Climate change, with global temperatures rising over the past decades, is a primary driver of sea level rise through the thermal expansion of seawater and the melting of the Antarctic and Greenland Ice Sheets (AIS and GIS). These ice sheets are crucial for predicting future sea level changes, as increased melting forms supraglacial lakes. These lakes can induce hydrofracture, leading to ice shelf instability and accelerated ice flow into the ocean, further elevating sea levels and affecting global climate systems. This study focuses on the AIS and GIS, emphasizing the development and application of a deep learning model to detect and classify the behavior of summer supraglacial lakes using Sentinel-1 SAR satellite data.

The methodology involved normalizing SAR imagery data to enhance data consistency, training a deep learning model using a U-Net architecture on a labeled dataset of Greenland lakes for semantic segmentation, and evaluating its performance on both Greenland and Antarctic datasets using metrics such as accuracy, precision, recall, F1-score, and SSIM. The model is trained to distinguish between draining and refreezing lakes based on backscatter intensity patterns captured in the satellite images. Furthermore, a sensitivity analysis is conducted by creating ten different perturbations of the testing dataset, which included variations in intensity, rotation, and zoom levels. The trained model is then applied to Antarctic data to create an Antarctic-wide map of lake behavior.

The deep learning model exhibited high performance, achieving an accuracy of 90.6%, precision of 90.9%, recall of 90.6%, and an F1-score of 90.0%. It showed high classification accuracy for non-lake pixels (97.5%) and draining lake pixels (84.4%), but lower accuracy for refreezing lake pixels (55.4%) due to class imbalance. Sensitivity analysis revealed optimal performance on the 'zoomed out' dataset with an overall accuracy of 92.9%. Applying the model to Antarctic data successfully identified regions of draining and refreezing lakes, providing a starting point for monitoring ice sheet dynamics and their implications for climate change.

This study underscores the potential of deep learning models to enhance supraglacial lake monitoring, contributing to a better understanding of ice sheet stability and the impacts of climate change. Future work should address class imbalance and explore further model optimizations to improve classification accuracy across both lake types.

Historical aerial imagery serves as a valuable data source for observing Antarctica, facilitating an extended temporal scale of observation and enabling comparisons to deepen understanding of glacier dynamics. However, many historical aerial datasets, including the Antarctica Single Frames dataset utilized in this study, lack geo-referencing and orientation metadata essential for spatial analysis. One method of geo-referencing these historical images involves image matching to establish Ground Control Points (GCPs). This study focuses on the prerequisite for image matching: ensuring alignment between unreferenced historical images and already geo-referenced images in terms of scene and approximate resolution, a process termed 'geo-localization' herein.

Geo-localization is achieved by comparing the historical image with positions within a predefined geo-referenced Area of Interest (AoI). Two predefined remote sensing datasets are used: Sentinel-2 and Quantarctica Rock Outcrop Mask, from which AoIs are generated. Positions within the AoI exhibiting the highest similarity to the historical image are likely to correspond to the same ground area, thus providing the location of the historical imagery.

This similarity assessment employs two Siamese Networks: SigNet and ResNet-50. SigNet, originally designed for signature verification tasks, consists of four convolutional layers. In contrast, ResNet-50, initially developed for image classification purposes, is characterized by its deep architecture comprising approximately 50 convolutional layers, as suggested by its name. In this study, these two models are initially pre-trained on cross-domain datasets and subsequently adaptively trained with task-specific datasets created in this study. The adaptive training datasets comprise triplets of similar and dissimilar images pre-processed using methods devised in this study. An evaluation methodology based on confidence level is developed to assess the model and workflow performance, which is then applied to 51 test historical image samples.

Overall, the results indicate that the ResNet-50 based network outperforms SigNet, achieving a 95.5% average confidence level. However, the method does not meet the initial expectation of directly providing the location of the historical image within the AoI. Instead, it identifies potential locations. Nevertheless, this outcome is valuable as it streamlines the search process for subsequent image matching steps. For instance, a 95.5% average confidence level for the ResNet-50 based network correlates with an approximate 95.5% reduction in processing time for geo-referencing when integrated with image matching in subsequent steps.

