Earth observation satellites rely on sunlight reflected by the Earth to retrieve information about atmospheric constituents and clouds. However, radiative transfer models used in satellite algorithms often neglect shadow effects, introducing potential biases. This thesis investigates the impact of both lunar (solar eclipse) and cloud shadows on satellite-based atmospheric measurements and proposes correction strategies to improve data accuracy.
We develop a method to correct for lunar shadow effects during solar eclipses by calculating wavelength-dependent solar obscuration, including solar limb darkening. Applying this method to TROPOMI/S5P data, we correct ultraviolet Aerosol Absorbing Index (AAI) anomalies during the annular eclipses of December 2019 and June 2020, restoring geophysical features like sunglint and desert dust. No evidence of eclipse-induced aerosol changes is found. Similarly, we apply the correction to cloud retrievals, revealing that shallow cumulus clouds begin to dissipate at solar obscurations as low as ~15%, a phenomenon supported by large-eddy simulations. Ignoring this effect may lead to overestimates of eclipse-induced radiative forcing.
We also introduce DARCLOS, the first cloud shadow detection algorithm for a spaceborne spectrometer. Using geometric and spectral criteria, DARCLOS identifies potential and actual cloud shadow pixels in TROPOMI data. We validate this method using VIIRS imagery, though cloud evolution limits validation accuracy. Applying DARCLOS to eight months of TROPOMI AAI data over Europe, we quantify cloud shadow signatures and simulate them with our new 3D radiative transfer model, MONKI. Despite cloud shadows significantly altering the UV reflectance ratio, the AAI computation inherently compensates for most effects. Remaining second-order biases are small and dependent on cloud properties, making a full correction challenging and possibly unnecessary.
This work improves our understanding of shadow impacts on satellite retrievals and provides tools to mitigate or study their influence, enhancing the reliability of atmospheric observations from space.

Atmospheric shallow convection (SC) is characterised by cumulus clouds that extend between a few hundred meters and a few kilometres. These clouds are more than passive tracers of the air flow in which they are embedded. They play an active role in atmospheric dynamics and the climate system by redistributing moisture, temperature, and momentum in the troposphere.
We focus on two recent field campaigns. 1. EUREC4A took place between January and February 2020 over the North Atlantic trade wind region, which provides perfect conditions for shallow cumuli. This campaign measured clouds and the wind flows on mesoscales. 2. CMTRACE took place in September 2021 around Cabauw, The Netherlands. During CMTRACE, a collocated wind lidar and cloud radar allowed, for the first time, to measure winds from the surface up to 2 km, with a frequency of about 1 minute (∼500 m).
Our results show that CMT crucially influences wind speed and direction throughout the lower troposphere, it also controls cloud organisation and mesoscale circulations. Observations reveal an excessively pronounced diurnal cycle of the winds in the IFS, with too strong winds at night and too weak winds during the day. The role of shallow convection parameterisation (SCP) in kilometre-scale models is explored using the HARMONIE-AROME model. Experiments with inactive SCP produce stronger mesoscale circulations, altering cloud cover and wind profiles.
Finally, the thesis examines novel momentum flux observations from the CMTRACE campaign. These data confirm errors in IFS momentum flux parameterisation, particularly under active convection.
The findings outlined in this thesis hold important implications for advancing our understanding of CMT and its role in atmospheric dynamics

