Refine
Year of publication
Document Type
Language
- English (29)
Keywords
- waves in rotating fluids (6)
- Climate (5)
- COSMO-CLM (4)
- Regional Climate Model (4)
- boundary-layer structure (4)
- geophysical and geological flows (4)
- direct numerical simulations (3)
- Climate Change (2)
- Meteorology (2)
- Centrifugal (1)
Institute
Horizontal spatial schemes of third order and above used for discretization of COSMO (Consortium for Small Scale Modeling) Euler equations can be described as quasi-higher order schemes since interpolation of the advecting velocities and differencing of the pressure gradient term remain second order accurate. For NWP and Regional Climate modeling, upwind schemes of either third or fifth order have been recommended combined with an explicit numerical diffusion. We have implemented fully fourth order central difference horizontal schemes for the model’s Euler equations with two types of discretization of the advection terms: the first is a natural extension of the COSMO fourth order scheme by introducing fourth order interpolation
of the advecting velocity, and the second is a symmetric type discretization which is shown to conserve the rotational part of kinetic energy. We combine both advection schemes with fourth order discretization
of the pressure gradient term. To make the schemes completely fourth order, we consider all metric terms
resulting from coordinated transformations. Theoretical analysis of the new schemes compared to the model’s
existing third order upwind scheme exhibits: a slightly increased group velocity error due to a wider stencil,
a similar dispersive error, a significant reduction of the amplitude error, a significantly minimized aliasing error due to symmetric advection-discretization, and a significant increase in effective Courant number which
potentially allows longer time steps. Using 20-year climate simulations, we show that the new symmetric fourth order scheme is more stable than the extended COSMO fourth order scheme and third order upwind scheme, and that an explicit numerical diffusion can be avoided when using the symmetric scheme. We show that a 20% dispersive (phase) and diffusive (amplitude) errors limit result to the model’s effective resolution of approximately 5Δx for all 4th and 3rd order schemes. Considering the same error limit for simulated kinetic energy spectra show that the horizontal numerical diffusion is reducing the model’s effective resolution to
more than 10Δx and thus using the symmetric 4th order scheme without explicit horizontal diffusion increases
the effective resolution by a factor of two to approximately 5Δx. We further show that both implicit diffusion in upwind schemes and explicit numerical diffusion necessary for the current model’s stable runs have effects of equal magnitude on the model’s predicted climatologies. Climatologies show that fourth order schemes
enhance vertical turbulence mixing in the planetary boundary layer which reduces parameterized convection.
This consequently results to approximately 20% peak reduction of summer precipitation and an increase of
approximately 0.5 degrees Kelvin in summer 2m air temperature.
We present the prototype of a regional climate system model based on the COSMO-CLM regional climate model coupled with several model components, analyze the performance of the couplings and present a strategy to find an optimum configuration with respect to computational costs and time to solution.
The OASIS3-MCT coupler is used to couple COSMO-CLM with two land surface models (CLM and VEG3D), a regional ocean model for the Mediterranean Sea (NEMO-MED12), two ocean models for the North and Baltic Sea (NEMO-NORDIC and TRIMNP+CICE) and the atmospheric component of an earth system model (MPI-ESM). We present a unified OASIS3-MCT interface which handles all couplings in a similar way, minimizes the model source code modifications and describes the physics and numerics of the couplings. Furthermore, we discuss solutions for specific regional coupling problems like handling of different domains, multiple usage of MCT interpolation library and efficient exchange of 3D fields.
A series of real-case simulations over Europe has been conducted and the computational performance of the couplings has been analyzed. The usage of the LUCIA tool of the OASIS3-MCT coupler enabled separation of the direct costs of: coupling, load imbalance and additional computations. The resulting limits for time to solution and costs are shown and the potential of further improvement of the computational efficiency is summarized for each coupling.
It was found that the OASIS3-MCT coupler keeps the direct coupling costs of communication and horizontal interpolation small in comparison with the costs of the additional computations and load imbalance for all investigated couplings. For the first time this could be demonstrated for an exchange of approximately 450 2D fields per time step necessary for the atmosphere-atmosphere coupling between COSMO-CLM and MPI-ESM.
A procedure for finding an optimum configuration for each of the couplings was developed considering the time to solution and costs of the simulations. The optimum configurations are presented for sequential and concurrent coupling layouts. The procedure applied can be regarded as independent on the specific coupling layout and coupling details.
