Microearthquakes have come into high public awareness due to being induced by the development and exploitation of enhanced and natural geothermal fields, hydrofracturing, and CO 2 sequestration sites. Characterizing and understanding the faulting process of induced earthquakes, which is generally achieved through moment tensor inversion, could both help in risk prediction and in reservoir development monitoring. However, this is a challenging task because of their lower signal-to-noise ratio at frequencies typically used in earthquake source analyses. Therefore, higher-resolution velocity models and modeling of seismic waves at higher frequencies are required. In this study, we examine both the potentials to obtain moment tensor solutions for small earthquakes and the uncertainty of those solutions. We utilize a short-period seismic network located in the Geysers geothermal field in northern California and limit our study to that which would be achieved by industry in a typical reservoir environment. We obtain full moment tensor solutions of M~3 earthquakes using waveform modeling and first-motion inversions. We find that these two data sets give complimentary but yet different solutions. Some earthquakes correspond possibly to complex processes in which both shear and tensile failures occur simultaneously or sequentially. This illuminates the presence of fluids at depth and their role for the generation of these small-magnitude earthquakes. Finally, since first motions are routinely obtained for all magnitude earthquakes, our approach could be extended to small earthquakes where noise level and complex Green's functions prohibit using waveforms in moment tensor inversions.