Multifrequency analyses of 2011–2012 Peace River SWIPS frazil backscattering data

A four-frequency SWIPS (Shallow Water Ice Profiling Sonar) instrument was deployed in the Peace River during the November 2011–May 2012 period near the Town of Peace River. The deployment site was previously used for deployments of one- and two-frequency SWIPS instruments. Acoustic volume backscattering coefficient and ice thickness data were collected at frequencies 125 kHz, 235 kHz, 455 kHz and 774 kHz. The data, acquired as a function of height in the water column, were processed to test particle size, concentration and acoustic frequency restrictions identified in a closely-related laboratory study of backscattering in pseudo-frazil suspensions (Marko & Topham, 2015) prior to applying optimal processing strategies on the 2011–2012 frazil data. Although one of the transducers (235 kHz) displayed anomalous returns under certain environmental conditions, simultaneous measurements at 125 kHz, 455 kHz and 774 kHz were found to yield consistent measures of particle concentration, size distribution and fractional volume with accuracies limited, primarily, only by the 1 dB transducer calibration uncertainties. Fractional volumes were found to vary by less than an order of magnitude from interval to interval, with maximum values being approximately 10− 4 (0.01%) except within, roughly, 0.75 m of the river surface where interpretations were complicated by the frequent presence of surface ice. These results confirmed the effectiveness of multifrequency acoustic backscattering sonar (ABS) technology in applications to frazil monitoring problems and expanded its range of applicability beyond that conservatively estimated in the laboratory. The obtained data also provided an unprecedented basis for evaluating current river ice modeling assumptions. Comparisons with corresponding results from CRISSP model runs appear to be indicative of massive over-prediction of suspended frazil despite good model performance for simulating surface ice quantities. The origins of the possible discrepancies are discussed and accompanied by suggested options for improving model performance.