Computing erosion modeling inputs from a DEM

The digital elevation model is the fundamental data layer needed for erosion modeling.

Types of DEMS

Assessing Quality

Deriving Erosion Parameters

 

Types of digital elevation data and their suitability for erosion modeling

• The United States Geological Survey (USGS) has produced several different digital elevation data sets that can be downloaded free either from the USGS web site or from the geo-community web site. The latest product, National coverage DEM (NED) merges the best elevation data available into a seamless, 30m resolution raster format. The DEM for a user specified area can be downloaded through the Seamless Data Distribution System (SDDS) website in a variety of formats including ArcGrid, Floating Point and BILS. Older DEMs produced by methods that are now obsolete have been filtered during the NED assembly process to minimize artifacts. However, some of those artifacts are still visible: 

  

  For example, this DEM has been assembled from at least 3 sets of elevation data with a wide range of quality, and most of it would be unsuitable for erosion modeling. Also note that the 30m resolution makes this data suitable only for watershed, regional scale modeling.

  For some areas, a new 10m resolution DEM is now available. It is derived from the same contours as the 30m DEM, however more accurate interpolation was used and elevations are available with vertical precision of 0.1m rather than 1m, reducing the "step" effect. In spite of using the same data source, the quality of this DEM is superior compared to the 30m DEM (see USGS documentation about the accuracy). Besides the increased precision, accuracy and resolution, the DEM provides a much more detailed description of terrain features. A few new problems due to the TIN-based interpolation method were observed, such as waves along contours in steep areas, noise in flat areas, leading to noisy and inconsistent contours and artificial dams in valleys due to triangulation (see 10m DEM sample data). However the magnitude of these artifacts is much smaller than in previous models. The negative impact on modeling can be reduced by smoothing and/or reinterpolating the data. The 10m DEMs have to be ordered directly from the USGS. 

   

• Interferometric Synthetic Aperture Radar for Elevation (IFSARE) uses interferometric radar techniques aboard an airborne platform to collect and record phase history data, which is then ground processed into Digital Elevation Models. A by-product of  DEM generation is a Synthetic Aperture Radar (SAR) imagery with an image resolution of approximately 3 meters. The data are georeferenced using a Differential Global Positioning System (DGPS). The main advantage of the IFSARE technology is that the data can be collected in all weather, day/night, and through atmospheric obscurants. The areas of insufficient radar return, such as water bodies and slope-induced shadows, are represented as null data. Less accurate radar data are obtained from a satellite by RADARSAT, see the Canadian Space Agency web site.  It is important to note that the data represent the terrain surface including vegetation, buildings and other features. This Digital Surface Model is not suitable for applications where bare soil surface is needed, such as erosion modeling. Radar data are available at 2.5m raw resolution, however, due to the speckle noise, the effective resolution is 10m (although smaller features can be recognized), with a  vertical accuracy of about 1m. These radar data are adequate for areas without vegetation, however it is unsuitable for erosion modeling for forested landscapes. Example.

• Light Detection and Ranging (LIDAR). LIDAR is an active remote sensing system using pulses of light to illuminate the terrain. LIDAR data collection involves mounting an airborne laser scanning system onboard an aircraft along with a kinematic Global Positioning System (GPS) receiver to locate an x and y position and an inertial navigation system to monitor the pitch and roll of the aircraft. Ideally, LIDAR systems should result in DEMs with 1-5m horizontal resolution, and 15cm vertical. By accurately measuring the round trip travel time of the laser pulse from the aircraft to the ground, a highly accurate spot elevation can be calculated. The major advantage compared to IFSARE is that LIDAR provides both DSM (first return) and bare soil surface DEMs by post-processing the last return. Note that bridges are captured as dams, but this is an issue for hydraulic modeling of rivers not overland flow. LIDAR collection methods result in data voids (no return) from asphalt, water, or other highly reflective surfaces. The DEMs include artifacts, contours are noisy, and some smoothing may be required. DEMs often include a large number of depressions - both automatic and manual post-processing may be needed. There are few documented applications of these data in modeling.  Example.

• Digital photogrammetry (points/TIN, contours) provides DEMs at 5-10m resolution which are usually  generated from aerial photography. The quality and suitability depends on the interpolation method and resolution. For example, the original DEM for Ft. Hood had a vertical precision of 1m, which is insufficient for correct estimation of slope pattern (and consequently also the erosion/deposition), especially in areas with relatively flat terrain. The negative effect of insufficient vertical resolution can be partially reduced by reinterpolating the DEM. There is a dramatic difference between the erosion estimates from the 1m integer DEM and floating point precision (real number) DEM derived from the same data. The elevation data obtained for Fort Polk were in the form of contours and TIN points so we interpolated the DEM using the RST method (see Mitasova et al. 1999).

• Real Time Kinematic Survey data provides high resolution data with centimeter vertical accuracy. These data are excellent for high resolution erosion modeling for smaller areas and assessments of terrain dynamics caused by  erosion. This is a fast efficient and accurate method for obtaining data with great detail for smaller areas.

• Ground survey data are the most accurate, but data collection is much more labor intensive than the previous methods.

Assessing the quality of DEMs

• One of the most important data layers used in evaluating erosion risk is the digital elevation model, so it is important to know how accurate and precise your data are so you can better judge the validity of results.
• Besides standard accuracy assessments using control points, the geometry can be evaluated using shaded 3D views, aspect, slope and curvature maps and flow tracing.
• Vertical precision in cm is needed.  Meter precision is not sufficient as it  creates a "step" effect that leads to zero slopes and a contour like slope pattern, which ultimately underestimats erosion.
• Waves along contours create artificial erosion/deposition patterns.
• Artificial peaks and pits distort flow.
• Reinterpolating, smoothing and stream enforcement can improve the properties of a DEM.

Deriving the model parameters:

• Slope and Aspect are components of elevation surface gradient, representing its magnitude in terms of steepest slope angle and its direction. Gradient controls the direction and magnitude of flow velocity, as well as the rate of soil detachment, therefore an accurate estimate is crucial. Various algorithms are used in GIS to estimate these parameters, however weighted average of differences computed from a 3x3 neighborhood of each grid cell is the most common and usually sufficient approach (e.g. SLOPE in ArcGIS,  r.slope.aspect in GRASS)

Slope computed from a 3m DEM

Aspect computed from a 3m DEM

• Upslope area can be computed using several different algorithms.
  • D8 or single flow direction (SFD) distributes flow to a single cell with the lowest relative elevation. This is the most common and now obsolete method. It was suitable for watershed delineation and stream network extraction from low resolution DEMs. It is used in flowacc in ArcView, r.watershed in GRASS, and Rivertools. It is not suitable for hillslope erosion modeling, as it produces artificial flow patterns because it is incapable of simulating dispersal flow and provides only a crude estimate of flow direction (FIGURE)
  • Multiple flow direction  (MFD) distributes the flow proportionally to all cells with lower elevation; it adequately simulates the dispersal flow and is suitable for hillslopes. However the concentrated flow is often too wide (a switch to SFD for concentrated flow is possible as implemented in TERRAFLOW). Examples of implementation: r.mapcalc tutorial through map algebra script (Westervelt and Shapiro 1992), TERRAFLOW (Arge et al. 2001)
  • Vector-grid algorithm (with diffusion option) 

  
  Upslope area computed from a 3m DEM

   

Up
Computing erosion modeling inputs from a DEM
Deriving erosion modeling inputs from a land cover
Erosion modeling inputs