4-9 September 2022, Bonn, Germany
Enter zoom meeting

UP1.7

SPARK session
Methodological Advances for the Accurate Representation of Physical Processes on Multiple Scales

Numerical modeling has become a cornerstone of atmospheric science. Applications cover all relevant scales, from microphysical processes (e.g., radiation, chemistry, cloud physics) to planetary-scale dynamics (weather and climate prediction). However, models are usually developed to represent a specific process on a specific scale, and processes on larger or smaller scales that may affect the problem of interest (e.g., due to scale interactions or scaling cascades) are simplified or neglected. Thus, a hierarchy of models and methods is often used to consider the effects of unresolved processes in a computationally efficient yet realistic way. Parameterization schemes often consider the mean effects of smaller, unresolved physical processes through effective descriptions and empirical correlations but with low fidelity and limited predictability. Larger scales are considered by boundary conditions that drive the development on the scale of interest. While increasing computational resources enable high-fidelity models to directly simulate complex physical processes on multiple scales, this is only possible for idealized configurations and a small fraction of the relevant parameter space. Thus, new and advanced methods are required to address the challenge of regime-overarching modeling and to bridge the gap between scales and processes. In this context, autonomous stochastic and machine-learning tools are increasingly explored, combining computational efficiency with validity across multiple scales and parameter regimes. Ranging from traditional statistical emulation to deep learning, data-driven approaches are expected to allow for qualitatively new atmospheric modeling and building digital twins of the atmosphere.

To provide a platform for the interdisciplinary exchange on accurate and economical representations of multi-scale physical processes in operational and research applications, the focus of this session is on novel modeling approaches in fluid dynamics, radiative transfer, cloud physics, atmospheric chemistry, and related disciplines, as well as their coupling in more complex frameworks, joining the UP and OSA sections of the meeting. We invite contributions on meteorological applications of fluctuation modeling, stochastic dynamics, data-driven and machine learning approaches, model coupling, nesting, grid adaptation, and regime-independent modeling, while welcoming promising approaches that have not yet been used in atmospheric science.

Co-organized by OSA1
Conveners: Fabian Hoffmann, Marten Klein, Franziska Glassmeier, Livia Freire
Orals
| Mon, 05 Sep, 14:00–15:25 (CEST)|Room HS 5-6
Posters
| Attendance Mon, 05 Sep, 16:00–17:30 (CEST) | Display Mon, 05 Sep, 08:00–18:00|b-IT poster area

Mon, 5 Sep, 14:00–15:30

Chairpersons: Fabian Hoffmann, Marten Klein, Franziska Glassmeier

14:00–14:20
|
EMS2022-103
|
Online presentation
Gustavo Abade et al.

Turbulent clouds are challenging to model and simulate due to uncertainties in microphysical processes occurring at unresolved subgrid scales (SGS). These processes include the transport of cloud particles, supersaturation fluctuations, turbulent mixing, and the resulting stochastic droplet activation and growth by condensation. In this work, we apply two different Lagrangian stochastic schemes to model SGS of cloud microphysics. Collision and coalescence of droplets are not considered. Cloud droplets and unactivated cloud condensation nuclei (CCN) are described by Lagrangian particles (superdroplets). The first microphysical scheme directly models the supersaturation fluctuations experienced by each Lagrangian superdroplet as it moves with the air flow. Supersaturation fluctuations are driven by turbulent fluctuations of the droplet vertical velocity through the adiabatic cooling/warming effect. A second more elaborate scheme uses both temperature and vapor mixing ratio as stochastic attributes attached to each superdroplet. It is based on the probability density function formalism that provides a consistent Eulerian-Lagrangian formulation of scalar transport in a turbulent flow. Both stochastic microphysical schemes are tested in a synthetic turbulent-like cloud flow that mimics a stratocumulus topped boundary layer. It is shown that SGS turbulence plays a key role in broadening the droplet-size distribution towards larger sizes. Also, the feedback on water vapor of stochastically activated droplets buffers the variations of the mean supersaturation driven the resolved transport. This extends the distance over which entrained CNN are activated inside the cloud layer and produces multimodal droplet-size distributions. Finally, our simulations suggest that stochastic Lagrangian SGS models may expand the ability of Large Eddy Simulations to represent cloud-top entrainment and associated microphysical details at cloud top.

How to cite: Abade, G., Waclawczyk, M., Grabowski, W., and Pawlowska, H.: On Lagrangian stochastic condensation models in turbulent cloud simulations, EMS Annual Meeting 2022, Bonn, Germany, 5–9 Sep 2022, EMS2022-103, https://doi.org/10.5194/ems2022-103, 2022.

14:20–14:30
Questions & Discussion on the presentation: "On Lagrangian stochastic condensation models in turbulent cloud simulations"

14:30–14:50
|
EMS2022-165
|
Onsite presentation
Robin Stoffer et al.

