Mapping potential off-channel habitat
Mapping valley-floor landforms without a machete.
Here I provide an informal, ongoing progress report for collaborators, potential collaborators, interested parties. Feel free to post comments, ask questions.
I've found that integrating HecRas with NetMap is going to take a bit of time, and present several data-availability challenges (which I intend to describe in subsequent blog posts), so in the interim, we seek simpler approaches to identify off-channel habitats that can be implemented now. In a NOAA webinar at the end of March, Jill Ory described her efforts to delineate off-channel features using a combination of aerial photography and the height-above-channel rasters provided in NetMap. The height-above-channel maps show elevation above the channel, either in meters or normalized by estimated bank-full-channel depth (here's a post that describes a bit about how we make these maps). Jill pointed out that many off-channel features are delineated by low values in these maps, which made me think that perhaps we can start with some simple analyses using height-above-channel maps to highlight potential off-channel habitats throughout a river network.
Height-above-channel maps have not yet been widely used for characterizing valley floors. Several groups have used such maps to estimate extent of inundation for mapping flood hazards (e.g., Nobre et al., 2016). Jones et al. (2007) showed that valley-floor landforms can be mapped manually using elevation ranges in high-resolution LiDAR DEMs. Jones (2016) used height-above-channel (he called it relative elevation), along with orthophotographs, to map side channels for a reach of the Dosewallips River on the Olympic Peninsula in Washington. For his Masters Thesis at UW, Chris Vondrasek found good correlation of inundation extent modeled with HecRas to a specific height above the channel determined using a LiDAR bare-earth DEM, which then aided in identification of off-channel areas flooded during high flows. Likewise, Fleenor (2015) in his Masters Thesis used Lidar-derived "height above river" to map floodplain extent to estimate current and potential abundance of off-channel habitat for Lagunitas Creek in northern California. These studies are encouraging enough to motivate moving forward.
To interpret height-above-channel values for off-channel habitat mapping, we need a way to translate elevations in meters to potential for inundation. One approach is to use a hydraulic model to estimate stage for discharges spanning a range of recurrence-interval floods to create rating curves for cross sections spanning a range of channel sizes (e.g., Nardi et al., 2006). The stage gives the height-above-channel value corresponding to the corresponding recurrence-interval flood. We will be able to take this approach once we've implemented HecRas network wide, but for the near term we need another approach.
A widely applied geomorphic rule of thumb is that channel "bank-full depth" stage corresponds to 1- to 2-year recurrence-interval floods (e.g., Leopold, Wolman, Miller, Fluvial Processes in Geomorphology, 1964), and it is widely observed that bank-full channel geometry (depth and width) vary systematically with contributing area, from which regional equations are developed for hydraulic geometry (bank-full channel width and depth, e.g., NRCS). Thus it seems natural that bank-full depth provides a scaling relation for flood stage; that is, stage for a given recurrence-interval flood corresponds to a certain multiple of bank-full channel depths. However, this is not a relationship that has been widely explored. In his channel classification scheme, Rosgen (1994) uses a height above the channel of two bank-full depths to identify the "flood-prone area", subject to "frequent flood (50 year return period) or less", although he does not provide empirical evidence to support this assumption. For the Oregon Coast Range, Clarke et al. (2008) found that the FEMA 100-year floodplain was best matched at five bank-full depths above the channel (using 10-m DEMs). Dodov and Foufoula-Georgiou (2006) show that the "geomorphic floodplain", based on topographic criteria, tends to lie at 0.6 bank-full-depths above the bank-full channel over a large range of basin sizes and in different geographic locations. This isn't a lot to go on, but these studies suggest that bank-full depth provides a logical method for scaling height-above-the-channel to stream size.
In the left figure above, I show height above the channel, in bank-full depths, for a portion of the South Fork Siletz. (The South Fork is one of few areas in the Siletz basin with much of an active floodplain). As Jill had pointed out, off-channel landforms are visible as low-lying features (blue), and to some extent, also visible in aerial photography. To better highlight these potential off-channel features, the image below shows the depth of inundation for floods of 5, 2, and 0.6 bankfull-depth stages, with the topographically defined channel (more on that later) colored purple.
Recall, Clarke et al. (2008) found that for the Oregon Coast Range 5 bankfull depths corresponded to the FEMA 100-year floodplain, Rosgen (1994) suggests that 2 bankfull depths delineates the flood-prone area, and Dodov and Foufoula-Georgiou (2006) found that the geomorphic floodplain typically lies at 0.6 bankfull depths above the channel. Perhaps we can use the extent of inundation for different bankfull stages as an indicator of current floodplain off-channel habitat (e.g., areas inundated at 0.6 bankfull depths) and of potential off-channel habitat (e.g., areas that could be inundated if the channel were aggraded, barriers removed, or low-lying terrain excavated).
