Premium
A comparison of selection at list time and time‐stratified sampling for estimating suspended sediment loads
Author(s) -
Thomas Robert B.,
Lewis Jack
Publication year - 1993
Publication title -
water resources research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.863
H-Index - 217
eISSN - 1944-7973
pISSN - 0043-1397
DOI - 10.1029/92wr02711
Subject(s) - sampling (signal processing) , stratified sampling , storm , environmental science , sampling design , sediment , statistics , hydrology (agriculture) , sample size determination , variance (accounting) , population , mathematics , geology , computer science , oceanography , geotechnical engineering , geomorphology , demography , accounting , filter (signal processing) , sociology , business , computer vision
Time‐stratified sampling of sediment for estimating suspended load is introduced and compared to selection at list time (SALT) sampling. Both methods provide unbiased estimates of load and variance. The magnitude of the variance of the two methods is compared using five storm populations of suspended sediment flux derived from turbidity data. Under like conditions, the SALT coefficient of variation was 1.4–7.7 times that of time‐stratified sampling. Time‐stratified sampling performs well if the range of sediment flux in each stratum is small. This requirement can be met by using small sample sizes in many short strata. Theoretically, SALT sampling has the potential for smaller sampling variance; however, it is difficult to select an auxiliary variable that predicts flux well under diverse flow conditions. An “optimum” auxiliary variable formed from the largest storm performed about as well as time‐stratified sampling for the larger storms. Time‐stratified sampling ensures that specimens are collected in all storms, facilitating load estimation for individual storms. In contrast, SALT can better allocate sampling resources over different size storms, enabling efficient estimation of the total load for longer periods. Because time‐stratified sampling is less sensitive to the way measurements are allocated to different parts of the population, it is preferred for estimating storm loads of multiple constituents from the same sample.