Good vibrations: big data impact bedload research

The fact that bedload transport is an inherently time-variant and location-sensitive fluvial process was revealed by systematic sampling, nine decades ago. Subsequent stream-wide measurements, that frequently incorporated lengthy collection periods, as well as the adoption of standardized sampling procedures, averaged out some temporal and spatial variability. However, continuous, highly resolved, long-period records of transport activity generated by active and passive bedload monitoring on diverse rivers have recently brought this variability into sharp focus. A defining characteristic of these ‘big data’ is that there are many possible bedload transport rates for each discharge and a wide range of discharges associated with each transport rate. Crucially, this incoherent scatter, which is generated by the various factors that affect bedload transport, can no longer be viewed as ‘noise’ that can be averaged out. We demonstrate that, even for small data sets, different methods of reporting and analyzing bedload transport records provide different perspectives on the bedload transport rate – flow relation. The inclusion/exclusion of zero values, present in all data that capture the intermittent nature of bedload transport, also affects the relation. To unlock the potential of big data and facilitate the development of bedload transport — flow relations from field measurements, it is essential to employ modes of analysis that are robust to outliers and do not assume associations between the independent and dependent variables are the same at all levels.