1), F. Filipitsch(2), A. Hoshyaripour(3), F. Wagner(4), A. Wagner(4), D. Lassahn(5), H. Vogel(3), B. Vogel(3), A. Seifert(1), and D. Majewski(1) (1) Deutscher Wetterdienst, Frankfurter Str. 135, 63067 Offenbach, Germany, (2) Deutscher Wetterdienst, Am Observatorium 12, 15848 Tauche, Germany, (3) Institute for Meteorology and Climate Research, Karlsruhe Institute of Technology, Hermann-von-Helmholtz-Platz 1, 76344 Eggenstein-Leopoldshafen, Germany, (4) Deutscher Wetterdienst, Meteorologisches Observatorium Albin-Schwaiger-Weg 10, 82383 Hohenpeißenberg, Germany, (5) meteocontrol GmbH, Spicherer Str. 48, 86157 Augsburg, Germany With the decision to phase out nuclear power until 2022, Renewable Energies play an essential role for the energy supply. Due to the weather influence on wind power and photovoltaic energy, weather forecasts become increasingly important for the energy sector. There are critical weather situations during which forecast errors arise due to known deficiencies in numerical weather prediction (NWP) models. Saharan dust outbreaks represent such a challenge because the actual mineral dust distribution is not considered in conventional NWP models.The project PerduS (March 2016 - February 2020) aimed at improving weather and PV power forecasts during Saharan dust episodes with an emphasis on the region of Germany. Therein, the global NWP model system ICON-ART was developed and evaluated in close collaboration with the IMK at KIT and is used to generate, study, and improve daily mineral dust forecasts. A quasi-operational mineral dust forecast system has been established at DWD and operates since January 2018. Daily 00 and 12 UTC mineral dust forecasts for the next seven and a half days starting from an own data assimilation cycle (“EnVar” mode) are running on a global 40 km grid. For a forecast time of five days a large 20 km EU-NA2-nest covering Europe, North Africa and the North Atlantic region is included by a two-way nesting approach. Both, the forecasts as well as the assimilation cycle account for the mineral dust-radiation interactions. This contribution will address the major outcome of PerduS. Specifics of the system, designed for

22 operational use, will be discussed. Since summer 2019 the ICON-ART forecasts are part of the multi-model ensemble and intercomparison of WMO’s SDS-WAS (Sand and Dust Storm Warning Advisory and Assessment System, see https://sds-was.aemet.es/). Furthermore, the ICON-ART forecasts including prognostic mineral dust are compared to pure NWP reference forecasts using ICON on the same domain configuration to study the differences and benefits of prognostic dust for now already three quasi-operational years. The benefit of the ICON-ART forecasts for the subsequent PV power prediction is evaluated within the system of the meteocontrol GmbH. An outlook on the plans for the follow-up project PermaStrom (May 2020 - April 2024) will be given. Here the main topics are the inclusion of additional aerosols (soot from vegetation fires and sea salt), the treatment and parameterization of aerosol-cloud interactions, studied in a high-resolution LAM (limited area mode) setup, and the assessment of their relevance by the means of ensemble predictions with ICON-ART.

Raikoke eruption 2019: Improving ash and SO2 forecasts by coupling ICON-ART to a plume rise model J. Bruckert(1), G. Hoshyaripour(1), A. Horvath(2), L. Muser(1), F. Prata(3), C. Hoose(1), and B. Vogel(1) (1) Karlsruhe Institute of Technology, (2) University of Hamburg, (3) AIRES Pty. Ltd

Explosive volcanic eruptions inject large amounts of ash and volcanic gases like SO2 into the atmosphere. The emitted material is hazard to aviation and modifies climate and weather. Modeling volcanic ash and SO2 dispersion in the atmosphere depends on the representation of the sources and the sinks. Here, we focus on improving the eruption source parameters (ESPs) by coupling ICON-ART with the 1D-volcanic plume model FPlume. The advantage of FPlume is that the effects of e.g., air entrainment, wind, water phase changes, and particle aggregation on the total MER are taken into account. Furthermore, we are able to define phases in which FPlume is active and calculates the mass eruption rate (MER) online. Empirical relationships derive the amount of very fine ash (particles ¡32µm) which is relevant for long range transport in ICON-ART and define the emission profile. In this presentation, the general framework is presented for the 2019 Raikoke eruption, which was characterized by several “puffs’ instead of a continuous eruption period. A comparison to Himawari-8 satellite data reveals an improvement of the ash forecast especially during the eruption and in the first hours after the eruption. For SO2, the validation of the simulation with the satellite data in the first hours is hampered, because of the dense ash cloud around the volcano leading to an underestimation of SO2. However, we find a good agreement on the second and third simulation day. We argue that representing the plume phases and ESPs in ICON-ART by FPlume enhances ash and SO2 predictability in the first days after the eruption, especially in case of non-continuous volcanic eruptions like the Raikoke eruption 2019.

23 The effect of spring 2020 lockdown in Moscow on the cloud characteristics according to COSMO-Ru simulations and measurements Y. Khlestova(1), N. Chubarova(2,1), M. Shatunova(1), and G. Rivin(1,2) (1) Hydrometeorological Research Center of Russian Federation (2) Lomonosov Moscow State University The air pollution in Moscow megacity has significantly decreased during spring 2020 lockdown due to coronavirus pandemic. The changes of air pollution reflected in the composition, concentration and physical properties of the aerosol. The aerosol-cloud interaction and its effects in the Moscow typical weather conditions were investigated in [Chubarova et al., 2020] using measurements and results of COSMO-Ru (2.2 km) configuration with cloud-radiation coupling scheme CLOUDRAD [Blahak et al., 2016]. Considering the ’typical’ results we studied the physical characteristics of clouds and cloud effects in various synoptic conditions during the spring 2020 lockdown. The ground-based and satellite measurements of radiative fluxes, cloud parameters and standard meteorological characteristics were used. The model experimental runs were carried out with different setup of cloud condensation nuclei concentration: from low polluted (hundreds particles per cm3) to high polluted (thousands particles per cm3) conditions.The research was supported by the Russian Science Foundation (grant #18-17-00149). The model experiments were partially supported by grant of Federal Service for Hydrometeorology and Environmental Monitoring (AAAA-A20-120021490079-3). Chubarova N.Ye., Zhdanova E.Yu., Androsova E.E., et al., 2020: The aerosol urban pollution and its effects on weather, regional climate and geochemical processes: Monograph / Edited by N.Ye. Chubarova - Moscow, MAKS Press. - 339 p. ISBN: 978-5-317-06464-8. Blahak, U., Muskatel, H. and Khain, P., 2016: Documentation of new optical properties of hydrometeors as function of effective size (radius or diameter) or mean axis ratio. DWD, Offenbach.

How did they get there? The Importance of the Particle History for the Ice Habit J.-N. Welss(1), A. Seifert(1), and C. Siewert(1) (1) Deutscher Wetterdienst

Understanding the detailed microphysical processes that determine the formation and growth of ice crystals is still one of the most challenging problems in the physical modeling of clouds. With Lagrangian Particle Models like the DWD-developed McSnow (Brdar and Seifert 2018, JAMES, DOI: 10.1002/2017MS001167), ice crystals can be treated as individual particles and explicit physical assumptions based on the particles’ characteristics can be used to predict their evolution. Recently, McSnow has been extended to include a habit prediction scheme for monomers. The monocrystals are approximated as spheroids. For each particle the aspect ratio for the primary crystal habit and the apparent density for secondary crystal habits are predicted. These feedback, e.g., into the sedimentation, depositional growth, and riming.It is shown that the atmospheric conditions especially at and shortly after nucleation can be decisive for the development of the ice crystal’s shape by vapor deposition because the early growth phase deﬁnes the habit of the

24 particle depending on the local temperature regime. As we found, this initial habit can hardly be changed or undone by deposition later on. The diﬀerences in particle shape alter its hydrodynamic properties and therefore decide upon the further path the crystal takes through the cloud. While typical bin models cannot represent the shapes of ice in such detail and often categorize them into classes, McSnow provides the opportunity to track individual crystals, learn about their history, compare their evolution and predict how they will transform while striding through diﬀerent thermodynamic environments.

An evaluation of kilometre scale ICON simulations of mixed-phase stratocumulus over the Southern Ocean during CAPRICORN V. Ramadoss(1), K. Pfannkuch(1), A. Protat(2), Y. Huang(3), S. Siems(4), and A. Possner(1) (1) Institute for Atmospheric and Environmental Sciences, Goethe University, Frankfurt, Germany, (2) Australian Bureau of Meteorology, Melbourne, Victoria, Australia, (3) The University of Melbourne, School of Earth Sciences, Melbourne, VIC, Australia, (4) Monash University, Melbourne, VIC, Australia Stratocumulus (Sc) clouds cover between 25% to 40% of the mid-latitude oceans, where they substantially cool the ocean surface. Many climate models poorly represent these marine boundary layer clouds in the lee of cold fronts in the Southern Ocean (SO), which yields a substantial underestimation of the reflection of short wave radiation. This results in a positive mean bias of 2K in the SO. The representation of stratocumulus clouds, cloud variability, precipitation statistics, and boundary layer dynamics within the ICON-NWP (Icosahedral Nonhydrostatic - Numerical Weather Prediction) model at the km-scale is evaluated in this study over the SO.Real case simulations forced by ERA5 are performed with a two-way nesting strategy down to a resolution of 1.2 km. The model is evaluated using the soundings, remote sensing and in-situ observations obtained during the CAPRICORN (Clouds, Aerosols, Precipitation, Radiation, and Atmospheric Composition over the Southern Ocean) field campaign that took place during March and April 2016. During two days (26 th to 27 th of March 2016), open-cell stratocumuli were continuously observed by the shipborne radars and lidars between 47 deg S 144 deg E and 45 deg S 146 deg E (South of Tasmania). Our simulations are evaluated against the remote sensing retrievals using the forward simulated radar signatures from PAMTRA (Passive and Active Microwave TRAnsfer). The initial results show that the observed variability of various cloud fields is best captured in simulations where only shallow convection is parameterised at this scale. Furthermore, ICON-NWP captures the observed intermittency of precipitation, yet the precipitation amount is overesti- mated. We further analyse the sensitivity of the cloud and precipitation statistics with respect to primary and secondary ice-phase processes (such as Hallett-Mossop and collisional breakup) in ICON-NWP. Both processes have previously been shown to improve ice properties of simulated shallow mixed-phase clouds over the Southern Ocean in other models.

25 Prognostic Ozone for ICON: Enabling UV Forecasts S. Weber(1), R. Ruhnke(1), P. Braesicke(1), and C. Scharun(1) (1) Karlsruhe Institute of Technology

Stratospheric Ozone (O3) absorbs biologically harmful solar ultraviolet radiation (most of the UV B radiation) and keeps it from reaching the surface. Such UV radiation is destructive of genetic cellular material in plants and animals, as well as human beings. Without the ozone layer, life on the surface of the Earth would not be possible as we know it. As part of its work the German Weather Service (DWD) provides UV index maps to warn the population in Germany of excessive UV exposure [1]. For this purpose, global ICON data, external ozone data and an external UV model is used. This study aims to create a self-consistent framework to generate UV index maps entirely from the non-hydrostatic global modelling system ICON [2]. For this purpose, a linearized ozone scheme (LINOZ) [3] will be optimized and the forecast functionality of ICON-ART [4][5] (ICOsahedral Non-hydrostatic - Aerosols and Reactive Trace gases) will be extended. For the derivation of UV radiation ﬂuxes and indices a radiative transfer model for solar radiation (Cloud-J) [6] shall be implemented and extended. Since the entire framework is to be used at the DWD during ongoing operations, a functionality with very low computational eﬀort is required.

Here we present the first results of the UV radiation flux through the atmosphere and its diurnal variation. Furthermore, the influence of clouds on the UV radiation flux is considered. [1] https://kunden.dwd.de/uvi/index.jsp [2] Zängl,G., et al. (2014), The ICON (ICOsahedral Non-hydrostatic) modelling framework of DWD MPI-M: Description of the non-hydrostatic dynamical core. Q.J.R. Meteorol. Soc., doi:10.1002/qj.2378 [3] McLinden, C. A., et al. (2000), Stratospheric ozone in 3-D models: A simple chemistry and the cross-tropopause flux, Journal of Geophysical Research: Atmospheres, doi:10.1029/2000JD900124 [4] Rieger, D., et al. (2015), ICON-ART - A new online-coupled model system from the global to regional scale, Geosci. Model Dev., doi:10.5194/gmd-8-1659-2015 [5] Schröter,et al. (2018), ICON-ART 2.1: a flexible tracer framework and its application for composition studies in numerical weather forecasting and climate simulations. Geosci. Model Dev., doi:10.5194/gmd-11-4043-2018 [6] Prather, M.J. (2015), Photolysis rates in correlated overlapping cloud fields: Cloud-J 7.3c. Geosci. Model Dev., doi:10.5194/gmd-8-2587-2015

Efficient generation of synthetic near-infrared satellite images F. Baur(1,2), L. Scheck(1,2), C. Köpken-Watts(1), and R. Potthast(1) (1) Deutscher Wetterdienst, Offenbach, Germany (2) Hans-Ertel-Center for Weather Research / Ludwig-Maximilians-Universität,Munich, Germany Satellite images in the solar spectrum provide cloud information with high spatial and temporal resolution and could be an important observation type for convective-scale data assimilation and model evaluation. Visible channels contain information on the cloud distribution, cloud optical thickness and cloud structure. In addition to that, near-infrared channels are sensitive to cloud microphysical properties and can be used to distinguish between water and ice clouds.However,

26 current operational data assimilation systems only use clear sky thermal infrared and microwave radiance observations. Those observations mainly provide temperature and humidity information. Furthermore, mainly infrared radiances have been used for model evaluation. The reason is that suﬃciently fast and accurate forward operators for visible and near-infrared radiances are not yet available, which is related to the fact that multiple scattering makes radiative transfer at solar wavelengths complicated and computationally expensive. The recently developed MFASIS, a 1D radiative transfer method based on compressed look-up tables, allows for similarly accurate but orders of magnitude faster calculations as compared to conventional radiative transfer solvers for the visible spectrum. Here we discuss the limitations in the current version of MFASIS that prevent it from simulating near-infrared channels accurately and present an alternative approaches that increase the accuracy signiﬁcantly for near-infrared channels. The new approach is tested using IFS and ICON model output.

