Analysis of rain-on-grid modeling tools for simulating flood risks in urban-coastal landscapes

ABSTRACT

Floods are increasing globally, causing significant economic strains on cities. This study evaluates the trade-offs between computational efficiency and model accuracy in flood prediction using rain-on-grid approaches. We compared two open-source rain-on-grid models (LISFLOOD-FP and Itzi) with a commercial one (InfoWorks ICM) to assess flood risk for Cambridge, Massachusetts, at spatial resolutions of 1, 3, 5, and 10 m, with and without drainage networks. Results showed that ICM achieved the fastest computation time at 1 m resolution (11 min). Itzi and LISFLOOD were efficient at 5–10 m resolutions but significantly slower at 3 m due to adaptive time-stepping. Both open-source models demonstrated substantial agreement with ICM (kappa >0.6) across all resolutions while reducing computation time by 88–97% at 10 m resolution. Incorporating drainage networks reduced predicted road inundation from 53.8 to 31.5% for 10-year storms (22.3% reduction), though efficiency gains declined to 10.3% for 500-year storms. Itzi and LISFLOOD overpredicted road inundation by 15–28% compared with ICM, likely due to simplified shallow water equations. ICM demonstrated superior accuracy in representing tidal and storm-surge boundary conditions critical for coastal flooding. This study provides a practical assessment of trade-offs between model fidelity, hardware demands, licensing constraints, and simulation times, informing decisions in resource-constrained environments.