Over the period 1901-2012, the average global sea surface temperature has increased by 0.89 °C due to climate change and it is expected to increase consistent with global warming. As a consequence, marine ecosystems have become more susceptible to change in species and ocean chemistry. Due to the increase in sea surface temperature, the ocean pH has decreased in all regions. The main driver for this development is the uptake of approximately 30% of anthropogenic carbon dioxide (CO2) by oceans (IPCC, 2014). In a warming ocean, populations of warm-water species grow and populations of cold-water species decline. Furthermore, declines in coral growth are observed (Poloczanska et al., 2016).

Marine cloud brightening is a proposed way to counteract climate change. This geo-engineering technique brightens clouds by spraying aerosols in the air which act as cloud condensation nuclei in order to form cloud droplets. As a result, the albedo of the brightened cloud increases and more sunlight is reflected, reducing the global mean temperature (Latham et al., 2012). Although it is a potential way to counteract climate change, marine cloud brightening has never been applied on a large scale. Also, little is known about its feasibility.

We determined how physical properties of sea salt aerosols, specifically the aerosol mean geometric radius, the aerosol number concentration and the modal standard deviation of the initial aerosol spectrum, relate to the albedo of the brightened cloud using a numerical cloud parcel model with input parameters (Glassmeier et al., 2020). For six different values of the initial aerosol mean geometric radius, the initial aerosol number concentration and the modal standard deviation of the initial aerosol spectrum, i.e. the variability in aerosol radius, three output variables were generated in time and height. These output variables were the mean radius of cloud droplets, the cloud droplet number and the relative supersaturation. The differences caused by varying the input parameters were interpreted for the output variables. Also, we generated the albedo for the six different values of each input parameter mentioned.

The cloud droplet number depends on the aerosol number concentration. The higher the number, the more cloud droplets are formed. The cloud albedo does not depend on the initial aerosol mean geometric radius. The moment of activation depends on the initial aerosol mean geometric radius and the initial aerosol number concentration. The growth of the cloud droplets is mainly driven by condensation when variability in aerosol radius is small. The higher the initial aerosol number concentration, the higher the albedo of the brightened cloud. When the variability in radius is bigger, collision-coalescence occurs as the spectrum of cloud droplets consists of bigger radii which tend to collide and coalesce. This leads to bigger and heavier droplets which precipitate due to an increased mass, resulting in a decrease in cloud droplet number and a decrease in cloud albedo.

To conclude, three relations between physical properties of sea salt aerosol and the cloud albedo were found. First, the aerosol mean geometric radius at start does not affect the cloud albedo. Secondly, we found that the higher the aerosol number concentration, the higher the cloud albedo. Furthermore, a higher modal standard deviation of the initial aerosol spectrum leads to a lower cloud albedo.

We recommend further research on other types of aerosols, like organic matter and algae, and the effect of external forcing, like wind, on cloud brightening.

Uncertainty in the radiative forcing from anthropogenic activities since the Industrial Revolution is dominated by how clouds respond to aerosol. Climate projections are limited by this uncertainty. The cloud response to aerosol is influenced by the meteorological conditions of the atmosphere wherein the cloud is suspended. Understanding the covariation between meteorological state and cloud response to aerosol is a path forward to improve our understanding of aerosol effects on the climate. In this thesis we study aerosol-cloud interactions while controlling for meteorology using clustering techniques. This allows us to study the interactions per meteorological regime and gain deeper understanding of the effect of meteorology on aerosol-cloud interactions. Cloud-controlling factors are clustered using k-means clustering. Six meteorological clusters are found and satellite observations over ocean between -60 and 60 degrees latitude are used to study cloud response to changes in aerosol concentrations per cluster. The choice of clustering does not create significant variability in the sensitivity and the radiative forcing of the cloud albedo effect, but the sensitivity of cloud liquid water path and cloud fraction adjustments do show variability between clusters. This indicates that controlling for meteorology is specifically important for the adjustments to the cloud albedo effect. Our results show that the effective radiative forcing from aerosol-cloud interactions over our study domain since 1850 is -1.0 watts per square meter with a 90% confidence interval of [-1.6, -0.48] watts per square meter. There are variations in the forcing estimates depending on the number of clusters, but this signal is small compared to other sources of uncertainty. Our findings corroborate recent findings and present a novel method to control for meteorological covariation using cluster analysis.

