06. Impacts on Bridges and Roads
Flood Inundation Mapping (FIM) based on the HAND method relies on USGS 3DEP Digital Elevation Models (DEMs), which often exclude bridge decks due to bare-earth processing. This “bridgeless” elevation data can result in bridges being falsely identified as inundated, potentially leading to unnecessary emergency response actions. To address this issue, a “bridge healing” method has been developed to more accurately represent bridge elevations in HAND-derived products. In addition, a new roads flood impact (FIMpact) method has been introduced to identify flooded road segments and estimate flood depths.
The bridge healing is implemented by updating the HAND grid values at the location of bridges. Bridges are identified using OpenStreetMap (OSM) and, where available, enhanced with USGS Entwine lidar elevation data. Bridges are classified into:
- Lidar-informed bridges: 48% of bridges, where classified lidar data enable deck elevation extraction
- Non-lidar bridges: 52% of bridges, where either lidar is unavailable or usable points are lacking
The bridge healing workflow (i.e., updating the HAND grid) for lidar-available bridges involves:
- Generate lidar-based elevation rasters
- Subtract DEM values from lidar-derived rasters to produce DEM correction rasters
- Add correction rasters to HAND grids
Plot below shows the steps to generate lidar-based elevation raster files for OSM bridges.
Figure 1: Methodology for generating elevation raster files for OSM bridges using Lidar
The main limitation of this method is the lack of reliable well classified lidar points.
The bridge healing workflow for non-lidar bridges involves:
- Buffer OSM bridge lines by 10m and overlay it on HAND grid
- Record max HAND value within the buffer
- Apply this max HAND value in place of the original HAND grids across the buffered area
This method assumes a uniform HAND value for the entire bridge buffered area. Therefore it can overestimate or underestimate flood impacts, especially for non-flat or perched bridges.
Two bridges were analyzed to compare workflows:
Here, we compare results between the two healing methods for two bridges, highlighting the improvements achieved using lidar. Figure 2 displays results for the first example.
Figure 2. A comparison in results between the two healing workflows for bridges
Figure 2-A shows the inundation map using Non-lidar workflow. The workflow generates a 10m buffer around OSM bridge lines (red ellipse) and assigns the maximum HAND grid value within the buffer to all pixel centers that intersect it. This method underestimates inundation near the bridge by overgeneralizing the elevated values within the buffered area.
Figure 2-B shows DEM correction values obtained from lidar-informed workflow. In this approach, only the three HAND grid cells directly intersecting the bridge are updated using the DEM correction values, which represent the difference between the lidar elevation and the elevation from the USGS DEM. Figure 3 describes the steps to generate these DEM corrections for this bridge. .
Figure 2-C shows the inundation map using lidar-informed workflow. The elevated values (DEM correction values) applied to the HAND grid result in non-inundation only for the three bridge cells.
Figure 3. The steps to generate DEM correction using lidar dataset
Note that the original DEM exhibits an approximate 1-meter dip in elevation for the middle cell, whereas the lidar dataset shows similar elevations across all three cells. The necessary elevation correction is applied through DEM correction values, with the middle cell requiring a greater elevation adjustment than the others to compensate for the error in the original DEM
The second comparison features the Watsontown Bridge in PA, shown in Figure 4 with its plan and side views, highlighting that the bridge deck is significantly higher than its approaches.
Figure 4. The plan and side views for the Watsontown Bridge in PA
The lidar-informed FIM (Figure 5, left) shows the entire bridge as non-inundated whereas the non-lidar workflow marks the entire bridge as inundated (Figure 5, right) likely due to using a HAND grid value from the lower elevation approaches to heal for the entire bridge.
Figure 5. Lidar-informed FIM (left) shows the entire bridge as non-inundated whereas the non-lidar workflow marks the entire bridge as inundated
Road flood inundation mapping (FIMpact) has been developed to identify inundated roads for a given flood event and to calculate maximum flood depths along those roads. The workflow consists of three main steps:
-
Ingest OSM roads into the FIM pipeline
Five primary OSM road types are processed: motorway, trunk, primary, secondary, and tertiary. To improve spatial accuracy, long road segments are split using NWM catchment boundaries before being used in the FIM workflow.
-
Derive inundation threshold discharges per road segment
To further improve accuracy, each road segment is split at HydroID boundaries, and the minimum non-zero HAND value within each resulting sub-segment is extracted as the inundation threshold stageThreshold discharge values (for inundation) are then interpolated from HydroTables based on these threshold stages.
-
Identify inundated roads and calculate flood depth
For each flood event, the workflow checks whether the flow at a segment’s intersecting feature ID exceeds its threshold discharge. If so, the segment is flagged as inundated. Flood depth is computed by subtracting the evaluated stage from the segment’s threshold HAND value. Since road segments may intersect multiple HydroIDs or branches, only the record with the maximum flood depth is retained per segment.
Figure 6 illustrates the final output, showing inundated roads with their depths and non-inundated roads in black over a FIM raster.
Figure 6. Roads FIMpact outputs