Quantifying Recent Erosion and Sediment Delivery Using Probability Sampling: a Case Study

Estimates of erosion and sediment delivery have often relied on measurements from locations that were selected to be representative of particular terrain types. Such judgement samples are likely to overestimate or underestimate the mean of the quantity of interest. Probability sampling can eliminate the bias due to sample selection, and it permits the explicit expression of uncertainty with estimates of variance and confidence intervals. In this case study of the Van Duzen River Basin of northwestern California, erosion and sediment delivery originating after 1955 were estimated using a ratio estimator for air-photo-identifiable features greater than 3800 m3 and a stratified random sample (STRS) of field plots to quantify smaller erosion features. A method is also illustrated for estimating the variance of the ratio of two STRS estimates. Variance estimation permits one to partition the uncertainty associated with different kinds of erosion measurements. The most reliable estimates were obtained from discrete hillslope features, while earthflow yield estimates had extremely high variance. The variance estimates presented can be used to determine sample sizes and allocations required to meet the objectives of future investigations in similar terrain types. Published in 2002 by John Wiley & Sons, Ltd.