Few rare, circular, concentric enclosure ditches called rondels were discovered in Slovakia, a country in Europe; within the ditches, material traces of Neolithic European culture can be excavated. For exca- vations to happen, archaeologists must first locate these rare structures. Most rondels were spotted on agricultural fields or searched for manually during aerial surveys in a time-consuming manner. With the release of a high-resolution multispectral aerial orthophotomosaic data set of Slovakia, detailed sites containing rondels may be discovered using machine learning techniques.

Machine learning techniques using convolutional neural network (CNN) models can be applied to the field of archaeology to search for excavation sites automatically. An obstacle remains: CNN models require a lot of training image data to be efficient at their task, which is to classify whether areas contain rondels or otherwise. There are only 20 visible rondels on the orthophotomosaic that can be used as training images for model input, creating an imbalanced data set of a class with a minority class of aerial images of rondels and a large majority class of aerial images without rondels. Sketches and recorded characteristics of rondels from current images and from archaeological publications were used to automatically and randomly replicate rondel appearances from above, resulting in a created balanced data set with sufficient rondel examples for CNN training.

Multiple ResNet-34 models and a ConvNeXt model with differing hyperparameters were trained. The most promising model, a modified ResNet-34, was selected based on validation loss from the cross- entropy loss function and on the number of correctly and incorrectly classified labeled images from a test set. The selected model is used to classify data from the orthophotomosaic for rondels using a sliding window technique. Over 9510 square kilometers of agricultural land cover in western and eastern Slovakia was selected from the CORINE land cover map for classification. 7 suspected rondel sites were found, and 2 were determined to likely be rondels, based on their circular ditch-like appearance in 4 sets of multispectral images and in LiDAR elevation data. Results indicate that exact rondel layouts can be delineated with high-resolution orthophotomasics, however identifying circular elevation patterns of ditches proves to be challenging without using additional LiDAR data.

Accurate classification of ice particles in clouds is essential for improving the understanding of cloud microphysics and improving weather and climate models.
This thesis investigates the use of spectral polarimetry in millimetre-wavelengths, combined with a Discrete Dipole Approximation (DDA) and Gaussian Mixture Model (GMM) scattering database, to classify ice particles through fuzzy logic. Utilizing a dual-wavelength (94 and 35 GHz), dual-polarized cloud radar installed in Cabauw, this study analyses two non-precipitating ice cloud events. Spectral polarimetric variables, including differential reflectivity (Z_DR), Slanted Linear Depolarization Ratio (SLDR), backscattering phase (φ_bs), and Dual Spectral Ratio (DSR), were derived from radar measurements and compared with modelled values from the scattering database. Results indicated that different ice particle types exhibited distinct polarimetric characteristics, but a lot of overlap between particles remained.

A fuzzy logic classifier was developed, incorporating both 1D and 2D membership functions to improve differentiability between particle types. Adding temperature and liquid water path as variables was necessary to distinguish between branched planar, aggregates and graupel particles. The classification results were mostly consistent and as expected, though there was a high dependence on temperature, suggesting areas for further refinement. Through fuzzy logic outputs Q and Q-gap, the most probable type of ice particles is identified and a first assessment on the quality of this identification is given.

This study demonstrates that combining spectral polarimetric variables with an advanced scattering database has potential to improve the classification of ice particles. In particular, the proposed technique could allow the classification of possible different ice particle types for each radar observation volume. The method lays the basis for future developments in cloud microphysics and radar-based ice particle classification.

The melting of the Antarctic ice sheet is anticipated to play a significant role in sea level rise over the upcoming decades. Long-term mass and volume changes of the Antarctic ice sheet are predominantly caused by changes in the movement of the ice layer, referred to as ice dynamics. Mass and volume changes of the Antarctic ice sheet are monitored by gravimetry and altimetry satellites. Their data are corrected for glacial isostatic adjustment and changes in the surface climate, namely cumulated surface mass balance anomalies and firn thickness changes, to obtain ice-dynamical mass and volume time series. These time series are modelled using dynamic state-space models in this thesis.

