Boundary Layer Clouds

The project S2 aims at exploiting the unique research opportunities created during Phase I of HD(CP)² to improve the parameterization of boundary-layer physics, radiation and low level cloud microphysics in climate models and to investigate the response of clouds to climate change. The aim is to gain more insight into the inner workings of the system of interacting fast physics, and its response when subject to an idealized external forcing of CO2 quadrupling.

To this purpose the ICON model system will be employed at a range of resolutions and domain sizes, from local cloud-resolving to global "climate-permitting". High-resolution runs of ICON over different geographical areas and multiple months will be confronted with operational supersite observations as well as advanced satellite products in order to assess for parameterization development and climate sensitivity studies.

A first sensitivity simulation with the HD(CP)² model at LES resolution with perturbed CO2 concentrations has been prepared and is currently run at the Forschungszentrum Jülich to assess fast cloud responses to CO2 increase. This is done in work package 6 of this project.

The figure shows a comparison (from Kühne et al. 2016) of the change in potential temperature (top), relative humidity (second row from top), specific humidity (third) and cloud fraction (bottom) in response to a doubling of CO2. This is an analysis over Central Europe for one year of simulation. The CO2 change is imposed globally for the results in the left column, and only over Europe for the right column. The similarity between the two results suggests that a simulation with the limited-area HD(CP)² model is meaningful.

Generally, aircraft (Opens external link in new windowHALO) observations of persistent, moist layers (far above the boundary layer) detected during the Opens external link in new windowNARVAL campaigns are used as a proxy in S2. To better understand the impact of these layers on the atmospheric boundary layer, case studies are first performed with LES and then assessed.

Confronting ICON-LEM with supersite observations

Work package 1 is exploiting ground based observations from supersites to evaluate high resolution ICON columns output. Boundary layer classification for model data is being developed, following the structure of the one used for the observations. Boundary layer height is determined in the model output by means of the Richardson number following Opens external link in new windowSeidel et al. (Journal of Geophysical Research, 2012). Statistical comparisons will be performed on a statistical ensemble of simulated days (HOPE period). Also, qualitative case-study based comparison with the observations is being developed with the aim of comparing cloud characteristics for evaluating various parametrizations for microphysics.

The composite shows different panels: panel a) shows the variance of the vertical velocity calculated over an interval of 30 minutes adopting a running mean over 5 minutes step. The white line represents the height of the boundary layer calculated according to Seidel et al. (2012). Panel b) shows the derived boundary layer product, based on a synergy of variables, identifying convective, non turbulent planetary boundary layer (PBL) and cloud presence. Further improvements to introduce turbulent PBL induced by wind shear and decaying PBL are currently under test. Panel c) shows a comparison of the liquid water path (LWP) time series on the same time resolution and the bias between them.

Cloud Top Height (CTH) retrieval of different cloud types according to CloudSat cloud classification* for CloudSat overpasses over Germany for the year 2013. Box-Whisker plots: the mid point corresponds to the median, the middle box represents the middle 50% of the data; stars stand for the extreme values. Note: The CloudSat classification is based on height, temperature, precipitation, and radar reflectivity.

Satellite and Ground-Based Synergy and Radiative Impacts for ICON-LEM Evaluation

The aim of the work package 2 is to assess the capability of the ICON-LEM to realistically simulate different cloud types both in terms of cloud properties and occurrence frequency and to accurately represent the resulting cloud radiative effects. As an initial step, long term cloud observations from the A-Train constellation (i.e., CloudSat, CALIOP) and supersites (i.e., Cologne, Lindenberg, Leipzig) over Germany will be employed in order to perform an observational-based cloud climatology study and radiative transfer models (RRTMG and SPARTA) will be utilized to quantify the resulting cloud radiative effects. In the next step, the combined observations of cloud and radiation fields will be compared in a statistical manner with ICON-LEM simulations. Deviations between observations and simulations will be interpreted in order to identify potential shortcomings in the cloud representation in the ICON-LEM. Ultimately, the observed and simulated cloud fields will be used as an input for a radiative transfer model and a radiation closure study will be performed.

