Speakers and Abstracts 

The cost-effective unit’s design includes all necessities for independent operation: a solar charger, a microcontroller and a 4G modem connection used for data transmission and remote connection. 

We present our system’s design and first positioning results. The results are validated by analysing data of one cost-effective unit which was co-located with a permanent GNSS station. We show that the cost-effective GNSS units are well suited to extend existing volcano monitoring networks. We encourage the application of such units also in areas where the installation of conventional GNSS is deemed too costly, or in risk-prone areas where rapid installations are necessary. 

In safety of life applications, such as civil aviation and autonomous driving, a reliable position estimator is needed, generally obtained with GNSS positioning. 

Reliability is expressed as the integrity risk: the probability that the position estimator provides a solution outside of an alert region of interest around the user-receiver's true, yet unknown, position. 

To quantify the integrity risk, information about the types and frequency of faulty measurements is needed. When GNSS systems are used for positioning, different types of faults may occur. Two of these fault types are a single satellite failure and a constellation failure. These failures occur when the broadcast satellite orbit and clock offset contain faults, thereby introducing a bias in the ranging measurements larger than a certain threshold. 

At the Galileo Reference Center, data are stored on the broadcast and precise satellite orbit and clock offset. A stochastic analysis will be carried out to estimate the variance of the Galileo broadcast orbit and clock errors, the variance of the ranging observable as a result of these components, and corresponding uncertainties. Consequently, the satellite and constellation failure probability can be estimated, which allows computing the integrity risk, providing a reliability measure for the position estimator.  

Back in 2017, the only satellite system in Low Earth Orbit (LEO) with global coverage was the Iridium constellation, primarily used for communications. In recent years, the LEO region has become densely populated with thousands of satellites deployed for broadband services (e.g., SpaceX Starlink, Eutelsat OneWeb) and remote sensing (e.g., Spire Global, Planet Labs). Future mega-constellations are planned to enhance Position, Navigation, and Time (PNT) capabilities, currently enabled by Global Navigation Satellite Systems.  

In this presentation, we explore the emerging role of future LEO-PNT systems in supporting Navigation, Geoscience and Remote Sensing applications. We start with an overview of the latest developments for currently planned constellations such as Xona’s PULSAR, Future Navigation’s Centispace, and ESA’s LEO-PNT demonstrator. Using numerical simulations, we then evaluate the expected performance of user positioning in terms of convergence time, while considering various system architectures. 

This analysis highlights the main benefits, e.g. large geometry variability, alongside the key challenges, e.g. limited coverage. Finally, we discuss a few potential applications in geodesy and remote sensing, including ionosphere monitoring, and water vapor tomography. We emphasize the pivotal role of Machine Learning (ML) in these applications, thus presenting new research opportunities enabled by the integration of LEO-PNT systems. 

Actively monitoring ground motion is of the highest importance in The Netherlands, a country in which many of its regions lie below sea level. Water tables in the country have been managed for centuries by using a system of dams, dikes and canals through which excess water can be pumped away to allow for the prevention of flooding, and for the reclamation of submerged land. However, the effects of centuries of active water management in the region have resulted in significant land subsidence, and its effects are being felt as it is becoming a significant threat to the future of the country as sea levels continue to rise. 

This has created the need to monitor land surface motion at large spatial scales with frequent temporal sampling. While InSAR is a promising candidate for such a task, the technique often suffers from drastic losses of signal quality in the spring and summer months when used to produce time series observations of peatlands. This significantly limits the effectiveness of InSAR as a tool to monitor peatland surface dynamics. 

We present the preliminary results of peatland surface motion using a novel InSAR processing method which is designed to overcome the issues which have prevented its application over northern peatlands in the past. This work is the first large scale analysis of the Dutch Green Heart region made with InSAR, providing land surface motion time series data at the parcel scale for a 2000 km2 region with sub-weekly sampling over the period Jan. 2015 to Oct. 2023. Our presentation will briefly outline the results, validation efforts and the various challenges involved. 

Based on a breakpoint analysis of tide gauge data along the Dutch coast, Steffelbauer et al. (2022, hereon S22) provided the first instrumental evidence of a sea level acceleration from the early 1990s, while Keizer et al. (2023, hereon K23) showed that the acceleration had started already in the 1960s. 

We update the breakpoint study by S22 and reconcile their result with the K23 findings. By using slightly longer timeseries, an additional tide gauge station in the Wadden Sea, and the K23 model for the nodal tide, we find that stations along the west coast and the Wadden Sea behave differently. In particular, the analysis of the western stations reveals a clear breakpoint in the 1970s, consistent with the K23 results, while the inclusion of stations in the Wadden Sea leads to the emergence of a breakpoint in the 1990s, as in S22.

We also perform a systematic analysis of all possible combinations of start and end years for the breakpoint detection, which allows us to estimate that the average pre-1990s sea level acceleration along the west coast has been about 0.04 ± 0.03 mm/yr2, which is again consistent with the K23 estimates. 