State-space models decompose the data into several components. These components consist of parameters that are constant with time and states that vary with time. This thesis considers two modelling approaches to estimate the parameters and states, referred to as the the Frequentist and Bayesian approaches. The Frequentist approach entails estimating model parameters using maximum likelihood estimation and subsequently determining the states at each epoch using the Kalman Filter and Smoother. The Frequentist parameter estimates are deterministic. As a result, potential stochasticity of parameters is not accounted for when determining the states. This may lead to overconfident small uncertainties in Frequentist models. The Bayesian approach remedies this by considering model parameters to be stochastic. Following the Bayesian approach, the marginal posterior distributions of the parameters are estimated using Markov Chain Monte Carlo methods. Samples from these distributions are used to estimate the conditional distributions of the states at each epoch using the simulation smoother. Because parameter samples are used to estimate the states, stochasticity of the parameters is accounted for when sampling the states following the Bayesian approach.

Three distinct Antarctic ice drainage basins are investigated in this thesis. For their ice-dynamical ice mass data, significant differences between the uncertainties of Frequentist and Bayesian models are found. The Bayesian uncertainties are consistently larger than the Frequentist uncertainties. The largest differences between Frequentist and Bayesian uncertainties are found for the slope components of state-space models. Depending on the complexity of the data that are being modelled, the Bayesian uncertainty of a slope component can be up to 4 times larger than its Frequentist uncertainty.

Frequentist and Bayesian methods to combine the trends of gravimetry-based and altimetry-based state-space models of ice-dynamical ice mass in Antarctic ice basins are also investigated in this thesis. Epoch-wise weighted averaging, with weights based on the uncertainties in the gravimetry-based and altimetry-based models, is done to combine the trends. It is found that the gravimetry data is of significantly higher quality than the altimetry data, having much smaller uncertainties. As a result, the weighted average of the trends aligns closely to the gravimetry-based trend. Finally, several choices that have to be made when working with Bayesian Markov Chain Monte Carlo methods for state-space modelling and their impact on the results are discussed.

Current urban tree inventories rely heavily on time-consuming manual work and often fail to capture all trees. To effectively monitor the impact of urban trees on their environment and vice versa, an automated method for detecting and grouping trees based on their characteristics is crucial. This research aims to expand current urban tree inventories and cluster trees based on their characteristics. Existing inventories typically include species, age, and height, but trees of the same species and age can vary significantly due to environmental factors. This study focuses on a 500x600 meter area in Delft, encompassing 641 recorded trees from the municipal inventory. An automatic tree detection method using airborne LiDAR (AHN4) point cloud data combined with Random Forest classification was implemented, achieving an accuracy of 85-90%. Individual trees were identified using an existing tree segmentation algorithm, detecting 70% of recorded trees with a mean location difference of 0.71 meters and identifying an additional 460 trees, including those on private land. Despite promising results, limitations include the undetection of small trees (below 3 meters) and classification errors leading to missed detections, multiple identifications for single trees, and false positives. Geometric and reflectance features were extracted. Highresolution, 30cm, spectral images from the SuperView Neo satellites, acquired across three seasons, provided spectral features like tree color and NDVI. Overall resulting in a total of 50 features. This comprehensive inventory allows for clustering of individual trees based on geometric, reflectance, and spectral similarities using a K-means algorithm. The approach enhances urban tree inventories by incorporating new features from airborne Li- DAR and spectral images, such as tree height distribution, crown sphericity, and density. Seasonal changes from spectral images provide insights into tree behavior. These detailed features significantly improve clustering, effectively grouping similar trees together. The findings reveal that trees of the same species in seemingly similar environments exhibit significantly different characteristics. This research offers a method to enhance urban tree inventories and supports long-term studies to reveal how trees respond to urban development, climate change, and ecological dynamics. Correlating tree growth and health with specific locations and climatic conditions can aid in developing sustainable urban planning and conservation strategies.

