9. Known Shortcomings and Opportunities for Improvement
This page lists some, but not all, of the known issues and opportunities for HAND improvement.
The 10m currently used by Inundation Mapping does not include a detailed representation of channel bathymetry. When the 10m DEMs were collected, pulsed laser beams from an aircraft or satellite could not penetrate the surface water and image the channel bottom. As a result, the water surface elevation at the time of the DEM's collection constitutes the elevation values within rivers. This lack of channel bathymetry information can result in an overprediction of flooding by HAND-derived inundation maps and is most pronounced at lower flows.
Figure 1: Lack of channel bathymetry data due to the inability of laser beams to penetrate the surface water.
SRCs are used to interpolate a stage from a discharge value. There are several known shortcomings associated with the 10m HAND-derived SRCs, including a tendency to overpredict stage at low flows (see Lack of Channel Bathymetry Data). To combat this issue, there is an ongoing effort within Inundation Mapping to estimate and represent bathymetry in the SRCs.
Figure 2: Examples of synthetic rating curve (SRC) limitations before calibration (a) and due to Manning's n limitations (b).
Manning's Equation is used to convert modeled streamflow values into stage values, which are then applied to the HAND grid to yield a flood inundation map. This equation is very sensitive to Manning's n, used most commonly as a value approximation of channel and floodplain roughness. Because roughness is dynamic across space and time, with features appearing and disappearing with increasing distance from the channel, and with seasonal changes in vegetation, it is very difficult to settle on a Manning's n value that yields consistently good performance.
Figure 3: Example of lack of Confidence in Manning's n (Channel/Floodplain Roughness). n value is dynamic across space and time, making it difficult to settle on a number that yields consistently good performance across river networks.
Some of the ongoing efforts to optimize Manning's n include, but are not limited to:
- A multi-n approach, where different N values are used in different parts of the channel
- Using n values established by studies of landcover
- Calibration using priori knowledge, observed FIM boundaries, or higher-fidelity models
In hydro conditioning, DEMs are treated so that thalwegs elevations are forced to decrease in the downstream direction. Occasionally, a large thalweg drop is encountered and the following thalweg pixel elevations may be unrealistically lower than its adjacent pixel elevations. There are some checks to protect against some of this behavior, but it is still a lingering issue.
Future Inundation Mapping work seeks to experiment with HAND production using LiDAR, where available.
In general, the Inundation Mapping codebase is modular and relatively computationally efficient. But there is always room for improvement. Furthermore, with an eye toward future LiDAR development, we expect that some of the existing Inundation Mapping code will need to be refactored to handle the larger LiDAR datasets.