Height above channel.
Height above channel.
Which point in the channel?
I’ll use the South Fork of the Siletz River to illustrate these approaches. This is the only location I could find in the Siletz Basin with much of a flood plain. The south fork was dammed around 1922 to form a log-holding pond (Valsetz Lake), which was drained in 1988. The river now traces a meandering path across the former lake bed and is occupying an active flood plain that serves well to illustrate some of the issues encountered in seeking to characterize valley-floor landforms.
How does one determine height above a channel using a DEM?
This may seem straight forward: there are well-established methods for tracing
channel courses over a DEM, and once we have the channel locations, we can measure
the difference in elevation between the channel and any point on the valley
floor. In implementing such a scheme, however, several questions arise:
- Which point in the channel should we use?
- DEM data are noisy, particularly for lidar in wide channels where there are no reflections (water absorbs the infrared laser signal) and areas with tree cover. What is the appropriate elevation for a point in a channel with noisy DEM data?
Which point in the channel?
In NetMap, we currently use a distance-weighted mean of all channel elevations
within a certain distance of each valley-floor point, which I’ll illustrate,
but first I want to discuss several other approaches that might seem at first
like good options, but don’t always work as well as we would like. These are:
- Nearest channel location based on Euclidean (straight-line) distance.
- Channel location intersected by DEM-based flow path (e.g., D-8).
- Cross-valley transect.
I’ll use the South Fork of the Siletz River to illustrate these approaches. This is the only location I could find in the Siletz Basin with much of a flood plain. The south fork was dammed around 1922 to form a log-holding pond (Valsetz Lake), which was drained in 1988. The river now traces a meandering path across the former lake bed and is occupying an active flood plain that serves well to illustrate some of the issues encountered in seeking to characterize valley-floor landforms.
 |
| The major rivers draining Watershed Boundary Database HUC 17100204: Salmon to the north, Siletz in the center, and the Yaquina River in the south. |
Elevation data for these comparisons are from a LiDAR DEM for these basins, sampled to 2 meters horizontal resolution. We used the DEM to trace the channel networks, and then use the traced channel locations to reference elevation above the channel.
 |
| Shaded relief (left) and aerial photograph for a section along the South Fork Siletz River. Shaded relief derived from LiDAR bare-earth DEM. Blue dots show DEM cells identified as channel flow paths. |
Nearest channel location.
I'll start by using the closest channel point to estimate elevation above the channel. The figure below shows valley-floor DEM cells color coded to the nearest (by straight-line distance) DEM channel cell.
Estimating height above the channel based on the closest channel location can create abrupt and unrealistic changes that will confound interpretation of the results.
Flow path
By tracing the downhill flow path to a channel, each DEM cell can be associated with the channel location that - based on the DEM flow path - would receive surface flow originating from that cell. The elevation of that channel location can them be used to estimate height above the channel for the originating cell. This approach is used for the "HAND" (Height Above Nearest Drainage) method, as described in Nobre et al., 2011 and 2015. NetMap has routines to follow flow paths and delineate the area draining to every point in a channel (Mindi Sheer at NOAA Fisheries call the resulting polygons "drainage wings", a term that's stuck) . The figure below illustrates the contributing area to each channel cell using the D-8 flow-direction algorithm.
This is a distinctly different pattern for associating DEM cells to channel locations than obtained using the closest euclidean distance shown previously. Shape of the drainage wings is controlled by terrain topography; for the low-relief terrain of a floodplain, these patterns may not have a lot of bearing on actual flow paths. Using the elevation of the channel cell receiving flow from each drainage wing gives the height-above-channel map below.
As with closest euclidean distance, using the DEM flow path can result in adjacent valley-floor cells being associated with widely separated channel cells and consequent abrupt changes in estimated height above the channel.
