Introduction: Urban pluvial flooding, caused by short and intense rainfalls that exceed drainage system capacity, is among the most frequent and costly hydrometeorological hazards in Mediterranean cities, and its impact is expected to grow with climate change. Reliable rainfall thresholds are essential for operational early-warning systems, yet many empirical approaches rely on fixed rainfall durations that poorly reflect catchment specific response times. This study adapts the Transfer Function (TF) model, originally developed for rainfall-induced landslides to urban pluvial flood forecasting, and tests it in the urban area of Naples, Italy.
Methods: The TF model convolves the observed rainfall intensity with a normalized exponential filter to produce a continuous floodability signal, whose characteristic response time is calibrated against recorded flood events. A critical threshold is also calibrated and set on this signal to separate flood from non-flood conditions. The model was calibrated and cross-validated separately for two basins of differing characteristics, and for two seasons, using rainfall series derived from six rain gauges and a crowdsourced flood-event catalogue.
Results: The TF model achieved good classification performance in the season with enough flood records, with F1-scores of approximately 0.7-0.8 in both basins, besides consistently low false alarm rates. The calibrated response time was notably longer in the larger, flatter basin than in the smaller, steeper one, matching independent physical characteristics of the catchments. Performance decreased in the flood-scarce season, reflecting the limited number of recorded events rather than a structural weakness of the model.
Conclusions: The TF model offers a simple, physically interpretable, early-warning tool for urban pluvial flooding, requiring only the calibration of urban response time and one flooding threshold. Its performance is competitive with more complex methods while remaining transparent and easy to operate, making it well suited to catchments where long, high-quality flood records are scarce.