B-QualMT - A software quality management tool to select borehole records for 3D models

Exploration data (e.g. borehole records, geophysical sections) represent the essential input data for any geological model. Borehole records often play a decisive role in geological modeling. Usually, they contain descriptions and interpretations on petrography, lithology and stratigraphy. This information is crucial for modeling the spatial distribution of lithostratigraphic layers in three dimensions. However, these interpretations can be inconsistent or error-prone. The reasons include for instance, the date of recording (reflecting the prevailing state of knowledge at the time), the specific exploration target and techniques employed, the quality of digitalization, and the potential for human interpretative bias. Separating adequate from unsuitable borehole records is of great importance, yet rather difficult, especially evaluating large datasets. While visual inspection of the inferred geological model is a viable approach, it results in numerous iterations to identify inadequate drilling profiles, which is time-consuming and expensive.
               In order to streamline the process of testing the quality of the data from borehole records, we developed the Python-based software package B-QualMT (borehole quality management tool) that can filter borehole records based on user-adjustable standards. The tool has a given set of deterministic tests depending on the user’s auxiliary information (e.g. previous 3D-models) and knowledge of the regional geological settings (e.g. sequence of geological layers), which can be used to select divergent drilling profiles for geologically comparable regions.
               For our pilot study, we selected a former lignite mining area of Lusatia in the southeast of Germany bordering the Federal State of Brandenburg and the Free State of Saxony (Freistaat Sachsen). Here, the goal is to improve the previous geological model with 3000 additional borehole profiles from various exploration surveys spanning several decades. We show the evaluation process, how the deterministic tests work and will additionally give an outlook on the planned integration of machine learning algorithms identifying geological patterns in previously quality-tested borehole records.