Clouds and precipitation in ICON-LAM from radar and SEVIRI observations A. de Lozar(1), A. Seifert(1), R. Posada(1), U. Blahak(1), S. Geiss(2), and L. Scheck(1,2) (1) Deutscher Wetterdienst, (2) LMU M¨unchen

We aim to improve the representation of clouds and convection in ICON-LAM using combined measurements from different observational systems. We mainly concentrate on the observational systems that directly observe clouds and precipitation: radar and SEVIRI satellite observations. Both systems complement themselves very well: while large hydrometeors (snow, rain, graupel, hail) are sensitive to the radar signal at C-band, small hydrometeors (ice and cloud water) mostly determine satellite observations in visible and near-infrared spectral bands. In the presentation we show how to evaluate the model in hindcast runs using these observations. In addition, sensitivity studies will be presented that reveal the impact of different choices in the model physics. The challenge is to determine which variant of the ICON model physics produces the most realistic clouds and convective events.We compare simulations using two different microphysical schemes (operational one-moment vs. Seifert and Beheng two-moment). All simulations are driven by the same boundary conditions from the ICON-EU analysis fields. The simulated period is characterized by summerly convection over Germany. Simulated reflectivities are calculated with the forward operator EMVORADO. EMVORADO mimics the scanning strategy of the DWD radar stations, which allows for a direct comparison between model and observations. Simulated and observed reflectivities are analyzed using the DWD Nowcasting cell-detection algorithm KONRAD-3D, which exploits data from different heights (three-dimensional approach) and uses an adaptive threshold for the detection of convective cores. This allows for a direct comparison of simulated and modeled cells in terms of lifetime, VIL, maximum intensity and other cell properties. Simulated reflectance in visible channels are calculated using the forward operator MFASIS (Method for FAst Satellite Image Simulation) and infrared channels are calculated with RTTOV. While visible channels mostly depend cloud cover, infrared channels are mostly sensitive to cloud height. The combination of both observations provides a fingerprint for the three-dimensional cloud field, which can be used to evaluate our model.

27 Improved Ice Aggregation Formulation in the Seifert-Beheng Two-Moment Microphysics Scheme M. Karrer(1), D. Ori(1), V. Schemann(1), A. Seifert(2), and S. Kneifel(1) (1) Institute of Geophysics and Meteorology (2) Deutscher Wetterdienst (DWD)

The simulation of ice microphysical processes constitutes a major challenge for numerical weather prediction (NWP) and is important for the prediction of precipitation. One of these processes is the aggregation of ice and snow particles, which we investigate by model-observation comparison.The observations were collected during winter 2015/2016 at the Jülich Observatory for Cloud Evolution with a novel multi-frequency radar configuration. Besides, we simulated 43 with significant clouds with a nested version of the ICON-LEM and the Seifert-Beheng two-moment microphysics (SB). A radar forward operator was applied to the model output to simulate radar measurements which allow the comparison in the observational space. The statistical comparison between the synthetic and actual observations reveals that the cloud model overestimates the Doppler velocity at lower and warmer temperatures. Also, an overestimation of the snow particle sizes at warmer temperatures is found, which can be identified by large dual-wavelength ratios. To overcome these inconsistencies we derived alternative particle property relations by combining a realistic snow aggregation model and hydrodynamic theory. We implemented these new particle property relations and updated other microphysical parameters that affect the aggregation rates (e.g. formulation of the aggregation kernel, sticking efficiency). This allows us to thoroughly investigate the sensitivity of aggregation to these parameters. Finally, we repeated the radar forward simulations of the ICON output with an improved setup. The new simulations show a reduction in the biases of the Doppler velocity and snow particle size. At the same time, the frequency of the larger precipitation rates is underestimated in the new simulations. This underestimation is likely due to an underestimation of the riming rates close to the melting layer.

Investigation of aircraft icing sensitivities using a parameterization respecting droplet spectra S. Werchner(1), C. Kottmeier(1), C. Hoose(1), H. Vogel(1), and B. Vogel(1) Karlsruhe Institute of Technology - Institute of Meteorology and Climate Research The aggregation of ice upon an aircraft poses a serious safety hazard in aviation. The additional weight and the deterioration of the aerodynamic properties of the iced surfaces reduce the overall performance of the aircraft. The source of icing occurring are supercooled cloud and rain droplets that spontaneously freeze when contacting a suitable surface. While the ﬁrst instances of icing stays “dry” (i.e. no supercooled liquid water remains) the latent heat released by the freezing process can partially melt the ice. The resulting semi-liquid water (“wet”) can ﬂow downstream the aircraft and freeze upon less exposed parts of the aircraft, further increasing the performance loss induced. Since icing is a crucial topic in aviation, warning systems are developed and operated to inform pilots and other decision makers about the future state of the atmosphere regarding icing conditions. ADWICE, one of these systems, is operated by the

28 DWD and covers the European airspace. This study aims to investigate the role of certain sensitivities regarding aircraft icing. For this a new icing intensity parameterization is developed. Contrary to other systems no empirical method for the determination of the icing intensity is applied, but a description of the underlying physical processes is implemented, namely impingement of the liquid water upon the surface and the actual freezing of the impinged water. The parameterization is able to consider the actual cloud droplet spectrum of the supercooled cloud droplets for the estimation of the total impingement eﬃciency and the variable properties of the rime ice generated during the “dry” icing phase. With this, a more complete view on the aircraft icing process is possible. First applications of the parameterization (implemented into ICON-ART) show that the impingement eﬃciency as well as icing related quantities show sensitivities regarding the way the horizontal resolution and the number concentration of cloud droplets.

Options and extensions for the stochastic shallow convection scheme in ICON M. Ahlgrimm(1), D. Klocke(1), E. Machulskaya(1), M. Sakradzija(1), and A. Seifert(1) (1) Deutscher Wetterdienst

When ICON is used in limited area mode, resolutions are typically on the order of a few kilometers horizontal grid spacing. Deep convective transport is partially resolved at these scales, but shallow convection remains poorly represented without a parameterization.A stochastic shallow convection scheme was developed in collaboration with the Max Planck Institute for Meteorology to provide a more appropriate parameterization framework at kilometer -scales. The scheme is scale-adaptive and should render resolution-dependent tuning of the convection parameterization unnecessary. Over the past 18 months, the original scheme has been extended and optimised. Alongside the original explicit stochastic scheme an approximation using stochastic differential equations (SDE) has been developed. Equivalence of the two versions can be demonstrated by running one version interactively, the other passively (“piggy-backing”). While the SDE approximation is computationally more efficient, the explicit version of the scheme can be easily extended to keep track of additional properties of the shallow cloud ensemble. For example, the convective updraft core fraction can be calculated for use in the diagnostic subgrid cloud scheme. Or knowledge of individual clouds’ depth can be used to derive a more realistic lateral detrainment profile than is currently in use. We demonstrate the performance of the scheme and illustrate options and applications in single column mode, case studies and month-long hindcasts.

Quantiﬁcation of the eﬀects of aerosol deposition on snow by ensemble simulations A. Rohde(1), S. Werchner(1), G. Hoshyaripour(1), J. Bruckert(1), H. Vogel(1), and B. Vogel(1) (1) Karlsruhe Institute of Technology

29 The quantification of the effects of improved processes in a model raises problems when the signal is small or superimposed by model internal variability. Here, we quantify the effects of aerosol deposition on snow albedo and radiation. Due to the absorption properties of these impurities, the snow albedo is darkened, resulting in a positive radiative forcing at the surface. This effect builds up gradually and is only modestly pronounced in a simulation over a few days. Nevertheless, the aerosols on the snow can already trigger far-reaching consequences in the atmosphere. Steady decrease of snow depth, warming of the surfaces and 2m temperature in regions with thin and patchy snow cover are some examples.The ICON built-in feature of generating an ensemble simulation was applied to reduce the influence of natural variability. Performing ensemble simulations, the signal generated by the aerosols could be separated from the interfering influences. In this way, the relationship between the snow albedo change and the feedbacks could be quantified. An average positive radiative forcing of 19.15 W m-2 was found for a mean decrease in surface albedo of 2.93 %. The feedbacks are, however, dependent on the respective region and show an elevation dependency.

Using Microphysical Piggybacking in ICON to explain the sensitivity of simulated convection to the choice of microphysics scheme A. I. Barrett(1), and C. Hoose(1) (1) Karlsruhe Institute of Technology

The choice of microphysics scheme can have a very large effect on the strength and organisation of explictly-simulated convection. We use Microphysical Piggybacking (Grabowski, 2014, 2015, Grabowski and Morrison, 2016) to help identify which particular aspects of the microphysics schemes are responsible for these differences. Without piggybacking, such analysis is challenging because of any change to the microphysics scheme results in changes to the latent heating distribution through condensation, freezing, melting and evaporation. The changes in latent heating modify the buoyancy and therefore also the wind circulations. In the end, the dynamics in two simulations with different microphysics schemes become quite different, which prevents an apples-to-apples comparison of the microphysical processes.With microphysical piggybacking this problem is bypassed. Piggybacking allows 2 microphysics schemes to be run in parallel, both driven by identical winds. This is done by adding a second set of thermodynamic variables (theta, qv, hdyrometeor mass and number concentrations) to ICON that only respond to the wind fields (i.e. they do not feed back changes through latent heating-¿temperature profiles-¿buoyancy-¿vertical velocity). This allows the separation of pure microphysical effects (e.g. of changing microphysics scheme, microphysical processes or aerosol concentrations) to those affected by changes to the wind fields. This presentation will include: • an overview of the piggybacking methodology and benefits • a short description of the technical implementation • an example showing that surface precipitation is sensitive to direct microphysics changes whereas surface hail is sensitive to microphysics-induced-wind changes • a first look at causes of differences between the 2-moment Seifert & Beheng microphysics scheme and the Predicting-Particle-Properties (P3) scheme.

30 Data Assimilation

Evaluation of ICON’s model cloud ﬁelds using simulated and observed visible satellite images C. Stumpf (1), L. Bach(1), C. Koepken-Watts(1), L. Scheck(1,2), and R. Potthast(1) (1) DWD, (2) Hans-Ertel-Zentrum / LMU M¨unchen

MFASIS is a novel fast radiative transfer method for the simulation of visible satellite images that is fast enough to cope with the computational constraints of operational data assimilation systems and has therefore been implemented into RTTOV since version 12.2. First evaluation and data assimilation experiments using MFASIS in combination with the COSMO and the new ICON-LAM regional models at DWD have demonstrated its value by improving the representation of cloud cover and precipitation as well as short term forecasts of surface variables. As a further step towards using visible satellite images in operational data assimilation, we perform a detailed validation of the accuracy of MFASIS in RTTOV v13 and an evaluation of the representation of clouds in DWD’s global NWP system ICON+EnVAR. For evaluating MFASIS, its forward computation results are compared to results of the DISORT implementation in RTTOV (RTTOV-DOM) for a range of model proﬁles and stratiﬁed according to model cloud situations.

For evaluating the model clouds, we focus on an experimental period in April and June 2020, comparing RTTOV-MFASIS forward simulations based on global ICON model fields to visible channel observations of SEVIRI on Meteosat. This setup allows for an evaluation of the model equivalents in a large variety of atmospheric situations and at different local times. Additionally, we make use of level-2 cloud products, such as EUMETSAT’s Optimal Cloud Analysis product OCA for SEVIRI, to analyse results and systematic errors classified by cloud types. This aims at validating the accuracy of the model cloud fields, also in conjunction with all-sky simulations of corresponding IR channels. Here, the visible channel information is complementary especially for the analysis of the representation of low clouds and has a higher sensitivity with respect to some model physics aspects like sub-grid scale cloud representation.

Assimilation of remote sensing proﬁler observations at MeteoSwiss C. Merker(1), D. Leuenberger(1), B. Crezee(1), S. Monhart(2), A. Haefele(3), M. Hervo(3), G. Martucci(3), and M. Arpagaus(1) (1) MeteoSwiss, Z¨urich,(2) MeteoSwiss, Locarno, (3) MeteoSwiss, Payerne, Switzerland

The current atmospheric observing systems fail to provide a satisfactory amount of spatially and temporally resolved observations of wind, temperature and humidity in the planetary boundary layer (PBL) despite their potential positive impact on numerical weather prediction (NWP). This is particularly critical for humidity, which exhibits a very high variability in space and time, and for the vertical proﬁle of temperature, which determines the atmosphere’s stability. Therefore, the analyzed thermodynamical structure of the PBL can be prone to errors, leading to poor

31 forecasts of warnings for relevant phenomena, such as severe summer convection or winter fog and low stratus. One approach to improve the model representation of the PBL is to include novel, ground-based remote sensing proﬁler observations in the data assimilation system to improve the forecast initial conditions. This also improves the quality of downstream applications relying on a good representation of the boundary layer in the model, such as dispersion modelling for emergency response after nuclear, chemical or biological incidents. In this contribution, we present progress of the MeteoSwiss eﬀort to include observations from Doppler wind lidar, microwave radiometer and Raman lidar devices into the 1km mesh-size data assimilation system KENDA-1. We will present results of observation minus background statistics and impact case studies.