Large-eddy simulation (LES) is an often used technique to simulate atmospheric boundary layers. In LES, the effects of the unresolved turbulence scales on the resolved scales have to be parameterized with subgrid scale (SGS) models. These SGS models usually require strong assumptions about the relationship between the resolved flow fields and the terms correcting for the unresolved physics, which are often violated in reality and potentially hamper their accuracy. Also, they tend to interact with the discretization errors introduced by the popular LES approach where a staggered finite-volume grid acts as an implicit filter.

We therefore developed an alternative LES SGS model based on artificial neural networks (ANNs) that allows to compensate for both the unresolved physics and instantaneous spatial discretization errors associated with a staggered finite-volume grid, without the need for strong assumptions. To this end, we used a test case of turbulent channel flow (with friction Reynolds number Reτ=590) simulated with the computational fluid dynamics code MicroHH (v2.0). We trained the ANNs based on instantaneous flow fields from a direct numerical simulation (DNS) of the selected channel flow. By applying an explicit spatial filtering procedure on the high-resolution DNS fields, we generated millions of samples to train the ANNs in a supervised manner.

In general, we found that the ANNs were well able to predict the correct values for flow fields not seen during training. In addition, our ANN SGS model was able to generalize towards multiple coarse horizontal resolutions, in particular when these resolutions were located within the range of the training data. This shows that ANNs have potential to construct highly accurate, generalizable SGS models. Several open challenges do remain though before this potential can be successfully leveraged in actual LES applications: we observed an artificial build-up of turbulence kinetic energy when we directly incorporated our ANN SGS model into a LES simulation of the selected channel flow, eventually resulting in numeric instability. We hypothesize that error accumulation and aliasing errors were both important contributors to the observed instability.

How to cite: Stoffer, R., van Leeuwen, C., Podareanu, D., Codreanu, V., Veerman, M., Janssens, M., Hartogensis, O., and van Heerwaarden, C.: Large-Eddy Simulation Subgrid Scale Modelling using Artificial Neural Networks, EMS Annual Meeting 2022, Bonn, Germany, 5–9 Sep 2022, EMS2022-165, https://doi.org/10.5194/ems2022-165, 2022.

14:50–15:00
Questions & Discussion on the presentation: "Large-Eddy Simulation Subgrid Scale Modelling using Artificial Neural Networks"

15:00–15:05
Poster Pitch (3min) & questions (2min): EMS2022-471 - "The Present and Future of Lagrangian Cloud Modeling: From the Centimeter to Kilometer Scale"

15:05–15:10
Poster Pitch (3min) & questions (2min): EMS2022-224 - "A stochastic model of mixed-phase cloud micro-physics"

15:10–15:15
Poster Pitch (3min) & questions (2min): EMS2022-239 - "Quantifying the role of uncertainty in microphysical processes for cloud and precipitation formation in an extratropical cyclone "

15:15–15:20
Poster Pitch (3min) & questions (2min): EMS2022-610 - "Parametrising nearest neighbor interaction in a convection scheme: an idealized squall line test case."

15:20–15:25
Poster Pitch (3min) & questions (2min): EMS2022-701 - " Gaussian-process emulation for integrating data-driven aerosol-cloud physics from simulation, satellite, and ground-based data "

Posters

P31
|
EMS2022-471
|
Onsite presentation
Jung Sub Lim et al.

Clouds are one of the most complex systems in the atmosphere, as the processes that constitute a cloud are of inherent multi-scale nature. Among the many challenges in cloud physics, representing the effects of turbulence on cloud microphysics is one of the most intriguing, with implications for precipitation development, cloud optical properties, and hence the role of clouds in the climate system. Lagrangian cloud models (LCMs) are a new approach for simulating cloud microphysics, which has been developed over the last decade. LCMs overcome the limitations of many previous cloud microphysical models, and enable new insights on important cloud microphysical questions, such as aerosol activation and regeneration, the role of giant aerosol particles in precipitation formation, and the effects of turbulent supersaturation fluctuations. In this presentation, we will introduce a new LCM approach for representing the effects of unresolved turbulence on cloud microphysics, focusing on the entrainment and mixing process. As the spatial scales of entrainment and mixing are usually smaller than the resolution of most dynamical models used for the simulation of clouds, unresolved mixing tends to be too fast, and the entire process may be misrepresented. Our new approach represents these unresolved scales by tracking air parcels, allowing for the explicit representation of entrainment and mixing on scales as small as a few centimeters, while entire clouds on scales as large as a few kilometers are covered simultaneously. Thus, we are able to represent different mixing scenarios such as homogeneous and inhomogeneous mixing naturally, applying only a minimum of assumptions. By highlighting how the mixing type changes during the lifecycle of a cloud, we outline how this scale-interactive approach can play a pivotal role in solving future problems in cloud physics.

