Comparative Hydrodynamic Simulation and Field-Based Validation of an Extreme Urban Flood Event in Southern Bali

Golan Geldoffer Mauregar , Robby Yussac Tallar, Alfrendo Satyanaga, Oscar Karnalim, James Dwikurnia Sihonggi, Febrian Timotius Sugiarto

A severe rainstorm on 9–10 September 2025 inundated Denpasar City, Badung Regency, and Gianyar Regency in South Bali, causing 15 fatalities across 188 reported locations. This study simulates the event over 14 sub-catchments (1,146 km²) with HEC-RAS 2D and compares two rainfall inputs: a lumped upstream boundary hydrograph based on Snyder’s synthetic unit hydrograph (Method A) and watershed-scale rain-on-grid (Method B). Rainfall came entirely from satellite products, SITHALA for daily totals and PERSIANN-CCS for a 37-hour storm duration, distributed with the Alternating Block Method. The SCS Curve Number method, with a five-class land cover map, Antecedent Moisture Condition I, and an initial abstraction ratio of 0.05, produced 33.1–108.8 mm of effective rainfall (68.6 million m³ of runoff), applied to a 10 × 10 m mesh; depths below 0.20 m were excluded from mapping. Method A flooded 101.92 km² (8.89%), with 50.6% of this area (51.53 km²) deeper than 5 m and hydrograph peaks arriving after the storm window. Method B flooded 80.83 km² (7.05%), of which about 93% was 0.20–2.00 m deep and only 0.39 km² exceeded 5 m. Against 32 field survey points, Method A overestimated depths at its nine inundated points (mean bias +189.8 cm; MAE 249.3 cm) and left 23 points dry. Method B underestimated depths at its six inundated points (bias −74.8 cm; MAE 74.8 cm); for the 26 points it left dry, 81% had an inundated cell within 100 m (median 55.8 m). Rain-on-grid was more suitable for satellite-driven, ungauged applications.