Cross-valley transects
A transect drawn perpendicular to the downslope valley trend can be assigned the elevation of the channel-crossing, from which height above the channel can be determined along the transect. Height above the channel can then be determined for many closely spaced transects and interpolated to areas in between. This is the method used by Jones (2006) and by Vondrasek (2015). This method appears to work well, and I devised an automated method for generating such transects, used to estimate valley widths across the Oregon Coast Range as described in Clarke et al. (2008). However, we've since abandoned that approach, because meandering channels and wide valleys present too much ambiguity for generating unique height-above-channel values from cross-valley transects.
Distance-weighted mean
Currently, we estimate height above the channel based on the distance-weighted mean elevation over a channel segment(s). First, for each DEM cell, we find the minimum horizontal distance to a channel cell, as shown in the figure below. We then find all channel cells within a certain multiple of that distance. For example, if we use a multiple of 1.5 and the nearest channel location is 20 meters from a valley-floor cell, we find all channel cells within a circle of radius 30 meters centered over the cell. A weighted mean of the channel-cell elevations is then assigned to the valley-floor cell. Weighting is calculated as 1-D/R, where D is distance to the channel cell and R is the circle radius.
This produces the height-above-channel results below:
This covers our choice of which point in the channel to use for calculating height above the channel.
 |
| Same location, with valley floor DEM cells color coded to the closest DEM channel cell. |
See how adjacent points on the valley floor can be associated with widely separated channel locations? There may be substantial differences in elevation between these two points along the channel. Associated channel elevations are shown for the two points illustrated with white dots in the upper left; a difference of over one meter.
 |
| Height above the channel, based on the closest channel node. |
Flow path
By tracing the downhill flow path to a channel, each DEM cell can be associated with the channel location that - based on the DEM flow path - would receive surface flow originating from that cell. The elevation of that channel location can them be used to estimate height above the channel for the originating cell. This approach is used for the "HAND" (Height Above Nearest Drainage) method, as described in Nobre et al., 2011 and 2015. NetMap has routines to follow flow paths and delineate the area draining to every point in a channel (Mindi Sheer at NOAA Fisheries call the resulting polygons "drainage wings", a term that's stuck) . The figure below illustrates the contributing area to each channel cell using the D-8 flow-direction algorithm.
This is a distinctly different pattern for associating DEM cells to channel locations than obtained using the closest euclidean distance shown previously. Shape of the drainage wings is controlled by terrain topography; for the low-relief terrain of a floodplain, these patterns may not have a lot of bearing on actual flow paths. Using the elevation of the channel cell receiving flow from each drainage wing gives the height-above-channel map below.
As with closest euclidean distance, using the DEM flow path can result in adjacent valley-floor cells being associated with widely separated channel cells and consequent abrupt changes in estimated height above the channel.
Cross-valley transects
A transect drawn perpendicular to the downslope valley trend can be assigned the elevation of the channel-crossing, from which height above the channel can be determined along the transect. Height above the channel can then be determined for many closely spaced transects and interpolated to areas in between. This is the method used by Jones (2006) and by Vondrasek (2015). This method appears to work well, and I devised an automated method for generating such transects, used to estimate valley widths across the Oregon Coast Range as described in Clarke et al. (2008). However, we've since abandoned that approach, because meandering channels and wide valleys present too much ambiguity for generating unique height-above-channel values from cross-valley transects.
Distance-weighted mean
Currently, we estimate height above the channel based on the distance-weighted mean elevation over a channel segment(s). First, for each DEM cell, we find the minimum horizontal distance to a channel cell, as shown in the figure below. We then find all channel cells within a certain multiple of that distance. For example, if we use a multiple of 1.5 and the nearest channel location is 20 meters from a valley-floor cell, we find all channel cells within a circle of radius 30 meters centered over the cell. A weighted mean of the channel-cell elevations is then assigned to the valley-floor cell. Weighting is calculated as 1-D/R, where D is distance to the channel cell and R is the circle radius.
This produces the height-above-channel results below:
This covers our choice of which point in the channel to use for calculating height above the channel.
Comments
Post a Comment