Our planet is a busy place—full of peaks, valleys, natural habitats, and human-made objects. When navigating Earth’s varied terrain, digital elevation data brings these highs, lows, and features into view.
By providing you with a visualization of the landscape in question with elevation data, you can estimate areas most vulnerable to sea-level rise, spot vegetation encroachment, and avoid eyesores when urban planning.
There are many ways to model elevation, and we’re going to focus on three digital elevation dataset in particular:
- DEM - Digital Elevation Model
- DSM - Digital Surface Model
- DTM - Digital Terrain Model
A Digital Elevation Model, also known as a DEM, is a type of raster GIS layer. They are raster grids of the Earth’s surface referenced to the vertical datum—the surface of zero elevation to which heights are referred to by scientists, insurers, and geodesists.
At most scales and environments, a generic term like DEM can be used because the differentiation between the bare-earth and a surface object is not significant, with DEMs commonly having spatial resolutions of 20 m or more.
The smaller the grid cells are, the more detailed the information within a DEM data file is. So, if you’re looking to model with lots of detail, then small grid spacing (or small cell size) is the one to go for.
DEMs are usually generated from remotely sensed data collected by satellites, drones, and planes. This variety of DEM source data means that it’s possible to fill data gaps where little data is available over remote regions, for example.
Automatic DEM extraction from stereo satellite scenes means that data from satellite sensors such as SPOT-5 (5-10m resolution) can be used.
Some remote sensing methods for obtaining DEM surfaces are:
- SAR interferometry (aka. InSAR): synthetic aperture radar (SAR) data collected, for example, by the Shuttle Radar Topography Mission (SRTM), uses multiple radar images from antennas captured at approximately the same time to create a DEM. Additionally, some have developed DEMs from InSAR and then applied deep learning to correct for urban influence, such as with the CoastalDEM.
- Stereo Photogrammetry: in both aerial photography and with satellite imagery, photogrammetry uses images from at least two, but more often 3, different vantage points of the same area. In this way, similar to how our vision works, we can obtain depth and perspective where the images overlap.
- LiDAR (aka. Laser Altimetry): as with DSMs, using light, LiDAR measures the reflected light that has bounced off the ground to determine the elevation of the Earth’s surface.
- Digitizing contour lines: with a contour map, DTMs (a subset of DEMs) can be easily digitized and then interpolated programmatically with geospatial software.
- DGPS measurements: differential GPS or DGPS, people do field campaigns with specialized devices that use satellite information to survey points across an area and determine their position. Thus this method is discrete and has to be interpolated to yield a continuous raster.
- Ground surveying: by assessing known XYZ positions, neighboring areas are measured using a device called a theodolite. This requires very skilled labor, and similarly to DGPS, all the points need to be interpolated to yield a continuous raster.
It is also important to note that with one of the most common methods to derive DEMs, InSAR, persistent objects such as buildings are often measured while non-persistent objects are removed. Thus, in urban areas, they more or less represent the surface, while in rural areas without dense vegetation, they represent the terrain.
DEMs can be segmented into Digital Surface Models (DSMs) and Digital Terrain Models (DTMs), which we’ll delve into further in subsequent sections.
Although elevation data is represented in DEMs in a regular grid of columns and rows, which is a very efficient means of storing lots of data, elevation data can also be conveyed in a vector-based format.
These datasets are called Triangulated Irregular Networks (TINs), and they use a variable point location strategy to put elevations at critical locations.
TINs promise to decrease storage enough to make up for having to store x,y, and z coordinates, and for the overhead of triangle indexing while still maintaining all key elevation information—yet the wide availability and popularity of DEMs indicates otherwise.
Triangulated Irregular Network (TIN)
However, there is another, more popular than TIN, DEM-alternative, which is elevation point clouds.
Due to the rise of LiDAR and LiDAR processing, direct manipulation of point clouds has become more common, with some even deriving point clouds from stereo images.
