Regional seismic networks typically monitor and locate seismic sources using only kinematic (i.e., traveltime) observations. We propose an automated “sliding box” procedure to improve source locations by leveraging dynamic (e.g., full waveform) information via Geometric-mean Reverse-Time Migration (GmRTM). We demonstrate the proposed method’s efficacy by selecting and relocating some of the cataloged tectonic earthquakes recorded by a network of ∼250 three-component seismic sensors deployed across 8,000 km2 of mountainous terrain in Central Asia. Each event is recorded with sufficient signal-to-noise ratio (≥ 4) by at least 8 sensors inside a 20 km × 20 km box centered on the cataloged location. For each event, we extract a local 3-D velocity model, multiply the waveform data by their corresponding STA/LTA ratios to suppress coda waves and other coherent noise, and apply GmRTM to reconstruct high-resolution source images. The automatic procedure reasonably relocates 280 earthquakes.