Fast growing atmospheric Rossby wave trains (RWT modes) are solutions of a Singular Vector (SV) analysis of a damped barotropic vorticity equation for northern winter basic flows. Using 40 DJF basic flows RWT modes are found over four regions of the globe only. Their propagation paths are remarkably constant for development times up to 96 h, but spread for longer times. The RWT mode with the largest 4 day growth rate develops over the North-Pacific (NPAC) region for each of the observed DJF basic flows considered. This mode is referred to as NPAC-mode. The eigenvalues, which are the 4 day kinetic energy growth factors of the NPAC mode, range from 10 to 24 for nearly all of the basic flows. In some exceptional years they are close to 30. The NPAC modes with moderate growth factors could be shown to be approximate solutions of the nonlinear model equation. Hereto a new formulation of the advection term provided by the energy-vortex theory was used. It appears that the constancy of propagation paths and the mode's shapes, the large growth factors and the validity of the linearization assumption up to development times of 96h make the NPAC mode a candidate to explain substantial parts of large scale interannual atmospheric variability in the North Pacific region.
The COSMO model in CLimate Model (COSMO-CLM or CCLM) is one of the most advanced Regional Climate Model with respect to its model dynamics, numeric, and physical parameterization options. It is widely used for climate research, climate mitigation and adaptation studies, as the model is specifically designed for simulation at a convective-permitting, meso-γ spatial resolution on the order of 1 km which:
• has a well-tested range of applicability, encompassing operational numerical weather prediction (COSMO),
regional climate modelling for retrospective and future projection (COSMO-CLM), idealised studies (ITC) and
the dispersion of trace gases and aerosol (ART) on weather forecast to interannual time scales,
• was successfully applied in several regions of the world, and especially widely within Europe
• has widely been used to downscale the results of global climate models and,
• is well documented.
The special issue “Recent developments in Regional Climate Modelling with COSMO-CLM” aims to present 7
model developments of high relevance for the climate mode, model evaluation of new model versions and results 8
of model application down to local scale. In the first part, published in May 2016 (http://www.schweizerbart.de/ 9
papers/metz/list/25#issue2) eight articles are presented. In this second part further key aspects of high resolution 10
regional climate modelling are addressed in five articles.
For the first time all radiation components TOA and at the surface of the regional climate model COSMO-CLM have been systematically compared with the CMSAF products aiming to analyse the potential of such a comparison for evaluation of the model and of the CMSAF data.
The group Environmental Meteorology at BTU Cottbus conducted a systematic comparison of CMSAF products for Europe with the model output of two regional climate model simulations with
COSMO-CLM on the model grid of 0.165◦. COSMO-CLM was forced with the analysis of the global model GME provided by the German Weather Service. The discussion of the significance of the deviations between model and data for all radiation components and integral cloud properties for selected regions confirmed the potential of the method.
First, different types of differences could be identified and hypotheses on possible origins could be formulated. Second, specific properties of CMSAF variables could be identified which should be improved in order to substantially increase the relevance of the CMSAF products for climate model evaluations.
This study is based
• on the accuracy statements of the SAF data and
• on the comparisons of absolute values and differences between model results and between
model results and SAF products for
– annual means,
– monthly means and
– annual cycles of spatial averages for 35 selected regions.
All differences exceeding the accuracy on space scales larger than 200 × 200 km2 are regarded as significant and the systematic analysis is restricted to the significant differences. The comparison of second moments and other statistical quantities like the spatio-temporal correlations remains for future work.
The analysis of the results aims to determine the consistency of the provided SAF products with the accuracy statements, to quantify the inevitable uncertainties, and to draw conclusions about the origin of the differences between model and data taking into account the assumed typical spatial structures of the differences originating in model deficiencies and/or Satellite data bias. It is assumed that
• a SAF data bias typically exhibits
– a North-South contrast,
– an annual cycle,
– a land-sea contrast,
– a day-night contrast and/or
– a cloud free to cloudy contrast in the differences, 115
• a typical model bias exhibits
– a land-sea contrast,
– a spatial structure defined by a typical local climate,
– a seasonal structure,
– a cloud free to cloudy contrast and/or
– a model domain boundary symmetry.
However, the land-sea contrast, the annual cycle, the daily cycle and/or the cloud free to cloudy contrast may have its origin in both, the SAF data bias and/or a model deficiency. As additional informations the model evaluation results obtained from comparisons with ground station observation products are taken into account. Furthermore, the inter comparison between the results for variables, which have a direct physical relation, like e.g. the albedo the incoming and the outgoing short-wave radiation at the surface is considered.
All discussions and analyses presented in this study are restricted to a previously selected set of parameters, regions, and time periods. The conclusions drawn in this report for a certain parameter in a specific region cannot generally be assigned to other parameters, regions and model configurations.
Additionally to the results presented in this report the annual cycles for the other regions are provided as a supplement.
In the following the main results of this first study are summarized.
