[Readers should be aware that due to the different nature of size dependence of optical and acoustic backscatter sensors, changes in grain size distribution can cause the two data to not track. Neither sensor holds calibration with changing grain size. The LISST-ABS does hold calibration for grains above about 30 microns, while optics continue to lose sensitivity as 1/diameter.]

The largest dam removal project in the United States was completed in 2012 when the Elwha dam in northwest Washington State, US was removed to restore salmon runs. The Elwha dam had been built in 1913 to supply power to a paper mill downstream. An upstream dam, Glines Canyon, built in 1927 was removed subsequently also, thus making the Elwha a free flowing river again. [For a time-lapse of the dam removal, click here.]

Fig.1 The Elwha dam before removal, the sediment plume that followed removal, and the monitoring station. LISST-ABS is 2^nd from left.[click on images to see larger.]

The LISST-ABS single-point acoustic backscatter sediment sensor was installed at the USGS gage below the site of the Elwha dam (photo below). The sensor was programed to deliver acoustic estimates of sediment concentration at 15 minute intervals. Simultaneously, 2 optical turbidity meters and an ISCO water sampler provided independent estimates of suspended sediment concentration.

Fig.2: Intercomparison of LISST-ABS (Blk), and two turbidity sensors. Also shown are physical samples, and the river stage (Blu). (click to enlarge). Preliminary Data courtesy of USGS.

The data show general agreement between the 3 sensors. Only the LISST-ABS and Analite sensors could cover the full dynamic range of sediment concentration, the DTS-12 saturated part way. Some mismatch in the early part of the time series between all sensors and physical samples is seen in the first of two large peaks. Following the first main peak, all sensors show good agreement. Some of the disagreement between acoustic and optical backscatter data may well be attributed to changes in particle size distributions (PSD). We show 2 example PSD’s. The PSD on left (R24, unlabeled)) is from the last sample at end of first small peak, and the second is the physical sample used as calibration point (B24, unlabeled).

Fig.3: The PSD for the R24 data point (left) and B24 (right).

The changes in PSD shown in Fig.3 affect the calibrations of OBS and ABS significantly and are the likely the cause of disparity in Figure 2. In reality, the calibration for both sensors varies continuously over time as the PSD changes. The data show only minor disagreement.

These storm events left dramatic changes to the landscape in the photos taken by a webcam at the station.