Flash floods are the most dangerous processes in Seine-Maritime (northern France) due to their sudden onset and rapid rising time. Such floods shortly follow high rainfall, ranging from 50 to up to 100 mm in less than 6 hours, and occur in small (< 20 km2) and dry valleys, in which a surge rushes down through the main valley just a few minutes after rains have peaked. Phenomena currently produce damages to buildings and road infrastructure, but rarely threaten human lives. Anticipating their occurrence, therefore, is crucial. However, such prediction remain delicate, especially at larger scales, due to the quantitative errors of radar disturbing rainfall-runoff simulations, the infrequency of events and the gentle return periods rendering statistical analysis and calibration of models far from obvious. In this study, a susceptibility analysis is then carried out, without depending on the meteorological predictions, by applying RUICELLS, a cellular automaton model driven according to a set of three deterministic hydrological rules of flow pathways. The 2,368 simulations launched for 16 different rainfall intensities on 148 basins of small size (less than 20 km2) permit to simulate the peak-flow discharges (Q), specific peak-flows (Qs) and lag times (T), but also to estimate the critical rains necessitating increased vigilance. The authors' simulations show that the number of basins susceptible to flash flooding greatly increases with higher rainfall intensity, cultivation of sensitive crops (sugar beet, corn, maize, flax), and basin morphology. Moreover, the authors show that certain valleys are more prone to flash flooding, since several susceptible basins are located in close proximity. The modelling results also question the effectiveness of a specific flash flood alert system for this region.