Probabilistic seismic hazard assessment (PSHA) is a widely used statistical approach to estimating where and when earthquakes are likely to occur based on the statistics of past seismicity patterns. Key parameters like the magnitude-frequency distribution (MFD) and the b-value are integral to hazard forecasting as they express the proportion of small to large earthquakes in a catalog. However, these statistical parameters are heavily influenced by the accuracy of earthquake magnitude estimates. This research addresses the critical need for high-quality magnitude measurements for small earthquakes by using relative amplitude methods. I then use these improved magnitude estimates to examine spatiotemporal variations in b-value for multiple earthquake sequences to improve our understanding of short-term seismic hazard forecasting. I first introduce a generalized methodology to determine relative magnitudes for earthquake sequences which is only dependent on relative amplitude differences between interlinked pairs of waveforms, as well as methods for determining the b-value from the distribution of magnitude differences between successive events. In chapter 3, I investigate the uncertainty of magnitude results produced from the relative magnitude method by conducting a parameter study on critical variables including thresholds for signal-to-noise ratio and cross-correlation, frequency content filtering, and seismic station selection. I show that signal-to-noise and cross-correlation thresholds limit the number of magnitudes that can be recalculated while bandpass filtering has the largest effect on the variability of magnitude results. In chapter 4, I develop a set of coda-envelope moment magnitudes (MW) as a benchmark data set for the relative magnitude method, allowing us to align our relative magnitude measurements to an absolute moment magnitude scale for small earthquakes. I produce moment magnitudes for approximately 80% of the events in the Delaware Basin and demonstrate the capabilities of this method to provide moment magnitude for small earthquakes in regional earthquake catalogs. In chapter 5, I use an uncalibrated relative magnitude method to reevaluate magnitude estimates for the 2011 Prague, Oklahoma earthquake sequence and calculate the temporal and spatial variations of b-value. I show that b-values during the aftershock sequence are consistently low which demonstrate that the aftershock distribution is skewed towards producing earthquakes of higher magnitude for at least 5 months following the mainshock. Additionally, we show a trend of decreasing b-value along the Meeker-Prague fault as distance from the mainshock increases suggesting that tectonic stress may still exist in areas of low b-value. Finally in chapter 6, I apply the relative magnitude method to 6 foreshock sequences in southern California and focus on an in-depth exploration of the spatial and temporal variations in b-value and their sensitivity to parameters such as spatial binning and window length. I show that approximately half of the sequences exhibit a drop in b-value in the months or days prior to a mainshock. I also show that mainshocks frequently occur in areas of low foreshock b-value for single-fault or dense seismicity. This research demonstrates the importance of reliable and transportable magnitude estimation for small earthquakes. With these improved magnitude estimates, we also gain valuable insights into the behavior of seismic sequences through analysis of the spatiotemporal variability of the MFD and b-value.