Evaluation of the precipitation-evaporation parameterization from supersite observations

Within work package 3, the evaporation of precipitation will be analysed in detail, using temperature and humidity profiles from supersites. The availability of passive and active remote sensing data for a longer time period will be used to find systematic features which should lead in improved parameterizations for evaporation in models. Especially cloud radar, ceilometer as well as the synergistic profiles from microwave radiometer, cloud radar, ceilometer and Raman lidar will be used for this work. So far, the cooling due to evaporating precipitation is parameterized in models and can only be estimated from standard observations. With this method, it is intend to derive these cooling rates directly from remote sensing observations which will allow developing improved parameterizations.

Temperature profile of lower troposphere (top) and heating rates (bottom) on 2 July 2016 at LACROS supersite in Leipzig, derived from microwave radiometer observations. Rainy periods are marked with white hatching. Largest cooling in lowest 500 meters is associated with precipitation.

The first steps in the course of this work package are to gather as many data as possible and check them for consistency of the used retrieval methods. In the next step we will estimate the heating rates and set them into relation with precipitation observations from radar and ground.

PDF Cloud Schemes for ICON and their interaction with radiation

The overall aim of work package 4 is to improve cloud representation and cloud radiative interactions in the ICON Global Climate Model by implementing and developing a so called probability density function (PDF) schemes. The core idea of PDF schemes (also sometimes referred to as statistical schemes) is to represent the sub-grid scale variability of one or multiple variables through an analytical PDF for each model cell. Our approach is to use a beta function to describe the total water in each cell. From this beta function we can deduce various parameters such as cloud fraction.

One of two main tasks is to implement and further develop the beta function PDF scheme in the ICON Global Climate Model and to test it using various simplified and realistic test cases. The second main task is to use the many available LES results for reference, inspiration, and evaluation. For this we use both small ­domain prototype LES simulations and the big­ domain ICON simulations. The prototype simulations serve to study the basic total water PDF evolution in the boundary layer while the big­ domain ICON simulations make it possible to study the link between total water variability and clouds and their radiative impact for a wide range of realistic conditions.

Left: Sketch of a beta function describing the total water distribution in a cell. The cell mean (blue) and saturation (red) values are included. The cloud fraction of the cell can be derived from the red
area showing the area of the PDF above saturation. Right: Histogram of the total water for a single vertical level of a 12.5x12.5 km DALES LES simulation. The blue histogram shows all cells while the red histogram represents only cells which contain liquid water.

Parameterizing 3D Heating and Cooling Rates

Within work package 5 we aim at systematically characterizing the effect of radiation on boundary layer cloud development on the ICON-LEM resolution, and at studying the relevance for the fast cloud feedback using ICON-GCM. In Phase I of HD(CP)², accurate 3D radiativ etransfer parameterizations of these effects have been developed (Opens external link in new windowJakub and Mayer, 2015a;Opens external link in new window Klinger and Mayer, 2016) which are fast enough to be included into LEM runs, at least for limited time periods.
Next to the systematical characterization of the radiative effects on boundary layer clouds in model studies, we will compare the model results to observations (in collaboration with work pachage 1 and 2). This allows not only testing if there is an effect, but also if this improves the accuracy of the LEM simulations. While radiative effects are fully resolved at the ICON-LEM scale, a parameterization of subgrid-scale cloudiness is required for ICON-GCM: In particular shading of the ground as well as radiative heating and cooling depend on cloud fraction, and in particular on the non-resolved area of the cloud sides. In work package 5, we will firstly adapt our newly developed 3D radiative transfer parameterizations for the use in ICON-LEM. Further steps require the systematic characterization of the radiative effects and the development of a GCM parameterization, considering systematic sub-grid scale radiative effects.

The figure shows a shallow cumulus cloud field simulation (liquid water mixing ratio) above an ocean surface simulated with 1D and 3D thermal radiation and with a constant cooling approach. We used the UCLA-LES at 100m horizontal resolution. The figure shows the different development of the cloud field in terms of organization (Klinger et al., 2016, ACPD).