In this study, we propose an innovative processing chain designed for low coherence areas to monitor landslide displacement. 

This processing chain includes a pre-processing feasibility assessment with the integration of the maximum detectable deformation gradients, vegetation coverage and R-index, and landslide displacement monitoring using distributed scatterer interferometry (DS-InSAR). 

Using DS-InSAR, the coherence of 85 Sentinel-1 SAR used (2021-2023) can be dramatically improved. We tested our methods over the small-sized Tianxi landslide located in Guangxi province, China, surrounded by an NDVI value higher than 0.5. The results show that the pre-processing InSAR feasibility assessment can predict the potential distribution pattern of the InSAR measurement points. 

This helps determine whether the Tianxi landslide labeled as low feasibility can be monitored by DS-InSAR, as opposed to small baseline subset interferometry (SBAS-InSAR). The results also show that DS-InSAR provides more comprehensive and detailed monitoring of landslide displacement and identifies five times as many InSAR measurement points compared with SBAS-InSAR. 

Time series analysis of displacement, combined with rainy factor, indicates that DS-InSAR can detect precursory movements. It is likely that the combined effects of human activity and rainfall contributed to the failure of the Tianxi landslide. 

Recent advancements in open-source Python libraries like Xarray and Dask have opened new possibilities for handling larger-than-memory datasets. 

In this talk, we will share an example case of scaling up a Python function designed for processing InSAR data, based on our recent experiences. Initially, this function was created to analyze small subsets of raster interferogram stacks, and we aimed to extend its capabilities to handle much larger datasets. 

Throughout this process, we faced and overcame challenges when leveraging the functionalities provided by Xarray and Dask. We distilled our findings into a set of best practices, which we will discuss and demonstrate using our function as a practical example. We believe our insights will inspire other researchers facing similar challenges. 

Wayfinding is an intricate human behavior that requires spatial cognition to perform optimally. 

Various variables are used during wayfinding varying from the person's ability and spatial interpretation to the spatial condition itself. 

Up until now, the current study of wayfinding mainly concerned variables in a single environment without acknowledging the possibility of variable correlation between environments. Furthermore, wayfinding simulation studies mainly use digital modeling for the simulation environment, causing a lack of real-life ambiance. 

Therefore, this ongoing study aims to address the potential correlations of variables between environments. Furthermore, the result will be gathered through a virtual reality simulation with a point cloud scan-building model. The progress of this study showed the promising potential of combining the technology of VR and Point Cloud as a wayfinding study method. Which is expected to provide a more accurate environment representation for the study of wayfinding. 

InSAR is an advanced satellite-based geodetic technique and it has become one of the most important tools for measuring earth surface deformation that can have a wide range of scientific and engineering applications. High-quality deformation information is playing a crucial role in helping with energy transition and climate adaption activities, ranging from energy storage, CO2 sequestration, mining, groundwater resources management to land subsidence and hazard mitigations. 

Here, we present: 

1. the development of a PSDS-based multi-temporal InSAR workflow, where we incorporate the DSI and PSI techniques into the framework of SBAS and 3D phase unwrapping at full resolution, aiming to improve the capability of identifying high-quality measurement points as well as data processing precision; 

2. InSAR post-processing for building damage assessment, with the focus on characterizing damage probabilities using InSAR data; 

3. showcasing of InSAR technique for land deformation monitoring at different spatial scales from nationwide to cities, with applications to large-scale groundwater dynamics and resources, shallow-deep aquifer deformation and the mechanism of induced seismicity, as well as city building damage risk mapping. 

Building use information is important for urban planning, city digital twins, and informed policy formulation. Prior research has predominantly focused on mapping building use in broad categories, offering general insight into their actual use. 

Our study investigates the extraction of hierarchical building categories, encompassing both broad and detailed classifications while accounting for mixed-use. To achieve this, we explore the fusion of building function information from satellite images, digital surface models (DSM), street view images, and point of interest (POI) data. 

We propose a novel multi-label multimodal transformer-based feature fusion network, which is capable of simultaneously predicting four broad categories and 13 detailed categories. 

Experimental results demonstrate the efficacy of our method, as it maps most of the building use categories, with the weighted average F1 score for four broad categories and 13 detailed categories of 91% and 77%, respectively. 

Our experiments underscore the critical role of satellite images in building use classification, with the inclusion of DSM data and POI significantly enhancing the classification accuracy. By considering detailed use categories and accounting for mixed-use, our method provides more detailed insights into land use patterns, thereby contributing to urban planning and management. 

Satellite-based interferometric synthetic aperture radar (InSAR) has shown its potential in landslide detection. 

Yet, to detect slow-moving landslides in a rapid and automatic manner, on a large scale, is challenging, especially over mountainous regions. 

