Although Graph Neural Networks (GNNs) have made significant advances in spatio-temporal traffic forecasting, their performance is limited when they rely solely on sensor proximity or road-network topology. This paper presents a spatio-temporal prediction framework, developed to incorporate knowledge in various forms. This framework aims to improve sensor-level, contextual understanding of the environment. A general-purpose knowledge graph (e.g., Wikidata) is used to create semantic subgraphs around traffic sensors and generate knowledge graph embeddings that capture meaningful relationships, such as nearby points of interest, administrative hierarchies, and the functional roles of locations. These embeddings are then fused with conventional traffic sensor graphs to provide additional adjacency matrices informed by semantics. This allows GNNs to learn the semantic context beyond physical connectivity. This study differs from previous research in two key ways. Firstly, rather than proposing a novel GNN architecture, it demonstrates the general impact of external knowledge on prediction accuracy. Secondly, experiments with well-established traffic forecasting approaches show that external knowledge provides additional information that street network data alone cannot convey. The results show that integrating data from general-purpose knowledge graphs and sensor networks through data fusion can enhance the prediction accuracy of traffic forecasting models, and offers a potential pathway toward improved interpretability.

Abstract The discovery of patterns in spatial and spatio-temporal data is crucial across scientific disciplines studying natural phenomena to enhance our understanding of the real world. These phenomena display complex patterns, necessitating novel specialized pattern mining techniques. In this paper, we introduce Vortex Correlation Clustering which aims to identify a subgroup of such complex pattern, namely correlated groups of objects oriented along a vortex. This can be achieved by adapting the Circle Hough Transform, already known from image analysis. The presented adaptations not only allow to cluster objects depending on their relative location next to each other, but also allows to take the orientation of individual objects into consideration. A multi-step approach allows to analyze and aggregate cluster candidates, allowing a certain deviation from the reference shape in the final clusters. Further adaptations allow to analyze clusters along a third dimension, which allows to reflect the shape of real-world objects in a three dimensional space. We evaluate our approach upon a real world application, to cluster particle simulations composing such shapes. Our approach outperforms comparable methods for this application, both in terms of effectiveness and efficiency. Additionally, we discuss how the adaptation enables further analysis capabilities. For instance, in the presented use case, the introduced approach allows to additionally analyze clusters throughout the depth of the water. So far, this is not feasible with existing approaches.

Interdisciplinary research projects aim to answer overarching questions combining research of multiple disciplines. Besides FAIR requirements this requires effective Research Data Management (RDM) to support the synthesis necessary to provide such desired answers and involves integrating data from various disciplines. Research projects benefit from early decision-making on core aspects for organization and management of data. We propose a framework to identify such aspects which require early decisions for successful planning and execution. Following up on our activities and experiences in the interdisciplinary research project CRC 1266 (Scales of Transformation), we propose an approach that focuses on reliability of external repositories and combines it with a project-specific wrapper-like infrastructure consisting of roles, processes and a software system deeply integrated into the research project for FAIR and enriched research data management, presentation and synthesis.

Mining spatio-temporal correlation patterns for traffic prediction is a well-studied field. However, most approaches are based on the assumption of the availability of and accessibility to a sufficiently dense data source, which is rather the rare case in reality. Traffic sensors in road networks are generally highly sparse in their distribution: fleet-based traffic sensing is sparse in space but also sparse in time. There are also other traffic application, besides road traffic, like moving objects in the marine space, where observations are sparsely and arbitrarily distributed in space. In this paper, we tackle the problem of traffic prediction on sparse and spatially irregular and non-deterministic traffic observations. We draw a border between imputations and this work as we consider high sparsity rates and no fixed sensor locations. We advance correlation mining methods with a Sparse Unstructured Spatio Temporal Reconstruction (SUSTeR) framework that reconstructs traffic states from sparse non-stationary observations. For the prediction the framework creates a hidden context traffic state which is enriched in a residual fashion with each observation. Such an assimilated hidden traffic state can be used by existing traffic prediction methods to predict future traffic states. We query these states with query locations from the spatial domain.

A special focus in data mining is to identify agglomerations of data points in spatial or spatio-temporal databases. Multiple applications have been presented to make use of such clustering algorithms. However, applications exist, where not only dense areas have to be identified, but also requirements regarding the correlation of the cluster to a specific shape must be met, i.e. circles. This is the case for eddy detection in marine science, where eddies are not only specified by their density, but also their circular-shaped rotation. Traditional clustering algorithms lack the ability to take such aspects into account.

Research papers are often the primary source of scientific information dissemination, as researchers encapsulate their findings in these documents. Generally such findings are of complex types, diverse expressions and also carry rich context. The traditional approach for extracting certain scientific information from these documents is manual extraction, which is very time consuming. Due to the rapid increase in number of publications, using the full potential of these rich data sources by manual extraction is becoming infeasible. In this paper, we propose a framework for the automatic extraction of targeted (user defined) quantitative information, e.g. temperature sensor values, with its geo-spatial context from scientific documents. Given a database of scientific documents and a targeted user-defined geo-tagable measurement variables, mass accumulation rate (MAR) and sedimentation rate (SR), the problem we are addressing is to retrieve all the values together with their geo-spatial information respectively. Though there has been done a lot in information retrieval, to the best of our knowledge, this problem has not been explored, yet. We design a novel heterogeneous linking solution, that links measurements with locations, which are found by our tailored extraction pipeline. In experimental studies based on our novel dataset of Marine Geology papers, we showcase the capabilities of our linking framework using common geo-tagable Marine Geology measurements.

The ocean is an important part of the global system. Tracking the connectivity between bodies of water is crucial for understanding local, regional and global changes in the ocean dynamics that mediate the spreading of nutrients and influence the marine ecosystem and ocean productivity. We developed a Data Fusion approach that enhances and automates the existing methods for the analysis of this connectivity. This approach combines and condenses two different data sources in two stages, Data Enhancement followed by Data Reduction. The Data Enhancement stage fuses equidistantly gridded data containing physical measurements and trajectories representing movement data. The Data Reduction stage aggregates the fused data into a Markov Model representation of the transition probabilities between ocean regions. We applied this framework to an exemplary analysis for the connectivity between two oceanic areas using real ocean data stemming from marine research. We show that this method directly tackles the limitations of existing marine data analysis methods and furthermore introduces new means to answer questions that had no quantitative answers up to now.