Semi-automated porosity mapping in carbonate reservoirs using borehole images

Reservoir quality in carbonate systems is commonly controlled by secondary porosity associated with fractures and karst. Accurate characterisation of these features is critical for predicting fluid storage and permeability distribution, yet remains challenging using conventional downhole geophysical logging techniques. Interpretation is typically performed manually using resistivity borehole images (BHIs), which resolve rock texture at millimetre scale. However, this approach is time-consuming and yields only a limited and largely qualitative representation of the true porosity distribution, as only a small number of features can be mapped.

To obtain a more comprehensive picture of the macroscopic porosity distribution, we developed a semi-automated workflow for high-resolution pore space mapping and classification in BHIs. We focus on greyscale-converted images rather than raw resistivity data because they are more commonly available for legacy wells. Our workflow applies simple thresholding to generate binary porosity maps from both static (linear conversion of resistivity to brightness) and dynamic images (with histogram equalisation). The dynamic map is grafted onto the static map in areas identified as dark or bright in the blurred static image, resulting in a millimetre-scale porosity map of the borehole wall. Following interpolation between the imager pads and/or flaps, geometric properties are extracted for each connected cluster of mapped pixels, allowing classification of pore types as fractures, vugs, or karst features. The workflow performs well in limestone and dolostone sequences with high resistivity contrast between matrix and pore space, but is less reliable in marly intervals and in sections affected by poor borehole or data quality. We are currently developing an updated, fully automated workflow leveraging machine learning algorithms for pore space segmentation and classification.

As a first application, we analysed BHIs from the North Alpine Foreland Basin in Bavaria, where the Upper Jurassic hosts a hydrothermal reservoir. Of the 16 good-quality BHIs analysed, visual inspection indicates that 13 produced reliable results. By combining the derived macroscopic porosity with available matrix porosity measurements, total porosity can be estimated along the well path. Integration with additional well data enables us to define porosity–permeability trends for active flow zones, elucidate controls on pore space distribution, and derive realistic porosity ranges for reservoir model parameterisation.