How to cite: Lim, J. S., Kainz, J., and Hoffmann, F.: The Present and Future of Lagrangian Cloud Modeling: From the Centimeter to Kilometer Scale, EMS Annual Meeting 2022, Bonn, Germany, 5–9 Sep 2022, EMS2022-471, https://doi.org/10.5194/ems2022-471, 2022.

P32
|
EMS2022-224
|
CC
|
Onsite presentation
Daniel Gomes Albuquerque and Gustavo Coelho Abade

Mixed-phase clouds, i.e., clouds that contain both super-cooled water droplets and ice crystals, are ubiquitous in the atmosphere and play an important role in the climate system. The mixture of liquid and solid water in sub-zero temperatures leads to a condensational instability, in which ice particles tend to grow at the expense of droplet  evaporation. Nonetheless, mixed-phase clouds are unexpectedly long-lived. Earlier mean-field stochastic models are based on the picture of turbulence-induced large-scale dynamical forcing of cloud parcels to explain the longevity of mixed-phase clouds. We claim that small-scale turbulence is key to explain the persistence of such systems. Due to limited computational resources, weather simulation on a global scale is limited to coarse grids with a resolution of kilometers at best. On the other hand, a typical  turbulent flow inside a cloud will display an intricate structure of eddies down to the scale of millimeters. A recent study using the linear eddy model showed that small  scale turbulence does play a role in slowing down cloud glaciation. We propose a more computationally tractable Lagrangian stochastic micro-physical scheme to account for sub-grid fluctuations in velocity, temperature and water vapor fields. The impact of our scheme on phase partitioning is tested in a synthetic, turbulent-like flow that mimics an Arctic mixed-phase stratocumulus (AMPS) cloud. Results are confronted with idealized reference simulations that use Eulerian bulk micro-physics based on an assumed (temperature-dependent) phase partitioning function. Our study suggests that accounting for local variability in a turbulent cloud is important for reproducing steady-state mixed-phase conditions.

How to cite: Gomes Albuquerque, D. and Coelho Abade, G.: A stochastic model of mixed-phase cloud micro-physics, EMS Annual Meeting 2022, Bonn, Germany, 5–9 Sep 2022, EMS2022-224, https://doi.org/10.5194/ems2022-224, 2022.

P33
|
EMS2022-239
|
Onsite presentation
Annika Oertel et al.

The characteristic large-scale cloud band near extratropical cyclones is associated with the so-called warm conveyor belt (WCB), which is a coherent airstream that ascends cross-isentropically from the boundary layer into the upper troposphere within two days. This transport of air into the upper troposphere can influence the large-scale flow evolution. The cross-isentropic ascent is influenced 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 extratropical circulation. The parameterization of microphysical processes in numerical weather prediction models introduces uncertainties related to cloud characteristics, which can also feed back on the larger-scale flow. In particular, ice formation and the phase partitioning are often poorly represented in numerical weather prediction models.
We analyze the role of uncertainty related to the representation of microphysical processes in a two-moment microphysics scheme for the detailed WCB ascent, the cloud characteristics, as well as changes in diabatic heating contributions from the individual parameterized processes. Furthermore, the propagation of uncertainty from the microphysical processes to the larger-scale flow is investigated. Systematic sensitivity experiments for a two-way nested convection-permitting simulation with the Icosahedral Nonhydrostatic (ICON) modeling framework are performed for an extratropical cyclone case study in the North Atlantic. The experiments include systematic perturbations to environmental conditions relevant for cloud formation (concentrations of cloud condensation nuclei and ice nucleation particles as well as sea surface temperature) and microphysical parameters (capacitance of ice and snow as well as maximum supersaturation in the saturation adjustment scheme). To quantify the effect of individual microphysical process rates for WCB ascent, we aggregate heating rates from each parameterized microphysical process along online trajectories.
First results indicate that the perturbations not only substantially modify the cloud properties but also influence the WCB ascent behavior and ascent-integrated diabatic heating contributions. The perturbed parameter ensemble further shows a growth of spread of the larger-scale flow with increasing lead time. To disentangle and quantify the contributions of the five perturbed parameters, we perform a variance-based sensitivity analysis using computationally inexpensive statistical surrogate models based on Gaussian process emulation that are developed from the large, computationally expensive perturbed parameter ensemble. For example, this indicates that WCB-related surface precipitation is most strongly influenced by changes to sea surface temperature and cloud condensation nuclei concentration, while the vertical distributions of snow and ice are also substantially influenced by perturbations to their capacitance. This contribution shows how the parameter perturbations affect microphysical processes which subsequently modify characteristics of the large-scale WCB cloud band, and illustrates how statistical emulation can be used to quantify uncertainty from parameter perturbations in a real-world case study.

