Deep moist convection and the associated rainfall frequently occur on global coasts, especially on the windward sides of coastal orography in monsoon regions, playing a vital role in energy and hydrologic cycles Xie et al., 2006). They threaten the public safety and property with flash floods, high winds, lightning, hail, and occasionally tornadoes. The genesis of convection, namely convection initiation (CI), is one of the most important issues in quantitative precipitation forecast (QPF) because rainfall strongly depends on the timing and location of CI. The improved understanding and prediction of CI in coastal areas are thus integral parts of the disaster prevention and mitigation for severe weather.CI has long been known to be a result of the interaction of multiscale atmospheric processes which contributes to the locally enhanced lifting and moisture pooling. Here, the moisture pooling indicates the development of an area in the planetary boundary layer (PBL) where moisture values become higher than in the surrounding region (American Meteorological Society, 2021). The synoptic-scale circulation often primes the mesoscale environment for CI by large-scale mean ascent, moistening the atmosphere and thus increasing the convective instability (Markowsk & Richardson, 2010). Local convergent zones in the PBL usually play a crucial role in triggering convection primarily by the forced lifting. The common enhanced PBL lifting is often located on mesoscale boundaries, such as gust fronts, drylines, surface wind shift lines, sea-breeze fronts, gravity waves, and horizontal convective rolls (e.g., Saito et al., 2018; Wilson & Roberts, 2006, and references therein). These boundaries can be identified by a variety of observational platforms and are carefully assessed by forecasters to help the CI nowcasting.In addition to the PBL convergent boundary, orography is also a key factor for triggering convection. The resultant convection often repeatedly initiates in the region with some particular topographical features rather than more or less random as in flat areas (Banta & Schaaf, 1987). The orography-related dynamics such as