Comparative Analysis of Open-Source vs. Commercial Models in Predicting Flood Risk under Climate Extremes

The world is experiencing a growing frequency of floods in recent years, resulting in billions of dollars in damage with each event. Numerical models can be an effective tool to predict the flood risks, but the underlying approaches can significantly influence the accuracy of the model. While the catchment approach reduces computation costs by neglecting the geospecific details on the extent and propagation of flooding, rain-on-grid approach allows for a complete representation on the spatial variability of flooding at higher computation costs. This study evaluates the trade-offs between computational efficiency and model accuracy in flood prediction. We compared two open-source models (LISFLOOD-FP and Itzi) and one commercial model (InfoWorks ICM) to assess flood risk in Cambridge, Massachusetts. First, all models simulated surface flooding without subsurface drainage networks at spatial resolutions of 1m, 3m, 5m, and 10m using rectangular grids (Itzi and LISFLOOD-FP) and triangular meshes (ICM). In second phase, subsurface drainage networks were incorporated into the Itzi and ICM models. Model performance was evaluated using Cohen’s Kappa statistics based on maximum flood depth and the exposure of critical urban infrastructure. While ICM yielded higher accuracy at finer resolutions, it required high-performance computer hardware with graphical processing unit (GPU) support. Results showed that coarser grids reduce accuracy but dramatically cut computational cost. This study offers a practical assessment of the trade-offs between model fidelity, hardware demands, licensing constraints, and simulation times, helping urban flood modelers make informed decisions in resource-constrained environments.