The weather significantly influences daily life, which is predominantly due to short-term weather phenomena occurring in the atmospheric boundary layer (ABL). The HARMONIE-AROME (HARMONIE) model, used by the Royal Netherlands Meteorological Institute (KNMI), simulates the ABL by discretizing the atmosphere into a three-dimensional grid. Processes occurring at scales significantly larger than these grid can be resolved by the model, but processes occurring at scales smaller than the grid (subgrid) are parameterized by theoretical frameworks. At the current horizontal grid resolution of the HARMONIE model, both shallow convection and smaller-scaled diffuse turbulent transport are parameterized by the Eddy-Diffusivity (ED) and the Mass-Flux (MF) scheme, respectively, coupled in the EDMF-framework. In which the MF is described separately for the dry and the moist (cloudy) updraft.
Increasing the model’s resolution promises an increase in atmospheric representation, yet introduces challenges in the so-called Grey zone of turbulence, where the scale of the turbulent motions are in the same order of magnitude as grid size, making
them neither fully resolved nor fully subgrid.
This study aims to investigate the scale-adaptivity of the HARMONIE EDMF-scheme in the Grey zone of turbulence, for the shallow cumulus boundary layer. To this end, high-resolution Large Eddy Simulation (LES) results of two shallow-cumulus cases are coarse-grained to quantify the partitioning of resolved and unresolved turbulence. It is reviewed how well these partitionings scale against the resolution, normalized with height of the dry (h) and the cloudy boundary layer (h+hc), to investigate the potential of scale-adaptivity of the EDMF-scheme with this height. Additionally, the HARMONIE model is run for one of these cases at three EDMF settings: without scale adaptations, with a scale-adaptive scheme based on both h and hc, and with an additional vertical velocity threshold.
In the dry boundary layer, the partitionings of turbulences showed to scale well with the height of the dry boundary layer h, but also implied additional large scaled turbulent transport not carried by strong updrafts. The scaling down of the dry MF in the HARMONIE run showed significant reduction, with increased resolved transport. However, the unresolved partitioning to the total flux still was higher than expected by LES results, which may be explained by these large scaled turbulent transport not accounted for with the mass-flux. In the cloud layer, scaling the resolved and unresolved partitioning of the total turbulence with the height of the cloud h + hc showed not as effective, and indicated that it may not sufficiently represent the strength of convection in the cloud layer. This is supported by the
HARMONIE run, that showed too much decrease of the moist updraft. The addition of the vertical velocity threshold showed a too strong decrease of mass-flux, both in the dry and in the mixed layer. An additional figure from LES results suggest that this threshold was set too low and scale-adaptivity of this threshold may be needed.

Globally, coastal communities face increasing risks from climate-related hazards such as flooding, shoreline erosion, and salt intrusion. These hazards pose threats to both people and their environment, with extreme sea level events increasing these risks. Satellite altimetry allows for global observation of the sea level, reaching remote regions that are not covered by unevenly distributed tide gauges, as these are concentrated in densely populated regions of Western cultures. However, their 10- to 35-day repeat cycles complicate the capture of extreme sea level events. Machine learning offers a promising approach to combine direct satellite altimetry observations with ERA5 pressure and wind speed fields into a data-driven model. As opposed to global and regional numerical models, which require substantial time and expertise to develop, machine learning models are time efficient and require relatively low effort to develop and expand.

