Quality Digital Elevation Models are important in flood modeling, and are helping the wider geospatial community in data-scarce regions.
The use of mathematical models to help engineers understand likely flood risk has been common practice for decades.
Over the last twenty years or so, developments in computing power and increased access to higher-quality data have meant that hydraulic modeling approaches have changed significantly. The use of 1D, steady-state river models has now largely been replaced with models which solve for flow and depth in both one and two dimensions, and these models can be used to understand the risk of not just fluvial sources of flooding but also pluvial and coastal. In particular, the increases in computing power have led to larger geographical areas being modeled far more quickly and at a lower cost than ever before.
The quality and efficacy of any flood or hydraulic model will only ever be as good as the data used to build it. For small detailed models, in the absence of LiDAR or other sources of high-resolution terrain data, the ground surface can be generated through spot height surveys. However, when looking at flood risk over larger geographical areas, this data needs to be generated more efficiently.
Access to LiDAR
In some countries, access to good quality LiDAR is available at no cost to the end user. In the U.S., elevation data is freely available via the USGS’ 3D Elevation Program, which provides coverage for much of the country.
Similarly, in England, the Environment Agency’s National LiDAR Programme provides a LiDAR Digital Elevation Model product, free of charge, for approximately 70% of the country. The eventual aim of both the USGS and the EA is for complete geographical coverage, which will greatly benefit the modeling communities in these regions, as areas with sparse data coverage will no longer be an issue. This will take time and investment.
The Problem of Data Scarcity
However, building and running hydraulic models to produce high-level flood maps is more challenging for locations with no reliable source of elevation data. This either requires undertaking a LiDAR survey for the area of interest or using other terrain data sources available to produce higher-quality maps. Carrying out LiDAR surveys is expensive, and this cost is likely prohibitive for projects with a large geographical footprint.
In terms of other terrain data sources, many DEMs are available globally that are based on surface elevations measured from space using near-infrared, radar and/or visible sensors.
These DEMs exist at both 3 arc-second (~90m) grid spacing, such as the widely used Shuttle Radar Topography Mission, and 1 arc-second (~30m) grid spacing, including the recent Copernicus GLO-30 model.
A range of factors may affect DEM data suitability, including acquisition period, sensor type, data processing, or even geographical extents. However, as these data represent the Earth’s surface (i.e., they are Digital Surface Models), they will also contain many features that need to be removed to be ready for use in building hydraulic models.
Forests, Buildings and FABDEM
Hydraulic models require Digital Terrain Models that represent the bare earth’s surface. A particular challenge in processing DSMs is the removal of forests and buildings, as both the location and height of these features are required in order to correct them. However, with data sets relating to global forests and buildings becoming more widely available, there is great potential for using machine learning techniques to remove these artifacts from the surface data.
Indeed, scientists at Fathom and the University of Bristol have done just that. By applying random forest machine learning algorithms to the Copernicus GLO-30 DEM, and making use of reference LiDAR DTM data sets to train these algorithms, a 1 arc-second (~30m) resolution DTM for the entire globe has been produced. This “first of its kind” global DTM is called FABDEM.
A comparison between Copernicus (top) and FABDEM (bottom). Copernicus GLO-30 is a DSM, meaning that it contains noise, such as buildings and forests. In contrast, FABDEM represents the bare earth’s surface. (Images courtesy of Fathom)
The uses of this data within hydraulic models are far-reaching. Having a 1-arc-second resolution DTM for the entire globe can help improve our understanding of flood risk at a large scale over extensive areas. This understanding will not replace the need for detailed flood models in high-risk areas but can allow engineers to understand where those high-risk areas are in the first instance.
Uses of FABDEM Beyond Flood Modeling
When first created, FABDEM was intended to improve Fathom’s flood model capabilities. However, customer demand has driven us to question whether this data set would be useful outside of flood modeling. The simple answer is yes. Since the release of FABDEM, Fathom has received significant interest from many industries, demonstrating the broad range of applications where FABDEM is proving helpful.
The same engineering consultancies that use Fathom’s flood hazard layers have been keen to explore how FABDEM can be used in other engineering domains. Their interest in FABDEM has been centered on how it can improve the wide-ranging services they offer to their clients, and there is particular demand for using the data in the early stages of many projects.
For example, having a more accurate understanding of the terrain during the master planning stages of projects covering large geographical areas can introduce significant efficiencies. We are seeing this in use cases ranging from new water or wastewater networks, road and rail layouts, and energy and renewables infrastructure planning.
Minimizing construction costs for water and wastewater networks is only one factor of a network's design. Engineers must also ensure that pipe fall gradients in networks are self-cleansing and optimize operational costs by reducing pumping and maximizing gravity systems.
Likewise, cut and fill management impacts the construction costs for road and rail projects; however, it must be balanced by ensuring that the geometric design complies with safety criteria. Using FABDEM enables the efficient generation of preliminary designs.
Beyond flood risk and construction, those involved in humanitarian aid and disaster response have also been keen to assess how this new data set can help identify the logistical challenges faced in the distribution of supplies. Often, disasters occur in remote areas where a detailed understanding of the terrain is limited, and this improved data set allows decisions to be made more confidently.
Finally, interest in using the data to visualize terrain within gaming platforms and flight simulation have also been new areas for Fathom to explore. Improved data provides a more realistic user experience that is highly sought after. We hope that the uses of the FABDEM data will continue to grow as we work to provide the most accurate and cost-effective solutions to a wide range of geospatial challenges.