Two model simulations have been conducted with different initial conditions for the soil variables temperature and soil moisture. The internal model variability due to this difference appeared to be of minor importance in comparison with the stated accuracy of the SAF products. It increases slightly
the total accuracy and will not be discussed in the follwing (see the discussion of each variable for details). One year data over Europe appeared to be sufficient for identification of different patterns of significant differences and to make suggestions on its possible origins like model deficiencies and/or SAF-data bias exceeding the stated accuracy.
In the following the main hypotheses derived from the results are listed.
• Boundary conditions inconsistent with cloud conditions in the model causing:
– Negative differences of CLCT in the boundary shown in 24.
– negative differences of ASWU_T shown in 8,
– positive differences of ASWG_S shown in 36,
– positive differences of ASOB_S shown in 49,
– negative differences of ATHB_S shown in 60.
• Overestimation of CLCT in the model over central to northern Europe of up to 0.3 in summer in Scandinavia causing:
– Positive differences of ASOU_T shown in 9 116
– Positive differences in ATHB_T shown in 17
– Negative differences of ASWG_S shown in 37
– Negative differences of ASWDIFU_S in summer shown in 45
– Negative differences of ASOB_S shown in 49
– Positive differences of ALWD_S in SCA shown in 53
• Overestimation of formation of clouds over land in the model or bias over land in CMSAF data causing
– negative differences of TQV over land shown in 32
– Neutral to positive differences in CLCT shown in 25
– CLCT and TQV show a weak negative correlation in winter (33 and 25 Jan.) and weak positive correlation in summer (33 and 25 Jul..)
• Overestimation of deep convection in the model over land in summer or overestimated NorthSouth gradient in CMSAF data causing
– positive differences of HTOP_CON in summer shown in 29 and 31,
• Underestimation of formation of clouds in the model over water or CMSAF data bias over water
surfaces
– negative CLCT shown in 24 and 26
– negative differences (downward positive!) and weaker land-sea contrast in SAF120, especially in winter in ToA LW net shown in 16 and 18
– negative differences in ALWD_S shown in 52 and 55 especially in late winter to early
spring.
– negative differences in ATHB_S over WAS and Italy shown in 60
– negative TOT_PREC shown in 75
• Overestimation of the model surface albedo in regions covered by snow causing:
– Positive albedo values over ALP and SCA in winter and spring shown in 41
– Positive ASWDIFU_S over ALP and SCA in late winter shown in 45
• Inconsistency of 2m temperature of the model and of the outgoing LW radiation in CMSAF
indicating a bias in the GME analysis used in the SAF product.
– negative differences in the 2m temperature shown in 68 and 69 all over the year and
– neutral to positive (in SUE and POE in spring and autumn) differences of ALWU_S shown
in 53
• Additional summer cold bias of GME008 configuration causing
117
– negative differences in central and northern Europe in summer in T_2M shown in 69
– negative differences in central and northern Europe in summer in ALWU_S shown in 57
– no differences in central and northern Europe in summer in ALWD_S shown in 53
– positive differences in central and northern Europe in summer in ATHB_S shown in 53
• Weaker variability on regional scale in CCLM in
– monthly and annual means of ALB_RAD in 40 and 41
– monthly and annual means of surface up SW in 44 and 45
• CMSAF data bias:
– higher north-south gradient in SIS than in ASOG_S in January shown in 37
– stronger land-sea contrast in CMSAF data CLC (24 and 25)
– strong land-sea contrast in SDL shown in 52
The results show that the availability of independent SAF products for complementary variables allows to identify different space-time patterns of differences and to draw up hypotheses on its possible
origins. However, the typical SAF product time scale of 1 or 2 years appears to be a limiting factor of
the analysis. In this sense the results confirm the potential of the CMSAF data for model evaluation
and the analysis of the quality of the CMSAF data. The availability of many variables can not replace
the limited length of the time serieses.
From physical point of view the cloud cover and the atmospheric content of water vapour, liquid and
ice water in dependence on the height appear as the primary variables and it is suggested to focus in
the future on these variables and on the consistency of the results obtained for these variables with the
results for the radiation components.
Assuming the accuracy stated for each SAF variable the provided SAF data are of satisfying quality
for state of the art climate model evaluation. The SAF product accuracy values are substantially
larger than the internal model variability and they are in many regions and seasons smaller than the
differences between model results and SAF data.
Assuming the accuracy stated as not certain, most of the difference patterns can be attributed to
SAF product bias or to a model deficiency. Additional investigations are necessary to find the right
answers. In this sense this report exhibits the potential of the method and invites to make additional
contributions.
Substantial progress can be expected from the analysis of the vertical structure, daily cycle and cloud
free/cloudy states focussing on the cloud properties and related variables. Furthermore, a specification
of the SAF accuracy limits in space and time and the extension of the time serieses are important for
clarification of the applicability of the SAF products for model evaluation.