A database-driven research data framework for integrating and processing high-dimensional geoscientific data

Abstract

This paper introduces a modular research data framework designed for geoscientific research across disciplinary boundaries. It is specifically designed to support small research projects, providing a bottom-up solution that empowers individual teams that need to adhere to strict data management requirements from funding bodies, but often lack the financial and human resources to do so. The framework supports the transformation of raw research data into scientific knowledge. It addresses critical challenges, such as the rapid increase in the volume, variety and complexity of geoscientific datasets, data heterogeneity, spatial complexity, and the need to comply with the FAIR (Findable, Accessible, Interoperable, and Reusable) principles. The framework uses a dual-component architecture. First, an Online Transaction Processing (OLTP) system features a user interface and a persistent relational database, ensuring accurate and consistent data storage when capturing and managing diverse geoscientific research data. Complementing this, an orchestration layer manages automated data pipelines to process the stored data and generate dynamic in-memory Online Analytical Processing (OLAP) databases that allow flexible, high-performance analysis. It is adaptable to evolving research requirements and supports various data types and methodological approaches, such as machine learning and deep learning, that place high demands on the data and their formats. A case study in Western Romania demonstrates the application of the data framework in an interdisciplinary geoarchaeological research project by processing and storing heterogeneous datasets, thereby reducing data management efforts, improving findability, replicability, and reproducibility, and streamlining the integration of high-dimensional data for small, interdisciplinary teams.