Book excerpt: With the abundance of global seismic waveform data across a broad range of temporal and spatial scales, ample opportunities exist to investigate seismic source and propagation effects. In this dissertation, I present results from three inductive, observation-driven analyses that exploit local-to-teleseismic distance seismic observations to explore seismic source characteristics across a broad range of spatial scales. Each investigation emphasized one or more seismological challenges that remain despite advances in global instrumentation, source characterization approaches, and seismic wave propagation modeling. For the first investigation, we exploited local-distance P- and S-wave observations generated by mining-related and small-magnitude events at a gold mine in South Africa to explore the robustness of P-to-S-wave amplitude ratios. P/S amplitude ratios are traditionally used in discrimination studies between earthquakes and explosions recorded at regional and teleseismic distances (greater than 200km). Fewer studies have explored the variability of P/S amplitude ratios using data recorded at local distances, distances less than 200 km, where more scrutiny of wave propagation, near-surface geology, and source and strain release patterns are required. We took advantage of the dense seismic network at the Klerksdorp mine to investigate the variability of low-yield earthquake and mining-related event P/S amplitude ratios at local distances. Our results showed that most locally recorded low-magnitude events in the Klerksdorp region have comparable shear-wave energy to low-magnitude earthquakes. Our time-domain rms-based measurement of P and S amplitudes resulted in stable event average P/S ratios that are likely separate from explosive sources. We used the observations to demonstrate the expected variability of the ratios with smaller seismic networks (3-, 5-, 7-station) to show the amplitude ratios remained relatively stable across a bandwidth of 1-30 Hz, but ratio variability decreased with increasing station number. In the second investigation, we conducted an earthquake relocation study of moderate-to-large magnitude events along the Southwest Indian Ridge (SWIR) from 1990 to 2022 using global surface wave observations. Earthquake locations are essential parameters used in seismology to investigate earthquake processes, tectonics, and the subsurface. Although seismologists have constrained the location of an earthquake reasonably well using arrival-time measurements, important tectonic regions, such as mid-ocean ridges (MOR), remain less well-constrained because of the lack of nearby seismic stations. We leveraged an existing surface-wave relative relocation approach to estimate precise relative locations along the SWIR, investigate updated seismicity patterns, and relate those patterns to tectonic processes occurring along the ultra-slow-to-slow spreading ridge system. To handle the large data set from a broad region we used an artificial neural network to identify the highest-quality observations and reduce the computational burden associated with cross-correlating millions of surface waveforms. Our results show a 3-second reduction in location misfit and precise locations well within 5 km of bathymetric features demonstrating the ability to use global surface-wave observations to improve the precision of earthquake locations in a remote region. We observed (with greater clarity) many expected general tectonic patterns including single- and multiple strands active along transform faults, the gradual transition from normal to strike-slip faulting along ridge-transform intersections, clustering of normal-faulting along the ridge segments (including doublet and multiplet sequences), occasional migration of events along transforms and ridge segments, and clustering of events around ridge volcanoes or regions of unusual bathymetry. For the final investigation, we utilized abundant mining-related seismic events that occur each year in Pennsylvania and the wealth of local to regional seismic networks to evaluate short-period seismic-wave propagation across the region. Pennsylvania hosts many industrial seismic events and experiences a small number of naturally-occurring earthquakes, but industrial seismic sources are currently excluded in the generation of ground motion prediction equations. We performed a linear regression of peak ground-velocity measurements for industrial explosions in Pennsylvania. We examined the data across a frequency band from 1 to 16 Hz, but the best results with the most observations were those from the 1-2 Hz band. The results show that the small-magnitude explosion data are unable to resolve differences in simple models of the decrease in amplitude with distance. But the results demonstrate that applying the geometric spreading corrections used for magnitude estimates across the eastern US is reasonable for the short distances used to estimate the size of small mining events across the region. Regression produced a set of relative source sizes that exhibit a complicated relationship with locally estimated magnitudes. The difference between regression estimated source size and traditional analyst-based magnitudes correlates with local magnitude. Careful examination of the data suggested that observations from greater distance are hard to screen using signal-to-noise measurements and simple magnitude-dependent distance cutoff relationships. Thus, the traditional approach to magnitude estimation that relies on human or AI-expert-trained systems to separate signal and noise remains important, especially for the smaller events.