Brashear Island Volumetric Interpolation
We selected a section of the Clinch River near Brashear Island, roughly between river mile 9.5 and 11.5, for additional intensive volumetric interpolation of sub-bottom sediments. This portion of the river contains several shallow sections, and is potentially contaminated with cesium-137 and mercury.
The Rivers and Harbors Act of 1899 and the Tennessee Valley Authority Act of 1933 charge the Army Corps of Engineers and the Tennessee Valley Authority, respectively, with maintaining sufficient hull clearance to permit navigability to boat and barge traffic. Assuming a winter draw-down water elevation of 735 feet, the mandated hull clearance requires that the navigable portion have a bottom elevation of 724 feet or less.
We prepared several 3D views of the interpolated bottom model for the Brashear Island area; one view to the northeast, the view to the north shown above, and one to the southwest. In these views, the interpolated bottom is shown surrounded by the Digital Elevation Model (DEM) for the area, and the DEM is draped with composite colors from a recent Landsat image. The colors on the bottom reflect hull clearances at a water level of 735 feet. Only yellow and green areas exceed the required hull clearance for legal navigability; red areas do not meet the legal requirement according to the interpolated bottom model.
These views indicate that the shallow bottom in this area would make it difficult for barge traffic to pass through this area at low water while remaining in yellow or green areas. Moreover, the construction of a new barge terminal just upstream of this area, at the K-770 site, is being considered, and would greatly increase barge traffic. Therefore, this area may be a likely candidate for future bottom dredging operations to maintain legal hull clearances.
Light blue spheres indicate locations of Army Corps of Engineers sub-bottom acoustic sounding data. The Army Corps of Engineers boat Waterways Explorer made 3 sounding passes through this portion of the river, resulting in 3897 discrete acoustic density data points for sub-bottom sediments. These are a portion of the data that have been presented visually in the ``tennis'' animation.
The portion of the river under consideration consists of 408 rows x 374 columns of cells, where each cell is 5 m on a side. Because the acoustic sounding reflections produced discrete density data points at about 0.5 foot vertical increments, it was decided that vertical resolution should be 0.5 foot. Thus, the Brashear area under consideration contains
408 x 374 x 100 = 15,259,200 voxels
A voxel is a volumetric pixel, and, in this case, each voxel contains
5 m x 5 m x 0.5 ft x 0.3048 m/ft = 3.81 m3/voxel
At this resolution, then, the volume contains more than 15 million voxels.
We employed a regularized spline with tension and smoothing to interpolate the sub-bottom densities in 3 dimensions. This algorithm is a part of the GRASS GIS as s.surf.3d, a variant of s.surf.tps. The regularized spline algorithm was developed originally by Mitasova and Mitas, and was converted to a GRASS utility and extended into 3 dimensions by Helena Mitasova and Irina Kosinovsky.
A mask was employed within GRASS to constrain the interpolation to areas within the vertical projection of the shoreline only. However, all voxels within this ``cookie-cutter'' volume, including water, were interpolated. Thus the spline interpolator was free to extrapolate sediment densities up into the water column. Obviously, then, the river channel had to be ``erased'' back out of the volume in a subsequent step.
An alternative approach would have been to ``pad'' the data supplied to the interpolator with data points in the water column whose values were set to 1.0 g/cm3, the density of water. Under this approach, there would presumably have been no need to subsequently ``remove'' the water in the channel. This latter approach would have maximized the presence of softer sediments at the sediment-water interface, since the interpolator would force a smooth transition from sediment to water.
Because these soft sediments preferentially bind contaminants like cesium-137 and mercury, we were particularly interested in getting an accurate representation of the quantity and location of these softer classes. For this reason, we elected not to ``pad'' the data set with water values, but allowed the spline to interpolate up into the water column. The sub-bottom densities resulting from this approach, therefore, will show soft sediments only near places where soft sediments were actually measured, and the interpolation will be conservative with regard to amount of soft material present.
In order to remove the water and the extraneous sediment from the interpolated volume, we used a variant of standard GIS map algebra extended into the third dimension. Software from Advanced Visual Systems (AVS) was used to manipulate and visualize the results of the volumetric interpolation.
Starting with the interpolated bottom model shown in the ``pinhead'' animation, we created a volume mask, consisting of only ones and zeros. We used UNIX shell programming to create a ``bathtub-like'' volume mask for the river sub-bottom. Because the deepest acoustic soundings penetrated about 15 feet into the sub-bottom sediments, we included in the volume mask all voxels less than 15 feet below the interpolated river bottom elevation. This helps to prevent the inclusion of sediment which lies below the extent of the sampled data. Each voxel within this mask was given a value of 1, and each voxel outside the mask was given a zero. Then, within AVS, we multiplied the entire interpolated ``cookie-cutter'' volume by the mask volume, thus retaining only the desired portion of the interpolated volume and preserving the bottom contours.