This study presents a shallow neural network that effectively estimates hourly non-tidal water levels in the Dutch coastal zone, using X-TRACK retracked and reprocessed satellite altimetry observations and ERA5 hourly pressure and wind speed fields. Reprocessed satellite altimetry observations from 11 missions are used to provide more accurate coastal observations. Tide gauge records are used as ground truth. Both tide gauge and satellite altimetry data are corrected for harmonic tidal signals before training. A 48-hour time window is applied, using all data from 48 hours to 1 hour prior to the estimates as input into the network. The area of interest covers most of the North Sea, from the Strait of Dover to the northern North Sea, excluding the Danish and Norwegian coasts. The neural network is trained and tested at three locations: Scheveningen, Vlissingen, and the Europlatform.

Results show that the neural network can estimate hourly non-tidal water levels with mean squared errors ranging from 0.011 to 0.018 m, mean absolute errors from 0.078 to 0.101 m and standard errors from 0.100 to 0.134 m. K-fold cross-validation with K=4 indicates high robustness, with mean squared errors varying by 0.004 m, mean absolute errors by 0.012 m and standard errors by 0.017 m. The model performs best for hourly and high water levels at the Europlatform and worst for high water levels at Scheveningen. This is partly due to the location of the Scheveningen tide gauge in a harbour with more localised disruptions of the water level compared to the tide gauge at the Europlatform. The ERA5 longitudinal wind speed component contributes most to the estimation of non-tidal water levels, accounting for $\pm$18\% of all weights corresponding to the input variables. Key regions for the estimation of non-tidal water levels include the Dutch coast and the northern North Sea.

When compared to a local numerical model, the developed neural network does not perform with the same accuracy. However, several upsides of the model are identified, such as high computational efficiency for single locations and easy implementation options for refinement of the model. Recommendations for future research focus mostly on improving the model's performance on high water levels and applicability to different regions.

The city of Amsterdam faces the challenge of monitoring and assessing 200 kilometers of historic quay walls, of which much is deemed to be in poor condition. A key monitoring technique used is photogrammetry resulting in deformation testing. The fundamental data source forming the basis of this deformation analysis is a collection of overlapping images acquired of the masonry quay walls. Solely focusing on deformations overlooks a potential wealth of information which could be retrieved from this imagery, like the existence of cracks in the quay walls, a key sign of potential deformation of the structure.
As manual visual inspection of this imagery is very time-consuming, this work proposes a methodology based on fully-supervised deep learning-based segmentation techniques with the goal of detecting and localizing cracks in the masonry quay walls. For this purpose, two neural networks are trained, one for the segmentation of quay walls in images, and one for the segmentation of cracks.
The neural network architectures which are considered in this work are DeepLabV3+, FPN, MANet and LinkNet, together with different encoders and loss functions. For quay wall segmentation, we adopt transfer learning on a network trained on masonry walls and fine-tune it for quay walls specifically. Here, DeepLabV3+ with ResNeXt-50 was found to be most effective, achieving a F1-score of 96.3 % on the test set. For crack segmentation, FPN with ResNeSt-50 performed best, resulting in a test set F1-score of 78.8 %.
The inference of the crack network is done with a multi-level scheme to detect cracks at different image scales and increase output confidence.
The inherent photogrammetric properties of the imagery have proven to be vital for further post-processing steps, like aggregating overlapping predictions, resulting in more prediction confidence.
Photogrammetry also enables converting pixel-wise predictions to crack length and crack width in the units of meters and millimeters respectively. The methodology additionally proposes photogrammetric image processing methods to transform neural network predictions to a 3D representation and a true-to-scale orthographic 2D image.
Additionally a concise visual evaluation has been conducted to assess the prediction performance on an otherwise unlabelled dataset.
This thesis presents an engineering effort for fully-supervised crack localization within the context of photogrammetric processed images, with generalization in mind for automatic assessment.