EnVAR for ICON-LAM: observations and quality control M. Burba(1), S. Ulbrich(1), S. Hollborn(1), and R. Potthast(1) (1) Deutscher Wetterdienst, FE12 (data assimilation)

The German Weather Service (DWD) introduces the regional NWP model ICON-LAM (ICON Limited Area Mode) in 2021 to replace the COSMO model. For the ICON-LAM data assimilation, a novel EnVAR (Ensemble VARiational data assimilation) setup is evaluated in comparison to the operational deterministic KENDA-LETKF (Local Ensemble Transform Kalman Filter). This includes in particular the diﬀerent observation processing in the EnVAR and KENDA, as well as the setup of the variational quality control for the EnVAR. We will give an introduction to the observation processing in DWD’s data assimilation framework (DACE).For future development, we will give an outlook on how a regional EnVAR can be used for a regional deteministic analysis by using a global ICON ensemble in combination with a regional deterministic ICON-LAM run. This is potentially of interest for DWD’s partners with less excessive computational capacities, because a regional EnVAR analysis is less computationally expensive than running a full KENDA assimilation cycle.

Adaptive non-linear bias correction for visible reﬂectance data A. T. Deppisch(1), and B. L. Bach(2) (1) DWD, (2) MetBW

Solar reflectances as measured by Meteosat’s SEVIRI instrument provide insights into the various processes happening in clouds and the upper athmosphere and therefore may complement conventional sources of information for numerical wheather prediction. The upcoming Seamless INtegrated FOrecastiNg sYstem (SINFONY) of Germany’s national wheather service (DWD) aims at assimilating these reflectances in an effort to improve short time forecasts of up to 12h.This endeavour is challenged by various sources of bias, coming from both cloud microphysics and athmospheric radiation modelled by forward operators. In order to reduce the negative influence of these biases in data assimilation, a method of adaptive bias correction is introduced: Each pixel of the satellite image is corrected based on the distribution of reflectances of all pixels. The non-linear bias correction function is updated at each analysis cycle using a 3dVAR algorithm and thus constantly learning about the current wheather situation.

32 This talk focuses on the basic concepts of adaptive non-linear bias correction for reﬂectance data as well as on presenting results from experiments with the system for numerical wheather prediction at DWD.

Assimilating 3D radar information at convective scales at DWD K. Stephan(1), U. Blahak(1), C. A. Welzbacher(1), K. Khosravian(1), R. Potthast(1,2), C. Schraﬀ(1), and K. Vobig(1) (1) DWD, (2) University of Reading, Department of Mathematics

At Deutscher Wetterdienst (DWD), research about the assimilation of 3D radar data (reflectivity, cell objects and radial winds) has been intensified within the project SINFONY, which is related to the future seamless ensemble prediction system for convective-scale forecasting for up to 12 hours. This system integrates Nowcasting techniques for radar data with numerical weather prediction (NWP based on the new ICON-model) in a seamless way with initial focus on severe summertime convective events and associated hazards such as heavy precipitation, hail and wind gusts. In parallel, assimilation of 3d radar data became operational in March 2020 (radial winds) and in June 2020 (reflectivity) in our current short range ensemble numerical weather prediction (SRNWP) system with forecast ranges of 27 hours (COSMO-D2-KENDA LETKF system). For both systems, SRNWP and SINFONY, the usage of 3D radar data is not only advantageous but crucial to improve the forecast skills related to convection and precipitation. We will present results of the currently operational COMSO-D2-KENDA system, where we will focus on operational experiences assimilating 3d radar data. An appropriate quality control is key, as well as super-observations, proper localization and a proper specification of the observation error model for the LETKF. At time ICCARUS will take place COSMO model will be already replaced by ICON model, extensive month-long assimilation experiments within the SINFONY project and its SRNWP counterpart have been performed using the ICON-D2-KENDA ensemble system. We present and compare results of different techniques and their mutual combinations to assimilate radar data: “traditional” Latent Heat Nudging, LETKF-assimilation of 3D radial winds and 3D reflectivity. Furthermore we will show results of ongoing project to further improve the benefit of radar data assimilation.

Improving radiation forecasts by assimilating visible satellite images in ICON-D2 L. Scheck(1,2), S. Geiss(1), L. Bach(2), A. de Lozar(2), and M. Weissmann(3) (1) Hans-Ertel-Centre / LMU Munich, (2) Deutscher Wetterdienst, (3) University Vienna

The share of renewables in the German net electricity generation has exceeded 50% for the first time in 2020 and photovoltaics power production has contributed more than 10%. A high percentage of uncontrollable and intermittent power generation poses problems for the network safety and makes it harder to find the optimal balance between different types of power generation. Accurate radiation and PV power generation forecasts are required to meet these challenges.In the

33 framework of the project MetPVnet satellite images are used to achieve better radiation forecasts in two ways: Firstly, they are used to evaluate and improve the representation of clouds in ICON. And secondly, observations from the visible 0.6µm channel of the SEVIRI instrument onboard Meteosat Second Generation satellites are assimilated to improve the clouds in the initial state. Compared to infrared channels, observations from visible channels can provide more information on the optical thickness clouds and are much more strongly correlated with the surface radiation. Here we report on assimilation experiments using ICON-D2 and the local ensemble transformation Kalman filter (LETKF) implemented in DWDs data assimilation coding environment (DACE). The Method for FAst Satellite Image Synthesis (MFASIS) is used to generate model equivalents for the satellite images. Results for test periods of several weeks indicate that errors in the cloud distribution and the surface radiation can be significantly reduced by assimilating visible satellite images. A beneficial impact is still present after 24 hours and also the agreement with most conventional observations is improved.

34 Dynamics and Numerics

A locally smoothed vertical coordinate to improve fog and low stratus forecasts S. Westerhuis(1,2), and O. Fuhrer (2,3) (1) ETH Z¨urich,(2) MeteoSwiss, (3) Vulcan Inc.

The accurate prediction of fog and low stratus (FLS) poses a major challenge for numerical weather prediction (NWP) models.In Switzerland, the Swiss Plateau is prone to FLS during wintertime. Previous studies and feedback from forecasters working at MeteoSwiss who rely on the COSMO model on a daily basis have reported that the model often dissipates FLS too quickly. We identify a mechanism so far neglected in the literature to be the main driver for the erroneous FLS dissipation: Terrain-following vertical coordinates exhibit sloping vertical coordinate surfaces in the lower atmosphere where FLS occur - even in regions with only moderate terrain such as the Swiss Plateau. The sloping coordinate surfaces intersect the flat cloud tops of FLS. In cases where horizontal advection is present, spurious implicit numerical diffusion of the horizontal advection scheme results in excessive vertical mixing which finally promotes FLS dissipation. To alleviate the problem of implicit numerical diffusion associated with sloping vertical coordinate surfaces we propose a new LOcally Smoothed Vertical Coordinate (LOSVEC). We demonstrate the advantages of LOSVEC based on a case study.

COSMO-EULAG dynamical core with long timestep capability Z.P. Piotrowski(1)(2) (1) Forschungszentrum J¨ulich (2) On the leave from the Institute of Meteorology and Water Management - National Research Institute, Warsaw, Poland The COSMO-EULAG dynamical core included in the recent COSMO releases may need to employ shorter timesteps due to the fully explicit vertical advection. In this presentation I discuss the extension of the COSMO-EULAG core with the capability to perform substepping of dynamics within a physics timestep, as well as the new split version of MPDATA advection allowing for the double Courant number in the vertical. Updated integration times and summary of another performance optimizations is provided.

ICON model on GPU X. Lapillonne(1), IMPACT and ENIAC team(1,2) (1) MeteoSwiss (2) C2SM, ETHZ

In view of further pushing the frontier of possible applications of the ICON modelling framework and to make use of the latest evolution in hardware technologies, parts of the model requires for numerical weather modeling is adapted to run on heterogeneous GPU hardware.Because of the low compute intensity of atmospheric model the cost of data transfer between CPU and GPU at every step of the time integration would be prohibitive if only some components would be ported

35 to the accelerator. We therefore present a full port strategy where all components required for the simulations are running on the GPU. For the dynamics, most of the physical parameterizations and infrastructure code the OpenACC compiler directives are used. For the soil parameterization, a Fortran based domain speciﬁc language (DSL) the CLAW-DSL has been considered. We discuss the challenges associated to port a large community code, about 1 million lines of code. We show ﬁrst performance comparison of the model for a subset of the components on CPU and GPU. Finally we discuss challenges and planned development regarding performance portability and high level DSL which will be used with the ICON model in the near future.

Towards a transient 3D gravity wave parameterization in ICON G. Bölöni(2), G.-S. Völker (1), Y.-H. Kim (1), S. Borchert (2), G. Zängl (2), U. Achatz (1) (1) Goethe UniversitätFrankfurt, (2) Deutscher Wetterdienst

Current gravity-wave (GW) parameterization (GWP) schemes are using the steady-state assumption, where an instantaneous balance between GWs and mean flow is postulated, thereby neglecting transient, non-dissipative direct interactions between the GW field and the resolved flow. These schemes rely exclusively on wave dissipation, by GW breaking or near critical layers, as a mechanism leading to forcing of the mean flow. In a transient GWP, without steady-state assumption, non-dissipative direct wave-mean-flow interactions are enabled as an additional mechanism. Idealized studies have shown that this is potentially important, so that the transient GWP Multi-Scale Gravity-Wave Model (MS-GWaM) has been implemented into the ICON model. In this implementation, MS-GWaM leads to a zonal-mean circulation well in agreement with observations, and increases GW momentum-flux intermittency as compared to steady-state GWPs, bringing it into better agreement with super-pressure balloon observations. Transient effects taken into account by MS-GWaM are shown to make a difference even on monthly time-scales: in comparison with steady-state GWPs momentum fluxes in the lower stratosphere are increased and the amount of the missing drag at Southern Hemispheric high latitudes is decreased to a modest but non-negligible extent. An analysis of the contribution of different wavelengths to the GW signal in MS-GWaM suggests that small scale GWs play an important role down to horizontal and vertical wavelengths of 50 km (or even smaller) and 200 m respectively.

36 Model Infrastructure and Data Processing

Zarr formatted Climate Data in DKRZ’s Swift Cloud Object Storage F. Wachsmann(1), M. Kul¨uke(1), and G. Siemund(2) (1) Deutsches Klimarechenzentrum, (2) Universit¨atHamburg

Earth system sciences handle large amounts of data. These data can be represented as multi- dimensional arrays and is traditionally stored in netCDF format on hierarchical file systems. NetCDF is a self describing data format with attached metadata.In recent years, cloud object storage systems became an alternative to traditional hierarchical file systems. Object storage systems are easily scalable and offer faster data retrieval, as compared to hierarchical file systems. However, the current netCDF-4 format is not yet optimized for object storage systems. NetCDF data transfers from an object storage can only be conducted on file level which results in heavy download volumes. An improvement to mitigate this problem can be the Zarr format, which divides arrays into chunks and compresses them. Corresponding Metadata are stored in lightweight JSON files. Zarr reduces data transfers due to the direct chunk and meta data access and hence increases the input/output operation speed in parallel computing environments. This contribution focues on the conversion of netCDF on a hierarchical file system into Zarr on a cloud object storage system. Moreover, the input/output performance between Zarr and netCDF will be compared.

The ICON single column mode I. B. Duran(1), M. Köhler(2),A. Eichhorn-Müller(2),V. Maurer(2), J. Schmidli(1), A. Schomburg(2), D. Klocke(2), T. Göcke(2), S. Schäfer(2),L. Schlemmer(2), and N. Dewani(1) (1) Goethe University Frankfurt (2) Deutscher Wetterdienst

The Single Column Mode/Model (SCM) is an important tool for research and model evaluation and development. The SCM is suitable for model development, because it offers a well-controlled environment, where the influence of large scale forcings, dynamics and selected physical parameterizations can be prescribed. Additionally, the SCM has a reduced computational cost and a low storage demand. We have implemented a new single column configuration into the ICON model. The ICON SCM can be run for well-established idealized cases already implemented, or semi-realistic cases based on analyses or model forecast. New cases can be easily prepared by the modification of the input NetCDF file.We will illustrate the usage of the ICON SCM for different idealized cases off shallow convection and stratocumulus and evaluate the performance of ICON SCM in semi-idealized cases.

