H21E-03 Comparative Analyses of Open-Source Vs. Licensed Commercial Models in Predicting Urban Flooding under Climate Change Extremes

The recent increase in extreme precipitation events around the globe resulting in severe urban flooding aligns with the projections of the impact of GHG (greenhouse gas) emissions on future climates. As urban flood analysts seek to design urban infrastructures more resilient to future climate changes, they face great uncertainty in projecting how climate change will impact future storm events. Dealing with this uncertainty requires the simulation of many potential storms. There is a growing use of high spatial resolution DEMs (digital elevation model) with direct rainfall approaches and 2-D surface hydraulics in urban flood analyses, which require high-capacity computer workstations and extended computing times. Simulating many storms with high computing requirements can become a resource constraint for appropriate analyses of climate change resilient urban flood management. There is a need to understand the trade-off between computing resources and accuracy of flood model outputs. This study compared two open-source models (LISFLOOD-FP and Itzi) with a licensed commercial model (InfoWorks ICM) in predicting flood risk for Cambridge, Massachusetts, using the direct rainfall (i.e., rain-on-grid) approach. First, the models were used to predict surface flooding for Cambridge without incorporating any sub-surface drainage structures by comparing spatial resolutions of 1m, 2m, 3m, 5m, and 10m for rectangular grids (Itzi and LISFLOOD-FP) and irregular triangular meshes (InfoWorks ICM). The performances were evaluated using the Cohen’s Kappa statistic on maximum flood depth and the vulnerability of key urban infrastructure assets. Second, InfoWorks ICM and Itzi were compared against each other by incorporating sub-surface drainage networks to the study site. InfoWorks ICM performed efficiently at finer resolution but requires high-end computing devices with GPU (Graphical Processing Unit) accessibility and only runs in a Windows environment. Results indicated that coarser resolution results in a loss of accuracy. However, what is the level of accuracy needed for risk assessments? This study will provide a trade-off analysis of model accuracy, hardware and license costs, and simulation run-times.