This thesis investigates the integration of algorithm unrolling and genetic algorithms (GA) for optimizing pump scheduling in water distribution systems (WDS), a critical component for ensuring energy-efficient water delivery. In the context of modern civilization’s reliance on clean, affordable water for diverse uses, the operation of a WDS, particularly through energy-intensive pumps, presents significant challenges. Traditional optimization techniques often resort to hydraulic solvers like EPANET, which, while accurate, are computationally intensive for large-scale applications. Our methodology introduces a meta-model based on algorithm unrolling, building upon prior work and extending it to address pump scheduling with a multi-objective function focusing on both cost and energy efficiency. This approach significantly reduces the computational load, offering a faster alternative to EPANET while maintaining considerable accuracy. The meta-model demonstrated promising results in the Fossolo network, achieving comparable schedules 20 times faster than traditional methods. However, its applicability to more complex networks and its ability to capture detailed system behaviors are limited, highlighting the need for further enhancements in model stability and reproducibility. Despite these limitations, the study emphasizes the potential of meta-models as a complementary tool to traditional methods, especially in scenarios requiring rapid decision-making under computational constraints. This research contributes to the broader field of water utility management, offering insights into more sustainable and efficient operation strategies.

Personal weather station (PWS) networks have the potential to supply precipitation data at high spatial and temporal resolution for urban hydrological modeling. Past research has shown promising results on the quality of PWS data, for example from Netatmo gauges, but studies on other PWS brands are limited. This thesis assesses the quality of precipitation measurements from the Alecto WS-5500 personal weather station. During a controlled experimental setup in an urban environment, the Alecto was found to overestimate rainfall due to incomplete emptying of the tipping bucket. Correcting this mechanical error by a 10 percent reduction factor lowered the relative bias to 0.00 or 0.06, when comparing the station to official KNMI gauge or KNMI gauge-adjusted radar, respectively. Correlations were high between stations with non-faulty setups, but at the 5 minute resolution, correlations were substantially lowered by sampling errors caused during the data transfer to PWS data platforms. A quality control method from de Vos et al. (2019) was adapted and applied to data from a citizen science project in Delft, the Netherlands, which had a 12-month period of measurements for 20 stations, and a 3-month period of measurements from 40 stations. The filtering of faulty zero measurements was improved by applying the filter on hourly accumulations, and the bias correction was stabilized. The variation over individual PWSs, however, remained high due to setup differences. The complex installation process for citizens and issues with software and data accessibility are limiting factors and warrant further research to improve the usability of PWS data for urban hydrological applications.

Mixed-phase clouds, which have a significant impact on the global climate, are complex systems where liquid water and various types of ice particles coexist at temperatures below the freezing point. A key process in mixed-phase clouds is riming which alters microphysical and scattering properties of ice particles. Cloud radar is a powerful instrument for observing and understanding the processes that occur within mixed-phase clouds. Observations from multi-frequency radars and simulation results were combined in recent research to retrieve microphysical properties of ice particles in snowfall and ice clouds. This report presents an ambitious attempt to retrieve all common microphysical properties of ice particles, such as maximum dimension, density, aspect ratio and number concentration in slight rime condition using Doppler spectra. Two mixed-phase cloud events with low liquid water path are studied for such purpose. Spectral dual-wavelength ratio is introduced to retrieve maximum dimension of particles. An iteration process is developed in order to retrieve aspect ratio and density of ice particles from observation of spectral differential reflectivity. The number concentration of particles is retrieved with additional spectral reflectivity. With all the retrieved microphysical properties, ice water content and particle size distribution can be further derived. Ice water content is compared with results from an empirical model. The retrieved properties obtained from using three distinct mass-size relations are compared. Also the bulk and spectral retrieved profiles are compared. The retrieval process can provide consistent microphysical properties of ice particles. It is found that the retrieved ice water content is generally smaller than that from empirical model. Besides, the mass-size relation has significant impact on all retrieved microphysical properties except maximum dimension. The resulting profiles from bulk retrieval are smoother, while spectral retrieval can provide values in regions where the former cannot. The possible error from different sources are discussed or estimated, including the effect on dual-wavelength ratio from the elevation angle of radar, the neglect of differential attenuation caused by liquid and the usage of soft spheroid model. Recommendations are discussed, which include the usage of the latest microphysical models for ice aggregates and Discrete Dipole Approximation for electromagnetic wave scattering simulation.

