Abstract. A common feature within coastal cities is small, urbanized watersheds where the time of concentration is short, leading to vulnerability to flash flooding during coastal storms that can also cause storm surge. While many recent studies have provided evidence of dependency in these two flood drivers for many coastal areas worldwide, few studies have investigated their co-occurrence locally in detail, nor the storm types that are involved. Here we present a bivariate statistical analysis framework with historical rainfall and storm surge and tropical cyclone (TC) and extratropical cyclone (ETC) track data, using New York City (NYC) as a midlatitude demonstration site where these storm types play different roles. In contrast to prior studies that focused on daily or longer durations of rain, we apply hourly data and study simultaneous drivers and lags between them. We quantify characteristics of compound flood drivers including their dependency, magnitude, lag time and joint return periods, separately for TCs, ETCs, non-cyclone associated events, and merged data from all events. We find TCs have markedly different driver characteristics from other storm types and dominate the joint probabilities of the most extreme rain-surge compound events, even though they occur much less frequently. ETCs are the predominant source of more frequent, moderate compound events. The hourly data also reveal subtle but important spatial differences in lag times between the joint flood drivers. For Manhattan and southern shores of NYC during top-ranked TC rain events, rain intensity has a strong negative correlation with lag time to peak surge, promoting pluvial-coastal compound flooding. However, for the Bronx River in northern NYC, fluvial-coastal compounding is favoured due to a 2–6 hour lag from the time of peak rain to peak surge.