How clouds change due to global warming is a major source of uncertainty in global climate models. A large part of this uncertainty originates from the feedback of low clouds over the subtropical ocean. The processes that shape these clouds range from microphysical processes of condensation and evaporation to the large-scale circulations spanning thousands of kilometers. While different models can resolve the large scale circulations and the small scale of individual clouds increasingly well, there is a gap in understanding the scale in between: the mesoscale with its fascinating patterns. They exhibit a large continuous variability that affects not only their visual appearance, but also the amount of sunlight they reflect.
This thesis aims to analyze the temporal evolution of the mesoscale cloudiness by following the clouds on their path over the Atlantic along the trade winds and characterize their patterns by cloud metrics, cloud cover and mean cloud object length, computed on geostationary satellite images. Both decrease as the cloud fields move over warmer waters. The most prominent feature is the diurnal cycle in the cloud metrics that can at least partially be explained by oscillations in the atmospheric stability and the surface wind speed of the large-scale environment. Following the line of evidence found from these cloud controlling factors, we conclude that the decrease in low cloud amount and size due to increase in sea-surface temperature will most likely overcompensate the rise in cloud object size due to increase of stability. Especially stratocumulus clouds might become much less and also smaller. However, there remains a large, seemingly random variability in the cloud metrics that has little mean contribution to the climatology. As we cannot expect this to be the case in a warmer climate, we still need to understand what causes this variability. Therefore, the trade wind cloudiness needs to be further investigated with higher temporal resolution of the cloud controlling factors, more trajectories to bin for a fixed large-scale and better understanding of the interplay of governing processes both from models and observations.

Large Eddy Simulation (LES) is a mathematical technique for performing simulations of turbulent flows, such as those found in the Earth’s atmosphere. Compared to traditional numerical weather and climate models, LES is more accurate in representing turbulent processes and cloud dynamics. The computational burden of LES, however, have histor- ically limited its application to relatively small domain sizes. In this work, part of the DALES atmospheric LES model was ported to Graphics Processing Units (GPUs) using the OpenACC programming model. GPUs, originally designed for accelerating computations related to 3D computer graphics, excel at parallel computations, which are abundant in LES models. The performance of the GPU port of DALES was measured on an NVIDIA RTX 3090 in a desktop workstation and an NVIDIA A100 in the Snellius supercomputer and compared to the existing CPU implementation. For the BOMEX intercomparison case, a speedup of 11.6 was achieved versus 8 CPU cores on the desktop system, while on Snellius a speedup of 3.9 was observed compared to 128 CPU cores. Furthermore, the existing MPI parallelization of DALES was adapted such that multiple GPUs can be used simultaneously. This thesis represents a step towards the enhancement of the scalability of DALES, enabling simulations on larger domains at higher resolutions. While a substantial acceleration of DALES was achieved, further efforts are needed to port more components of the model to the GPU to facilitate the simulation of increasingly realistic meteorological phenomena.

Shallow cumulus clouds interact with their environment and redistribute heat, moisture, and momentum (wind speed and direction) in the atmosphere. The same convective plume that forms the cloud is for a large extent responsible for this transport. Modeling the effect of shallow cumulus clouds is challenging because these clouds are smaller than can be directly simulated by models. Weather and climate models therefore rely on empirical functions to represent the effect of sub-grid processes such as the turbulent and convective transport. The transport of moisture and heat through cumulus convection has received a lot of attention from the atmospheric science community. Momentum transport has been studied far less, even though a few studies indicate that momentum transport may have a large influence on local weather as well as the large-scale circulation.