To do this, we can use the depth and extent of inundation for a specified flood stage, as shown above, as an indicator of existing off-channel habitat. For areas that are not inundated, we can examine the height of the surface above that flood stage as an indicator of areas that could potentially provide additional off-channel habitat. This is illustrated in the figure below for a flood stage of 0.6 bankfull depths (that required to reach the typical geomorphic floodplain) and with height of the surface above the flood stage indicated by the depth of "excavation" required to get to the flood height. Here I've translated elevations from bank-full depths to absolute meters and plotted depth of "excavation" only up to one meter.
Note that with this scheme, broad areas of relatively flat terrain are included for both inundation and excavation, as highlighted by the white oval in the upper left. Flooded pastures may not provide the best off-channel habitat, so I'd like to limit the mapped zones to locally low-lying terrain, like side channels and sloughs. For that, we can use a measure of topographic relief and focus on areas with low elevations relative to local terrain. But how "low" and how "local"? Whether or not to consider any particular DEM grid point "low" depends on average elevations in the surrounding area, which depends on how large a surrounding area is examined. To deal with the issue of how low, I use the "deviation from mean elevation" (DEV), a relative measure normalized by topographic roughness (described in the text edited by Wilson and Gallant, Terrain Analysis, 2000).
Low values of DEV indicate low-lying terrain relative to the average local elevation. We still need to decide on how local; i.e., over what radius to measure elevations. In the area pictured above, channel widths run 10 to 15 meters, so it seems reasonable to delineate low-lying terrain of about this dimension. I used a radius of 50 meters, which gives a diameter of about 10 channel widths, and plotted only DEV values less than -0.5. This appears to work well for highlighting the sort of terrain I think we're interested in. Limiting areas mapped for inundation and excavation to DEM points with DEV values less than -0.5 gives the map shown on the right below.
Filtering by a measure of local relative relief limits delineation of inundation and excavation areas to sites more likely to provide off-channel habitat.
It would be useful to have measures of off-channel habitat that could be summarized over portions of the channel network. With NetMap, the DEM-traced channel network is represented in ArcGIS as a polyline vector object with channels divided into segments of about 100 meters length. As an initial measure, we can find the area of off-channel features associated with each channel segment. To do so, off-channel zones are associated with the nearest channel segment (using straight-line distance, not flow distance) and the surface area inundated or within a specified excavation depth is summed (by DEM cell) for each segment. However, I suspect that a cell inundated to 0.1 meter may provide less potential habitat than a cell inundated to 0.5 meters. Likewlise, I suspect that a location 0.1 meter above the flood height has greater potential for becoming off-channel habitat than a location 0.5 meter above. So I provided the option to weight each DEM cell area based on the magnitude of inundation or excavation. For inundated zones (falling below the DEV threshold), DEM-cell area is weighted by inundation depth divided by flood stage. This weighting is zero for zero inundation depth and one for inundation depth equal to flood stage. For excavation areas (falling below the DEV threshold), DEM-cell area is weighted by one minus excavation depth divided by a specified maximum depth (e.g., I used a maximum of one meter in the maps above). This value varies from one for an excavation depth of zero to zero when excavation depth reaches the maximum. These values can be summed channel segment by channel segment to provide a network-wide view of existing and potential off-channel features.
We can now use height-above-channel and deviation-from-mean-elevation rasters to identify topographic indicators of off-channel features, and then integrate these indicators over area (weighted by depth of inundation and excavation) to provide estimates of relative existing and potential off-channel habitat. As shown above, these estimates indicate that certain portions of the channel have existing available off-channel features (for a flood stage of 0.6 bankfull depths), some portions have both current and potential off-channel features, and other portions lack currently available (at this flood stage) features but have locations conducive to becoming available (based on elevation of the surface relative to flood stage). Calculated inundation and excavation values for a section of channel in the lower center of the image above (in the box) are shown below.
Such maps might provide estimates of relative inundation frequency. Field visits will determine if the mapped features correspond to identifiable off-channel landforms (side channels, wetlands) on the ground.
Here I provide an informal, ongoing progress report for collaborators, potential collaborators, interested parties. Feel free to post comments, ask questions.
