Flood maps and data decisions: what are the real-life impacts?

In December 2020, the American Geophysical Union (AGU) hosted its annual Fall Meeting via a substantial virtual setup – 13 full days of live programming over three weeks and nine different time zones.

As part of the Advances in the Prediction and Application of Design Floods session, I explored the many decisions that must be made when creating hydraulic models and flood maps, and how these decisions in turn can have a significant impact on results, the final flood map, and the view of risk itself.

You can read some of the key takeaways from the presentation below.

Flood: a complex peril

Flood is a complex peril to model and varies significantly over small areas and timescales. As a result, it can behave in lots of different ways and, as modellers, we must make decisions around the data we use and the parameters we assign in the process of making national-scale flood maps for a range of design return periods.

However, decisions made regarding the fundamental aspects of flood map development can lead to significant differences in the flood maps that are ultimately generated.

River definition

You might think that defining a river is simple and done consistently by all scientists, but there is a surprising degree of subjectivity in something as fundamental as this. The choice made by the modeller in terms of which watercourses are modelled can yield a very different picture in terms of the resultant flood map and view of risk. Figure 1 shows how watercourses are classified in the US National Hydrography Dataset flowline.

Figure 1: A map of Houston with the various waterways as defined in the National Hydrography Dataset provided by USGS.

If all the flowline classifications are included in the river definitions, there are significantly more locations shown to have a fluvial flood hazard than when only the lines classified as stream/river are included, as seen in Figure 2.

Figure 2: The difference in fluvial flood hazard when different river classifications are used. Left: All flowlines used. Right: Only lines classed as stream/river lines used.

Hydraulic boundary conditions

How to configure the hydraulic model boundary conditions is another key decision in the flood map process and one that can significantly affect the resulting view of flooding and risk. This could be in terms of the volume of flow entering the model at the upstream extent or how flow behaviour is represented at the downstream boundary.

For example, the image in Figure 3 shows the city of Jacksonville in Florida. In normal weather conditions, the tidal range means that even at high tide, water in Trout River is able to discharge into St John’s River unimpeded. In hurricane conditions, however, the combined low pressure and high winds can cause extreme storm surges, with sea water being forced many kilometres inland up the St John’s estuary. The ability for Trout River to discharge is therefore limited by the backwater effect of the surge conditions and inland flooding from heavy rainfall is therefore exacerbated.

The extent to which these processes are represented in the hydraulic model design can have a significant impact on the resulting flood map, with the subtle nuances shown in the image making a real difference at an individual property level.

Figure 3: The difference in the fluvial hazard extent when normal tidal (light blue) and storm surge (dark blue) boundary conditions are used.

Hydrological losses

Furthermore, how hydrological losses are represented can make a difference at location-level – meaning the amount of rainfall that is lost to either natural or anthropogenic factors such as infiltration, evapo-transpiration and urban drainage, and therefore doesn’t contribute to surface flooding. For example, if and how storm drain capacity is represented in a hydrological model can determine the level of flood hazard represented in the resultant map, and subsequently whether a property is represented as at risk of flooding or not (Figure 4).

Figure 4: The difference in the fluvial hazard extent when an urban drainage allowance is used (light blue) and when no drainage capacity is accounted for (purple).

Hydraulic model resolution

The final decision I explored in my presentation was hydraulic model resolution and the importance of striking the right balance in terms of detail, to optimise model performance for the intended map use.

Flood maps which are hydraulically modelled at 5m resolution produce a very refined result. This is particularly noticeable when modelling pluvial flooding in an urban environment, with higher resolution modelling doing a better job of simulating the flow of water down roads, narrow streets and walkways, and between buildings. The flood map developer might then choose to resample this high resolution (5m) flood map to a coarser cell size, for example to minimise data size and file storage requirements. This resampled map will likely present a larger overall flood extent and can give a more cautious “worst case” picture, whilst still capturing the localised nature of flooding.

However, running hydraulic models at high resolution is computationally expensive, particularly for larger model areas. By modelling at coarser resolution, for example 30m, results can be obtained much faster and more cost-effectively. The compromise is that the hazard is not as well represented, as illustrated in Figure 5.

Model run times and expense have to be evaluated alongside the need for finer detail and precision, to make an appropriate choice, with the decision having a significant impact on the view of risk.

Figure 5: The difference in flood maps when modelled at different resolutions. Left: pluvial flood map hydraulically modelled on 5m grid. Middle: pluvial flood map hydraulically modelled on 5m grid and resampled to 30m. Right: pluvial flood map hydraulically modelled on 30m grid.

Impacts on real-life decisions

As I’ve explored, decisions made when creating a flood map can have a significant impact on the flood extents represented and the subsequent view of risk. This, in turn, can play a big role in the real-life decisions made across different sectors when it comes to preparing for, and managing, flood risk.

Figure 6 shows how two design flood maps for the same area can present two very different views of risk. The map on the left, from FEMA, indicates far fewer areas as having a flood risk than the map on the right from JBA, due to different decisions made when creating the flood map.

If reliant on flood maps that only provide a partial view of hazard, home and business owners could make decisions around purchasing flood insurance without knowing the full detail, and it could even impact whether homeowners can secure a mortgage. Similarly, local authorities are unlikely to be fully prepared for disaster response, and communities may have a false sense of security when it comes to flood resilience and preparedness.

Figure 6: The flood hazard extents of FEMA's RP100 fluvial flood map (left) and JBA's equivalent RP100 fluvial and pluvial flood maps (right).

To put this into context, a JBA study carried out using the high-water marks from Hurricane Harvey shows that only 48% of the marks were covered by FEMA’s 100-year flood map, whereas 68% were covered by JBA’s equivalent mapping and 83% were captured in JBA’s 1,500-year flood map - the closest modelled return period to the rainfall experienced during the event. Many communities were devastated by Hurricane Harvey, despite a widespread understanding that they were not at risk to flood. You can read more about this here.

This demonstrates a need for flood maps that represent low frequency, high severity scenarios to ensure that map users have a fully representative view of risk.

These decisions are only some of those which need to be considered by flood map developers and, in reality, there are many more. Ultimately, modellers must always have the end use in mind when creating flood maps to ensure flood risk is appropriately represented. It is imperative that both the modeller and end user understand the decisions made when creating the maps and the implications these decisions have.