In recent years, extreme rainfall events and attendant flooding continue to pose significant threats to life and property in China. For example, on July 19-20, 2016, an extensive extreme rainfall event swept across North China, causing 139 human fatalities, 141,000 house collapses, and 718,000 ha of crops damages over the region, leading to a direct economic loss of 10 billion RMB (Xia & Zhang, 2019). More severely, the "21•7" extreme rainfall in Henan caused devastating floods, with a record-breaking hourly rainfall rate of 201.9 mm hr −1 , leaving 398 human fatalities and a direct economic loss of 120.06 billion RMB (Luo et al., 2022;Su et al., 2021;Wei et al., 2022;Yin et al., 2022), which has brought unprecedented attention to extreme rainfall events. Early warning and forecasting are always a crucial part of disaster prevention and mitigation. However, the current meteorological forecasting system is still struggling to accurately predict the location and intensity of extreme precipitation (Fritsch & Carbone, 2004;Novak et al., 2011;Zhang et al., 2016), which urgently requires a deeper understanding of the mechanisms of extreme rainfall events.Previous studies have pointed out that sufficient moisture supply, favorable uplifting, and unstable atmospheric environments are generally the key factors that are conducive to heavy rainfall (Doswell et al., 1996;Lin et al., 2001). Subsequent studies revealed that mesoscale convective systems (MCS) play an important role in extreme rainfall events. Johnson (2005, 2006) found that about 65% of extreme rainfall events in the central and eastern United States were directly caused by MCS (weak synoptic-scale forcing), 27% were caused by synoptic-scale forcing (e.g., frontal, or extratropical cyclone), and the other 8% are caused by tropical cyclones and their remnants. Of importance is the organizational structure of MCS, which primarily controls the intensity, duration, and location of extreme rainfall. The so-called "back-building" and "echo-training" processes are the two major patterns. The back-building process occurs when convective cells repeatedly develop upstream of previous deep convection and pass over the same region (Schumacher & Johnson, 2005, 2009Zheng et al., 2013), forming a quasi-stationary MCS, while the echo training process occurs when a series of convective Abstract Extremely persistent heavy rainfall (EPHR) events frequently occur in southern China, but its mechanisms of initiation and maintenance are not well understood, leading to difficulties in accurately forecasting. Thus, we investigated a typical EPHR case that occurred on Hainan Island on 14 December 2013 (UTC) using convection-resolving WRF simulation and multi-source observations. Results revealed that the intrusion of the cold surge, the formation of the quasi-stationary mesoscale perturbed vortex (MPV) caused by deflected southerly winds and northeasterly winds, and local topographic forcing are three crucial factors for this EPHR event. Firstly, the invasion of cold surge provides a ...