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.

Cloud droplet number concentrations change due to perturbations in aerosol concentrations. The strength of this correlation covaries with meteorology. Using polarimetric aerosol estimates and MODIS-2 cloud retrievals we compute the interaction strength per meteorological regime, which we determined using clustering techniques on MERRA-2 reanalysis data. The clusters that were found are similar to other clustering studies. The clusters are however not well-separated. The resulting interaction strengths are slightly higher compared to previous satellite studies. The clusters which show large scale vertical movement, generally have higher interaction strengths.

In this thesis we are interested in distinguishing patterns of mesoscale cloud patterns in the trades. Specifically, whether Sugar, Gravel, Fish and Flowers patterns can objectively be identified using physical quantities. For this purpose, we use cloud fraction data attained by the CORAL Ka-Band cloud radar at the Barbados Cloud Observatory during the boreal winter seasons of 2018, 2019 and 2020. These cloud fraction data represents the curves up until a height of 4 km for a given 6-hour interval of time. We do this to see if these clusters match up with the labels assigned to each cloud fraction curve obtained from a classification model used in Schulz (2021). Firstly, we map the cloud fraction curves onto points on a finite dimensional space using functional principal component analysis. We subsequently apply K-means, Gaussian Mixture Models and Mean Shift clustering onto the pre-processed dataset to identify any robust clusters. We have been able to attain robust Sugar-like clusters for K-means for 3 and 4 partitions and Mean Shift with bandwidth λ ≈ 585. This provides evidence that we are able to use cloud fraction data to distinguish Sugar. However, the same can not be said for Gravel, Fish and Flowers as we have not been able to identify them in our analysis. It is suggested for future research to do sensitivity analysis in the height interval of the cloud fraction data, that outliers are omitted and that the labeled data from Schulz (2021) are used instead of the mean and spread of the data pertaining to those labels.

Clouds, specifically shallow clouds, are known as a major source of uncertainty in global climate models. Shallow clouds over the global oceans show different spatial patterns and organizations that may be influenced by climate change. Besides, the frequency of these patterns can change the climate feedback of marine clouds in the subtropics. Experts in atmospheric sciences have proposed different kinds of organization metrics. In a recent paper (Janssens et al., 2021), 21 cloud organization metrics were collected and computed over 5000 satellite scenes of shallow clouds located near the east of Barbados. Almost all of the existing metrics of cloud organization are bulk parameters of the cloud field. Accordingly, over the same data as Janssens et al. (2021), this project aims to define a metric that originated from the mutual arrangements of the individual clouds in the field. To this end, network theory as a mathematical tool is employed to define new cloud organization metrics to explore how cloud objects interact with and coordinate concerning each other. It should be noted that this research is the first in which cumulus cloud fields are considered as complex spatial networks. In this regard, an important question is whether a new network metric can distinctly explain a variability which has not been captured by previously defined metrics in cloud organization. To address the research question, we utilize a multivariate regression model to understand whether a linear combination of principal components of the existing metrics can encapsulate a variation in newly proposed network metrics. We found that degree standard deviation and mean of clustering coefficient are two network metrics that can distinctly capture a variability that has not been encapsulated by the existing metrics in cumulus cloud organizations. Degree standard deviation simultaneously measures the homogeneity of the cloud size distribution and the distance between nearest neighboring clouds distribution. A large value of the average clustering coefficient indicates that the network field consists of either large clouds located on corners of triangles or relatively small clouds closely coagulated. Additionally, two sensitivity analyses were performed to understand how the main results are influenced by either center-spacing or edge-spacing distance between clouds and the nodes (clouds) located close to the boundaries of the field. Finally, one should note that all of the results of network theory-based approaches are comprehensively affected by how the network is defined. Defining a network that combines both geometric and physical interaction between cloud objects will be the future work of this research.