A Digital Surface Model, or DSM captures a surface—including natural and human-made structure such as vegetation and buildings. They illustrate reflective surfaces of all features elevated above the ‘bare earth’.
In short, DSMs represent the Earth’s surface and all objects on it.
Because DSMs represent the bare-Earth and all of its above-ground features, they are particularly important in urban planning.
3D surface models can increase the understanding and explanation of complex urban scenarios, especially as built-up areas change with time due to urban expansion.
DSMs are ideal for runway approach zone encroachment in aviation, and urban planning to check how a proposed building may affect views. Beyond that, DSMs can be used for visualization, disaster management, navigation, vegetation management, decision-making, and much more.
A Digital Surface Model of Berlin with extruding features such as houses and trees visible (NEXTMap One)
A DSM paints a picture of the world, often using LiDAR (Light Detection and Ranging) technology or stereo photogrammetry.
Sometimes, specific radar wavelengths can be used to generate DSMs too.
In a LiDAR system, pulses of light travel to the ground from a LiDAR unit. The LiDAR pulses bounce off surrounding objects and return to the sensor. The sensor then uses the time taken for each pulse to return to the sensor to calculate the distance it traveled.
The sensor can also measure the intensity of the return to estimate the surface geometry and material composition of the reflecting surface.
LiDAR produces a huge point cloud of elevation values for a given area. But, height can be down to tree canopies, buildings, and other features.
That’s where the magic of DSMs comes in. A DSM captures both natural and built features on the Earth’s surface, such as tree canopy and vegetation changes.
This way, you can gain an eagle-eye view of all the extruding features an area contains.
Above left: The pulse of light emitted from the aircraft during LIDAR collection returns different information about the surface that it encounters. Source: USDA Natural Resources Conservation Service. Above right: Pulse backscatter sensed in the aircraft helps to classify return rank and eventually to aid creation of bare-earth terrain and first-return surfaces. Source: Gatziolis & Anderson (2008).
However, LiDAR can be extremely expensive andis typically flown just over the smaller, high-value areas like cities.
DSMs can also be efficiently generated via automated image matching of high-resolution optical stereo images or stereo photogrammetry.
Stereo matching of the images is used to find corresponding pixels in pairs of images, enabling 3D reconstruction via triangulation, given both the exterior and interior orientations are known.
These image pairs can be sourced from either aerial or satellites, but in either case, generally manually measured object heights from oriented images are used as a reference and leverage computer vision algorithms to yield the final results.
Various open-source and commercial tools can be used to programmatically derive elevation data from stereo images, making photogrammetry both accessible and scalable.
One of the most well-known algorithms of this sort is called the Semi-Global Matching (SGM) algorithm, which has a good trade-off between runtime and accuracy. (Source)
Digital terrain models or DTMs have different definitions depending on where you are in the world.
For our purposes, we recommend considering a DTM as a synonym of bare-earth DEM. DTMs are often confused with DEMs. The main difference between the two models lies in the fact that the DEM generally takes into account all persistent objects on the ground (vegetation, buildings, and other artifacts)—while the DTM shows the development of the geodesic surface.
Bare-earth refers to the fact that vegetation and human-made features such as trees and power lines are filtered out with DEMs. Each cell has a value corresponding to its elevation (z-values at regularly spaced intervals) in a DEM.
It’s worth noting that in some countries and fields of research, people refer to vector data sets composed of natural features such as ridges, breaklines, and spaced points, as DTMs. This definition refers to a DTM as something that augments a DEM by including linear features of the bare-earth terrain.
Regardless of the definition, a DTM is essentially a three-dimensional, digital representation of a surface, consisting of X, Y, and Z coordinates. Within a DTM, you’ll find heights and elevations as well as natural features such as rivers and ridge lines.
However, these subtle differences between DEMs and DTMs are most evident in urban areas where the prevailing high-rise buildings.
For example, the island of Manhattan or megacities like Hong Kong, can significantly influence how the terrain should be measured and how much elevation data may need to be corrected to remove the influence of objects on the surface.