Thanks to increasing computational resources, the grid sizes of global climate and weather prediction models are decreasing to scales at which subgrid processes, especially the evolution of clouds, can become spatially too variable for traditional deterministic parameterization approaches. We propose a stochastic parameterization in which clouds are represented by a Markov-Chain (MC) lattice model. The state of each lattice site is characterized by cloud top height (CTH) and cloud optical depth (COD). Based on these two parameters, clouds can be categorized into different cloud types (cumulus, stratus, cirrus, deep convection, etc.). In contrast to previous work, where the transition probabilities between different cloud types were calculated, we determine the transition probability densities in the continuous CTH-COD domain. We moreover explore the influence of neighbouring lattice sites on these transition densities to research whether it is possible to simulate cloud organization structures, without the use of extra external large scale variables or forcings. We estimate the parameters of the distributions using GOES-16 satellite images. Using these estimates, we perform simulations of the continuous state space Markov Chain over the lattice. This results in 2D spatial cloud evolutions, which spatial organization we compare to the original images.
The simulations did not show any spatial structures as were observed in the original images. All the simulations resulted in a granulated image of different cloud types after only a couple of hours.
A simulation from the MC model describing transitions from a finite number of cloud types converges to the observed occurrence frequencies, which is property of a MC. When describing the MC in a continuous state space the simulations still converged to the observed cloud state frequencies. However, introducing spatial dependence this property was not a given anymore.
Since we were unable to reproduce the organizational structure from previous work using the GOES-16 satellite images, we cannot conclude whether it is possible to simulate cloud organization using the MC lattice model. A better understanding of these type of models was achieved. Neighbouring sides could diverge too fast in the performed models. We believe adding a stronger coupling between the cloud state evolution of neighbouring cells can improve the spatial models. Therefore one method to simulate correlated variables was tried to implement on the CTH evolution. This showed promising early results for further research. Also a better correction for horizontal advection on the original images could be performed to observe a specific cloud over time, which could improve the observed data quality.

By combining satellite observations and model reanalysis winds, a global data set with higher temporal and spatial resolution ocean surface wind fields can be generated. These superior surface wind fields can have various applications, such as wind forcing of ocean models. The Royal Netherlands Meteorological Institute (KNMI) is responsible for the delivery of two different global data sets of combined surface winds, referred to as IFREMER and ERA5*. IFREMER’s approach focuses on satellite scatterometer observations which are complemented with model reanalysis winds (ERA5). On the other hand, ERA5* is based on ERA5 model reanalysis and corrected with scatterometer observations. The aim of this research is to evaluate global IFREMER and ERA5* surface winds in terms of spatial and temporal characteristics.

In order to validate the surface winds of IFREMER and ERA5*, a reference system of true winds is required. For this, quality-controlled wind observations from an independent Chinese scatterometer Haiyang-2B have been used. This comparison study has been carried out for the year 2019 focusing on zonal wind, meridional wind and wind speed. IFREMER and ERA5* wind fields have been collocated in time and space to HY-2B wind observations using a 6-hour and 1-hour window respectively. This methodology was applied to overcome differences between the data sets in grid definition, wind representation, spatial resolution and temporal resolution.

On a global scale, IFREMER shows larger average and standard deviation of wind differences with respect to HY-2B than ERA5*, specifically in coastal regions and at higher latitudes. The results of this research suggest that the ERA5* wind product generates more accurate surface ocean winds than IFREMER, except for in the tropics, assuming that HY-2B winds represent the true value. Moreover,
at higher wind speeds both global wind products underestimate HY-2B wind, whereby IFREMER already deviates at lower winds than ERA5*. The IFREMER winds showed suspicious satellite tracks and temporal wind inaccuracies, most likely as a result of its 6-hourly resolution. It is recommended that a follow-up investigation is carried out with validation of buoy data in the tropics to investigate the underlying causes of the observed local wind bias.

It is well understood that precipitation extremes will increase in a warming climate. On (sub-)hourly timescales, the apparent scaling of precipitation extremes is stronger than expected from the Clausius-Clapeyron relation. Here, we focus on a common hypothesis which states that feedbacks from the local dynamics of clouds are responsible.

The effects of climate change on intensification of extreme precipitation in the Netherlands are modeled using the Dutch Atmospheric Large Eddy Simulation (Lochbihler et al. 2019). The objective is to understand more about extreme precipitation events in a warming climate in the Netherlands. A feedback loop is proposed in which cold-pools increase precipitation intensification with super-CC scaling.

Atmospheric conditions from a composite of days with extreme precipitation are perturbed.The entire atmospheric temperature column is warmed and cooled by 4 degrees Kelvin under constant relative humidity. The convective precipitation is described with a 10 minute timescale. A period of organized convection in the simulations is selected for comparison between the simulations. In this timeframe of extreme precipitation, the yield increases 8.5\% per degree Kelvin.