37 NWP Model Applications and Case Studies

Case study of an Arctic atmospheric river with the ICON model H. Bresson(1), A. Rinke(1), V. Schemann(2), M. Mech(2), S. Crewell(2), C. Viceto(3), I. Gorodetskaya(3), and K. Ebell(2) (1) Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research, Potsdam, Germany, (2) Institute of Geophysics and Meteorology, University of Cologne, Cologne, Germany, (3) Department of Physics and CESAM, University of Aveiro, Aveiro, Portugal

The Arctic climate changes faster than the ones of other regions, but the relative role of the individual feedback mechanisms contributing to Arctic amplification is still unclear. Atmospheric Rivers (ARs) are narrow and transient river-style moisture flows from the sub-polar regions. The integrated water vapour transport associated with ARs can explain up to 70% of the precipitation variance north of 70◦N. However, there are still uncertainties regarding the specific role and the impact of ARs on the Arctic climate variability.For the first time, the high-resolution ICON modelling framework is used over the Arctic region. Pan Arctic simulations (from 13 km down to ca. 6 and 3 km) are performed to investigate processes related with anomalous moisture transport into the Arctic. Based on a case study over the Nordic Seas, the representation of the atmospheric circulation and the spatio-temporal structure of water vapor, temperature and precipitation within the limited-area mode (LAM) of the ICON model is assessed, and compared with reanalysis and in-situ datasets. Preliminary results show that the moisture intrusion is relatively well represented in the ICON-LAM simulations. The study also shows added value in increasing the model horizontal resolution on the AR representation.

Impact of diﬀerent external parameters on Turin UHI with COSMO at 1km F. Bassani(1), V. Garbero(2), and M. Milelli(2,3) (1) Polytechnic of Turin, Italy, (2) Arpa Piemonte, Italy, (3) CIMA Foundation, Italy

In an increasingly urbanized world, the numerical weather prediction models need to better represent the urban areas, in order to capture the micro-climate phenomena induced by the cities. The parameterization TERRA URB (TU) (Wouters et al., 2016), recently implemented in COSMO (Bucchignani et al., 2019), not only represents a novelty in this field but has proved to correctly reproduce the Urban Heat Island effect over different European cities (Garbero et al., 2020, submitted). TU provides a heterogeneous description of the urban-atmosphere interactions, through the definition of several urban external parameters, such as the anthropogenic heat flux (AHF), the impervious surface area fraction (ISA), and other urban canopy parameters such as the building area fraction (BF), the mean building height (H) and the height-to-width ratio (H/W). In this study we performed simulations with COSMO model at 1 km resolution with the aim of a better characterization of the UHI over the city of Turin. In particular, we compared the results by using AHF and ISA from the EXTPAR preprocessor and from the Local Climate Zones (LCZ) classification system (Stewart and Oke, 2012). Furthermore, we focused on the influence of the urban parameters BF, H and H/W by comparing

38 two different approaches: as a default, their values are assumed constant for all the urban grid points, while a different 2-D approach consists in deriving their values for each urban grid point based on LCZ classification (Demuzere et al., 2019). A sensitivity analysis was then performed to detect which of the 2-D urban parameters have a greater impact on the results, with an emphasis on the Surface Energy Balance (SEB). With the purpose of unravelling the driving mechanism behind the UHI, we analyzed the individual SEB components and evaluated how much each flux contribute to the urban heat island effect.

Towards understanding the role of uncertainty in microphysical processes for warm conveyor belt ascent using microphysical heating rates along online trajectories in ICON A. Oertel(1), A.K. Miltenberger(2), C.M. Grams(1), and C. Hoose(1) (1) Institute of Meteorology and Climate Research (IMK-TRO), Karlsruhe Institute of Technology (KIT), (2) Institute for Atmospheric Physics, Johannes Gutenberg University Mainz

The characteristic large-scale cloud band in extratropical cyclones is often formed by the so-called warm conveyor belt (WCB), a coherent and strongly ascending airstream in extratropical cyclones that typically ascends cross-isentropically from the boundary layer into the upper troposphere within two days. This transport of air into the upper troposphere can significantly influence the large-scale flow evolution and lead to ridge amplification downstream. The cross-isentropic ascent and the WCB outflow strength in the upper troposphere are strongly driven by latent heat release from the formation of liquid, mixed-phase and ice clouds. In this way, WCBs provide an environment where small-scale cloud microphysical processes are directly linked to the large-scale atmospheric circulation in extratropical cyclones. The need for parameterization of microphysical processes and convection in numerical weather prediction models introduces uncertainties in their representation which can feed back on the larger-scale flow. In particular, ice cloud formation and the phase partitioning are often poorly represented in numerical weather prediction models.We analyse the role of uncertainty in microphysical process rates in the ICON 2-moment microphysics scheme for (i) the detailed WCB ascent behavior and (ii) the large-scale flow evolution. Therefore, we run two-way nested simulations with two refined nested domains for a WCB case study in the North Atlantic. To quantify the effect of individual parameterized microphysical processes for (i) WCB ascent and (ii) the circulation, we implement temperature and associated potential vorticity tendencies for each process and aggregate their evolution along online trajectories implemented in ICON. Subsequently, we explore parameter uncertainty in microphysical parameterizations in an ensemble of sensitivity experiments with systematically varying microphysical parameters and quantify the effects for WCB ascent. Here, we present first results including the technical setup, the dominating microphysical processes for WCB ascent and associated temperature and potential vorticity tendencies, and an outlook for our diagnostic framework.

Lagrangian analysis of an Alpine Foehn event A. L. Jansing(1), and B. M. Sprenger(1) (1) Institute for Atmospheric and Climate Science, ETH Zurich

39 Foehn is a generic term for downslope windstorms in the lee of mountain ranges. These strong and gusty winds have attracted the attention of both scientists and the general public for more than a century because of their socio-economic (public health, wind damage, winegrowing, danger for aviation) and environmental (forest fires) impacts. One of the long-standing research questions concerns the mechanisms leading to the characteristic warming of Foehn air masses as they pass over the Alpine crest and descend into the northern Alpine valleys. Specifically, the importance of adiabatic descent and diabatic processes for the warming has been heavily disputed in the literature.Here, we readdress this question using a state-of-the-art methodology: We analyze a COSMO-1 hindcast with online trajectories for a long-lasting Alpine Foehn event during November 2016. A Lagrangian heat budget allows us to quantify the warming mechanisms for six different Alpine Foehn valleys. Thereby, we separate it into adiabatic and diabatic contributions, e.g., due to microphysical processes, turbulent mixing and radiation. A considerable day-to-day variability of the total Foehn air warming is identified. It correlates with the adiabatic contribution, which most strongly contributes to the air masses’ warming. With respect to the considered Alpine valleys, the magnitude of the diabatic warming exhibits a pronounced west-east gradient. It is especially enhanced during the central period of the event, where it leads to a net warming in the western Alpine valleys. In contrast, during the last phase of the Foehn event, diabatic processes predominantly cool air masses. Several diabatic processes play a role along the Foehn trajectories. The diabatic warming upstream of the Alpine crest can almost exclusively be attributed to cloud microphysical processes (condensation). After passing over the Alpine crest, Foehn air parcels experience diabatic cooling of smaller magnitude compared to the upstream diabatic warming. This cooling during the descent into the valleys results from evaporation of cloud and rain droplets and from turbulent mixing of the potentially warmer Foehn air masses with the residual cold air pools in Foehn valleys. Finally, we categorize the Foehn trajectories in the different valleys based on their diabatic warming contributions. In this way, airstreams with distinct thermodynamic properties are identified and further characterized in terms of air mass origin and altitude as well as their specific pathways when crossing the Alps. In future studies, we will generalize our findings to multiple case studies representing different Foehn flavors.

High-resolution Simulations of Atmospheric CO2 with ICON/MESSy B. Kern, and P. Jöckel Institut fürPhysik der Atmosphäre, Deutsches Zentrum fürLuft- und Raumfahrt (DLR) e.V., Oberpfaffenhofen Robust global observations for assessing CO2 emissions are important to reach the goals of the COP21 ”Paris Agreement”. An essential prerequisite to coordinate necessary mitigation measures is monitoring of emission fluxes of CO2 from point sources, like industrial and power plants. An important part is the ground-based monitoring network of in-situ measurements. However, the rather coarse spatial coverage of in-situ measurements may hamper precise determination of local point sources.In-situ measurements can be comlemented by high-resolved satellite observations to provide a continuous global coverage. Because of the small enhancements of CO2 mixing ratios in the plumes of point emission sources, this proves a challenge for the detection system. The DLR project CO2MON tries to assess the potential and explore the requirements of a space-borne CO2

40 monitoring system. The development and design process is supported by high-resolution modelling of the atmospheric CO2 distribution. At later stages of the project, the numerical modelling system may also be used to estimate emission ﬂuxes of dedicated point sources from observations. We present high-resolution simulations (O(100m)) with ICON of the atmospheric CO2 mixing ratio over limited areas in Northern America and Europe. Tracers and emissions are implemented in ICON using the Modular Earth Submodel System (MESSy). CO2 emissions are prescribed as gridded emission datasets and point sources.

A preliminary evaluation of the near-surface evolution of Foehn events in COSMO-1 Y. Tian(1), J. Schmidli(1), and J. Quimbayo-Duarte(1) (1) Goethe University Frankfurt

Foehn is a downslope wind with a large impact on society due to its gusty nature and the associated high temperature extremes. The accurate forecasting of the onset, strength, and decay of Foehn events is challenging as the near-surface evolution is the result of multi-scale interactions of the larger atmosphere with the mountain and the local valley topography. An important process influencing the Foehn onset (at the surface) is the interaction with cold pools which are often present in the Foehn valleys during the cold season. This study investigates the skill of the most current COSMO version (v5.7) at 1.1 km grid spacing in simulating the near-surface Foehn evolution for a set of events, representative of different Foehn types. The evaluation is based on a comparison to station data from the automatic monitoring network of MeteoSwiss. Significant biases in Foehn intensity, duration, and spatial extent are found. The sensitivity of these biases to several parameterization choices are investigated and presented.

How do mesoscale weather systems interact with oﬀ-shore wind farms: A study for the Kattegat (Denmark, Sweden) with the mesoscale model COSMO-CLM and satellite scatterometer data J. Neirynck(1), J. Van de Walle(1), A. Stoﬀelen(2), J. Meyers(3), and N. van Lipzig(1) (1) Department of Earth and Environmental Sciences KU Leuven, (2) R&D Satellite Observations KNMI, (3) Department of Mechanical Engineering KU Leuven

Accurately determining the wind speed distribution is indispensable for a thorough wind-resource assessment of a wind farm. Mesoscale weather systems are the main cause of variability in wind speed, and consequently wind farm energy production in the Kattegat region, located between Denmark and Sweden. In particular, this sea strait has an irregularly shaped coastline, thereby exhibiting strong coastal effects on the wind fields. One of the main wind farms in the region is the Anholt wind farm, located just 21 km away from the Danish coast, providing the opportunity to study mesoscale weather systems and their interactions with wind farms.Since these mesoscale weather systems are not sufficiently resolved in global reanalysis data such as ERA5, regional climate modelling can support the quantification of the variability in wind speed. Previous studies

41 have already shown that convective cells, gravity waves and sea/land breeze systems have a major impact on the surface wind speed variability. In this study we use a high resolution (horizontal grid resolution of 1.5 km) mesoscale atmospheric model (COSMO-CLM) to better understand how variability in wind speed is related to mesoscale weather systems in the Kattegat region. Moreover we aim at studying how wind farms interact with these mesoscale weather systems. In order to validate our simulation we’ve compared the model output with scatterometer data from the Metop-A satellite. A time series of the surface wind speed at one single point was compared with the scatterometer data. The model data satisfactorily clearly represents the wind evolution measured by the scatterometer satellite. Output of our simulations show substantial temporal and spatial variability, mainly related to frontal systems travelling across the Kattegat. In the presentation we will examine the eﬀect of spectral nudging on the surface wind speed variability, investigate downscaling strategies and study the interactions of mesoscale weather systems with the Anholt wind farm. The simulation period will be extended to a full year in order to chart the seasonal trends in wind speed variability. As more and more wind farms are being constructed it becomes important to study the two way interactions between wind farms and mesoscale weather systems for a more accurate resource assessment concerning future wind farms and this is a ﬁrst step towards this overarching goal.

Pollen forecasts using ICON-ART in a limited area mode C. Endler(1), S. Muthers(1), J. Förstner(2),T. Hanisch(2), H. Vogel(3), and A. Pauling(4) (1) Deutscher Wetterdienst, Stefan-Meier-Str. 4-6, 79104 Freiburg, Germany, (2) Deutscher Wetterdienst, Frankfurter Str. 135, 63067 Offenbach, Germany, (3) Institute for Meteorology and Climate Research, Karlsruhe Institute of Technology, Hermann-von-Helmholtz-Platz 1, 76344 Eggenstein-Leopoldshafen, Germany, (4) MeteoSchweiz, Operation Center 1, Postfach 257, CH-8058 Zürich-Flughafen, Switzerland Pollen information and pollen forecasts are playing a dominant role in the management of allergies. As an increase of allergenic diseases caused by pollen could be observed in the last decades, allergies are becoming one of the major health issues. Among others, health care costs are increasing and limitations in quality of life are linked with such diseases. Precise pollen forecasts can support diagnostics of pollen allergies, management of specific immune therapy, behavioral adaptation and targeted intake of medication.Until the mid-1980s the pollen forecasts provided by DWD are based solely on measurements and observations. Now we are able to forecast the pollen concentration using atmospheric dispersion models. Recent developments have been performed to run pollen forecasts using ICON-ART in a limited area mode (LAM), on a regional 6.5 km grid over Europe. ICON-ART-LAM simulates four of the most relevant pollen species alder, birch, grasses, and ragweed. Preliminary results for the pollen season 2019 and 2020 show a good agreement to observations. This LAM application will be operational in 2021.