Whether you are viewing a DTM as something that augments a DEM or as a “bare-earth” DEM, thanks to computing power in engineering—the DTM has become an integral tool for earth and engineering applications.
DTMs can be created through various methods, including digitized contours and even from DSMs using the difference between the height values for trees and buildings and their local neighborhood.
Therefore, DTMs can also be created from any of the methods used to generate DSMs, from LiDAR to stereophotogrammetry, as well as SAR, DGPS, and ground surveying, all at varying levels of detail.
Above: A DTM depicting a crater and possible clay beds in West Ladon Valles channels on Mars. Source: NASA/JPL/University of Arizona/USGS.source
The figures below illustrate how DSMs differ from DTM .
A DSM captures both natural and human-made features of the environment.
Whereas, as shown below, a DTM only retains features of the bare-earth terrain, such as rivers and ridges.
A DTM can be derived from a DSM, but the same is not true vice versa.
DSMs include objects on the Earth’s surface, whereas DTMs do not
For most LiDAR applications, the focus is placed on the DEM and DSM as defined above, with DTMs more applicable for GIS and cartographic representations.
Several factors influence the quality of DEM-derived products:
- Vertical resolution
- Terrain roughness
- Sampling density and resulting spatial resolution or pixel size
- Terrain analysis algorithm
- Interpolation algorithm
- Reference 3D products with quality masks containing information on the coastline, snow, clouds, water bodies, etc.
DEMs are critical in areas such as infrastructural management, hydrology and flow-direction studies, and land-use planning.
They are especially useful across greater spatial scales for the contouring of topographic and relief maps:
- Modeling water flow or mass movements (e.g., landslides)
- Creating physical models (such as raised-relief maps)
- Rectifying aerial photography or satellite imagery
- Rendering 3D visualizations
- Reducing (terrain correction) gravity measurements (e.g., gravimetry, physical geodesy)
- Analyzing terrain in physical geography and geomorphology
Before we move onto where to find elevation data, let’s recap the differences between the three types:
- A Digital Surface Model (DSM) is an elevation model that captures both the environment’s natural and artificial features. It includes the tops of buildings, trees, powerlines, and any other objects. Commonly, this is seen as a canopy model and only ‘see’s ground where there is nothing else above it.
-A Digital Elevation Model (DEM) is a generic term for an elevation model, which encapsulates both DSMs and DTMs, and can be generated from various methods. Often, because of scale and environment, differentiation between DSM and DTM is unnecessary, (e.g., SRTM derived DEM at 30 m or 90 m resolution). Note that sometimes people consider this term to be synonymous with DTMs, so always delve into the methodology of how it was derived.
- A Digital Terrain Model (DTM) is a bare-earth elevation model. DTMs do not contain any features above the bare-earth, even persistent ones. Thus, they can be paired with DSMs to derive height information regarding objects on the surface. Some consider DTMs as something that augments a DEM, a network of vector points of terrain elements instead of a continuous raster.
Some key terms to keep in mind when working with elevation models are:
- Ground: the solid surface of the Earth, such as the bottom of the sea
- Height: a measurement of elevation from base to top above the ground or a recognized level
- Elevation : the height above a given level, especially that of the sea or above the horizon
- Terrain: an extent of ground, region, or territory
There are plenty of places to find global DEMs. From free satellite data to LiDAR sources, here’s how to find the elevation data you need:
During its 11 day mission, the space shuttle Endeavour orbited the Earth 16 times and captured Earth’s topography at one arc-second (30 meters) for over 80% of the Earth’s surface.
SRTM used synthetic aperture radar and interferometry to collect one of the most accurate digital elevation models of Earth. The SRTM payload, launched in 2000, used two radar antennas and a single pass to generate a digital elevation model using the technique known as interferometric synthetic aperture radar (inSAR).
This data is freely available for you to use on the USGS Earth Explorer. It covers most of the world with an absolute vertical height accuracy of less than 16m.