Convection grows deeper and updrafts in the clouds become stronger with warming. The updrafts that are the most representative of precipitation yield increase in velocity by 4.7\% per degree Kelvin. At 7 km height, updraft speeds increase up to 20.7 \% per degree Kelvin.

Moisture transport occurs predominantly low in the clouds, near cloud base.
With warming, more moisture is transported higher up in the clouds. Moisture transport near the surface increases along organized gustfront lines. On average the increase in transport by updrafts in the sub-cloud layer is dominated by increased moisture rather than due to strong updrafts becoming stronger. The moisture that is concentrated into convective cores coincides with intenser precipitation cores. With warming the subsiding motions accelerate and a stronger drying effect is present around precipitation cores. Cold-pools become stronger, forming bigger gustfront structures.

Climate on Earth is changing rapidly. In order to take the right measures it is important to know with which magnitude and rate the warming of the climate continues in the future. However, climate models currently predict different scenarios. The differing in response of shallow cumulus clouds to a warming climate explains the largest part of the spread of climate sensitivity in these models. It is known that clustered clouds have a different climate feedback than randomly distributed clouds and that we can expect more clustered cloud fields in a warmer climate. It is however unknown how these clouds respond and what the exact difference in climate feedback is. Future research must provide us with more insights and eventually a better understanding of the response of shallow cumulus clouds to warming. However, in order to investigate this an objective measure of the degree of cloud organization is necessary, something that is currently lacking. In this thesis a dimensionless combined organization measure is introduced that can be applied to a large range of cloud field organizations on different field sizes. This is a combined measure, as it makes use of the existing organization index I_org and the size of clouds and clear sky areas in the cloud field. A different approach on calculating I_org was introduced taking into account the sizes of the clouds, resulting in an useful and more realistic values from this parameter. The results of the combined organization measure were compared to a visual inspection of 557 cloud fields with dimensions of 10x10 degree, which showed promising results. Sub-fields of 5x5 degree and 2.5x2.5 degree were also analysed, which showed that the combined organization measure could often be applied on smaller scales as well but yielded some problems when areas contained either no or little clouds or some very large ones. It was concluded that the newly developed method is an improvement of the already existing method, providing scientists with a better and more reliable index to quantify the degree of cloud organization.

In the context of climate modelling, convective clouds in tropical regions play a major role in determining the climate sensitivity. The vertical transport of mass and energy associated with this type of clouds is often represented with so called mass-flux parameterization schemes. In this work the aim is to evaluate the relationship between mass-flux and large scale environmental conditions using observations in a tropical region, over a period of 13 wet seasons. A uniquely comprehensive data set from the C-band polarimetric radar (CPOL) in Darwin, Australia, is used to estimate vertical velocity inside precipitating convective clouds. Ultimately, mass-flux is derived over a domain size similar to that of a general circulation model (GCM) grid box. Five parameters (RH_500, CAPE, CIN, ­Omega_500 and Chi_crit) are selected to describe environmental conditions and with these, the magnitude and shape of mass-flux is analysed. Chi_crit appears to be the most valid parameter to represent both the shape and magnitude of mass-flux. All other selected parameters strongly influence only one of the two aspects of the profile. Additionally, fractional entrainment is retrieved from mass-flux profiles and partitioned into two terms: one dependent on area fraction and the other on vertical velocity. Under all environmental conditions, the layer between 4.5 and 7 km experiences detrainment. It can be inferred that a stable layer, known as the freezing level, is present at 4.5 km. Below the freezing level, the vertical velocity and the mass-flux shape are most relevant to determine entrainment rates leading to the conclusion that the vertical velocity should not be disregarded when parameterizing convection.