42 Representation of the urban environment in ICON-LAM A. Valmassoi(1,2), and J. D. Keller(1,3) (1) Hans-Ertel-Centre for Weather Research, Climate Monitoring and Diagnostics (2) University of Bonn (3) Deutscher Wetterdienst, Oﬀenbach, Germany

Urban environments and its feedbacks on the climate are a focus of recent scientific and socio- economic perspectives. The work presented here aims to investigates the urban climate representation in ICON-LAM. Free simulations are carried out at 2.1 km resolution for the central European domain (ICON-LAM D2.1) for the reference period of June 2019. The chosen period includes the late-June heat wave, which provided favorable conditions for the Urban Heat Island (UHI) development. In our study, we asses the urban representation of the standard version of ICON-LAM and find that ICON-LAM does not reliably reproduce the UHI. In order to enhance the representation of the UHI, we use various approaches to modify ICON-LAM by using higher resolution land use data (CORINE at 100 m) or higher spatial resolution in the simulations (500 m). Further, we focus on two possible computationally inexpensive ways to improve the urban representation for the D2.1 simulations. First, we introduce two additional urban land use categories from the CORINE dataset. Second, we apply an urban correction to the surface sensible heat fluxes based on the high complexity of the cities shape. This method uses buildings location and heights to derive adequate parameters to include to the ICON-LAM land surface model scheme.

SINFONY - the combination of Nowcasting and Numerical Weather Prediction on the convective scale at DWD U. Blahak(1), and the SINFONY-Team(1) (1) Deutscher Wetterdienst

DWD’s new Seamless INtegrated FOrecastiNg sYstem (SINFONY) is about to come to life in the next two years,after 4 years of research and development. Iniitally it focuses on severe convective events in the very short time forecast range from minutes to 12 h. A regional ICON-ensemble model with extensive data assimilation of high-resolution remote sensing data and hourly new rapid update cycle forecasts (SINFONY-RUC-EPS) is one of its core components. There are different “optimal” forecast methods for different forecast lead times and different weather phenomena. Focusing on precipitation up to some hours ahead, radar extrapolation techniques (Nowcasting) show good skill up to about 2 h ahead (depending on the situation), while numerical weather prediction (NWP) outperforms Nowcasting only at later hours. Ensembles of both Nowcasting and NWP help to assess forecast uncertainties. ”Optimally” combining precipitation forecasts from Nowcasting and NWP as function of lead time leads to seamless forecasts. Different interdisciplinary teams work closely together in developing a) Radar Nowcasting ensembles for precipitation, reflectivity and convective cell objects, b) hourly SINFONY-RUC-EPS NWP on the km-scale, c) optimal combination of Nowcasting and NWP ensemble forecasts in observation space

43 (precipitation, radar reflectivity and cell objects), d) systems for common Nowcasting and NWP verification of precipitation, reflectivity and objects.

For b), new innovative and efficient forward operators for radar volume scans and visible satellite data enable direct operational assimilation of these data in an LETKF framework. Advanced model physics (stochastic PBL scheme, 2-moment bulk cloud mircophysics) contribute to an improved forecast of convective clouds. For c), the SINFONY-RUC-EPS outputs simulated reflectivity volume scan ensembles of the entire German radar network every 5’ online during its forecast runs. Ensembles of composites and cell object tracks are generated by the same compositing and cell detection- and tracking methods/software packages which are applied to the observations. To help evolve DWD’s warning process for convective events towards a flexible a seamless forecast (”probability objects”) in a pragmatic way. The gridded combined precipitation and reflectivity ensembles are targeted towards hydrologic warnings. The presentation will give an overview on the system concept, its status, and the roadmap towards operational implementation during the next two years.

Potential links between tropospheric and stratospheric extremes during the winter of 2019/20 P. Rupp(1), S. Löffel(1,2),H. Garny(1,2), and T. Birner(1,2) (1) LMU Munich, (2) DLR Oberpfaffenhofen

We present results from a set of coupled troposphere-stratosphere extended-range hindcast ensemble simulations with the global ICON set-up for the period February-March 2020. Our analysis is based on sets of perturbation experiments with corresponding perturbations implemented in various ways, e.g., as changes to in the initial conditions, constantly forced via global nudging of the total wind field or based on a post-simulation clustering of ensemble members. These different approaches pose a range of challenges in terms of numerics and dynamics, in particular in combination with the specific set-up of the ICON model.The aim of this research is to gain insights into the extent to which observed tropospheric and stratospheric extreme (magnitude and persistence) events in early 2020 occurred independently. It is found that February was mainly dominated by a prominent episode of planetary wave reflection that lead to an extreme increase in polar vortex strength and that the strengthened vortex did, in return, create stratospheric conditions that were favourable for tropospheric extremes in terms of a (persistently) strengthened mid-latitude jet.

COSMO-GHG simulations of 14C to support the design of a radiocarbon measurement network in Europe M. J¨ahn(1),and D. Brunner(1) (1) Swiss Federal Laboratories for Material Science and Technology (Empa), D¨ubendorf, Switzerland

44 When it comes to monitoring carbon dioxide (CO2) emissions caused by human activities, a large-scale observation network must be designed to separate the effects of anthropogenic emissions from the effects of the complex natural carbon cycle, since both influence atmospheric CO2 concentrations. To accomplish this, measurement of other trace gases besides CO2, such as radiocarbon (14CO2) or carbon monoxide (CO), which are less affected by the biosphere, could play an essential role.Here we present European scale atmospheric transport simulations of these gases at 5 km x 5 km resolution to support the design of such a surface network. The simulations were carried out with COSMO-GHG, which is an extension of the GPU-accelerated COSMO model (v5.8) allowing to simulate the sources and transport of greenhouse gases in the atmosphere. Different contributions to total atmospheric 14CO2, CO2 and CO were simulated as separate tracers including anthropogenic sources by type, biogenic uptake and release, ocean fluxes and background concentration from outside the limited domain. The tracers were additionally divided into 6 geographical regions across Europe, summing up to a total of 75 tracers. We analyzed the concentration fields of the three tracers with respect to the contribution from fossil fuel emissions versus other sources for two one-month periods in 2015 (January and July). Time series were evaluated in detail for two different stations to illustrate the different characteristics of the station types (e.g., rural vs. urban) and the contributions from different regions as well as different emission source types. Especially during winter, the fossil fuel emissions strongly dominate the 14CO2 variability (almost 100%) over most of Europe, suggesting that radiocarbon is an almost perfect tracer of fossil-fuel CO2 emissions. In summer, the fossil fuel signals is still the dominant contribution, but it drops to 70-80% due to the presence of stronger biospheric fluxes. Ocean fluxes and nuclear power plant emissions, in contrast, play only a minor role except in close vicinity of nuclear power plants. CO can be a valuable tracer of fossil fuel emissions especially during summer when the burning of biofuels (e.g. wood) for heating is low, but only in the absence of strong emissions from biomass or agricultural waste burning. As shown in previous studies, CO could also be used to interpolate between low-frequency 14C measurements. Overall, the results obtained by our COSMO-GHG simulations provide invaluable hints towards the design of a dense ground-based measurement network of CO2/14CO2 , which could ideally complement the planned Copernicus CO2 satellite mission CO2M.

Sensitivity of gravity wave drag in seasonal experiments with ICON: Stratospheric dynamics and pathways R. K¨ohler(1),D. Handorf(1), R. Jaiser(1), and K. Dethloﬀ(1) (1) Alfred-Wegener-Institut

Stratospheric pathways play an important role in connecting distant anomaly patterns to each other on seasonal timescales. As long-lived stratospheric extreme events can influence the large-scale tropospheric circulation on timescales of multiple weeks, stratospheric pathways have been identified as one of the main potential sources for subseasonal to seasonal predictability in mid-latitudes. The ability of atmospheric models to realistically simulate these processes strongly depends on stratospheric performance of the model. In this context, we investigate stratospheric processes in a suite of gravity wave drag adjusted seasonal experiments with the atmospheric model ICON-NWP. Although biases are reduced due to gravity wave drag adjustments, the strength of the stratospheric polar vortex remains underestimated in ICON. Furthermore, based on these seasonal climatologies, this work explores strongly discussed stratospheric pathways. Here, we focus on the effect of El

45 Niño-SouthernOscillation (ENSO) on the stratospheric polar vortex, and thus the circulation in mid and high latitudes in winter. The effect is simulated realistically by ICON and the results from the ensemble simulations suggest that ENSO has a significant effect on the large-scale Northern Hemisphere winter circulation. ICON and the reanalysis exhibit a weakened stratospheric vortex in warm ENSO years. Furthermore, in particular in winter, warm ENSO events favour the negative phase of the Arctic Oscillation, whereas cold events favour the positive phase. The ICON simulations also suggest a significant effect of ENSO on the Atlantic-European sector in late winter.

Postprocessing of COSMO and IFS ensemble predictions for providing seamless forecasts C. Spirig(1), J. Bhend(1), S. Hemri(1,2), J. Rajczak(1,3), D. Nerini(1), R. Keller(3,1), D. Cattani(1), M. Schaer(1), L. Moret(1), and M.A. Liniger(1)

(1) MeteoSwiss, Development of Forecasting, (2) University of Zurich, Dep. of Mathematics, (3) ETH Zurich, Centre for Climate Systems Modelling (C2SM)

MeteoSwiss is developing and implementing a NWP postprocessing suite for providing improved automated weather forecasts at any location in Switzerland. The aim is a combined postprocessing of high resolution local area and coarser global model ensembles with different forecast horizons to enable seamless probabilistic forecasts over two weeks leadtime. Further, the output should be coherent in space and provide predictions at any location of interest, including sites without observations. We use the full archive of MeteoSwiss’ operational forecasts of COSMO-1 and COSMO-E over the past four years and the corresponding IFS-ENS medium range predictions of ECMWF to develop postprocessing routines for temperature, precipitation, cloud cover and wind. Here we present our findings on the performance of various postprocessing methods we applied but also on practical aspects of their implementation into operational production.Both ensemble model output statistics (EMOS) and machine learning (ML) approaches increase skill of COSMO-E direct model output by up to 30% in terms of CRPS, with most pronounced improvements in mountainous regions. Particularly for temperature, the combined postprocessing of COSMO and IFS-ENS resulted in a skill benefit over postprocessing the COSMO models alone. Locally optimized postprocessing would allow further skill improvements, but only at sites where observations are available. However, the ability of non-local postprocessing approaches to provide calibrated forecast at any point in space is a key advantage for providing automated forecasts to the general public via the internet and smartphone app. Furthermore, the computational efficiency of these non-local approaches makes it attractive for operationalization in a realtime context.

Evaluation of ICON-LAM Forecasting of a Strong Rain Eevent at the Coast of Sao Paulo State in Brazil Reinaldo.B. Silveira(1), Gilberto.R.Bonatti (2), D. Rieger(3), J. M. D. Mol(2), and R. R. dos Santos(2) 1) Sistema Meteorológico do Paraná,SIMEPAR, Curitiba, Paraná,Brazil, (2) Instituto Nacional de Meteorologia, INMET, Bras´ılia,DF, Brazil, (3) Deutscher Wetterdienst, Offenbach.

46 The summer season of 2019 in Brazil was drier and hotter than the average for both conditions. However, the Southeast of Brazil suﬀered with several events of strong precipitation, most of them happened due to large convective clouds fed by humidity advection from either central part of Brazil and Amazon rainforest or from the ocean. We therefore studied an event that occurred on Sao Paulo state, at 27th of March of 2019, which caused ﬂood and property losses in the city hit by the storms. Thus, ICON is being used to simulate such challenging patterns of regional weather and an experiment was performed with the goal to examine the impact of increasing the exchange between land-atmosphere and sea-atmosphere. We then run a set of simulations for the selected places and respective date of occurrences, comprising one nested domain setup at 2.8km and 1.4km, covering Central and Southeast of Brazil. The results are compared to a merged observation analysis, comprised by rain gauge, satellite, as well as compared to radar observation and automatic weather station network. In the context of tuning the ICON model for simulation of such events in Brazil, the results show that ICON is capable of reproducing timing and amount of precipitation satisfactory.

Improving forecasts of wind resources by including two-way coupling between atmospheric ﬂow and oﬀshore wind farms S. Jamaer(1), J. Meyers(1), and N. P. M. Van Lipzig(1) (1) KU Leuven

Recent studies have found that the increasing sizes of offshore wind farms can cause a reduced energy production through mesoscale interactions with the atmosphere. For example, a recent study [1] showed that energy production over the North Sea could decrease of up to 6% annually and 30% hourly when taking these effects into account. Therefore, incorporating these mesoscale feedback mechanisms in wind resource forecasting algorithms and planning tools becomes increasingly important in the future. Currently, forecasting of wind energy resources is performed using operational weather forecasts, without taking into account the interaction. During the course of the FREEWIND project, we aim at developing a methodology to take mesoscale interactions into account and apply this to improve forecasts for the Belgian wind farms in the North Sea. In later stages of this project, two methods will be developed and tested: the first method uses a look-up table with the correction factors for different classes from a mesoscale weather type classification that is currently developed. These correction factors will be derived from model integrations with the mesoscale model COSMO-CLM including the full interaction between atmosphere and wind farms, for selected representatives of the different classes. The second method will use a recently developed fast atmospheric perturbation model [2], driven by data from the operational forecasts. The perturbation model will be evaluated using the COSMO-CLM simulation for the selected cases and it will also be used to study the variation in correction factors withing the classes that are not accounted for in the first method. In the current phase of this project, the mesoscale weather type classification is implemented as it is the first crucial element of this work. Because the two-way coupling is largely determined by the vertical structure of the atmosphere, the classification is based on vertical profiles of temperature and wind speed from ERA5. Characterisation of the vertical profiles is a key element in this work and several methods (physically-based, using line segments and machine learning approaches) will be tested. In the end, the clustering parameters and methods will be selected such that they optimally span the atmospheric states over the North Sea in terms of modulation to atmospheric flow induced by two-way interaction with the wind farm. This is evaluated by using

47 the perturbation model from [2], that is designed to calculate this modulation of the atmospheric flow in a very computationally efficient way. References: [1] D. Allaerts, S. V. Broucke, N. van Lipzig, and J. Meyers, ”Annual impact of wind-farm gravity waves on the Belgian-Dutch offshore wind-farm cluster”, Journal of Physics: Conference Series, vol. 1037, p. 072006, jun 2018 [2] D. Allaerts and J. Meyers, “Sensitivity and feedback of wind-farm-induced gravity waves,” J. Fluid Mech, vol. 862, pp. 990-1028, 2019.