To download, select your area of interest. Under the data sets tab, select Digital Elevation>SRTM>SRTM 1-ArcSecond Global. This handy guide by GIS Geography will help get you started.
Before September 2014, the best available SRTM DEM was 90-meter resolution. The 30-meter resolution SRTM is publically available on the USGS Earth Explorer, thanks to Space Shuttle Endeavour
The Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) is a joint operation by NASA and the Ministry of Economy, Trader, and Industry (METI) of Japan. As a part of this, the ASTER Global Digital Elevation Model (GDEM) was born.
With a global resolution of 90 meters and 30 meters in the United States, ASTER GDEM has high resolution and wide coverage—around 80% of the Earth.
How did the ASTER GDEMs come to be? Using stereoscopic pairs and digital image correlation methods. Based on two images at different angles, it measured elevation using stereo pairs and photogrammetry.
Something to note: some users expressed issues with its data, often in cloudy areas.
However, over time, ASTER DEM data has improved its products with artifact corrections leading to considerable improvements.
Some now consider ASTER GDEM-2 to be a more accurate representation than the SRTM elevation models in rugged mountainous terrain. But, go ahead and take a look and see for yourself.
ALOS Global Digital Surface Model, or ALOS World 3D, is a global DSM dataset by the Japanese Aerospace Exploration Agency (JAXA).
It is generated from images collected using the Panchromatic Remote-sensing Instrument for Stereo Mapping (PRISM) aboard the Advanced Land Observing Satellite (ALOS) from 2006 to 2011. The DSM dataset has a horizontal resolution of approximately 30-meter mesh (1 arcsec) and is available to the public for free.
Based on the DSM dataset (5 meter mesh version) of the World 3D Topographic Data—the most precise global-scale elevation data with world-leading precision of 30 meters.
To access this accurate DSM, you’ll have to register online through the JAXA Global ALOS portal to download it.
As the world has moved towards a global LiDAR map, LiDAR data sources are available online for free—if you know where to find them.
Why LiDAR? Its spatial and vertical accuracy is often unparalleled. After you filter ground returns, you can build an impressive DEM from LiDAR data. You can often use the different returns to determine the height of vegetation vs. the height of the ground surface, depending on the LiDAR instrument and the density of the vegetation- thus, you can yield both a DTM and a DSM from the same dataset!
If you’re still looking for LiDAR data over your area of interest, try contacting your local or regional government. As long as you tell them what you’re using it for, they may share their LiDAR data for free.
If you’re unsure which provider to go for, geospatial marketplaces collect multiple datasets from multiple providers and data types.
UP42 brings together elevation data alongside satellite and aerial imagery, weather data, AIS, and more.
The UP42 marketplace contains Intermaps’ digital elevation models with up to 1-meter resolution. Intermap’s NEXTMap 3D elevation products are available as DSMs and DTMs—enabling you to build 3D solutions with or without surface objects, such as vegetation or buildings.
Software tools such as QGIS enable you to open and read DEM files
With the abundance of elevation data available online, once you’ve found the right one for your needs, it’s time to dive right in.
As we now know, DEMs are files that contain either points (vector) or pixels (raster), with each point or pixel containing an elevation value. They come in a variety of file formats, from .csv and .tif to .flt and .dem.
GeoTIFFs allow location information to be embedded within a TIFF file.
Here’s a helpful guide by Carleton University on DEM formats and how you can open them in various tools.
Speaking of tools, let’s look at those next.
You’ll need a Geographic Information System (GIS) or other special application software since elevation data is not directly viewable in a browser. Some software programs that recognize DEM files include:
- ArcGIS—here’s a helpful guide on exploring DEMs using ArcGIS—create elevation layers and more
- QGIS 3 —QGIS 3 is in 3D and brings a whole new set of cartography possibilities. It’s also free and open-source, making it easy to understand how its algorithms work
- [QGIS 2](https://qgis.org/en/site/forusers/download.html "QGIS 2")—jam-packed with functions such as automating map production for free
- [gVSIG](http://www.gvsig.com/en/products/gvsig-desktop "gVSIG")—another free tool, this time with CAD tools, a NavTable, and a mobile application
- [GRASS GIS](https://grass.osgeo.org/ "GRASS GIS")—a free option with an intuitive UI and over 350 vector and raster manipulation tools
For most of these GIS software programs, you can drag and drop the .tif file from your browser directly into the program.
