Can we estimate flood frequency with point-process spatial-temporal rainfall models?

Stochastic rainfall models are commonly used in practice for long-term flood risk management. One of the most widely model types is based on point processes. Despite widespread use such models, whether their known simplifications describing space–time structure will affect accuracy estimation has not been quantified. In this study, we quantify biases introduced by limitations to estimates two medium-sized river catchments (717 km2 and 844 km2) South East UK. To achieve this, nine years hourly radar data, a dense network rain gauges, spatial–temporal stochastic processes, fully distributed hydrological model. We modelled corresponding catchment water dynamics using observed simulated then assessed errors propagate flow dynamics. Our results show that properly captures point-scale statistics, including extremes cross-site spatial correlations. However, bias areal an overestimation reduction factor, extreme mean precipitation, fraction (wet area ratio). Using as input continuous simulations, find duration curves well preserved, particularly high seasons (relative less than 7%). The also reproduces frequency at daily scale with averaged relative 0.36–16.9% 10-year return levels, confirming its ability infer catchments. summer-season peak discharge highly overestimated over 163.5% same level. summer explained dominating convective systems misrepresented