Interaction between stratospheric equatorial waves and gravity waves and its implication in QBO simulation Y.-H. Kim(1), and U. Achatz(1) (1) Goethe-Universit¨atFrankfurt am Main

We report an interaction between stratospheric Kelvin waves and gravity waves (GWs) in the tropics, using ICON with the R2B4 horizontal resolution (grid spacing of 160 km). In ICON, the Multi-Scale Gravity Wave Model (MS-GWaM) is used as a subgrid-scale parameterization to represent GWs instead of the operational non-orographic GW scheme. MS-GWaM is a prognostic model based on the Lagrangian technique, which simulates the evolution of GW action density in phase space. The simulation is initialized on 1 May 2010, when the easterly jet of the stratospheric quasi-biennial oscillation (QBO) is placed at 20 hPa, so that Kelvin waves freely propagate in the vertical throughout the lower stratosphere. It is found for the first time that Kelvin waves with typical amplitudes (about 10 m/s for the zonal wind) affect the distribution of GW drag significantly, by perturbing the local wind shear. Furthermore, this effect of Kelvin waves does not vanish in zonal average of GW drag, implying that it may have an implication in the QBO progression. The area of a strong tropical convective system, where large-amplitude GWs are emitted, is small compared to the zonal wavelength of large-scale Kelvin waves, so that it is covered by a certain phase of Kelvin waves in the stratosphere. Accordingly, the coupling of Kelvin waves and GWs in the stratosphere is asymmetric between the phases of Kelvin waves, resulting in the zonal-mean effect. The effect on the zonal-mean GW drag seems to be large when the strong convective system propagates eastward together with a phase of the Kelvin wave aloft. In our simulated case, such an interaction provides a zonal-mean GW drag of 8 m/s/month at 20 hPa for 7-8 days during an early phase of the easterly-to-westerly transition of the QBO there. The result revisits the importance of a proper representation of large-scale waves, subgrid-scale GWs, and their association with convective systems in QBO simulations.

First steps towards 1 km horizontal resolution over Germany V. Maurer(1), A. De Lozar(1), and G. Zaengl(1) (1) Deutscher Wetterdienst

Usually, an improvement of numerical weather prediction can be expected at higher horizontal resolution. On the one hand, higher resolved surface data such as orography are more realistic. On the other hand, physical parametrizations, especially convection, which is only an approximation of reality, should be less active as resolved processes are taking over.However, ICON-D2 has a horizontal resolution of about 2 km, which is lying in the gray zone of convection, and a part

48 of the forecast performance, especially for summertime clouds and precipitation, depends on a tuning of parametrized shallow convection. Thus, moving towards 1 km is not necessarily an improvement. Moreover, at 1 km also the gray zone of turbulence is approached. Possibly, this is the reason why near-surface wind speed shows a significantly different mean error and RMSE. We investigated differences between 2 km and 1 km simulations with ICON in a hindcast setup with a 1-km nest inside of the ICON-D2 domain. Apart from near-surface winds, bias differences can also be seen for precipitation and 2m temperature. These differences have to be reduced before 2-way nesting can be used properly. When focusing on summertime convective periods as in SINFONY and using output from the radar forward operator, an increased hit rate as well as ETS can already be shown at 1 km resolution. Thus, even if it is not possible in near future especially due to limited data storage capacities, a horizontal resolution of 1 km is a longer-term aim for SINFONY.

49 Planetary Boundary Layer

A preliminary analysis of the impacts of small-scale orography on the stable atmospheric boundary layer J. Quimbayo-Duarte(1), and J. Schmidli(2) (1) Hans Ertel Centre for Weather Research and Institute for Atmospheric and Environmental Sciences, (2) Goethe University Frankfurt, Frankfurt/Main, Germany

The representation of small-scale turbulent processes in the stable boundary layer (SBL) remains a challenge for weather prediction systems, especially over complex terrain where variations in the orography dramatically affect the development of the flow. To improve the representation of the SBL over complex terrain, the impacts of small-scale orography on the SBL needs to be better understood and parameterized. As a first step, an idealized set of experiments has been designed to explore the capabilities of ICON-LES to represent turbulence processes in the stably-stratified atmosphere. Initial experiments testing the model performance over flat terrain (GABLS experiment, Beare et al., 2006) and moderate complex terrain (U-shaped valley, Burns and Chemel 2014) have been conducted. The results demonstrate that ICON-LES adequately represents the SBL for the investigated cases in comparison to the literature. In a second step, an idealized set of experiments of atmospheric flow over idealized multiscale terrain has been designed to study the impact of the orographically-induced gravity waves on the total surface drag and the vertical flux of horizontal momentum, as well as the impact of stratification on the turbulent orographic form drag. The influence of different atmospheric conditions is assessed by varying the background wind speed and the temperature stratification at the initial time.

A Budget-Based Turbulence Length Scale Diagnostic I. B. Duran(1), J. Schmidli(1), and S. Reilly(1) (1) Goethe University Frankfurt

The most frequently used boundary-layer turbulence parameterization in numerical weather prediction (NWP) models are turbulence kinetic energy (TKE) based schemes. However, these parameterizations suffer from a potential weakness, namely the strong dependence on an ad-hoc quantity, the so-called turbulence length scale. The physical interpretation of the turbulence length scale is difficult and hence it cannot be directly related to measurements or large eddy simulation (LES) data. Consequently, formulations for the turbulence length scale in basically all TKE schemes are based on simplified assumptions and are model-dependent. A good reference for the independent evaluation of the turbulence length scaleexpression for NWP modeling is missing. We propose a new turbulence length scale diagnostic which can be used in the gray zone of turbulence without modifying the underlying TKE turbulence scheme. The new diagnostic is based on the TKE budget: The core idea is to encapsulate the sum of the molecular dissipation and the cross-scale TKE transfer into an effective dissipation, and associate it with the new turbulence length scale. This effective dissipation can then be calculated as a residuum in the TKE budget equation (for horizontal sub-domains of different sizes) using LES data. Estimation of the scale dependence of the diagnosed turbulence length scale using this novel method is presented for several idealized cases.

50 Evaluation of thermally driven local winds in the Swiss Alps simulated by COSMO-1 J. Schmidli, and J. Quimbayo-Duarte Hans Ertel Centre for Weather Research and Goethe University Frankfurt, Frankfurt/Main, Germany In fair weather conditions, thermally driven local winds often dominate the wind climatology in deep Alpine valleys resulting in a unique wind climatology for any given valley. The accurate forecasting of these local wind systems is challenging, as they are the result of complex and multi-scale interactions. Even more so, if the aim is an accurate forecast of the winds from the near-surface to the free atmosphere, which can be considered a prerequisite for the accurate prediction of mountain weather. This study investigates the skill of the most current COSMO version (v5.06) at 1.1 km grid spacing in simulating the thermally driven local winds in the Swiss Alps for a month-long period in September 2016. The study combines the evaluation of the surface winds in several Alpine valleys with a more detailed evaluation of the wind evolution for a particular location in the Swiss Rhone valley, the town of Sion. The former is based on a comparison with observations from the operational measurement network of MeteoSwiss, while the latter uses data from a wind proﬁler stationed at Sion airport.

Slope winds in the convective boundary layer over mountainous terrain: LES results and parameterization approach J. Weinkaemmerer(1), I. B. Dur´an(1),andˇ J. Schmidli(1) (1) Goethe University Frankfurt

The results of idealized large-eddy simulations studying the vertical exchange of heat and mass over mountainous terrain are presented. The amount of heat and mass exported out of a valley is important, e. g., for cloud formation or pollutant dispersion. Local thermal circulations developing over heated valley slopes on fair-weather days strongly influence the flow in the convective boundary layer. Here, the decomposition of the flow into a local turbulent part, a local mean circulation capturing the slope winds, and a large-scale part leads to a better process understanding. In the idealized case of an infinitely long alpine valley, the thermal winds export moisture out of the valley while turbulent transport is mainly responsible for the heat exchange. The higher temperatures in the valley compared to the rest of the atmosphere are in accordance with the so-called valley-volume effect. These processes affect the mean temperature and moisture profiles even over relatively shallow topography and also lead to additional advection and shear terms in the budgets of the turbulent kinetic energy and the turbulent fluxes. However, an upper-level synoptic wind alters the local flow and reduces the mass export out of the valley effectively while leaving the temperature distribution almost unmodified. The impact of an upper-level wind was also tested in a more realistic diurnal-cycle experiment.

51 Simulations of the Arctic atmospheric boundary layer around the MOSAiC drift track using ICON-LEM D. Littmann, W. Dorn, H. Bresson, M. Maturilli, A. Rinke, and M. Rex Alfred-Wegener-Institute, Helmholtz Centre for Polar and Marine Research, Potsdam, Germany The icosahedral non-hydrostatic large eddy model (ICON-LEM) is applied for the first time in the central Arctic around an area of the multidisciplinary drifting Observatory for the Study of Arctic Climate (MOSAiC) that was running from September 2019 to September 2020. The model domain is set up for a storm event in November 2019 with the drift track of the Polarstern vessel as a central point, horizontal resolutions between 100 m and 800 m, and radii between 22 km and 140 km. ICON-LEM is driven at the lateral boundaries by analysis data from the German Weather Service (DWD), downscaled to 3 km resolution using ICON-LAM. First ICON-LEM experiments showed that the sea-ice surface is not realistically represented for the central Arctic using the default ICON-LEM configuration in which the surface conditions are just downscaled values of the coarse-scale analysis data. Sub-grid-scale structures within the sea ice, especially open water leads, are thus not adequately considered and resulted in too cold and too dry surface conditions in ICON-LEM and a decoupling of the near-surface layers from the upper atmosphere. To obtain a more realisticrepresentation of the surface, sea-ice thickness, fraction, and temperature are set manually for the ICON parent domain, whereby small-scale variations in the surface conditions are made possible. A first comparison reveals that the previous model biases are greatly reduced with the result that the simulations agree now much better with the measurements from MOSAiC. Also the strong decoupling of the near-surface layers from the upper atmosphere is not present anymore. The results indicate the need of a more sophisticated treatment of the ice surface conditions for ICON-LEM applications in the central Arctic.

An improved sea ice parameterization and tile approach in CCLM G. Heinemann, L. Schefczyk, and R. Zentek Environmental Meteorology, University of Trier The parameterization of ocean/sea-ice/atmosphere interaction processes is a challenge for regional climate models in polar regions, particularly for wintertime conditions, when small fractions of thin ice or open water cause strong modifications of the boundary layer. Thus, the treatment of sea ice and sub-grid flux parameterizations are of crucial importance. CCLM was adapted to polar regions by implementing a two-layer sea ice model, a tile approach for sea ice, and modifications for the stable boundary layer (SBL). The modifications include an improved computation of the sea ice energy budget, a new parameterization of the subgrid-scale ice thickness (thin ice in leads and polynyas), non-linear averaging for the tile approach, new parameterizations for the roughness lengths of momentum and heat as well as a TKE-dependent asymptotic mixing length for the SBL. In-situ data of the Transarktika expedition for thick sea ice conditions in late winter 2019 and Moderate Resolution Imaging Spectroradiometer (MODIS) data are used for the verification of the CCLM simulations with 5km horizontal resolution.

52 The two-energies turbulence scheme I. B. Duran(1), and J. Schmidli(1) (1) Goethe University Frankfurt

A turbulence scheme with prognostic equations for two turbulence energies is presented. The scheme is an extension of a Turbulence Kinetic Energy (TKE) scheme with and additional prognostic energy, which is a simplified equivalent of the two scalar variances. The additional turbulence prognostic energy is used only for the calculation of the stability parameter. Thus the turbulent fluxes in the two-energies scheme are down-gradient as in a conventional TKE scheme. However, the energy dependent stabilityparameter is not anymore strictly local and has a prognostic character. These characteristics enable the scheme to model both turbulence and clouds in the atmospheric boundary layer. To ensure a consistent treatment of boundary layer clouds and turbulent, the Assumed Probability Density Function (APDF) method is used for the computation of the buoyancy flux and the cloud fraction. The higher order moments, which are required to determine the shape of the trivariate probability density function, are diagnosed from the turbulent fluxes and the two prognostic energies. The scheme was implemented into the ICON model. We will present results of the new scheme for selected idealized cases in a one-dimensional and three-dimensional model setup.