Keep in mind: beforehand, digital elevation data and images are usually unedited and intended for scientific use and evaluation.
They are outputs directly from the data source itself, so they may contain:
- Numerous areas without data
- Ill-defined coastlines
- Water bodies that may not appear flat
- Other errors already discussed.
The USGS DEM format is an open standard for raster-based DEMs.
So, once you’ve downloaded data, chosen a software tool, it’s time to visualize the DEM in all its glory.
There are many ways to do this. Let’s look at a few examples and tutorials that are readily available online.
QGIS 3.0 comes with a 3D layer view. This enables you to visualize GIS data in 3D—providing you with a more vivid visualization of data that contains elevation or height.
Here’s a tutorial that walks you through visualizing DEMs in 3D with QGIS 3.0, using QGIS 3.0 master candidate. Here’s another one that focuses on visualizing DEMs using QGIS 3.0 in comparison with ArcGIS Pro.
3D visualization of DEMs (Source: Geodose)
Visualization techniques also depend on your purpose. When it comes to landform mapping in earth sciences, for example, it’s important to produce both complete and unbiased results.
Here’s a study that looked into five different visualization methods when using DEMs for landform mapping (Smith & Clark, 2005). It found that no single visualization method provides complete and unbiased mapping—prone of azimuth biasing.
Although subtle landforms could be highlighted, researchers recommend curvature visualization for initial mapping as this provides a non-illuminated (and therefore unbiased) image. Then, this can be supplemented with data from relief-shaded visualizations.
If you’re looking to visualize elevation contours from raster DEMs in Python, here’s a tutorial on doing just that using packages such as GDAL and Matplotlib.
A contour plot of Mount Shasta, California (Source: Earth Lab)
DEM analysis includes four essential components, namely:
- Data acquisition: capturing terrain images or scanning the earth surface
- Data modeling: interdisciplinary approaches such as image processing, photogrammetry, interferometrics, etc.
- Data management: data coding, data structuring, spatial database technique, computer graphics
- Application development: urban planning, mine management, surveying, geomorphology analysis, facility management, civil engineering, resource management, geological engineering, landscape design, hazard identification and monitoring, and even computer games and missile/airplane navigation
The uses and applications of DEMs are even more varied than how they are acquired. Relevant and useful for almost any industry or sector that uses location data, some of the general uses include:
- Slope analysis
- Aspect analysis
- Delineating drainage networks and catchments
- Identifying geologic structures
- Viewshed analysis
- 3D simulations
- Change analysis
- Contour mapping
Above: Slope analysis of a volcano in the Galapagos, Ecuador. Source: doi:10.1146/annurev.earth.28.1.169
The key to building DEMs is, you guessed it, elevation data (Z) that’s defined spatially (in X and Y).
This elevation is always normalized in reference to some arbitrary datum in the landscape, generally mean sea level.
This means there has to be a known and consistent reference point of observation and a consistent method of measurement of the area being assessed.
How the elevation data points were acquired (see capture methods in the previous section) determines what corrections need to be made and how the points should be interpolated.
Satellites offer both the known observation point—with their consistency of orbit and all the metadata that comes with each image, such as the orbit file with track information and the angle of incidence—as well as the consistent methodology of acquisition, scanning relatively continuously over an area and uniformly measuring elevation.
When it comes to DEMs derived from satellite data, a crucial differentiation is made between the methods of using sets of optical imagery for stereoscopic analysis vs. using radar information for interferometric analysis.
