Geomechanical modeling of hydraulic stimulations requires data to build and calibrate models to predict decline curves and other aspects of long-term reservoir performance. Initialization of these models requires knowledge of the preexisting fractures and geologic properties of the media. Orientations of different fracture sets, their intensities, and spacing, along with characterization of their size scales, critically impact geomechanical predictions of the stimulations in terms of proppant and fluid placement and the decline of reservoir productivity. With sufficient sampling of azimuths around the stimulation, the mechanisms and associated fracture planes and stress/strain conditions can be reconstructed through seismic-moment tensor inversion (SMTI) of recorded microseismic-event waveforms. At smaller scale lengths, signal-to-noise ratios can be low, and events might not be observed at high enough quality to permit SMTI. To extend the characterization of these fractures to smaller scales, a stochastic optimization algorithm is used, designed to search for optimally placed fractures in the reservoir that intersect with event locations while constraining their orientations from the same distribution observed at larger (SMTI-resolvable) length scales. Effectively, this technique allows for extension of the power law governing fracture distribution to smaller scales by invoking observed trends in self-similar behavior. In turn, characterization of the wider band of fractures in the reservoir provides necessary inputs into geomechanical models to predict fluid and proppant distributions from the full band of generated microseismicity and long-term behavior of the reservoir through decline-curve estimation.