Natural earthquakes occur on faults ranging from 0 to 700 km beneath Earth's surface in different tectonic settings, such as along major subduction zones in Japan and the arc-continent collisional environment in Taiwan. Recent studies suggest that earthquake activities can be affected by various Earth processes, including extreme weather events, large earthquakes, water/snow/glacier loading and unloading, erosion and sedimentation, etc. The Gutenberg–Richter magnitude-frequency statistics suggest that the number of earthquakes decays as a power law with the increase of earthquake magnitude, which means most earthquakes are of small magnitudes, i.e., microseismicity. Studying the behavior of microseismicity and their response to the Earth’s surface process can help us to better understand fault structures at depth as well as the physics of earthquake nucleation, and to mitigate seismic hazards. However, our understanding of microseismicity may be limited by the incompleteness of standard earthquake catalogs, especially during the noisy period following extreme weather events and large earthquakes. During my Ph.D. study, I develop machine-learning and template-matching tools to improve earthquake catalogs by detecting microearthquakes and calculating their focal mechanisms. Based on the improved high-resolution catalogs, I then perform a detailed analysis of the microseismicity behavior and their response to Earth processes. Specifically, I build a deep-learning network to automatically determine P-wave first-motion polarity, which can be used for calculating earthquake focal mechanisms. Next, I use template-matching and deep-learning methods to build a more complete earthquake catalog in Taiwan before and after the 2009 typhoon Morakot. I observe no other significant seismicity changes that can be attributed to surface changes induced by typhoon Morakot, but a clear reduction in seismicity rate near the typhoon’s low-pressure eye center in northeastern Taiwan during the typhoon passed by. Last, I perform a systematic detection of intermediate-depth earthquakes (IDEQs) in the Japan subduction zone using the template-matching technique. I obtain a more complete IDEQ catalog before and after the 2011 magnitude (M) 9 Tohoku-Oki earthquake (TOEQ). The newly built template-matching catalog does not show any significant increase in IDEQs in the two months prior to TOEQ. But following the TOEQ, I find a significant increase in the rate of IDEQs in both upper and lower planes of the double seismic zone beneath 70 km depth. These results suggest that like seismic activity at shallow depth, IDEQs in the double seismic zone also respond to stress perturbations generated by the 2011 M9 TOEQ, highlighting a a sustained seismic hazard associated with these intraslab events in the next decades.