DEM accuracy is most commonly estimated by calculating the Root Mean Square Error (RMSE) of elevation computed by comparing DEM points and reference points.
However, there’s much more to a DEM’s accuracy than just the elevation (vertical) component.
The DEM quality depends on various interrelated factors such as data acquisition methods, the nature of input data, and techniques employed to develop the DEM.
Different acquisition methods of elevation data, from manual methods like DGPS to passive methods like stereoscopic satellite images to active RADAR or LiDAR acquisitions, have their own biases and sources of error to look out for.
For example, manual methods will be prone to sampling bias and rarely have atmospheric influence or other common biases seen in other acquisition techniques.
Let’s look at some of the key factors that affect DEM accuracy for non-manual methods:
- Atmospheric and ionospheric influence
- Temporal decorrelation
- Coregistration errors
- Phase errors and signal decorrelation
- “Shadow” effects
Above: Common errors encountered in InSAR measurements. Source: doi: 10.1146/annurev.earth.28.1.169
There are three main types of resolution that one should always consider when assessing the fit of a DEM for a given project or application: spatial resolution and vertical resolution.
Spatial resolution is determined by the distance between sample points, which can be relatively uniform, such as in the case of stereoscopic imagery, somewhat uniform, such as with RADAR and LiDAR, or highly variable, such as with DEMs obtained with manual methods.
One of the most important aspects of a DEM is its vertical accuracy or vertical resolution.
The vertical resolution of elevation data is defined as the possible height difference between the modeled or detected elevation and the actual or ground-truthed elevation of the surface.
Each of the various aforementioned methods of obtaining elevation data, such as radar, LiDAR, or photogrammetry, produces differing accuracy levels. Of these methods, LiDAR generally yields the best spatial and vertical resolutions, yet, it is often prohibitively expensive at scale.
You may find that different DEM providers will also have different definitions of what vertical resolution or vertical accuracy is.
Above: Table of global DEMs and their respective spatial resolutions (resolution) and vertical resolution (vertical accuracy). Source
One last resolution one may want to consider before choosing a DEM is the temporal resolution, namely, how recently was the elevation data used to generate the DEM acquired.
This is particularly relevant if you'd like to conduct some change analysis or if you are using a DSM to study something fairly temporally variable like vegetation or new construction.
Vertical errors in DEMs are usually classified as sinks or peaks.
A sink is an area surrounded by higher elevation values. It’s also referred to as a depression or a pit—an area of internal drainage.
Where do they come from? Some sinks may be natural, particularly in glacial areas, although many sinks are often imperfections in the DEM.
On the other hand, a peak, also known as a spike, is an area surrounded by cells of lower value.
These are usually natural features and are less detrimental to the calculating flow direction, as mentioned above.
The number of sinks in a given DEM usually is higher for coarser-resolution DEMs.
Sinks are also commonly caused by storing the elevation data as an integer number. This can cause issues in areas of low vertical relief.
Often, you may find 1% of cells in a 30-meter resolution DEM to be made up of sinks. This can increase as much as 5% for a three arc-second DEM.
You may notice another kind of error in DEMs known as striping artifacts—contained within DEMs resulting from systematic sampling errors when creating the DEM itself. This is also most noticeable on integer data in flat areas.
When faced with sinks and peaks in DEMs it’s important to remove or fill them—to create a *depressionless DEM*.
A DEM free of sinks is the derived input to the flow direction process, for example. This is because the presence of sinks may lead to an erroneous flow-direction raster.
Using a depressionless DEM is key to ensuring accurate analysis.
Many GIS applications include tools for you to create a depressionless DEM, enabling you to:
- Identify sinks
- Fill sinks
- Find sink depth
Here’s a guide by ArcGIS Pro on how to do just that by using the ArcGIS Spatial Analysis extension toolset.
DEMs can be used to perform many geospatial and hydrological modeling.
Ranging from flood prediction and the physical development of urban and rural areas to watershed delineation and flood impact analysis for emergency preparedness.