The Greenland Ice Sheet (GrIS), which stores freshwater equal to more than seven meters of potential sea level rise, strongly interacts with the global, Arctic and North Atlantic climate. In a warming climate, the GrIS has been losing mass and is projected to lose mass at an increasing rate. The interactions between the GrIS and the climate have the potential to amplify or dampen GrIS mass balance responses to a CO2 forcing. We investigate the impact of ice sheet-climate interactions on the mass balance and climate of the GrIS using the Community Ice Sheet Model version 2 (CISM2) coupled to the Community Earth System Model version 2 (CESM2). We compare idealized simulations with a non-evolving and evolving ice sheet topography in which we apply an annual 1 % increase until we reach four times pre-industrial (PI) CO2 concentrations. Furthermore, we analyze an idealized simulation in which we first apply a 4x PI CO2 forcing and thereafter annually reduce atmospheric CO2 by 5 % until PI concentrations are reached. By comparison of a 1- and 2-way coupled simulation, we find significant changes in atmospheric blocking, precipitation and cloud formation over Greenland as the GrIS topography evolves, acting as negative feedback mechanisms on mass loss. Besides, we find that a uniform temperature lapse rate misrepresents temperature changes in the ablation area, leading to an overestimation of the positive melt-elevation and melt-albedo feedback in 1-way coupled simulations, resulting in an overestimation of mass loss. During a 350 year 4xPI CO2 forcing period, the ice sheet loses a total mass of 1.1 m sea level equivalent, and part of its margins retreat land inward. When applying an annual 5 % decrease in CO2 to 1xPI CO2 concentrations, melt reduces rapidly. The small discharge concerned with the retreated state of the ice sheet allows for halting the GrIS mass loss, despite a surface mass balance that is only slightly positive under a remaining global warming of 2 K. During a complex transitional phase towards a colder climate, the GrIS, Arctic and North Atlantic ocean strongly interact, causing the area south of the GrIS to transition from a ’warming hole’ towards a ’cooling hole’. Elevated atmospheric temperatures, larger ocean heat transport and a poorer state of the snowpack, compared to the initial pre-industrial state, result in limited regrowth of the ice sheet under reintroduced PI CO2 conditions.

Icebergs drifting through the Southern Ocean release fresh water and nutrients. This has local impacts on surrounding ecosystems and sea ice formation. On a global scale, salinity patterns and ocean circulation are affected. In addition,
tudying icebergs as a proxy for ice shelves in a warming climate can help predict future climate impacts and sea level rise. Furthermore, drifting icebergs can pose a threat to ship navigation and offshore projects. In the past, icebergs have been tracked mostly manually, a time-consuming and labour-intensive task. The most widely used data source for this is Synthetic Aperture Radar (SAR), as icebergs often have a much higher backscatter than their surroundings. A few attempts have been made to automatically track icebergs, but these methods do not allow tracking of icebergs that are only partially visible in a satellite image. In this study, a new method is proposed based on partial contour recognition using the contours’ curvature, a technique derived from the matching of ancient pottery fragments. Since the automatic tracking of multiple icebergs requires a large amount of data and computational resources, the web-based environment of Google Earth Engine is used. The new method, called the Contour Curvature (CC) method, is based on three main steps. (1) Detection of icebergs using Simple Non-Iterative Clustering (SNIC) in combination with a threshold function. (2) The icebergs targets are filtered using an area and solidity filter. (3) Among the remaining targets, the best match is selected by comparing the curvature function of the contour with the reference iceberg. The performance of the algorithm is tested by automatically tracking 15 icebergs and comparing
the results to the existing Centroid Distance Histogram (CDH) method. The overall performance of the CC method can be attributed in large part to the inclusion of the area and the solidity filter, with the latter serving as an overall shape filter. For small icebergs (< 10 km2), both the CC and CDH method perform poorly, due to the abundance of icebergs in this range. For medium to large icebergs (10 to 1000 km2), the methods show similar performance with one method occasionally outperforming the other method. For large icebergs (> 1000 km2), the CC method performs better. Since these icebergs are often only partially visible, this leads to strong deviations in the histogram used in the CDH method, making this method less suitable for these situations. Since the CC method allows for partial contour recognition, these icebergs can still be identified. Furthermore, due to the wide variety of backscatter conditions, the detection method occasionally fails to distinguish icebergs from their surroundings.