In the coming decade, the Dutch government plans to broadly expand its wind farm real estate over the North Sea region. This is a significant undertaking because, from literature, it is understood that wind turbines can affect the local environment. These impacts include, but are not limited to, changes in variables such as wind speed and turbulent kinetic energy. To accurately capture these effects and improve power forecast accuracy, weather prediction models numerically represent the physical effects of a wind farm (i.e., a wind farm's behaviour as a momentum sink and a source of turbulent kinetic energy) by incorporating a wind farm parameterization in the model. This study looked at two simulations (with and without a wind farm parameterization) from HARMONIE-AROME, an operational weather model, and assessed the performance of the wind farm parameterization by using the simulation results for the Belgian wind farm zone in the North Sea, and comparing it with observational data from Sentinel-1 SAR, floating LiDARs, measured power as well as results from a large-eddy simulation. The data was also composited based on wind direction and atmospheric stability and analysed. Most importantly, it was found, after comprehensive analysis, that the wind farm parameterization caused a marked improvement in the results of the model especially in the region downwind of a wind farm where the effects are most severely experienced. Specifically, the simulation with the wind farm parameterization reduced the overall wind bias to -0.028 m/s from 0.602 m/s when compared with observational data for the selected region. The overall power bias was also found to be approximately 1.92\%. Several hypotheses based on existing literature were also tested, and along with the composite analysis, indicated that the simulation with the wind farm parameterization performed well in most stability regimes with the exception of stable conditions. It was recommended that this be further explored to determine the cause. Recommendations for expanding the study to look at wind farm effects on surface fluxes in offshore regions have also been made.

The height of wind turbines continues to increase, making the need for more and higher wind measurements for wind turbine model calibrations also increase. The Energy-research Centre of the Netherlands (ECN) and the Danish Technical University (DTU) have conducted the ScanFlow campaign in the winter of 2016/2017 to study the inflow wind field of one of ECN's research turbines by deploying multiple lidar instruments. One of the instruments used in this campaign was a SpinnerLidar. A SpinnerLidar is a forward looking, nacelle-mounted, continuous wave wind lidar system. It measures Line-of-Sight components of the wind in a plane 60 meters in front of the turbine, which need to be transformed into 3D wind vectors. To do so necessary assumptions were made about the free inflow periods, namely that a vertical shear is present and that a wind direction misalignment is more likely than horizontal deviations in wind speed. With these applied assumptions the 3D wind components were determined and used in a validation study with a pulsed lidar instrument, the WindCube V2. The proposed method seems robust as a high correlation in wind speeds at hub-height between two distinctly different lidar systems was found. Using the validated SpinnerLidar measurement to find the turbulent characteristics of the free inflow wind field resulted in turbulence intensity plots showing a higher turbulent component in the lower regions of the measurement plane. Also indications for a induction zone are visible in the SpinnerLidar measurements when compared to the WindCube measurements.

The coarse grid of numerical weather prediction and climate models requires parametrization models to resolve atmospheric processes that are smaller than the grid size. For parametrization development, these processes are simulated by a high resolution model. At the Royal Netherlands Meteorological
Institute, the Dutch Atmospheric Large-Eddy Simulation (DALES) is used. This three-dimensional high resolution model uses advection schemes that are too diffusive when steep gradients are present. In this thesis, an advection scheme based on the Discontinuous Galerkin (DG) method is implemented
for DALES.

The DG method is known to be dispersive. To remove those non-physical oscillations, the moment limiter of Krivodonova is used. Krivodonova constructed the limiter for one- and two-dimensions. In this thesis the moment limiter and limiting order are derived for three-dimensions.

DALES is a model based on the finite difference method and uses operational splitting. Therefore, the DG advection scheme needs a mapping from each cell average to all nodal values that are needed for one DG cell, and a mapping back, which we called mapping a and b respectively. Mappings a that are discussed are taking the cell average as value for all nodal points of the DG cell (cell average a), and taking the L -projection of the cell average to the continuous finite element space (L -projection). This thesis describes mappings b that calculate cell averages of nodal DG values (cell average b)
and calculate the cell averages of the tendencies of DG values (cell average of tendency). Using cell average a combined with cell average of tendency, made the DG method as diffusive as the first order upwind scheme. Substituting the cell average a method with the L -projection, the DG method became
very dispersive, meaning that there was not enough diffusion. At last, cell average b was tested with the L -projection. Its numerical results showed that the speed of the advection was slower than the theoretical velocity. Therefore, a method is suggested which does not need mappings. An option could
be a supergrid that takes multiple DALES cells as a DG cell.