Flood inundation modeling or flood inundation mapping is required to understand the effects of flooding in a particular area and on important structures such as streets, buildings, roads, and railways.
Quantifying the risk of flooding through flood models predicts inundation extents. This can be a crucial source of information for flood risk studies—especially as our world warms and sea levels rise.
Flood inundation models provide us with important information such as depth and the spatial extent of flooded zones—required by local authorities to inform citizens about significant flood-prone areas and adopt appropriate flood management strategies.
Accurate flood models require high resolution and highly accurate DEMs. According to this paper from 2019, current global DEMs do not capture the topographic details in floodplains—often leading to inaccurate prediction of flood extends by flood models (Shastry & Durand, 2019).
These data-scarce regions can be studied by creating flood inundation maps produced by combining flood extents with prediction modeling and modified DEMs.
Learn more about utilizing flood inundation observations to obtain floodplain topography in data-scarce regions in the paper.
Flood inundation modeling (Source)
It’s important to consider the DEM data source when using DEM to predict flood risk. This study looks at assessing coastal flooding, seal-level rise, or erosion risk and explores the role of DEM data source.
As mentioned, data-scarce areas exist, and to fill these voids, data modeling can be done. In recent years, advances in machine learning algorithms, affordable computing power, and big data availability has spurred the deep learning revolution across domains.
Machine learning techniques such as image inpainting can be used to fill data voids (Gavrill & Muntingh et al, 2019).
Image inpainting, similar to the term used in the art world to conserve damaged or incomplete images, image inpainting algorithms reconstruct terrain pixels in missing areas.
Machine learning and deep learning have advanced errors. For example, this study looks at CoastalDTM and the ability to reduce error when using DEM data from NASA’s SRTM.
It suggests that while DEM accuracy and spatial resolution are usually considered before being used for flood models, DEMs’ limitations arising from their original data source can often be overlooked during DEM selection (Coveney & Fotheringhham, 2011).
Elevation data can be used as an input to infrastructure projects, ensuring the construction of railways, pipelines, and power lines are not planned across high-slope areas.
Differences in elevation can also be tracked to ensure that geological hazards are monitored, predicted, and mitigated—minimizing damage and outages.
The National Geospatial-Intelligence Agency has teamed up with the University of Illinois, the University of Minnesota, and Ohio State University to produce digital elevation models of the world through Earth DEM.
The project feeds satellite images of an area from multiple angles into the Blue Waters supercomputer to create 3D models of the terrain.
As one of the most powerful and fastest supercomputers in the world, Blue Waters can perform over 13 quadrillion calculations per second.
EarthDEM will be a publicly available, 3D map of the globe and follows the complete mapping of the Arctic in 2017, as part of the ArcticDEM project—which helped scientists track changes, detect deforestation, ice cap collapse, and more.
Digital elevation models provide geologers with insights into tectonic plate boundaries. In the image above of East Africa, the outlines of the elevation highs demonstrating the thermal bulges and large lakes in East Africa are visible.
Scientists used DEM data to uncover and predict rift-to-ridge transition—leading to a new ocean formation as the African continent splits in two. The East African Rift system stretches from the Afar region of Ethiopia down to Mozambique. It’s an active continental rift that began millions of years ago and splits at 7mm annually.
In the recent study, the Victoria microplate, which lies between the eastern and western branches of the East Africa Rift System, was found to be rotating counterclockwise for the last two years with respect to the African Plate.
As well as these insights into the continents, regular eruptions of volcanoes along the rift adds to the belief that the continent may be splitting to form a new ocean.
ArcticDEM data supported the investigation of a possible second impact crater buried under more than a mile of ice in northwest Greenland.
Following the November 2019 announcement of a 19-mile wide crater beneath Hiawatha Glacier—the first meteorite impact crater ever discovered under Earth’s ice sheets—the second crater has a width of over 22 miles.
NASA glaciologist Joe MacGregor checked topographic maps of the rock beneath Greenland’s ice for signs of craters. He used imagery from NASA’s MODIS instrument and noticed a circular pattern around 114 miles southeast of Hiawatha Glacier.