Fast response I: Perturbed C02 Simulations

Work package 6.1 employs the ICON model system at a range of resolutions and domain sizes, from local cloud-resolving (ICON-LEM) to global "climate-permitting" (ICON-GCM). High-resolution runs of ICON over different geographical areas and multiple months will be confronted with operational supersite observations as well as advanced satellite products in order to assess for parameterization development and climate sensitivity studies.
A first sensitivity simulation with the HD(CP)² ICON-LEM with perturbed CO2 concentrations has been prepared and is currently run at the Forschungszentrum Jülich to assess fast cloud responses to CO2 increase.

The figure shows a comparison (from Nam et al., in prep.) of the change in effective radiative forcing in response to a quadrupling of CO2 using for two experiments in the ICON-GCM configuration. The similarity between the experiment with a locally imposed CO2 change over Central Europe to the changes over Central Europe from a globally imposed CO2 change suggest that the limited-area HD(CP)² simulation with ICON-LEM will be meaningful and representative. In fact, preliminary results of the perturbed ICON-LEM are imposed onto the plot and it is within the distribution found using the ICON-GCM. The ICON-GCM simulations will help identify the temporal & spatial scales which rapid adjustments statistically significant and guide the study of rapid adjustments with the ICON-LEM.

Fast response II: LEX Proxy Cases

The main aim of work package 6.2 is to determine the effects elevated moisture layers have on low level boundary layer clouds. Present day cases of observed elevated moisture layers are being used as proxy cases to represent future atmospheric conditions resulting from climate change. These cases are being developed using data recorded as part of the NARVAL I South Campaign in 2013. Currently the focus is on data recorded during dropsonde launches which were part of Research Flight 4 (RF04), which took place on December 14th 2013. Initial work included running a control large eddy simulation (control LES) for the time and the location at which the first dropsonde was launched. This LES was forced using ECMWF analysis data. The dropsonde observations were then interpolated into the ECMWF input file and a second simulation was run (nudged LES). Current work has been focused on assessing the impact of incorporating dropsonde data into the simulation.

The figure shows profiles of potential temperature and water vapour mixing ratio for the Control LES (dark blue), the LES incorporating the dropsonde data (light blue), the dropsonde observations (green), and the ECMWF analysis data (dashed blue). The LES simulations incorporating the dropsonde data better matches the observed thermodynamic state of the boundary layer. Current work focuses on the representation of other variables including clouds. To this purpose HALO observations will be used.

The next step will be to use this case as a benchmark for phase-space analyses of the impact of elevated moisture layers on boundary layer clouds.

Stochastic closure for turbulence under the influence of larger scale motion

Work package 7 focuses on data-driven investigations of interactions between turbulence, small-scale non-turbulent motions and boundary-layer clouds. At nighttime, boundary-layer turbulence is highly intermittent and poorly parameterized. The intermittency of turbulence is related partly to local forcing by small-scale wind accelerations that are typically not resolved in numerical models, but also on larger time scales to cloud cover and geostrophic wind.
Data-driven analyses of nocturnal boundary-layer states reveal the existence of several regimes with different dynamical characteristics, and show that regime occupancy statistics are strongly related to cloud cover and geostrophic wind. One of the tasks of this work package is to quantify the influence that the external forcing variables exert on non-stationary nocturnal turbulence by explicitly including them in a data clustering methodology. A subsequent model development will be based on the construction of a transition matrix between different turbulence regimes using data clustering from the observational supersites. The statistics of boundary-layer regimes will be compared with statistics obtained by ICON-LEM simulations.

Different regimes of nighttime flow features. Top: Scatterplot of the mean wind speed between 90 and 2 m, the wind speed difference between 90 and 2 m, and the potential temperature difference between 90 and 2m for one year of nighttime data from the RAO site. The colours show data points in two different regimes of interaction between non-turbulent motions and turbulence.
Bottom: Joined variability of different scales of horizontal wind velocity (u) and different scales of vertical velocity variance (w2). Left: case with stronger wind, low temperature stratification. Right: example with weak wind, strong temperature stratification.

We aim at developing a stochastic closure for use in ICON-LEM in stably stratified and nighttime conditions. On the other hand, intermittent bursts of turbulence can be critical to the formation of boundary-layer clouds. The data-driven clustering methodologies will also be applied to observational data to quantify the influence of isolated turbulent bursts on cloud formation. The stochastic closure would enable the representation of intermittent turbulence and could thus help representing cloud formation more accurately.