Net sediment transport patterns inferred from grain-size trends, based upon definition of “transport vectors”—comment

The use of spatial changes in grain-size parameters to determine net sediment transport paths has many applications, and any attempt to improve the concepts pioneered by Plumley (1948) and developed by, inter alia, McLaren and Bowles (1985) is welcomed. Gao and Collins (1991, 1992) proposed some modifications to the essentially one-dimensional approach of previous authors, allowing two-dimensional treatment of grain-size data. The new method purports to reduce implied predisposition in the selection of sampling lines inherent in the line-by-line approach of McLaren and Bowles (1985). However, not only does the proposed technique fail to identify the true transport directions, but also the way in which it is applied by Gao and Collins (1992) suggests that it may be open to bias. Data from Yangpu Harbour (southern China) were analyzed by Gao and Collins (1992) on the basis of two types of grain-size trends. Assuming that the net transport is from sampling sites 1 to 2 (denoted by subscripts), these are: Type 1:0-2 >/0"2; 1,1 /Sk 2 Type 2:0 -2 i> 0-22;/z 1 > #1,2; Sk 1 ~< Sk 2 where 02 is the sorting coefficient,/z is the mean grain size, and Sk is the skewness (in $ units). The method sets out by comparing each sampling site with its nearest neighbours on the basis of any of the two types of grain-size trends defined above, drawing vectors with unit length in the direction of decrease in the sorting coefficient. If more than one vector is identified at a sampling station, they are summed to produce a combined vector. To reduce the noise resulting from random fluctuations in the grain-size trends, a filtering or smoothing technique is employed in which the vector of each sampling site is averaged with those of neighbouring stations. The result is a residual pattern of transport vectors, which is subjected to a significance test to determine the extent of orderliness. To achieve this, Gao and Collins (1992) statistically compared the average length of the residual transport vectors with the average length of vectors derived from randomly rearranged samples. The main objection to the proposed method is that it ignores the low probability of defining true transport directions by comparing only two samples at a time. In Fig. 1, it is assumed that the net transport has taken place in the direction A to B, i.e. along a transport "front" rather than from point to point. If a vector is drawn between any two sampling sites (say from 1 to 3, or from 2 to 3) in the direction of improved sorting, the difference between the obtained and true directions can approach 90 ° , depending on the relative distribution of the sampling stations. Even if two or more adjoining sites are compared to the relevant