Although monitoring and maintenance of railways is important to ensure safety and avoid delays and financial losses, it is still mainly based on human inspection. The complexity of a railway along with the large area it extends makes manual monitoring difficult and time-consuming. The increasing availability of 3D acquisition technologies has made point clouds a widely used 3D data form. Thus, identifying the key components of a railway and its environment using 3D point cloud data can be the first step for automating this procedure. The past years, machine learning has become the most popular subfield of artificial intelligence used for various different applications, with its subfield of deep learning evolving dramatically. Although deep learning has been vastly researched on 2D data, applying deep learning on 3D point clouds can be challenging due to the irregularity, unstructuredness and unorderedness of such data. To overcome those challenges, recent approaches use projection or voxelization, while the latest methods focus on working directly on raw point cloud data.

In the present research, 2 machine learning methods are implemented to classify 3D point cloud data of railway environments into 7 categories of rails, sleepers, track bed, masts, overhead wires, trees and other. To this end, the ensemble method of Random Forest, and the deep learning method of DGCNN (Dynamic Graph Convolutional Neural Network) are implemented. While Random Forest is a simple and handy ensemble algorithm used for a wide range of applications, DGCNN is a deep learning method based on PointNet, the pioneer method on raw point clouds, and graph CNNs. The methods are validated under 3 case studies, produced by structure from motion photogrammetry, airborne laser scanning and terrestrial laser scanning, and locating in 2 different areas within the Netherlands. For each method, 2 scenarios are developed for classifying colored and uncolored point clouds, respectively. Finally, the 2 methods are combined for the first scenario, as a first attempt to further improve the final results.

The obtained results show that, in this work, DGCNN performs better than Random Forest, and both methods approximate state-of-the-art performance. The contribution of colors is important to improve both the overall accuracy of the models, as well as the classification results of the individual classes. More specifically, Random Forest scenario 1, Random Forest scenario 2, DGCNN scenario 1, DGCNN scenario 2, and the combination of Random Forest and DGCNN scenario 1 result in an overall accuracy of 88.65%, 83.39%, 89.19%, 88.20% and 90.57%, respectively. The corresponding per class F1-scores for all methods and scenarios range between 43% and 94%. Both methods meet difficulties in generalizing on data from different sensor systems and different areas with very different point density and missing data. Nonetheless, the individual methods are already very promising, while they are able to achieve the required accuracy of more than 90% when combined.

Airport operations are highly dependent on safe weather conditions for their daily operations. Especially the visibility conditions are crucial. Low Visibility Procedures (LVP phase M, A, B and C) are defined at Schiphol Airport, resulting in capacity and runway use restrictions. These Low Visibility Pro­cedures are based on a post­processing model of a Numerical Weather Prediction (NWP): the Terminal Aerodrome Forecast Guidance (TAFG). This model provides a probability for the occurrence of a given LVP phase, based on thresholds of the Runway Visual Range (visibility including back luminescence at airports) and the ceiling (cloud base height). Currently, the TAFG is operational for three NWP models (HARMONIE, HIRLAM and ECMWF). This research found that the HIRLAM TAFG has a better performance than the HARMONIE TAFG. This result is caused by the larger training data­set that is used for the HIRLAM TAFG (20 years) compared the training data­set of for the HARMONIE TAFG (3 years). The ECMWF TAFG has a better performance for the severe LVP phases B and C. This is remarkable, since ECMWF is not often used for airport operations. However, the standard error of the performance indicator of the ECMWF TAFG is higher compared to the other models and, therefore, a larger data­set should be studied to confirm these findings.