guest ::
TY - THES AU - Boemke, Bruno Johannes TI - Unlocking pattern extraction from geospatial big data - methodological innovations in remote sensing and geospatial analysis for environmental and geoarchaeological research PB - Rheinisch-Westfälische Technische Hochschule Aachen VL - Dissertation CY - Aachen M1 - RWTH-2024-00528 SP - 1 Online-Ressource : Illustrationen PY - 2024 N1 - Veröffentlicht auf dem Publikationsserver der RWTH Aachen University N1 - Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2024 AB - In recent decades, the amount of openly available geodata has increased exponentially. To a large part, this can be attributed to the technological progress in satellite remote sensing, producing world-wide coverage from a large variety of sensors on a sub-daily basis. In addition, improved accessibility and a growing user base have favoured methodological advances in geospatial analysis as well as an increase in the number of derived products such as, e.g., land cover classifications. While this opens up new opportunities for geospatial applications, the vast amounts of geodata also challenge studies in terms of selection, filtering, harmonizing, processing and interpretation. To fully utilize the opportunities that the increasing amount of geodata offers while tackling its challenges, new and innovative methodologies for different geospatial applications are needed. This cumulative dissertation is a contribution to this goal in the fields of environmental science and geoarchaeology. It presents three novel and experimental approaches on how to effectively utilize a large variety and/or long time series of geodatasets and analyse them sensibly within the framework of a specific research question. The first approach explores the possibilities and limitations of assessing complex aeolian dune field morphology and evolution using synthetic aperture radar (SAR) satellite data. This study relies on the Sentinel-1 mission, which acquires data in volumes of approximately 600 gigabytes per day. To analyse this continuous stream of geospatial big data, the study examines the key interaction mechanism between C-Band radar and sand dunes and introduces a visual pattern extraction method based on continuous wavelet transfer. This novel method is applied to the Western Mongolian dune field Bor Khyar. The results give new insights into the temporal and spatial dynamics of dune scales and their response to aeolian activity, revealing local differences as well as inter- and intra-annual variations in the dune morphology. The second approach is a methodological contribution to the field of archaeological predictive modelling. The main challenge of this study is the extraction of a thematic pattern from a small sample of 23 available Upper Palaeolithic sites in Lower Austria. This is achieved using a novel approach combining a classical deductive method with the capabilities of machine learning. This way, ten spatial predictors representing morphological, hydrological, and sedimentological factors of the paleo-environment are analysed for optimal, viable, and non-viable value ranges and combined mathematically. The resulting predictive model reveals several spatial dynamics of site probability and shows high compliance with known sites in the study area. In stark contrast to this study, which is challenged by the small number of available sites, the third study conducts geoarchaeological pattern extraction based on a substantially bigger dataset of close to 4200 European Upper and Final Palaeolithic sites. The main aim of this study is to explore whether the site distribution is representative of human distribution in the paleo-landscape or if sampling biases obscure this information. To this goal, eight Pan-European geodatasets representing both settlement-relevant factors of the paleo-environment and discovery-relevant biases of the modern to contemporary landscape are analysed using a combination of geospatial and geostatistical methods. The results show that the actual distribution of sites seems to be most strongly influenced by sampling biases. The influence of the settlement factor, however, is still significant when comparing site subsets from different regions, different Upper Palaeolithic periods, and, especially, between open-air and cave sites. The implications of this study are substantial for geoarchaeological approaches, as the sampling bias is often overlooked or underestimated as a factor actively influencing the distribution of known sites. All three approaches present a novel methodological approach in their respective field of study and outline a workflow that can be adapted and built on. For the availability to a broader audience, all studies are published as open access. In addition, the results of both geoarchaeological approaches are distributed as open data in universally usable geodata formats. As such, they can serve as foundation and inspiration for many future studies that utilize geospatial big data for environmental and geoarchaeological research. LB - PUB:(DE-HGF)11 DO - DOI:10.18154/RWTH-2024-00528 UR - https://publications.rwth-aachen.de/record/977017 ER -