Abstract
<title>Abstract</title> <p>Active and passive seismic data can both be used to derive subsurface velocity models, but they differ in acquisition geometry and inversion strategy. Active data are generated by controlled sources with known locations and origin times, providing relatively homogeneous coverage within the survey area. In contrast, passive seismicity occurs over a larger region, with unknown source parameters that must be estimated during inversion. Passive events also tend to cluster along fault zones, resulting in spatially heterogeneous ray coverage. Combining both data types can improve seismic tomography by exploiting their complementary strengths: active data provide robust constraints on the velocity model and event locations, while passive data increase the illuminated volume. We present a methodology that integrates both data sets while explicitly accounting for the heterogeneous distribution of passive seismicity. An optimized selection of passive data enables both data types to be treated consistently during inversion without requiring additional data weighting. The proposed approach also provides guidance on the amount of passive data needed to maximize model information without overemphasizing densely sampled regions of the volume. The methodology is applied to a data set from the BedrettoLab rock laboratory in Ticino, Switzerland, where both active and passive seismic data provide excellent coverage of a controlled rock volume. We compare the optimized approach with random passive data selection, active-only tomography, and inversion using the complete data set. The optimized combined model increases the illuminated volume by approximately a factor of three compared with the active-only model while preserving the essential information contained in the full data set without biasing the inversion toward densely sampled regions. Comparison with independent geophysical observations supports the plausibility of the resulting velocity model. A residual analysis further suggests the presence of seismic anisotropy associated with the predominantly parallel fracture network at the BedrettoLab.</p>