The debate on self-similarity of source physics between small and large earthquakes has been high on the agenda in earthquake science for decades. Seismic source parameters, such as corner frequency, seismic moment, and stress drop, to name only a few, have been analyzed and compared across the entire earthquake magnitude range, considering large natural events, induced seismicity, and laboratory acoustic emissions. Stress drop is an important source parameter that substantially affects ground motions and is often used as an earthquake scaling parameter. The majority of stress drop studies show on average independence on earthquake size and thus scale-invariance is assumed. Scaling breakdown has, however, also been reported for individual datasets in which stress drop scales with earthquake magnitude. Regardless of the debate on self-similarity, a large global stress drop scatter of 0.01-100 MPa has been observed throughout all earthquake sizes, which is not yet fully understood. The reasons for this scatter need to be investigated in more detail to better understand the similarities and differences between earthquake source characteristics and the corresponding physical processes in the earthquake source. This doctoral thesis focuses on small earthquakes, including induced seismicity and laboratory acoustic emission events, and addresses two groups of factors whose effects are assumed to cause scatter in stress drop estimates: 1) Non-physical factors that cover error propagation, and inappropriate assumptions made during the analysis of seismic data, can lead to bias in source parameter estimates. 2) Physical factors, which reflect the actual properties of earthquakes foci and rupture processes, are assumed to naturally influence the resulting static stress drop. In this thesis, non-physical effects on stress drop estimates are suppressed by considering only high-quality data, adequate methodologies, and carefully evaluated and selected parameters for analysis. One aspect that this thesis studies in more detail, is the effect of high-frequency wave attenuation on seismic records. Attenuation is expressed as the quality factor Q, which describes the quality of the medium that is in general unknown. Q is important for the correction of amplitude source spectra while estimating source parameters. Incorrect assumptions of Q can bias the high-frequency spectral fall-off leading to error propagation (a non-physical effect) during further data analysis. The first study analyzes the time-dependent decay of S-coda waves of induced seismicity at The Geysers geothermal field, California, to better assess attenuation effects for high-frequency seismic signals. The application of the moving window method enables us to estimate stable coda quality factors (QC) for pre-defined frequency bands in the frequency domain. An additional sensitivity analysis is conducted beforehand to evaluate the impact of different parameters used in the coda technique, which are the magnitude range, signal-to-noise ratio, maximum permitted uncertainties of QC estimates, moving window width, lapse time effect, and total coda length. Findings of the sensitivity analysis emphasize the relevance of parameterization and the usage of high quality data. Waveforms of in total 717 shallow earthquakes with duration magnitudes 1 < MD < 3 of two spatially separated locations in the northwestern and southeastern part of The Geysers were investigated. Both areas show clear differences in attenuation properties dependent on locally varying geological, structural and geothermal characteristics. The coda quality factor QC was estimated for frequencies up to 70 Hz, exceeding previous field data considerations. QC obtained from the northwestern dataset was also investigated in the context of temporal injection variations. The temporal stability of QC suggests that either the local fault network is on average constant over time or the injection-induced fractures are not detectable with the frequencies considered. Additionally, QC from the northwestern dataset is compared to the quality factor obtained by direct S-waves (Qb ). QC shows on average higher stability in mean estimates compared to Qb . Therefore, it is suggested that QC leads to fewer uncertainties of further estimated source parameters such as stress drop. The second study focuses on acoustic emission (AE) events of two laboratory triaxial stick-slip experiments on oven-dried Westerly granite samples. Source parameters, in particular static stress drop, of acoustic emission events from a rough and a smooth fault were analyzed taking advantage of the spectral ratio method based on a multi-eGf (empirical Green’s function) approach. Here, attenuation effects (path- and site effects) of linked, co-located events were suppressed and source parameters were directly obtained from the source spectra using the quasi-dynamic Madariaga source model. For the first time, a span of more than three orders of magnitude (-9 < MW < -5.6) was evaluated for laboratory AE events providing a reasonable base for scaling analysis. Obtained AE stress drops are comparable to the globally observed estimates. However, a clear scaling breakdown was observed for both faults putting the global self-similarity assumption into question. Here, the physical aspects of fault surface roughness, source radius, and rupture velocity were scrutinized. AE stress drop shows mainly no dependence on source size and only very little variation with rupture velocity changes. The observed stress drop–magnitude scaling therefore suggests differences of slip over rough and smooth fault surfaces. The complexity of rough faults might inhibit larger slips leading to lower stress drops, whereas smooth faults can result in larger slips and thus higher stress drops due to fault simplicity. These two studies provide new insights into the analysis of high-frequency seismic signals. More comparative studies based on similar approaches are necessary for further earthquake cross-scale investigations. The variety of contemporary analytical methods and models render the comparison of studies challenging. Therefore, data- and interdisciplinary expertise exchange, data standardization, careful data quality evaluation, and preservation of methodological applications are needed to facilitate and accelerate future research. Drawing on these crucial requirements for future investigations, the third part of this thesis complements the first two scientific studies. It shows the innovative development of an online research platform headed by the Thematic Core Service Anthropogenic Hazards (TCS AH) within the framework of the Implementation Phase of the European Plate Observing System (EPOS-IP). The TCS AH team introduces the genesis of the novel online platform IS-EPOS (IS = Induced-Seismicity) that consists of a large number of complete, comprehensive datasets and applications related to the exploration and exploitation of georesources and the underground storage of liquids and gases. The pioneering work of the EPOS project provides an example of how enhanced research could deal with the similarities and differences between individual studies, and thus lead to a more comprehensive understanding of seismic data and its sources.