Not all assumptions are equal: Ranking influence of parameters in US National Flood-Transport Model

Abstract

Road and freight networks increasingly face flood risk, and federal agencies rely on national-scale flood impact models to determine investment strategies for protection and recovery. Such large-scale models require dozens of non-data assumptions and parameters, such as values taken from earlier literature or expert judgment. However, it is not always clear how these values affect cost estimates. We address this using global sensitivity analysis. We adapted a flood-transport model (Li et al., 2026) for United States passenger and freight networks and examined 56 non-data parameters. We tagged each by how its value was sourced: taken from literature (15), re-estimated for the US (32), or structurally altered (9). Overall, US values were derived from USDOT benefit-cost guidance, EPA MOVES fuel-consumption curves, the Highway Capacity Manual, and ATRI freight data. We screened the parameters using the Morris elementary-effects method, computing each parameter’s influence (mu*) and interactions (sigma) separately to analyze direct damage and indirect rerouting and isolation costs. In a Sioux Falls test environment, indirect costs scaled almost linearly with demand, while direct damage was influenced most heavily by road length, damage ratio, and damage level. Furthermore, indirect costs were driven by flood depth, damage level, and road-width, and the most influential parameters spanned a 14.2% range in estimated cost. By identifying the assumptions with the greatest impact, federal agencies can focus intensive efforts where they increase accuracy the most.