The use of radars for remote sensing in atmospheric sciences has become increasingly popular over the past few decades. Weather radars play a crucial role in measuring, interpreting, and monitoring various atmospheric phenomena. However, accurate retrieval of vertical air velocities remains a challenging problem owing to certain deficiencies in radar data and errors in deriving numerous parameters of precipitation. This research aims to develop data processing and air motion retrieval algorithms that can improve the accuracy of this task. The project proposes a robust data processing pipeline to enhance data reliability. It also conducts the classification of rainfall types and hydrometeor classes in the different regions of atmosphere for more accurate parameter retrieval. In this project criteria for the quality of air-motion retrievals have been developed. The performance of the retrieval algorithms achieved by using exponential drop size distribution (DSD) for fall velocity calculations has been evaluated based on the proposed criteria. This research also utilizes data spanning six months for performance analysis of air velocity retrieval algorithms under various weather conditions over extensive periods of time.

Lightning is a natural phenomena that can be dangerous to humans. It is however challenging to study thunderstorm clouds using direct observations since it can be dangerous to fly into thunderstorm clouds. In this study, cloud radar with millimeter wavelength is used to study the properties and dynamics of thunderstorm clouds. It is based on a case of thunderstorm on 2021-06-18 from 16:10 to 17:45 UTC near Cabauw. Polarimetric radar variables are used to investigate possible hydrometeors in the clouds and look for vertical alignment of ice crystals that is expected due to electric torque. The technique of Doppler spectra analysis, which has not been used in previous studies about thunderstorms so far, is used to help understand the behaviours of different types of particles within a radar resolution volume. Due to challenges posed by Mie scattering, scattering simulations are carried out to aid the interpretation of spectral polarimetric variables. From the results, there is a high chance that supercooled liquid water and conical graupel are present in thunderstorm clouds. There is also a possibility of ice crystals arranged in chains at the cloud top. Ice crystals become vertically aligned a few seconds before lightning and return to their usual horizontal alignment afterwards. However, this phenomenon has been witnessed in only a few cases, specifically when the lightning strike is in close proximity to the radar's line of sight or when the lightning is exceptionally strong. Doppler analyses show that updrafts are found near the core of the thunderstorm cloud, while downdrafts are observed at the edges. Strong turbulence is also observed as reflected by the large Doppler spectrum width.

The Greenland Ice Sheet (GrIS) is an ice sheet situated on the island of Greenland. It has a surface area of about 1.74 million km² and contains a volume of ice equivalent to 7.4 m of global mean sea level rise. The GrIS is vulnerable to climate disruptions such as anthropogenic climate change. As a result of increased greenhouse gas emissions, the mass of the GrIS is observed to be decreasing starting in the 1990.
Our research uses output from a climate model called Community Earth System Model version 2.1 (CESM2.1). CESM is an earth system model, meaning that it tries to model the whole earth for its major physical, chemical and biological functions. Importantly for our research, it models the GrIS such that its shape can change, so that it can model changes in atmospheric flow. This is known as an “interactive ice sheet”.
This research three scenarios run on an interactive ice sheet to find its results. It uses a historical run to evaluate the model. Then it also uses two hypothetical scenarios called 3xCO2 and 4xCO2. Each starts at the pre-industrial concentration of 285 ppm CO2, and then increases the concentration by 1\% per year until reaching 3 and 4 times pre-industrial concentrations respectively, after which the concentration is maintained constant.
This thesis focuses on the mass balance (MB) of the GrIS, as these differ strongly between 3xCO2 and 4xCO2. In the scenarios we researched, this mass balance is dominated by the surface mass balance (SMB). SMB is in turn primarily driven by the ice melt. Negative SMB is called Ablation Area SMB
The thesis makes two important observations. Firstly, area distribution can be split up into a total surface area component, which depends on time, and a relative elevation distribution component, which depends on elevation. Secondly, the area normalised AASMB is linear with elevation, but the gradient varies with time.
Typically, increased melt is explained as a result of ablation area expansion. However, this does not capture the elevation distribution of the AASMB. Using the two results drawn from above, we create a new mental model which results in a more detailed explanation of the GrIS mass loss.