Remote mapping of off-channel habitats
A primary goal of mapping valley-floor landforms is to estimate the abundance of off-channel habitat within a basin and to identify opportunities for increasing access to potential off-channel habitats. Valley-floor landforms that can provide off-channel habitat for salmon include side channels, alcoves, sloughs, beaver ponds, and other floodplain features (Roni et al., 2016 provide a nice review). Off-channel areas that remain submerged over summer, generally from groundwater influx, provide important rearing habitat. In winter, features inundated during flood events provide areas with lower flow velocity than through the mainstem and offer refuge for overwintering juvenile salmon. For a coho-habitat workshop last September, Brian Cluer presented an analysis using a 2-D hydraulic model (HecRas) that shows how low-flow-velocity areas available to juvenile fish vary with flood discharge (his presentation is available to view here). His presentation sparked interest in developing routines within NetMap to facilitate running hydraulic models over entire river networks to aid in identifying existing and potential low-velocity refuge areas.I've found that integrating HecRas with NetMap is going to take a bit of time, and present several data-availability challenges (which I intend to describe in subsequent blog posts), so in the interim, we seek simpler approaches to identify off-channel habitats that can be implemented now. In a NOAA webinar at the end of March, Jill Ory described her efforts to delineate off-channel features using a combination of aerial photography and the height-above-channel rasters provided in NetMap. The height-above-channel maps show elevation above the channel, either in meters or normalized by estimated bank-full-channel depth (here's a post that describes a bit about how we make these maps). Jill pointed out that many off-channel features are delineated by low values in these maps, which made me think that perhaps we can start with some simple analyses using height-above-channel maps to highlight potential off-channel habitats throughout a river network.
Height-above-channel maps have not yet been widely used for characterizing valley floors. Several groups have used such maps to estimate extent of inundation for mapping flood hazards (e.g., Nobre et al., 2016). Jones et al. (2007) showed that valley-floor landforms can be mapped manually using elevation ranges in high-resolution LiDAR DEMs. Jones (2016) used height-above-channel (he called it relative elevation), along with orthophotographs, to map side channels for a reach of the Dosewallips River on the Olympic Peninsula in Washington. For his Masters Thesis at UW, Chris Vondrasek found good correlation of inundation extent modeled with HecRas to a specific height above the channel determined using a LiDAR bare-earth DEM, which then aided in identification of off-channel areas flooded during high flows. Likewise, Fleenor (2015) in his Masters Thesis used Lidar-derived "height above river" to map floodplain extent to estimate current and potential abundance of off-channel habitat for Lagunitas Creek in northern California. These studies are encouraging enough to motivate moving forward.
To interpret height-above-channel values for off-channel habitat mapping, we need a way to translate elevations in meters to potential for inundation. One approach is to use a hydraulic model to estimate stage for discharges spanning a range of recurrence-interval floods to create rating curves for cross sections spanning a range of channel sizes (e.g., Nardi et al., 2006). The stage gives the height-above-channel value corresponding to the corresponding recurrence-interval flood. We will be able to take this approach once we've implemented HecRas network wide, but for the near term we need another approach.
A widely applied geomorphic rule of thumb is that channel "bank-full depth" stage corresponds to 1- to 2-year recurrence-interval floods (e.g., Leopold, Wolman, Miller, Fluvial Processes in Geomorphology, 1964), and it is widely observed that bank-full channel geometry (depth and width) vary systematically with contributing area, from which regional equations are developed for hydraulic geometry (bank-full channel width and depth, e.g., NRCS). Thus it seems natural that bank-full depth provides a scaling relation for flood stage; that is, stage for a given recurrence-interval flood corresponds to a certain multiple of bank-full channel depths. However, this is not a relationship that has been widely explored. In his channel classification scheme, Rosgen (1994) uses a height above the channel of two bank-full depths to identify the "flood-prone area", subject to "frequent flood (50 year return period) or less", although he does not provide empirical evidence to support this assumption. For the Oregon Coast Range, Clarke et al. (2008) found that the FEMA 100-year floodplain was best matched at five bank-full depths above the channel (using 10-m DEMs). Dodov and Foufoula-Georgiou (2006) show that the "geomorphic floodplain", based on topographic criteria, tends to lie at 0.6 bank-full-depths above the bank-full channel over a large range of basin sizes and in different geographic locations. This isn't a lot to go on, but these studies suggest that bank-full depth provides a logical method for scaling height-above-the-channel to stream size.
 |
| Valley-floor elevation, in bank-full-depths above the channel. Aerial photo of the same area on the right. |
Recall, Clarke et al. (2008) found that for the Oregon Coast Range 5 bankfull depths corresponded to the FEMA 100-year floodplain, Rosgen (1994) suggests that 2 bankfull depths delineates the flood-prone area, and Dodov and Foufoula-Georgiou (2006) found that the geomorphic floodplain typically lies at 0.6 bankfull depths above the channel. Perhaps we can use the extent of inundation for different bankfull stages as an indicator of current floodplain off-channel habitat (e.g., areas inundated at 0.6 bankfull depths) and of potential off-channel habitat (e.g., areas that could be inundated if the channel were aggraded, barriers removed, or low-lying terrain excavated).
