Conference Agenda

Issues with “bad” analyses have long been observed by researchers working with geochemical compilations. Although many of these researchers perform “data cleaning” steps, few address the more fundamental questions: Why do errors occur, and how can we pinpoint their type and source?

Rare Earth Element (REE) pattern smoothness (excl. Eu and Ce) is commonly used to check the quality of a rock or mineral sample. Our assessment REE pattern smoothness (excluding Eu and Ce) in two large public databases (GEOROC and PetDB) reveals that the majority of irregular patterns originate from data handling errors in spreadsheets, rather than inherent analytical problems. Frequent issues include column/row misalignments, rounding, normalization inaccuracies, and unit scale discrepancies.

We have developed a Machine Learning (ML) framework using REE patterns that predicts if such errors are present and the most likely cause. The training dataset contains all originally smooth REE data from GEOROC and PetDB. We then simulate the different data management errors to use as prediction labels. Pairwise log-ratios are used as features to capture both the shape and zigzagginess of the REE pattern. The base ML model is a stratified (based on material classification) RandomForest and outlier detection pipeline. Our model is able to predict the labels with more than 90% accuracy on the testing dataset and can be applied on any geochemical rock, mineral and inclusion dataset compilation that contain REE data. It can thus be used to improve the publication quality and reuse potential of geochemical datasets.

Physical samples represent the origin for geochemical analyses and are often hard to track across publications. One reason is that they usually change their names when they move through institutions and laboratories, another that samples are highly granular and often not supported by digital data curation systems. The International Generic Sample Number (IGSN) is a globally unique and persistent identifier for physical samples and collections with discovery function in the internet. IGSNs connect the physical object (the sample) with an online sample description that follows a specific metadata schema which also includes the current location of the sample. As such, IGSNs enable to directly link scholarly publications, data and instruments with the samples they originate from. A sample assigned with IGSN may be found and reanalysed when new evidence and analytical techniques can open new lines of inquiry. Moreover, IGSNs ensure traceability and promote transparent data provenance across laboratories, institutions and scientific disciplines.

The challenges for IGSNs are manifold: Firstly, samples are highly granular; secondly, researchers working with samples are often neither metadata specialists nor dependent on databases and other digital workflows, Thirdly, while IGSNs can be cited in scholarly publications, the high number of samples contributing to publications, the inclusion of sample descriptions in references lists is very challenging. This presentation presents the benefits of IGSNs and also introduces the FAIR SAMPLES template that aids researchers to provide rich and customized descriptions for IGSN metadata.

The EC GSEU project is building a pan-European geological framework which encompasses a metadata system, a pan-European data model, methods to visualize 3-D models, and hierarchical, machine-readable vocabularies on anthropogenic deposits, lithotectonic units and lithology. The vocabularies are set up as SKOS-readable Excel sheets to be made available at the European Geological Data Infrastructure (EGDI) of EuroGeoSurveys. They are mostly based on existing standards such as the INSPIRE Directive Data Specifications, the OneGeology-Europe terminology or the CGI GeoScienceML vocabularies. The aim is the application to digital geological maps, assigning a term and description to a rock unit mapped in the field and to aid semantic harmonisation across political borders in Europe.

The lithological concepts are organised, to a large extent, in a systematic, scientifically sound but also pragmatic nomenclature and take into account the mapping geologist’s rock identification in the field.

Igneous rocks are e.g. classified geochemically mainly either according to the ratios of quartz (Q), alkali feldspars (A), plagioclase feldspars(P), and feldspathoids/foids (F) (Le Maitre et al. 2002), to the TAS classification (Total-Alkali-Silica) or simply according to its observable silica content (basic, intermediate, acidic) for a first identification in the field.

Correspondingly, the sedimentary rocks are classified according to their chemistry (e.g. carbonate, silica, phosphate, carbonate), too, but also according to other characteristics such as grain size (Wentworth, 1922).

This presentation will describe both the scientific and pragmatic approaches to the GSEU project lithology classification and will present test use cases of its application.

Water quantity and quality are increasingly stressed by climate change and human activities. Intensifying droughts and growing water use reduce river discharge, and in turn, the natural dilution capacity. Among human pressures, resource extraction, especially past and ongoing mining, has a pervasive footprint, altering regional water cycles and degrading groundwater and surface-water quality with large, long-lasting effects. In the Ruhr area, centuries of coal mining necessitate continuous pumping of groundwater with higher salinity, which elevates river salinity and offers a compelling case to quantify mining return flows and their contribution to river water quality.

The forever task to pump mining water into the river will influence the river water salinity in the Ruhr area in the future and may have an impact on regional aquatic ecosystems. We use the large-scale modelling framework WaterGAP3 to simulate how mining return flows influence river salinity in the Ruhr area. We further explore how projected climate change will modify river discharge and, consequently, salinity concentrations, considering plausible hydrological and management scenarios. The study advances process understanding of mining-related return flows at regional scale, provides scenario-based projections of salinity under combined mining and climate pressures, and informs water-quality management in heavily industrialized river systems.