How to cite: Oertel, A., Miltenberger, A. K., Grams, C. M., and Hoose, C.: Quantifying the role of uncertainty in microphysical processes for cloud and precipitation formation in an extratropical cyclone, EMS Annual Meeting 2022, Bonn, Germany, 5–9 Sep 2022, EMS2022-239, https://doi.org/10.5194/ems2022-239, 2022.

P34
|
EMS2022-610
|
Onsite presentation
Tobias Goecke

 Most convection schemes in operational models for numerical weather prediction
 operate in a single column that is connected to other columns only via
 the large scale dynamics. In particular the triggering of convection is usually
 independent of the convective activity in the surrounding of the grid cell.
 This is not realistic since convective activity is able to trigger
 convection nearby [1,2].
 Here, the intention is to work toward an implementation of nearest neighbor
 interaction (NNI) within the framework of a parametrization of deep convection 
 within the global ICON model [3].
 The main research question is whether a convection scheme equipped with NNI
 is able to interact with the large scale dynamics in a constructive way.
 In particular the focus is laid on an idealized squall line case in which the 
 standard convection scheme fails. It is shown how a propagating squall line can be recovered
 either in the large scale dynamics, the pure NNI or the coupled dynamic.

Observations of tropical convection suggest continous phase

transition between precipitating and non-precipitating phases [4] similar

to what is observed in simple two-dimensional lattice models.

Thus, using simple two-dimensional presciptions to descibe horizontal correlations

in convection schemes might be suitable to introduce a scale-consistent behavior.

Percolation  models are good canditates to describe spatio-temporal

correlations, critical scaling or cloud size distributions [5].

This motivates the current work, which  investigates, how a given NNI might

interact with the dynamic of the host model and to establisch a proof of principle

that such a mechanism might be beneficial.

 

[1]: Tompkins (2001), JAS 58 (13), 1650.

[2]: Seifert & Heus (2013), ACP 13 (11), 5631.

[3]: Zängl et.al. (2015), QJRMS 141 (687), 563.

[4]: Peters & Neelin (2006), Nature Physics 2 (6), 393.

[5]: Windmiller (2017), PhD Thesis, LMU Munich, https://edoc.ub.uni-muenchen.de/21245/

How to cite: Goecke, T.: Parametrising nearest neighbor interaction in a convection scheme: an idealized squall line test case., EMS Annual Meeting 2022, Bonn, Germany, 5–9 Sep 2022, EMS2022-610, https://doi.org/10.5194/ems2022-610, 2022.

P35
|
EMS2022-701
|
Presentation form not yet defined
Franziska Glassmeier et al.

Data-driven quantification and parameterization of cloud physics in general, and of aerosol-cloud interactions in particular, rely on input data from observations or detailed simulations. These data sources have complementary limitations in terms of their spatial and temporal coverage and resolution; simulation data has the advantage of readily providing causality but cannot represent the full process complexity. In order to base data-driven approaches on comprehensive information, we therefore need ways to integrate different data sources. 

We discuss how the classical statistical technique of Gaussian-process emulation can be combined with specifically initialized ensembles of detailed cloud simulations (large-eddy simulations, LES) to provide a framework for evaluating data-driven descriptions of cloud characteristics and processes across different data sources. We specifically illustrate this approach for integrating LES and satellite data of aerosol-cloud interactions in subtropical stratocumulus cloud decks. We furthermore explore the extension of our framework to ground-based observations of Arctic mixed-phase clouds.

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

References:

  • Glassmeier, F., F. Hoffmann, J. S. Johnson, T. Yamaguchi, K. S. Carslaw and G. Feingold (2019): “An emulator approach to stratocumulus susceptibility”, Atmos. Chem. Phys., 19, 10191- 10203, doi: 10.5194/acp-19-10191-2019
  • Hoffmann, F., F. Glassmeier, T. Yamaguchi and G. Feingold (2020): “Liquid water path steady states in stratocumulus: insights from process-level emulation and mixed-layer theory”, J. Atmos. Sci., 77, 2203-2215, doi: 10.1175/JAS-D-19-0241.1
  • Glassmeier, F., F. Hoffmann, J.S.  Johnson, T. Yamaguchi, K. S. Carslaw, and G. Feingold (2021): “Aerosol-cloud climate cooling overestimated by ship-track data”, Science 371, 485–489, doi: 10.1126/science.abd3980

How to cite: Glassmeier, F., Hoffmann, F., Feingold, G., Gryspeerdt, E., van Hooft, A., Yamaguchi, T., Johnson, J. S., and Carslaw, K. S.: Gaussian-process emulation for integrating data-driven aerosol-cloud physics from simulation, satellite, and ground-based data, EMS Annual Meeting 2022, Bonn, Germany, 5–9 Sep 2022, EMS2022-701, https://doi.org/10.5194/ems2022-701, 2022.

Supporters & sponsors