The use of deep learning in global weather forecasting has shown significant promise in improving both forecasting accuracy and speed. Traditional numerical weather prediction models have gradually improved forecasting skills but at the cost of increased computational complexity. In contrast, new deep learning models, trained directly on reanalysis data, have demonstrated significant gains in forecasting accuracy, achieving competitive levels of performance.

However, the potential of deep learning in predicting spatiotemporal chaotic systems, such as weather patterns, remains unexplored. To address this gap, we investigate the efficacy of a data-driven Fourier neural operator Markovian forecaster to replicate the intrinsic predictability and the characteristic Lyapunov spectrum of the Kuramoto-Sivashinsky system.
FNO reproduces intrinsic predictability and precisely estimates the characteristic Lyapunov spectrum, even with a small dataset.
They cannot represent one of the invariant symmetries, a zero characteristic Lyapunov exponent in the spectrum.

Our findings suggest that deep learning can not only enhance the speed and accuracy of traditional numerical forecasting models but also replicate the weather's chaotic nature.
This has significant implications for generating large ensembles and improving overall probabilistic forecasts.

The research is limited to a deterministic system defined on a single process time scale surrogated by FNO.
The future merits a similar study to investigate if FNO models can perform similarly on larger, stochastic, coupled, and/or multiple time-scale spatiotemporal chaotic systems.

This project is a consulting project for Community Cloud Forest Conservation (CCFC) on how to obtain and communicate to relevant stakeholders an understanding of the impact of land use change and climate change on the hydrological balance of the cloud forest ecosystem in the Sierra Yalijux. The outcomes of the project will be used by CCFC and partners in four areas: Rural water committee capacity building with municipal and village leadership groups, environmental education with the ministry of education, reforestation, and conservation carbon/water credit prioritization with the national forestry institute, and to create thesis topics for bachelors level students with Universidad Rafael Landívar and Universidad de San Carlos. In order to achieve this goal, we divided our efforts in four areas: First, a description of the situation and a review of literature to identify gaps in scientific and practical understanding of local cloud forest hydrology (Chapter 2). Second, an analysis of the situation at a regional scale using publicly available historical data such as remote sensing data and data from the national meteorological authority (Chapter 3). Third, identifying important hydrological processes in the Cloud Forest micro-climate (Chapter 4) and prototyping and testing measurement setups (Chapter 5). Fourth, making suggestions on how to apply the results to the intended impact areas that CCFC has (Chapter 6). Our recommendations to CCFC for capacity building with water committees are based on a literature re view, we found that the presence of Cloud Forest is expected to increase base flow in springs due to its ability to capture additional hydrological inputs in the dry season, increase moisture recycling after heavy rain events, and store water in the soil. We recommend working with water committees to outline the recharge zones of their springs, run some simple calculations on water availability based on precipitation, and develop manage ment plans for the area. Our recommendations for further research are based on the research approaches we describe at the regional scale and the prototyping of field methodologies that we tested. A more permanent setup for data collection is being developed jointly with the Universidad de San Carlos at CCFC’s nature preserve.

Traditionally, archaeological investigations, especially archaeological remains detection, mostly depend on human observation. In order to find the objects in large areas, a lot of fieldwork has to be done and it takes a long time for archaeologists to travel around. Nowadays, the development of LIDAR provides accurate 3D geometric information, which can be used for computer-based detailed terrain study. The application of deployment of computer vision methods also provides a new idea for the automatic object detection approach.

