Microseismic reconstruction of dominant hydraulic-fracture surfaces in a deep shale reservoir using two-stage DBSCAN and magnitude-weighted plane fitting

Passive microseismic monitoring provides a non-intrusive basis for reconstructing hydraulic-fracture geometry in deep shale reservoirs. However, event catalogs commonly contain low-density outliers and noise-like detections, while reconstructed geometry is sensitive to clustering parameters, event weighting, and boundary delineation. This study develops a sequential workflow combining two-stage DBSCAN for event denoising and clustering, magnitude-informed weighted least-squares plane fitting for fracture-orientation estimation, and plane-based two-dimensional alpha-shape boundary extraction. A silhouette-score-guided strategy supports DBSCAN parameter selection. The workflow is applied to a field microseismic catalog acquired during hydraulic stimulation of a deep shale gas reservoir in the southern Sichuan Basin, China. Three dominant fracture surfaces are reconstructed, with spatial distributions consistent with the clustered microseismic activity. Comparison with focal-mechanism interpretations indicates partial orientation consistency between the reconstructed surfaces and observed source characteristics. Synthetic microseismic catalogs and parameter-sensitivity analyses further evaluate reconstruction stability and limitations. The reconstructed surfaces provide quantitative geometric constraints for fracture interpretation, reservoir-stimulation evaluation, and engineering-geological assessment in deep unconventional reservoirs.