No abstract

The vast majority of continental sediment delivered to the world's oceans moves by suspension in rivers. Depth‐integrated or point‐integrated bottle samplings are the traditional methods used to determine the mean concentration of suspended sediment in rivers. While there has been some investigation of the error associated with depth‐integrated sampling, the representativeness of a point‐integrated bottle sample has not been addressed in the literature. Here we analyze continuous hour‐long measurements of suspended sediment and grain‐size fractions collected using a LISST‐SL in the sand‐bed portion of the Fraser River, British Columbia, to determine an appropriate sampling time. The 2σ uncertainty range of individual 30 s samples varied from ±3% to ±33% about the observed mean, with a systematic increase toward the streambed. Mean concentrations for suspended sediment and grain‐size fractions were computed over increasing time periods and compared with a long‐duration mean concentration to determine when a sample becomes representative. A cumulative probability distribution was generated from multiple iterations of this process. All suspended sediment load and grain‐size fractions bear a low probability of representing the mean concentration over standard bottle sample durations. A probability >90% of representing the mean concentration and grain size of various fractions requires ∼570 s (9.5 min) of sampling. Sampling for a shorter period of 264 s (4.4 min) can yield a sample with 73% probability of representing the mean concentration.