In this study, the neural network architecture "ResNet18" was applied to airborne LiDAR data from the Western regions of Slovakia for the automated detection of undiscovered Neolithic Circular Enclosures (also called rondel in the thesis). NCEs are mysterious stone hedge-like rings scattered through Central/Eastern Europe. The LiDAR data was processed into digital rater data and realized data enhancement by the visualization technique -- Simple Local Relief Model (SLRM). Since the positive samples were limited, expanding the training dataset was crucial and was realized by data augmentation methods based on the positive samples of rondels. The augmented roundels were created by cropping the real roundels and pasting them on the new empty areas after slight modification. After that, the positive image samples and the same number of negative image samples constructed the whole data set and it was divided into two parts -- training data and test data. After the training process of ResNet18, the performances of deep learning models with different combinations of parameters were evaluated, and the selected model was applied to a large area (44276 × 29984 m2), the spatial distribution of the probabilities could be observed and 32 possible new rondel areas were chosen for further validation.

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.

In this research the possibilities of the application of machine learning models at ‘Hoogheemraadschap van Delfland’ are studied. A random forest (RF) and an LSTM model are used for the prediction of the sum of the discharge in the next 2, 8 and 12 hours from the polders to the boezem canals. This research has showed the potential of machine learning models for the prediction of discharge for the considered pumping stations in the case area. This case area is clustered in the Sobek RR model as node 49. The RF and LSTM model are compared to the current Sobek RR model, the machine learning model of Delfland (ReRengAI) and a naïve model by calculating the root mean squared error (RMSE) for the last year of the dataset. For the prediction of the 2 hourly sum of Node 49 the RF model performs the best. Additionally, the performance of the RF model for the 12 hourly sum is satisfactory with a RMSE of 11,071 m3, though using a deep learning model (LSTM) the performance improved to a value of 10,181 m3 for the RMSE. Machine Learning models are known as black-box models and are hard to explain and interpret, which makes the practical implementation of these new models, despite good model results, challenging. Technical recommendations for implementation ML models are improving the quality and availability of the data, increasing the interpretability and explainability of the model, combining multiple objectives in the new model or combining a ML model with a physical model. Organizational recommendations are improving the knowledge about these models within the organization, studying the advantages of these models in comparison to the current model and involving different departments of the water authority in the development of these new models.

In September 2022, Greenland experienced an extraordinary late-season melt event, characterized by temperatures exceeding the melting point at Summit Station for the first time on record and surface melt appearing across one-third of the ice-sheet. This thesis investigates extreme melt events at the Summit in Greenland, focusing on the attribution of the September 2022 extreme melt event to human-induced climate change. The study combines observational data and climate model simulations to assess the influence of climate change on these events and project their likelihood in the future. The research involved identifying melt events in observational and model data. Subsequently, melt-event probability ratios were calculated between the pre-industrial, current, and future climates. These ratios were synthesized to form an attribution statement and provide insights into future scenarios. The study reveals that melt events in any month at the Summit in Greenland have become 20 times more likely in the current climate compared to the pre-industrial climate. This increase in likelihood of melt events in any month is significant and can be attributed to human-induced climate change. However, for melt events specifically in September, although unprecedented in pre-industrial and recent times, no significant increase is found due to a lack of data. Definitive conclusions are expected with more data. Projections based on climate models indicate a substantial rise in future melt event probabilities, reaching up to a 46% chance of Summit melt in September and a 83% chance throughout the remainder of the year. The findings suggest that, while the September 2022 event cannot definitively be attributed to climate change, it highlights the increasing likelihood of such events and their potential impact on sea levels. However, the analysis carries inherent uncertainties due to limited historical and climate model data usage and limited consideration of atmospheric river circumstances. Despite these challenges, these insights contribute to enhancing our understanding of extreme melt events and, in turn, inform the formulation of future climate mitigation and adaptation strategies.