Global climate change affects mountain regions such as the European Alps. Consequently, glacier extents decrease, and proglacial areas, the areas that recently lost their ice cover, increase in size. These proglacial areas are subject to a high frequency and magnitude of geomorphological activity and act as a sediment source for downstream fluvial systems, extending the influence of their activity beyond proglacial margins. Therefore, a good understanding of the geomorphological activity in these regions is important and subject to numerous studies. Challenges to access Alpine proglacial areas complicate collection of data with high spatial and temporal coverage. A permanently installed terrestrial laser scanner (TLS) overlooking the Hintereisferner glacier in the Ötztaler Alps (Austria) provides daily ranging observations of a proglacial area. The aim of this study is to assess the potential of this permanent laser scanning system for identifying geomorphological processes in proglacial areas. Point clouds of this TLS system were rasterised into range images. An automated registration method is developed to ensure alignment of large quantities of range images in the event of scanner movement. Scans of the 2020 and 2022 summer were combined into two 4D spatiotemporal datasets, allowing analysis of range change time series for each raster cell. A principal component analysis (PCA) of the 4D spatiotemporal datasets is used to explore spatial and temporal patterns of change in the observed proglacial area. The range image registration procedure performed slightly worse than conventional registration methods, but allowed for automated registration of large quantities of scans. Based on the patterns found in the PCA, as well as in the range time series and precipitation data, different geomorphological processes are identified in the observed proglacial area at daily time scales, with magnitudes of several meters. Among these processes are debris flows in gullies, and slumps on lateral moraines. These results demonstrate that characteristic patterns of topographical change can be distinguished using PCA, and PCA is a promising method to be used in other studies exploiting 4D spatiotemporal datasets. The identification of different geomorphological processes indicates that the permanent TLS system could be of use for further research on geomorphological activity in proglacial areas, such as reworking of lateral moraines, while the developed methodology could be used in other studies exploring 4D spatiotemporal datasets as well.

Knowledge of nearshore bathymetry is crucial for every aspect of the blue economy. Currently, it is expensive to obtain nearshore bathymetry at regional scales, and therefore this data is not available for many coasts. The data scarcity occurs especially in the global south and big ocean states that are at the highest risk from climate change. Currently the only global bathymetric dataset, GEBCO, provides depth data at approximately a 500 m resolution. This data is of limited accuracy in the nearshore zone because the source data is sparse in nearshore areas. Recent research has found that NASA’s ICESat-2, launched in 2018 to study the cryosphere, can incidentally capture bathymetric data. Some studies have been done in small sites, but the potential for a global product has not yet been investigated. This thesis proposes a method of using GEBCO data as a starting point and incorporating ICESat-2 data via a Kalman filter. This results in a product with a downscaled spatial resolution and improved accuracy without requiring any in-situ data. If other local bathymetry data is available, it can be added as input to the Kalman filter or used for validation of the method. Recent studies in lidar remote sensing have shown that ICESat-2 can capture nearshore bathymetry at depths of up to 20 me in tracks 0–3 km apart, provided atmospheric conditions are good and the water is sufficiently clear. Hence, this data could provide a source of high-resolution bathymetric depth profiles in tropical areas and possibly at higher latitudes. This research project proposes an automated processing chain, written in python, for extracting bathymetry points from the lidar data based on the density of the photon returns in the underwater zone. This method can reliably identify points containing bathymetric signal. To downscale the data using a Kalman filter, first global data from GEBCO is clipped to the area of interest, and then resampled bilinearly to 50 m resolution. Then, the ICESat-2 photon data for the area is processed to generate point estimates of bathymetric depth. To fill in the gaps between these point estimates, the bathymetric points are subsampled and interpolated to the same resolution as the GEBCO data using a universal kriging interpolator. This interpolator results in a raster of the estimated depth, and a raster of the estimated uncertainty. To update the interpolated GEBCO grid, the Kalman gain is calculated for each raster cell and using the Kalman state equation a new bathymetry grid is produced. If other data is available, the process can be applied recursively with other depth and uncertainty grids, allowing the Bayesian combination of any number of bathymetry datasets for the site. To validate the method, the improvement in RMSE is calculated between the resulting bathymetry grid and previously validated, high-accuracy survey data. The validation has been applied at several global test sites to verify that the method is generalizable to other regions. Validation sites are chosen based on the availability of validation data from either USGS survey or survey data from Van Oord projects. The approach was found to improve RMSE compared to validation data by up to 34% in some test sites. However, other sites showed marginal improvement, or increased RMS error. The results of this research could allow easier methods of characterizing nearshore bathymetry in remote areas. It could also be extended to include temporal variation – as more bathymetric data becomes available, it could be used to measure the dynamic changes of coastal systems. However, the method is limited by strict water clarity requirements and missing data due to atmospheric conditions. Despite these limitations, it could potentially be expanded to create a downscaled global bathymetry dataset of clearwater areas.

