Reaches of a shallow $(<1.0\mathrm{m})$ , braided, sand-bedded river were surveyed in 2002 and 2005 with the National Aeronautics and Space Administration’s Experimental Advanced Airborne Research LiDAR (EAARL) and concurrently with conventional survey-grade, real-time kinematic, global positioning system technology. The laser pulses transmitted by the EAARL instrument and the return backscatter waveforms from exposed sand and submerged sand targets in the river were completely digitized and stored for postflight processing. The vertical mapping accuracy of the EAARL was evaluated by comparing the ellipsoidal heights computed from ranging measurements made using an EAARL terrestrial algorithm to nearby ($<0.5\mathrm{m}$ apart) ground-truth ellipsoidal heights. After correcting for apparent systematic bias in the surveys, the root mean square error of these heights with the terrestrial algorithm in the 2002 survey was $0.11\mathrm{m}$ for the 26 measurements taken on exposed sand and $0.18\mathrm{m}$ for the 59 measurements taken on submerged sand. In the 2005 survey, the root mean square error was $0.18\mathrm{m}$ for 92 measurements taken on exposed sand and $0.24\mathrm{m}$ for 434 measurements on submerged sand. In submerged areas the waveforms were complicated by reflections from the surface, water column entrained turbidity, and potentially the riverbed. When applied to these waveforms, especially in depths greater than $0.4\mathrm{m}$ , the terrestrial algorithm calculated the range above the riverbed. A bathymetric algorithm has been developed to approximate the position of the riverbed in these convolved waveforms and preliminary results are encouraging.

The Colorado River in Marble and Grand Canyons displays evidence of annual supply limitation with respect to sand both prior to [Topping et al, this issue] and after the closure of Glen Canyon Dam in 1963. Systematic changes in bed elevation and systematic coupled changes in suspended‐sand concentration and grain size result from this supply limitation. During floods, sand supply limitation either causes or modifies a lag between the time of maximum discharge and the time of either maximum or minimum (depending on reach geometry) bed elevation. If, at a cross section where the bed aggrades with increasing flow, the maximum bed elevation is observed to lead the peak or the receding limb of a flood, then this observed response of the bed is due to sand supply limitation. Sand supply limitation also leads to the systematic evolution of sand grain size (both on the bed and in suspension) in the Colorado River. Sand input during a tributary flood travels down the Colorado River as an elongating sediment wave, with the finest sizes (because of their lower settling velocities) traveling the fastest. As the fine front of a sediment wave arrives at a given location, the bed fines and suspended‐sand concentrations increase in response to the enhanced upstream supply of finer sand. Then, as the front of the sediment wave passes that location, the bed is winnowed and suspended‐sand concentrations decrease in response to the depletion of the upstream supply of finer sand. The grain‐size effects of depletion of the upstream sand supply are most obvious during periods of higher dam releases (e.g., the 1996 flood experiment and the 1997 test flow). Because of substantial changes in the grain‐size distribution of the bed, stable relationships between the discharge of water and sand‐transport rates (i.e., stable sand rating curves) are precluded. Sand budgets in a supply‐limited river like the Colorado River can only be constructed through inclusion of the physical processes that couple changes in bed‐sediment grain size to changes in sand‐transport rates.

This paper describes a non-contact methodology for computing river discharge based on data collected from small Unmanned Aerial Systems (sUAS). The approach is complete in that both surface velocity and channel geometry are measured directly under field conditions. The technique does not require introducing artificial tracer particles for computing surface velocity, nor does it rely upon the presence of naturally occurring floating material. Moreover, no prior knowledge of river bathymetry is necessary. Due to the weight of the sensors and limited payload capacities of the commercially available sUAS used in the study, two sUAS were required. The first sUAS included mid-wave thermal infrared and visible cameras. For the field evaluation described herein, a thermal image time series was acquired and a particle image velocimetry (PIV) algorithm used to track the motion of structures expressed at the water surface as small differences in temperature. The ability to detect these thermal features was significant because the water surface lacked floating material (e.g., foam, debris) that could have been detected with a visible camera and used to perform conventional Large-Scale Particle Image Velocimetry (LSPIV). The second sUAS was devoted to measuring bathymetry with a novel scanning polarizing lidar. We collected field measurements along two channel transects to assess the accuracy of the remotely sensed velocities, depths, and discharges. Thermal PIV provided velocities that agreed closely ( $R^2$ = 0.82 and 0.64) with in situ velocity measurements from an acoustic Doppler current profiler (ADCP). Depths inferred from the lidar closely matched those surveyed by wading in the shallower of the two cross sections ( $R^2$ = 0.95), but the agreement was not as strong for the transect with greater depths ( $R^2$ = 0.61). Incremental discharges computed with the remotely sensed velocities and depths were greater than corresponding ADCP measurements by 22% at the first cross section and <1% at the second.