Clouds are chaotic, difficult to predict, but above all, magnificent natural phenomena. There are different types of clouds: stratus, a layer of clouds that may produce drizzle, cirrus, clouds in the higher parts of the atmosphere, and cumulus, clouds that arise in convective updrafts. Thermals, rising air that is often used by birds and gliders to gain height, are an example of atmospheric convection. When the sun heats Earth’s surface layer, warm and moist air rises in thermals to higher parts of the atmosphere. In this way, convection transports heat and moisture vertically in the atmosphere. This often leads to the formation of clouds and heavy rainfall. A major part of the rainfall on Earth, especially in the tropics, is produced by cumulus clouds. Furthermore, convection and cloud formation affect the large-scale planetary circulation. In the atmosphere, these processes are of major importance for Earth’s weather and climate. Convection and clouds also play a major role in numerical simulations of weather and climate. With general circulation models, the large-scale wind circulation and variables such as temperature and humidity are calculated on a three-dimensional global grid. The model grid resolution is low, and therefore, smaller-scale processes such as convection and cloud formation can not be calculated explicitly. The impact of these small-scale processes has to be determined in another way. They are represented by parameterizations that give an estimate of the effect of the smallscale processes on the large-scale model variables. For models with relatively large columns, the presence of a large number of realizations of the same small-scale process justifies the expression of their effect on the large-scale variables in terms of statistical properties. For example, the effect of a large number of clouds can be represented statistically. A problem arises from the fact that the resolution of operational weather and climate models tends to increase. Generally speaking, with higher model resolutions the atmosphere can be simulated more accurately. However, if resolutions keep increasing, the expression of the small-scale effects in terms of statistical properties can no longer be justified. In a small model column, there is for example only space for a small number of clouds. The chaotic behavior of convective clouds becomes an important factor and deterministic parameterizations no longer give accurate estimates. The increase of fluctuations and randomness is a motivation for using stochastic convection parameterizations. The central research theme in this dissertation is stochastic convection parameterization. Stochastic processes are used in the representation of convective clouds. Traditional deterministic parameterizations only give an estimate of the expected value of the effect of small-scale variables. Stochastic parameterizations can deviate from this expected value and can produce a range of convective responses. Especially in models with a relatively high resolution, it is important that parameterizations can represent fluctuations around the expected value. There are several ways of introducing stochastics. In this dissertation, Markov chains are examined, stochastic processes that are named after the famous Russian mathematician Andrei Markov (1856-1922). Markov chains have a finite number of states of which the transition probabilities can be estimated from data. By inferring transition probabilities from high-resolution data of convection, Markov chains mimic convective behavior.
A Large-Eddy Simulation model is used to construct a data set. Large-Eddy Simulation models are able to resolve clouds and convection in detail. After inference of the Markov chains, they are able to mimic clouds and convection as observed in a field-experiment near Barbados. The same method has also been applied for convective clouds in Brazil. These Markov chains only work for a very specific range of atmospheric circumstances. Therefore, another Markov chain model is constructed from a large observational data set from a rain radar in Darwin, Australia. A larger range of atmospheric circumstances is covered, and the Markov chains can be applied more generally. The Darwin Markov chains are implemented in a climate model to stochastically parameterize convection. This improves the variability related to convection as well as the distribution of the simulated tropical precipitation. The Markov-chain model is not perfect yet; however, a large step has been made in the development of this stochastic method for usage in state-of-the-art weather and climate models.