Creating Fluvial Flood Maps for New Jersey Using the National Water Model (NWM) and Height Above Nearest Drainage (HAND)

Traditionally, coastal flooding is considered the primary flooding hazard in the State of New Jersey, and the risk from inland fluvial flooding is less emphasized. This imbalance leaves a gap in inland flood modeling and information; current flood risk maps such as those created by FEMA cannot provide information about flood depth and may not reflect future flood conditions. These problems impede State agencies from 1) planning future infrastructure systems, e.g., a new road elevation cannot be determined due to a lack of flood water depth information, and 2) preparing for the future flooding risk. Thus, a new inland flooding tool is needed to provide updated maps to state and local agencies to inform decision making and infrastructure planning. Using the Height Above Nearest Drainage (HAND) method, a large-scale flood model called National Flood Inundation Experiment (NFIE) was developed for the continental United States, which translates flow data from the National Water Model into flooding depth using synthetic rating curves based on the channel geometry and the Mannings equation. The water depth information is then projected onto maps of HAND, showing the depth and spatial extent of the inundation. There is a potential to downscale this model to develop a local inland fluvial flooding tool. The challenges to develop such a tool are twofold. First, the field data is sparse, e.g., in the flooding event of Hurricane Irene, only 1% of the catchments in NJ have ground-truth data to compare with the simulation result to validate the model. Default NFIE rating cures were based on an assumed roughness of 0.05 for all channels, which yields inaccurate flow/stage relationships. Our study demonstrated that the roughness should be calibrated/tuned. Second, flooding isnt translated across the catchment boundaries, making large floods and tailwater conditions difficult to map. This problem compromises the accuracy of the modeling result and limits its application. To address these issues, a regression scheme was developed to predict the roughness based on the physical parameters and landscape data of the catchment. A smoothing scheme was further invented to enhance the result. The improved scheme shows a promising result compared with the direct down scaling. More validation results will be presented in the conference.