53 Predictability and Ensemble Systems

Choosing the Optimal Sub-Ensemble of Boundary Conditions to Drive Convection Permitting Ensemble P. Khain, A. Shitivelman, Y. Levi, E. Amitai, I. Carmona, A. Baharad, E. Vadislavsky, A. Savir, and N. Stav Israel Meteorological Service In order to represent the true uncertainty of NWP forecast, the model ensemble has to reflect the uncertainties in model dynamics, initial and boundary conditions. In this work we focus on the method for obtaining the optimal boundary conditions. The 50-members ECMWF ensemble is used to drive the convection permitting 20-members COSMO ensemble, and therefore a sub- ensemble of 20 ECMWF driving members needs to be specified. This selection is performed using the cluster analysis of ECMWF ensemble members (Molteni et al., 2001, Marsigli et al., 2011). Various ECMWF atmospheric fields at different forecast ranges are tested yielding the definition of the optimal method to select the sub-ensemble of boundary conditions.

Statistical and machine learning methods for postprocessing ensemble forecasts of wind gusts B. Schulz(1), and S. Lerch(1) (1) Karlsruhe Institute of Technology (KIT)

We conduct a systematic and comprehensive comparison of state-of-the-art postprocessing methods for ensemble forecasts of wind gusts. The compared approaches range from well-established techniques to novel neural network-based methods. Our study is based on a 6-year dataset of forecasts from the convection-permitting COSMO-DE ensemble prediction system, with hourly lead times up to 21 hours and forecasts of 57 meteorological variables, and corresponding observations from 175 weather stations over Germany. We ﬁnd that simpler methods such as ensemble model output statistics (EMOS), member-by-member postprocessing and a novel isotonic distributional regression approach, which utilize ensemble forecasts of wind gusts as sole inputs, already result in improvement in terms of the mean CRPS of up to 40% compared to the raw ensemble predictions. This can be substantially improved upon by more complex machine learning methods such as gradient boosting-based extensions of EMOS, quantile regression forests, and variants of neural network-based approaches that are capable of incorporating additional information from the large variety of available predictor variables.

Predictability analysis and veriﬁcation of the Lightning Potential Index (LPI) in the COSMO-D2 high resolution EPS M. Salmi(1), C. Marsigli(2), and M. Dorninger(1) (1) Universit¨atWien, (2) DWD

During the last decade, the constant improvement in computational capacity led to the development

54 of the first high resolution ensemble prediction systems (EPSs). The COSMO-D2 EPS has a spatial resolution of around 2km, thus permitting large scale, deep convective processes such as thunderstorms or heavy showers to be handled explicitly, without any physical parametrization necessary. Special parameters involving both clouds (micro-)physics and large scale lifting - such as the Lighting Potential Index, or LPI - have also been developed in order to try and bring the forecasting of deep convection and therefore also of lightning activity to a new level of spatial resolution. With such high-precision forecasts comes however also a much higher error potential, at least for gridpoint-verification. The use of this high resolution setup in an ensemble prediction system might however bring huge benefits in terms of accuracy. This work is a preliminary attempt to apply innovative verification approaches such as the dispersion Fractions Skill Score or the EPS-SAL to the LPI in the COSMO-D2 EPS. Our final aim is to assess the error-predictability relationship as well as the horizontal skillful scale of the high resolution COSMO-D2 EPS for the Lightning Potential Index.

An estimation of the intrinsic predictability limit and required improvements to approach it T. Selz, and G. Craig(1) (1) Meteorologisches Institut, LMU

The predictability of atmospheric turbulence and thus weather is thought to be limited by the short intrinsic predictability of moist convection followed by subsequent upscale error growth. To investigate this phenomenon we use ICON model simulations together with the stochastic convection scheme of Plant and Craig. This setup allows for a much more accurate simulation of upscale error growth from convection than obtained with deteministic schemes and is still inexpensive enough to perform larger numbers of ensemble members and cases. Our results conﬁrm earlier estimates for the intrinsic predictability limit of a bit more than two weeks. Furthermore, we used ensembles with rescaled initial condition uncertainty obtained from ECMWF together with the stochastic convection scheme. These experiments provide an assessment of the current practical importance of upscale error growth from convection and the required improvement of the current observational and data assimilation system to get close to the intrinsic limit.

55 Soil, Vegetation and Ocean

Marine extreme events in high-resolution coupled model simulations G. Eirund(1), M. M¨unnich(1), M. Leclair(2), and N. Gruber(1,2) (1) Environmental Physics, Institute of Biogeochemistry and Pollutant Dynamics, ETH Zurich, (2) Center for Climate Systems Modeling (C2SM), ETH Zurich

Extreme events on land, such as heatwaves and droughts, are well known to substantially shape the structure and metabolism of terrestrial ecosystems. However, our understanding of the role of such events in the marine environment remains limited. Many marine extreme events, and especially marine heat waves, result from the interaction of oceanic and atmospheric weather, yet, there is little consensus on how the interplay between atmospheric and oceanic forcings impacts the spatial and temporal scales of such extreme events and affects the marine ecosystem and ocean biogeochemistry.Given these complex interactions between the atmosphere, the ocean, and marine biogeochemistry, we developed a coupled regional high-resolution Earth System Model (ROMSOC). ROMSOC comprises the latest officially released GPU-accelerated COSMO version coupled to the Regional Oceanic Modeling System (ROMS). ROMS in addition includes the Biogeochemical Elemental Cycling (BEC) model that describes the functioning of the lower trophic ecosystem in the ocean and the associated biogeochemical cycle. Here, we present first simulations of our coupled model system for the California Current System (CalCS) at the US west coast.

Convergence of Richards Equation and Implications for Inﬁltration and Water Propagation D. Regenass(1), L. Schlemmer(2), and C. Sch¨ar(1) (1) Institute for Atmospheric and Climate Science, ETH Zurich, (2) Deutscher Wetterdienst

The exchange of water and energy between the land surface and the atmosphere is tightly coupled to the hydrological cycle. On annual and interannual timescales, catchment precipitation is roughly balanced by evapotranspiration and river discharge. The partitioning of incoming rainfall between runoff and infiltration, and thus availability of water for subsequent evapotranspiration, is a key task of the land surface model. Errors in one component of the hydrological cycle will inevitably propagate to the other components. In regional climate modelling and numerical weather prediction, one is mostly concerned with an accurate estimation of precipitation and evapotranspiration, but for the aforementioned reason, an accurate representation of the partitioning between infiltration and runoff is critical for the estimation of evapotranspiration. Our recent investigations using TERRA reveal that runoff formation is tightly linked to the representation of the infiltration process (Regenass et al., submitted), which is in turn subject to the representation of vertical water transport. In addition to physical processes determining the partitioning between infiltration and surface runoff, one is also faced with numerical challenges. In order to represent the infiltration process properly, the sharp gradients around the propagating wetting front must be resolved. In TERRA and many other land surface models, vertical water transport in the soil is represented by the one-dimensional Richards Equation, which is solved

56 using finite differences on a grid with typically around ten vertical layers. In this work, we use a stripped-down version of TERRA – reduced to its hydrological component – to examine the infiltration process and the formation of surface runoff with a focus on numerical aspects. The convergence of numerical implementations of the Richards Equation with respect to the spatial and temporal discretization is investigated. This is done for two different parameterizations of the hydraulic conductivity (Rijtema (1969) versus Mualem (1976) - van Genuchten (1980)). It is shown that in order to resolve the sharp gradients around the propagating wetting front – which determine infiltation capacity – relatively high resolutions in space and time are required. The propagation velocity of the wetting front decreases systematically with decreasing resolution in the vertical direction and time. While the convergence is better for the Mualem - van Genuchten parameterization, in both formulations the demands in terms of vertical resolution are higher than what is typically implemented in TERRA and other land surface models. Moreover it is shown that some formulations using a grid with uneven (’telescope-like’) vertical spacing introduce first-order errors to the solution.

PT-SAINT – Outcomes and outlook for multi-layer snow modelling in COSMO V. Sharma(1,2), S. Bellaire(3), L. Braud(3), M. Lehning(1,2), and J.-M. Bettems(3) (1) Snow Processes group, SLF Davos (2) Laboratory of Cryospheric Sciences, EPFL (3) Meteoswiss The PT-SAINT priority project of the COSMO-Consortium is focused on the development as well as the operational deployment of a newly developed multi-layer snow model for modelling seasonal snowcover in the mid-latitudes in general and in particular, in the Alps. The multi-layer snow model is a significant improvement over the existing single layer treatment of the snow cover in COSMO/TERRA that has severe issues, most prominently, under estimation and untimely melting of the snow water equivalent (S.W.E). The newly developed multi-layer snow model, named ’SNOWPOLINO’ has now been extensively tested and implemented in both the COSMO model as well as the closely related, Terra StandAlone (TSA) model. In this presentation, we shall first discuss the performance of SNOWPOLINO within the TSA framework when driven by meteorological data from high-altitude IMIS stations in the Swiss Alps. The IMIS stations have high quality measurements of both standard meteorological variables as well as snow height and are therefore valuable data sources for snow modelling. Comparing SNOWPOLINO to the current snow model in TERRA reveal significant improvements in both S.W.E as well as snow height predictions. Next, we shall show results from the new model driven by atmospheric analyses from COSMO-1 for the 2018-2019 winter season while comparing the results to the existing model in TERRA as well as multiple snow analysis products. Improvements in the spatial extent as well as heterogeneity of the snow cover are highlighted. Finally, focusing on the deployment for forecasting operations, we shall discuss performance numbers such as wall time and memory costs of running COSMO-1 with SNOWPOLINO as well the new snow analysis scheme based on TSA. The maturity of the SNOWPOLINO scheme opens new possibilities both for operational forecasting as well climate modelling. Closely related to the PT-VAINT project, improved modelling of snow interaction with vegetation and forest canopies can now be pursued. Impact of aerosols and dust can now be tackled following the work by Anika Rohde et al. at KIT. Snow cover remote sensing in complex topography remains a challenging problem and is closely related to cloud cover

57 detection. Using SNOWPOLINO within TSA could be of interest in this direction as well. These and other possible user scenarios are discussed in the concluding section of the presentation.

Implementing ICON in TSMP – Coupling strategy and applications S. Poll(1,2,3), A. Ghazemi(1,3), D. C. Voullieme(1,2,3), H.-J. H. Franssen(1,2,3), and S. Kollet(1,2,3) (1) SimLab Terrestrial Systems, JülichSupercomputing Centre, Research Centre Jülich,Jülich, Germany (2) Institute of Bio- and Geosciences, Agrosphere (IBG-3), Research Centre Jülich, Jülich,Germany (3) Centre for High-Performance Scientific Computing in Terrestrial Systems, Geoverbund ABC/J, Jülich,Germany

The ever increasing computational resources are leading to a refinement of grid spacing of atmospheric models as well as the possibility of large eddy simulation for real data applications. Thus, the land surface with its multi-scale heterogeneity related to e.g. land cover and soil moisture gains in importance in atmospheric modeling. The Terrestrial System Modeling Platform (TSMP) is a scale-consistent, highly modular fully integrated soil-vegetation-atmosphere modeling system for regional earth system modeling. TSMP is composed of an atmospheric model (ICON, COSMO), a land surface model (NCAR Community Land Model - CLM), and a subsurface flow model (ParFlow) coupled together using the OASIS3-MCT coupler. The model components can be configured to run standalone or coupled in various configurations and with different grid spacings for each model component. TSMP can be applied at scales ranging from field-scale to continental scale.In TSMP, we incorporated the numerical weather prediction and large eddy mode of the atmospheric numerical model ICON, developed by the German Weather Service (DWD) and Max-Planck Institute for Meteorology. Here, we present an overview about the development strategy along with technical and performance aspects arising from the coupling process. We provide insights in the boundary layer development originating from an improved physical treatment of the land surface as well as the 3D water transport in the (sub)surface model.

Projected future changes in Vb-cyclone precipitation and moisture source region characteristics P. K. Pothapakula(1), A. Krug(1,2), A. Obermann(1), T. Keber(1), and B. Ahrens(1) (1) Goethe University, Frankfurt am Main (2) Deutscher Wetterdienst

The Vb-cyclones propagating from the Mediterranean Sea to Central Europe caused several past extreme summer floods in Central Europe, such as the Elbe flood in August 2002. This study explores the characteristics of the Vb-events moisture sources contributing to the precipitation over Danube, Elbe, and Odra catchments and associated processes in coupled regional climate simulation under the historical (1950-2000) and possible future climate conditions (2050-2100 under RCP 8.5 scenario). Besides continental moisture uptake and other oceanic moisture source regions, the Mediterranean Sea plays an extraordinary role in the intensification of Vb-events

58 (Krug et al., 2020). Therefore, ﬁrstly, we evaluate our coupled regional climate simulation with a focus on the mean climatic behavior of Mediterranean sea surface temperature (SST) and its variability. Secondly, we present the characteristics of various moisture source regions during detected Vb-events in the historical period. We compare our ﬁndings with Vb-cyclone events under future climate conditions and discuss changes in Vb-cyclone precipitation intensity and moisture source region characteristics and linked processes.