To do this, we can use the depth and extent of inundation for a specified flood stage, as shown above, as an indicator of existing off-channel habitat. For areas that are not inundated, we can examine the height of the surface above that flood stage as an indicator of areas that could potentially provide additional off-channel habitat. This is illustrated in the figure below for a flood stage of 0.6 bankfull depths (that required to reach the typical geomorphic floodplain) and with height of the surface above the flood stage indicated by the depth of "excavation" required to get to the flood height. Here I've translated elevations from bank-full depths to absolute meters and plotted depth of "excavation" only up to one meter.
Note that with this scheme, broad areas of relatively flat terrain are included for both inundation and excavation, as highlighted by the white oval in the upper left. Flooded pastures may not provide the best off-channel habitat, so I'd like to limit the mapped zones to locally low-lying terrain, like side channels and sloughs. For that, we can use a measure of topographic relief and focus on areas with low elevations relative to local terrain. But how "low" and how "local"? Whether or not to consider any particular DEM grid point "low" depends on average elevations in the surrounding area, which depends on how large a surrounding area is examined. To deal with the issue of how low, I use the "deviation from mean elevation" (DEV), a relative measure normalized by topographic roughness (described in the text edited by Wilson and Gallant, Terrain Analysis, 2000).
Low values of DEV indicate low-lying terrain relative to the average local elevation. We still need to decide on how local; i.e., over what radius to measure elevations. In the area pictured above, channel widths run 10 to 15 meters, so it seems reasonable to delineate low-lying terrain of about this dimension. I used a radius of 50 meters, which gives a diameter of about 10 channel widths, and plotted only DEV values less than -0.5. This appears to work well for highlighting the sort of terrain I think we're interested in. Limiting areas mapped for inundation and excavation to DEM points with DEV values less than -0.5 gives the map shown on the right below.
Filtering by a measure of local relative relief limits delineation of inundation and excavation areas to sites more likely to provide off-channel habitat.
It would be useful to have measures of off-channel habitat that could be summarized over portions of the channel network. With NetMap, the DEM-traced channel network is represented in ArcGIS as a polyline vector object with channels divided into segments of about 100 meters length. As an initial measure, we can find the area of off-channel features associated with each channel segment. To do so, off-channel zones are associated with the nearest channel segment (using straight-line distance, not flow distance) and the surface area inundated or within a specified excavation depth is summed (by DEM cell) for each segment. However, I suspect that a cell inundated to 0.1 meter may provide less potential habitat than a cell inundated to 0.5 meters. Likewlise, I suspect that a location 0.1 meter above the flood height has greater potential for becoming off-channel habitat than a location 0.5 meter above. So I provided the option to weight each DEM cell area based on the magnitude of inundation or excavation. For inundated zones (falling below the DEV threshold), DEM-cell area is weighted by inundation depth divided by flood stage. This weighting is zero for zero inundation depth and one for inundation depth equal to flood stage. For excavation areas (falling below the DEV threshold), DEM-cell area is weighted by one minus excavation depth divided by a specified maximum depth (e.g., I used a maximum of one meter in the maps above). This value varies from one for an excavation depth of zero to zero when excavation depth reaches the maximum. These values can be summed channel segment by channel segment to provide a network-wide view of existing and potential off-channel features.
We can now use height-above-channel and deviation-from-mean-elevation rasters to identify topographic indicators of off-channel features, and then integrate these indicators over area (weighted by depth of inundation and excavation) to provide estimates of relative existing and potential off-channel habitat. As shown above, these estimates indicate that certain portions of the channel have existing available off-channel features (for a flood stage of 0.6 bankfull depths), some portions have both current and potential off-channel features, and other portions lack currently available (at this flood stage) features but have locations conducive to becoming available (based on elevation of the surface relative to flood stage). Calculated inundation and excavation values for a section of channel in the lower center of the image above (in the box) are shown below.
Such maps might provide estimates of relative inundation frequency. Field visits will determine if the mapped features correspond to identifiable off-channel landforms (side channels, wetlands) on the ground.
Comments
Post a Comment