In this sequence the river has been drained to make the riverbed visible. We are flying at a fixed elevation of 750 feet above MSL. The river bed elevation has been interpolated using a 3-D spline with tension (Mitasova and Mitas, 1993). The spline surface acts as a bending and twisting flexible surface which is trying to pass close to all known data points. A tension parameter controls the flexibility of the surface, changing its behavior from that of a thin rubber membrane to a thick inflexible plate. Riverbed colors are again the potential deposition/erosion filter results.

Virtual push-pins mark locations where actual river depth was supplied to the spline interpolation program. Two types of depth data were available; high quality ``primary'' data from the Army Corps of Engineers depth soundings, sedimentation cross-section readings, and depths recorded at actual sampling locations, and ``secondary'' depth data, such as contour lines scanned from pre-impoundment USGS quad maps.

In order to minimize dilution of the primary data with secondary information, a 20 m buffer zone was placed around all primary data points, and only secondary data points that fell outside this buffer zone were included in the input to the spline interpolator. Thus the secondary information was used to fill in spatial ``gaps'' for which no primary data existed. In addition, to minimize computation time, only points at least 10 m apart were included in the spline.

At each of these locations, the height of the interpolated surface was subtracted from the true depth. A green pin indicates that this delta was positive, and a red pin indicates that this difference was negative. The absolute value of this difference was added to the height of the surface, and the pin heads were positioned at this height.

Thus, the head of the virtual push-pin is green if the interpolated bottom surface should have been higher at the pin location, and red if the surface should have been lower at this spot. The surface should have passed through the center of the heads of the green pins; it should have passed through a spot as far below the surface of the red pins as the red pins are tall.

Short alternating red and green pins, which are often seen in the center of the bottom, indicate a nearly perfect fit of the interpolated surface with the actual known depths. Alternating red and green pins of greater height indicate that the spline has ``smoothed out'' a rough bottom (or a series of noisy depth measurements). A perfect match (to the nearest hundreth of a foot) between spline depth and actual depth data is indicated by white pins whose heads rest on the bottom.

Often throughout the sequence there are two rows of green pins on each bank, while the next inner rows of pins are red. Depths for the outermost rows of pins were established by the known water surface elevation at the shoreline. The red rows surrounded by green rows at the banks indicate that the tension setting used was too high to allow the surface to bend deep enough for the red row, then come up quickly enough to meet the outer green rows. This is exacerbated where the river is narrow and deep.

Moreover, smaller islands, like Grubb Island (frame 94) and Brashear Island (frame 470), have a mesa-like appearance, with straight vertical sides and a flat top. The flat top is because the portion ot these small islands which projects above the waterline is smaller than the 30m resolution of the DEM data which were used to provide the above-water vertical relief; the vertical sides result from a spline tension setting that is too low.