By studying high-resolution DEM data of the entire Arctic using ArcticDEM, he noticed the same circular pattern—leading him to suspect a possible second impact crater.
DEM data is incredibly useful for delving into the past. When archaeologists scoured the Nefud desert in northern Saudi Arabia, they examined 376 footprints left in the mud of an ancient lake bed.
Among the footprints left by animals such as giant extinct elephants, camels, buffalo, and ancestors of modern horses, they spotted human footprints that may testify to humans’ presence in the region some 115,000 years ago.
Analysis using digital elevation models of three selected hominin tracks argues that anatomically modern humans created the seven footprints. If confirmed, these would be the oldest traces of Homo sapiens ever found on the Arabian Peninsula.
The first human footprint discovered at Alathar (left) and a digital elevation model that helped researchers discern its details (right) (Stewart et al., 2020)
Mapping our own planet isn’t where it ends. Thanks to the Mars Orbiter Laser Altimeter (MOLA) instrument, you can view the rugged terrain of Mars.
The instrument onboard the Mars Global Surveyor (MGS), a spacecraft launched on 7 November 1996, collected altimetry data until 30 June 2001. Alongside this, a laser altimeter onboard MGS determined the height of surface features on Mars.
Starting in 1998, MGS made pole-to-pole observations of the red planet. It’s goal? To map the entire Martian globe, laying the foundation of over ten more years of NASA missions. To determine the geology and perhaps history of Mars and its climate.
Scientists used MOLA to map out ancient Martian streams and explore what might have been. MOLA works by measuring the time that a pulse of light takes to leave the spacecraft, reflect off the surface of Mars, and return to MOLA’s collecting mirror. By multiplying the reflection time by the speed of light, scientists calculated Surveyor’s altitude above the local terrain to within around 30 meters.
As the spacecraft flew over hills, valleys, and craters, its altitude above the ground continuously changed. Such detailed maps help us construct a topographical atlas of the planet and understand the geological forces that shaped Mars.
The topography of Mars—white and red features are the highest in relative elevation and blue areas are lowest. (Source: NASA)
To summarize, a Digital Elevation Model or DEM is a generalized term for a raster data set with a regular grid of elevation information. DEMs are popular for calculations, manipulations, and further analysis of an area, and analysis based on the elevation.
Digital Surface Models or DSMs capture a surface—including natural and human-made structures such as vegetation and buildings. They illustrate reflective surfaces of all features elevated above the ‘bare earth’.
Lastly, Digital Terrain Models or DTMs are a bare-earth elevation model, and therefore free of vegetation, buildings, and other above-ground objects.
There are plenty of places to find global DEMs. From free satellite data to LiDAR sources. They come in a variety of file formats, from .csv and .tif to .txt and .dem. You’ll need a Geographic Information System (GIS) or other special application software since elevation data is not directly viewable in a browser. Some software programs that recognize DEM files include ArcGIS and QGIS 3.
QGIS 3.0 comes with a 3D layer view. This enables you to visualize GIS data in 3D—providing you with a more vivid visualization of data that contains elevation or height.
Errors in DEMs are usually classified as sinks or peaks. A sink is an area surrounded by higher elevation values. It’s also referred to as a depression or a pit. On the other hand, a peak, also known as a spike, is an area surrounded by cells of lower value. These should be removed before attempting to derive any surface information, creating a depressionless DEM.
DEMs can be used to perform many geospatial and hydrological modeling—including flood prediction and flood impact analysis for emergency preparedness.
Machine learning algorithms can be applied to derive more from DEM data—such as image inpainting to fill data voids and complete the picture.
Elevation data can be used as an input to infrastructure projects, geology studies, archaeological findings, and exploring planets other than our own such as Mars.
Ready to access elevation data as well as satellite imagery and machine learning algorithms? Sign up to UP42 and build geospatial solutions today.