To take up this challenge, this study designs and demonstrates a workflow that integrates GPU (Graphics Processing Unit)-assisted computation, InSAR phase-gradient stacking, and an improved YOLOv8 (iYOLOv8), namely GIINWF, to facilitate a rapid and automatic detection of slow-moving landslides. 

GIINWF first employs GPU-assisted interferometry and phase-gradient stacking method to rapidly obtain phase gradients in the range and azimuth directions, and then utilizes an iYOLOv8 network to automatically and rapidly detect landslides. 

We tested GIINWF using Sentinel-1 SAR images between 2021 and 2022 in both ascending and descending orbital directions, over a landslide-prone area close to the city of Chengdu, China. 

The results show that we detected 298 slow-moving landslides that were validated by ground truth data. 

We conclude that our GIINWF is suited for rapid and automatic landslide detection, which opens a new perspective for its application in large-scale regions. 

Land subsidence induced by human activities is a well-known issue. In urban areas this inflicts damage to buildings and infrastructure, leading to high costs and hazardous situations. 

A complicating factor in urban areas  is that the presence of the built environment also itself influences the processes of subsidence. 

To address these challenges, we intend to fill out two big knowledge gaps. The final objective is to disentangle the relative subsidence contribution of the presence of building in one selected urban area in the Netherlands.

The first challenge in predicting subsidence in urban areas is the relatively small spatial-scale of the subsidence processes. This requires the spatial downscaling of existing modelling approaches, since damages to buildings are driven by small-scale spatial subsidence fluctuations. 

The second one consists of assessing the effect of urbanization itself on land subsidence. Our current subsidence models disregard the presence of the build environment and thus its potential additional effect. 

We plan to combine multiple data sources (building locations/weights/years of construction, InSAR, LiDAR, and Cone Penetration Tests) with physics-based and ML-based models. 

Polygonal building outline extraction has been a research focus in recent years. Most existing methods have addressed this challenging task by decomposing it into several subtasks and employing carefully designed architectures. Despite their accuracy, such pipelines often introduce inefficiencies during training and inference. 

This paper presents an end-to-end framework, denoted as PolyR-CNN, which offers an efficient and fully integrated approach to predict vectorized building polygons and bounding boxes directly from remotely sensed images. Notably, PolyR-CNN leverages solely the features of the Region of Interest (RoI) for the prediction, thereby mitigating the necessity for complex designs. 

Furthermore, we propose a novel scheme with PolyR-CNN to extract detailed outline information from polygon vertex coordinates, termed vertex proposal feature, to guide the RoI features to predict more regular buildings. 

Comprehensive experiments conducted on the CrowdAI dataset show that PolyR-CNN achieves competitive accuracy compared to state-of-the-art methods while significantly improving computational efficiency, i.e., achieving 79.2 Average Precision (AP), exhibiting a 15.9 AP gain and operating 2.5 times faster and four times lighter than the well-established end-to-end method PolyWorld. Replacing the backbone with a simple ResNet-50, PolyR-CNN maintains a 71.1 AP while running four times faster than PolyWorld. 

This presentation explores the effects of environmental factors—specifically illumination zenith, moisture content, and shadow conditions—on the spectral variation and robustness of hyperspectral mineral classification. The research assesses how varying conditions influence the spectral measurements and classification processes used in geological remote sensing.

Key findings demonstrate that averaged endmembers enhance classification robustness across multi-temporal images more effectively than individual image-extracted endmembers. Additionally, while spectral brightness and noise levels are impacted by changing illumination angles, the center position of spectral features remains relatively stable, aiding robust classification. The study also reveals that moisture significantly alters spectral absorption features and brightness, but its effect on the center position of spectral features is the least. Finally, shadows and moisture predominantly reduce reflectance albedo and weaken spectral features, with shadow conditions showing the most significant impact on classification robustness.

Overall, the thesis outlines the importance of considering environmental variability in improving the reliability of hyperspectral data interpretation for mineral mapping. 

The atlas narrates the activities of ITC, University of Twente. It resulted in a hybrid atlas, an atlas published on multiple mediums: digitally as an online edition and analog as a printed edition. 

The challenges faced were data, technical and design related. The data wrangling process took more time than anticipated. This was mainly due to dependencies on different sources, incompleteness and in consistencies. Additionally, the EU General Data Protection Regulation (GDPR) had an influence on the workflow since all student and staff data had to be anonymized. Technical issues were mainly related to the web mapping implementation. 

The components library offered us a considerable flexibility, but to create components for all our creative ideas proved impossible due to the project’s time budget. Design issues where mainly related to the (im)possibilities of both print and web environments. However, certain map types work well online but must be adapted for print to avoid for instance visual clutter. In some cases it might have been more effective to come up with new visualizations, instead of adapting visualizations which we originally designed for one environment for the other (e.g. make a map intended for the print atlas interactive). See https://atlas.itc.utwente.nl