59 Veriﬁcation (NWP) and Evaluation (Climate)

Preliminary tests with ICON-LAM and comparison with COSMO-LM at high resolution over Italy C. De Lucia(1), A. Mastellone(2), P. Schiano(1), and E. Bucchignani(1,2) (1) CMCC Foundation, (2) CIRA Italian Aerospace Research Center

The ICON (ICOsahedral Nonhydrostatic) is a joint project between the Deutscher Wetterdienst (DWD) and the Max-Planck-Institute for Meteorology (MPI-M) for the development of a unified next-generation global numerical weather prediction system.The CMCC Foundation and the Italian Aerospace Research Center (CIRA) as members of COSMO consortium are taking part to the process of transition from COSMO to ICON Limited Area Model (LAM). To this aim, ICON software package (icontools version 2.4.12, ICON version 2.6.2.2) has been installed on CMCC and CIRA supercomputing facilities, and preliminary runs are currently in progress. In particular, the ZEUS CMCC supercomputer is a Lenovo HPC cluster with an Intel Xeon Gold 6154 (18 cores), 348 dual processors nodes, for a total number of 12528 cores. The interconnection is Infiniband EDR (100Gbps). The ICON package has been compiled using Intel Fortran 19.5 compiler and the following libraries: Eccodes 2.12.5, HDF5 1.10.5, and NetCDF 4.7.2 (C language), 4.5.2 (Fortran), 4.3.1 (C++). The libraries have been generated using the Intel compiler. The main aim of this work is to show preliminary results obtained with ICON-LAM over the Italian peninsula using a R2B10 grid (328246 cells, about 2.5 km resolution) with boundary conditions provided by ICON global model (courtesy of DWD). The time step was equal to 24 s. The period considered is from 16th to 31 August 2020. Model evaluation has been conducted by means of a comparison with a combination of available ground observation provided by CIRA instrumentation data and by the SCIA system (national system for the collection, elaboration, and diffusion of climate data) developed by ISPRA (Istituto Superiore Protezione e Ricerca Ambientale). Numerical performances of the model have been analyzed in terms of parallel speed-up and efficiency by varying the number of cores employed. Moreover, a comparison with forecasts provided by the COSMO model (cosmo 5.05 urb5) has been performed, using data obtained with daily simulations over a domain located in the southern Italy, including the urban area of Naples and CIRA facilities, employing a spatial resolution of 0.009◦ (about 1 km) and driven by ECMWF IFS model, A preliminary analysis has shown that in a day with intense precipitation the convection is triggered close to the northern and eastern boundaries in both cases. The position of the individual convective cells, however, is different between the two models. In a day without precipitation, this null value is well reproduced by both models, while ICON better reproduces the daily temperature value.

Comparison of calculated by COSMO-SIB and ICON-SIB models temperature proﬁles in the boundary layer with available observation data for Novosibirsk city A. Gochakov(1), V. Tokarev(1), and A. Kolker(1,2) (1) Siberian Regional Hydrometeorological Research Institute, (2) Novosibirsk State Technical University

60 The accuracy and reliability problem of data assimilation, forecasting, and verification of models in the lower part of the boundary layer of the atmosphere are well known. This work attempts to make numerical estimates of errors in the model reconstruction of vertical temperature profiles of the COSMO-SIB and ICON-SIB in the layer from 10 to1000m in comparison with the synchronous measurements of the ultrasonic temperature profiler installed at Tolmachevo (UNNT) airport and the radiosounding (UNNN) measurements up to altitude 1 km. Synchronous profile data graphs and standard surface observations of the airport were analyzed. Statistics were calculated for a few months: the mean errors, root mean square errors, and correlation coefficients. For the comparison model and observed values, the nearest model grid points were used. The differences between nearest to ultrasonic temperature profiler and radiosounding model grid points data were analyzed. As well as between the ultrasonic temperature profiler and the radiosounding measurements.

Evaluation of a high-resolution dynamical downscaling of ERA5 with COSMO-CLM for the North Sea R. Borgers(1), J. Meyers(2), and N.P.M. Van Lipzig(1) (1) KU Leuven, Department of Earth and Environmental Sciences, (2) KU Leuven, Department of Mechanical Engineering The growing importance of offshore wind energy emphasizes the need for realistic projections of the energy yield of wind farms over their lifetime. An analysis of CMIP5 projections shows that, even though near-future wind speed changes over Europe and the North Sea are uncertain across models, these changes should be taken into account by wind industries due to the potentially large impact on the energy yield (Devis et al., 2018).In this study, we aim to improve estimates of lifetime yield by including the losses due to the interaction between the wind farms and the atmospheric flow for wind farms over the North Sea for the 2020-2050 period. This will be done through the dynamical downscaling of several CMIP6-model realisations to a spatial resolution of 2.8km using the COSMO-CLM regional climate model which includes the wind farm parametrization by Fitch et al. (2012) that was recently implemented (Chatterjee et al., 2016, Akhtar et al., 2020). The first step of the research will be to evaluate the quality of COSMO-CLM wind data at 2.8km resolution by downscaling a 30-year reanalysis dataset (ERA5). For ICCARUS, we aim to present (a subset of) this ERA5-driven output of COSMO-CLM, including a comparative analysis of the generated wind data with different observations. This analysis can then be used as a measure for the quality of the simulated wind fields, which is a fundamental uncertainty in an assessment of wind energy potential. While reanalysis-driven wind fields generated by COSMO-CLM for the North Sea have already been tested against observations in the past (Geyer et al., 2015), the present study will evaluate model results at a higher, convection-permitting spatial resolution and attempt to incorporate observational data from other measurement stations. This planned study frames within the FREEWIND project of KU Leuven (freewind-project.eu). References: Akhtar, N., & Rockel, B. (2020, May). Mesoscale resolving high-resolution simulation of wind farms in COSMO-CLM 5. In EGU General Assembly Conference Abstracts (p. 7178). Chatterjee, F., Allaerts, D., Blahak, U., Meyers, J., & van Lipzig, N. P. M. (2016). Evaluation of a wind-farm parametrization in a regional climate model using large eddy simulations. Quarterly

61 Journal of the Royal Meteorological Society, 142(701), 3152-3161. Devis, A., Van Lipzig, N. P., & Demuzere, M. (2018). Should future wind speed changes be taken into account in wind farm development?. Environmental Research Letters, 13(6), 064012. Fitch, A. C., Olson, J. B., Lundquist, J. K., Dudhia, J., Gupta, A. K., Michalakes, J., & Barstad, I. (2012). Local and mesoscale impacts of wind farms as parameterized in a mesoscale NWP model. Monthly Weather Review, 140(9), 3017-3038. Geyer, B., Weisse, R., Bisling, P., & Winterfeldt, J. (2015). Climatology of North Sea wind energy derived from a model hindcast for 1958-2012. Journal of Wind Engineering and Industrial Aerodynamics, 147, 18-29.

Polarimetric radar forward operator for model validation and data assimilation J. Mendrok(1), and U. Blahak(1) (1) Deutscher Wetterdienst, Research&Development, Data Assimilation (Oﬀenbach, Germany)

Radar observations play a crucial role in detecting and measuring precipitation and in short-term numerical weather prediction (NWP). Accurate quantitative precipitation estimation remains a challenge as relationships between radar reflectivity and precipitation rate are often ambiguous. Polarimetric measurements provide better constraints on, e.g. shape and orientation, hence size of precipitating particles. Beside improving precipitation measurements, polarimetric observations can be used to verify and improve NWP models. Calculating radar observables from prognosed NWP fields, forward operators (FOs) are a crucial link in comparing radar measurements to NWP output. DWD operates 17 polarimetric weather radar stations over Germany, providing 3D information about precipitation and its movements with five minute and sub-kilometer resolution. To exploit the rich, but underexploited information content of polarimetric measurements, it is crucial that FOs can accurately simulate corresponding observations, which is so far a major bottleneck. Here, we report on extending the Efficient Modular VOlume-scanning RADar Operator, applied in the COSMO and ICON models, with polarimetric capabilities.

Object based verification of radar-reflectivities on the convective scale G. Pante(1), M. Hoff(1), and U. Blahak(1) (1) Deutscher Wetterdienst, Offenbach, Germany

Summer thunderstorms are one of the high impact weather phenomena that can cause strong socio-economic impacts over Germany. Such convective events can be hard to predict and are in the focus of the project SINFONY (Seamless integrated forecasting system) at DWD. SINFONY has the goal to improve forecasts of such events in the short range up to 12 hours. Nowcasting systems currently are superior to NWP systems on the very-short range up to about three hours in predicting convective cells while NWP forecasts perform better afterwards. Within SINFONY products are developed that optimally integrate both approaches in a seamless prediction system. High-resolution reﬂectivities from the German radar network are used as observational data base. The respective reﬂectivities from NWP models are derived by employing the radar forward operator EMVORADO (Zeng et al. (2016), Quarterly Journal of the Royal Meteorological Society,

62 142, 701, 3234-3256). Convective cells are systematically identified from these reflectivity data sets using the KONRAD3D cell detection tool (Werner, M. (2017), Second European Nowcasting Conference, Offenbach, Germany, EUMETNET, 15-16).Here we present results of the object-based method for the verification of radar reflectivities in NWP and Nowcasting called Median of Maximum Interest (MMI, Davis et al. (2009), Weather and Forecasting, 24, 1252-1267). It potentially helps to circumvent the well-known double-penalty problem. Another advantage of the MMI is that an a priori matching between certain observed and forecasted objects, which is often unreliable, is not mandatory. It rather measures the similarity of all possible object pairs in observation and forecast and allows for an objective a posteriori matching. In return, the MMI requires a careful parameter selection and tuning. The method can be applied to all objects within the entire German radar composite. Recently the approach was extended to handle so called “gridded objects”. Therefore, the radar composite is subdivided into overlapping, regular grid boxes and the MMI is calculated for all grid boxes separately. Various statistics are calculated for matching object pairs. Low thresholds for the matching reveal fundamental differences between forecasts and observation, e.g., in the total number of objects. Higher matching thresholds select only object pairs that show a certain similarity with respect to the chosen MMI-parameters. Statistics then show the systematic disagreement between observation and forecast of these per se well forecasted convective cells. Gridded objects allow us to regionally differentiate these systematic differences. A selected case study period of one month in early summer 2016 with strong convective activity on different spatial scales is analyzed. Now- casting is compared to two sets of COSMO-DE forecasts using the 1-moment and the 2-moment microphysics scheme, respectively.

Temperature and radiation biases in COSMO-REA6 and their role for representation of extreme events D. Niermann(1), T. Spangehl(1), and F. Kaspar(1) (1) Deutscher Wetterdienst

High-resolution regional reanalyses enable a variety of applications based on spatially and temporally homogeneous data. While there are numerous advantages over purely observation-based datasets with comparable resolution, the uncertainties and limitations for the different variables must be known. COSMO-REA6 is the current regional reanalysis of DWD. Here we focus on a combined evaluation of near-surface temperature and radiation.Different reference datasets are used. Near-surface temperature is compared to station data and station-based gridded datasets. Short-wave radiation at the surface is compared to station data and satellite-based products such as SARAH-2. Moreover, the outgoing long-wave radiation (OLR) is assessed based on CERES. First results show an overestimation of the near-surface temperature over continental areas during the summer half year. Whereas the effect is partly compensated by an underestimation of the incoming short-wave radiation at the surface, there is evidence that deficiencies in the representation of high clouds lead to reduced OLR. A further step is the evaluation of these effects towards the representation of extreme events in COSMO-REA6, which is one focus of the DWD contribution to the BMBF funded research consortium ClimXtreme. Using different definitions for heatwaves, the detected events in station-based datasets and the regional reanalysis COSMO-REA6 are compared. In particular, the minimum and maximum

63 temperatures are examined. Results show a similar behaviour for the minimum temperature as for the mean temperature in the summer months, while the maximum temperature bias does not show a strongly pronounced yearly cycle. The connection with the mentioned eﬀects of the radiation components will be investigated.

Multilayer cloud conditions in trade wind shallow cumulus – confronting two ICON model derivatives with airborne observations M. Jacob(1), P. Kollias(1,2), F. Ament(3), V. Schemann(1), and S. Crewell(1) (1) Institute for Geophysics and Meteorology, University of Cologne, (2) School of Marine and Atmospheric Sciences, Stony Brook University, (3) Institute for Meteorology, Universität Hamburg How are models and forward simulations able to reproduce cloud observations over the tropical ocean? This presentation is based on airborne remote sensing observations in the North Atlantic trades. The observations were taken with the HALO research aircraft upstream of Barbados during the NARVAL 1 field experiment in the dry season. Clouds were observed by an airborne nadir-pointing backscatter lidar, a cloud radar, and a microwave radiometer. NARVAL 1 was supported by large-domain simulations using the ICON model on different storm- and cloud-resolving resolutions. The simulations were produced within the HD(CP)2 and HErZ projects. The ICON output is forward simulated into the observational space of radar and lidar observables considering the model microphysics. We compare the vertical cloud boundaries to study the vertically resolved cloudiness, which is a major parameter modifying the moisture structure and determining the heating rate profile.The observations reveal two prominent modes of cumulus cloud top heights separating the clouds into two layers. The lower mode relates to boundary layer convection with tops closely above the lifted condensation level, which is at about 700 m above sea level (asl.). The upper mode is driven by precipitating shallow convection, also contains shallow outflow anvils, and is closely related to the trade inversion at about 2 to 3 km asl.. Differences between the lidar- and radar-detectable clouds in both layers with different liquid water path (LWP) were seen in the observations. Their representation in both realistically forced models differs depending on model configuration. The kilometer-scale model (ICON-SRM) with one-moment microphysics reproduces a lower cloud mode of lidar-visible clouds and an upper mode of radar- and lidar-visible clouds in principle, but the observed gap between the two layers and the relative occurrence frequency of higher and lower clouds is simulated differently than observed. The hectometer-scale model (ICON-LEM) with two-moment microphysics reproduces the bimodal distribution of cloudiness seen in the observations better. However, neither model seems to account for drizzle sized drops within the cloud droplets, that do not precipitate but generate a stronger radar signal even in scenes with low LWP.