Using AVS, we produced this set of 3 x-ray views along each of the major axes in order to generate insight into the composition of the volume. This technique looks at the row of voxels ``behind'' each screen pixel, and creates a new pixel based on the mean of these voxels. The resulting images are similar to a colorized x-ray in each direction through the volume. Although red is most dense and blue is least dense, the color scale is arbitrary, reflecting x-ray density rather than sediment density.
In order to increase understanding of the contents of the Brashear sub-bottom volume, we generated a series of volume animations. These animations are in the mpeg format, and require that an mpeg player be installed along with your browser. If you have a Macintosh, Sparkle can display mpeg movies.
Click on this 3-D filmstrip to see an animation of the Brashear sub-bottom volume as an orthogonal cut plane iterates upwards from the bottom of the volume (2.4 MB MPEG movie). In this animation, the sediment density is mapped onto the surface of a horizontal plane as this plane is slowly moved from the bottom of the volume to the top. The original Army Corps acoustic sounding data are shown as stacks of spheres whose colors are mapped to their respective densities in order to indicate the locations of actual data points. The 3 sounding passes made by the Waterways Explorer are clear from this animation. As the animation proceeds, notice that most of the actual data are limited to deeper and more central portions of the river.
The second animation was designed to indicate the composition and arrangement of 5 sediment classes within the sediment volume. The sediment classes chosen reflect simple descriptive classifications of the continuous sediment density data. The class names and density ranges were recommended by the Army Corps of Engineers as follows:
Sed Type Wet Density Range (g/cm3) Fluid Muds < 1.2 Silty Clay 1.2 - 1.4 Clayey Silt 1.4 - 1.6 Silty Sands 1.6 - 1.8 Sand/Gravel/Rock > 1.8
These density ranges represent the same sediment classifications used in the ``tennis'' animation.
Click on this 3-D filmstrip to see particular sediment classes gradually become transparent and fade from view, revealing the internal arrangement within the volume (6.4 MB MPEG movie). This dissolution continues until only voxels in the darkest red Fluid Mud category remain. Then, in a different order, each of the sediment classes materializes back into place. The order and colors of these transitions are:
Clayey Silt yellow fades out Silty Sands lt green fades out Silty Clay lt red fades out Sand/Gravel/Rock dk green fades out Silty Clay lt red fades in Sand/Gravel/Rock dk green fades in Silty Sands lt green fades in Clayey Silt yellow fades in
A histogram decomposition of the Brashear volume reveals how many voxels are in each of the above sediment classes. From the number of voxels in each class, the total volume of that sediment class can be calculated. Moreover, because the average wet density for each class is known, an estimate of the wet weight of each sediment class can be estimated.
Thus, the volume and weight of each sediment class seen in the animation are given below:
Sed Type # Voxels Volume (m3) Wet Weight (kg) Fluid Muds 199 758 8.9E+05 Silty Clay 54,991 209,515 2.8E+08 Clayey Silt 468,531 1,785,103 2.7E+09 Silty Sands 268,098 1,021,453 1.7E+09 Sand/Gravel/Rock 51,336 195,590 3.6E+08 Totals 843,155 3,212,420
A statistical distribution of voxels by density reveals a bimodal density distribution within the volume. No statistical artifacts from the interpolation are apparent.
Finally, the second animation shows a series of vertical longitudinal slices through the Brashear area. These longitudinal slices clearly show the nearly complete sediment density sorting and layering. Also revealed is a large protruding ``hump'' of very dense material in the center of the bottom channel. This same hump can be seen in the first animation, where it is revealed that the hump contains several several acoustic soundings. The longitudinal slices then recede in slightly different locations as the Brashear volume is re-assembled.
Click on this 3-D filmstrip to see an animation presenting the Brashear volume presented in an end-on view, with the slicing plane passing across the river channel (3.3 MB MPEG movie). Notice the clear layering of softer, less dense material on top of more dense, harder sediments. Occasionally, a pocket of Sand/Gravel/Rock can be seen inside an area of Silty Sand.
Much of the Silty Clay and Fluid Muds are found along the surface of the bottom, often along the sides and in the deepest portions of the channel. Future dredging operations could disturb these softest sediments and potentially re-suspend contaminants. The Brashear volume dataset could be useful for planning such potential dredging operations. With the same technique used above, volumes and weights of material to be removed could be calculated for particular scenarios. In fact, dredging operations could be designed using the Brashear volume model to optimize particular parameters of interest. For example, the dredging path could be designed to maximize the amount of Fluid Mud and Silty Clay that is to be removed, or, alternatively, to minimize the amount of these soft materials to be newly exposed in the dredged channel.
For an interactive interface to the Brashear volumetric data, visit the Virtual Brashear Isosurfer and the Virtual Brashear Isomovie pages on this server.
Draped 3-D views of Brashear area were produced by sg3d, a GRASS graphics program specific to Silicon Graphics workstations written by Bill Brown, and were rendered on an Indigo 2 borrowed from Dr. Robert J. Luxmoore. All volumetric graphics were produced using AVS on the Indigo 2.
For additional information contact:
(865) 574-1902