Due to climate change, glaciers are melting on the Tibetan Plateau. These glaciers are the water source for big rivers as the Brahmaputra which stream towards the most densely populated area on Earth. Previous studies indicated that water surface heights of lakes connected to the glaciers are increasing in the north of the Tibetan Plateau, while they are decreasing in the south. Current literature does not provide a clear answer on the reason why the north and the south are behaving differently. For determining why the south is behaving different than the north, the processes in the basin should be identified and distinguished. For addressing what all the processes in the basin are, the waterbodies in the upper Brahmaputra should be monitored and their relation to the processes and the other waterbodies should be determined. For determining this relation it is of importance to first determine the lake surface heights and river surface heights in this area. This research focusses on determining water surface heights in the upper Brahmaputra and Nam Co basin (located in south of the Tibetan Plateau) using remote sensing, which is the preferred method since waterbodies in this area are difficult to reach for in-situ measurements.
In this study, data of ICESat-2 is used for water surface height measurements. The laser satellite mission ICESat-2 has been launched in 2018 and makes use of three pairs of beams. Lake surface heights are determined by first computing the water surface heights per beam and finally per ICESat-2 passing. Several additional steps to determine the river surface heights are required due to the slope of rivers. Therefore, the passing beams are clustered, resulting in several locations in the river which can be further analysed. It is found that river surface heights are harder to determine due to the presence of bars and shorelines. Therefore, the method for determining the river surface heights should only be used for non-braiding rivers.
A precision for both lake surface heights and river surface heights of 0.1 meters is found. This research found that ICESat-2 can have a contribution for determining lake and river surface heights due to its relatively good precision, but also due to its high spatial resolution. The ground-track spacing of ICESat-2 is lower than other satellite missions, resulting that ICESat-2 passes the most amount of lakes and river locations. In total the water surface heights of 299 lakes and 127 river locations are determined. This small ground-track spacing is at the cost of the temporal resolution resulting in large time gaps in the timeseries of the lakes and river locations. Therefore, a seasonal variation and generic water surface height curve will be difficult to assess. It is possible that ICESat-2 will miss the moments with the extreme water surface heights resulting in a misconception of the water surface heights. But, over a longer time period it may be possible to fit a generic water surface height curve through the water surface heights of lakes larger than 10 km² due to its higher amount of ICESat-2 passings.

The melting of mountain glaciers worldwide contributes significantly to global sea level rise, and is considered an indicator of climate change. The accurate modelling of glacier heights is crucial for determining their past, current and future state. The new lidar altimeters ICESAT-2 and GEDI were launched in 2018, which could provide valuable new data in this regard. ICESAT-2 uses a novel photon counting technique that delivers elevation data at an unprecedented 0.7 m along-track resolution. This also poses questions about how such data should be processed. GEDI uses a more conventional full waveform sensor, which creates a vertical profile of surface returns every 60 m representing a footprint with a 30 m diameter. In this thesis the potential of ICESAT-2 and GEDI for detecting glacier changes in Austria is examined. The vertical errors on the Austrian glaciers of the ATL03 geolocated photons and ATL06 Land ice elevation data products of ICESAT-2 were found to be 1.94 and 1.72 meters respectively, which outperforms most alternative sensors. The GEDI L2A product provides an error of 5.80 m for the same area. For both sensors, the horizontal geolocation error was found to be relatively large, which results in a strong relation between the surface gradient and the vertical error. The high resolution and accuracy of ICESAT-2 can provide valuable detailed information about glacier height differences worldwide. The inclusion of GEDI data can help provide observations in sparsely sampled areas, but considering its relatively poor resolution and accuracy better alternatives may be available. The satellite observations were confined to glacier inventory outlines and compared to a high quality Digital Elevation Model of Austria. The results show an average glacier height loss of 5.85 m for ICESAT-2 and 4.86 m for GEDI, with the majority of observations representing the period 2010-2019. The height differences were compared with their local elevation, slope, orientation and their distance to the glacier edge, in order to find what features could assist in extrapolating the observations. All features show some relation with the glacier height differences, but the elevation and distance to glacier edge were found to contribute the most information.
Spaceborne lidar data is confined to narrow profiles of the surface, so they require extrapolation to make conclusions about the total state of the glaciers. The total mass difference of Austria was estimated first by simply applying the average height difference to the total glacier area, and secondly by ordinary cokriging with the elevation. Both techniques agree on a total mass loss of 1.59 gigatons, while the cokriging method additionally results in a more detailed image of mass loss for individual glaciers. For the Ötztal region a mass loss of 77.07 megatons annually was estimated for the period 2010-2019.