Representativity and Accuracy of Measurements of Soil Loss from Runoff Plots

ABSTRACT RESEARCH designed to demonstrate soil loss sampling was carried out on a Submediterranean rangeland hillslope covering 8250 m2 on which there was no grazing or grass-burning. The study is based on interrill, pre-rill and rill erosion measurements using Modified Gerlach troughs and involves random, systematic and cluster sampling procedures. A statistical method is introduced to test the representativity of the various erosion samples and to calculate the accuracy of the plot data for different sample sizes. Both the representativity and the accuracy of the erosion data are evaluated for separate storms ranging from frequent low magnitude to occasional high magnitude. A large scatter was found in the erosion data: soil loss estimates from samples which deviate by 50 to 100% from the (sub)population soil loss are the rule rather than the exception. Systematic soil loss sampling is more accurate than cluster sampling. However, even if large systematic samples are used and sampling is applied separately to each source area, samples still deviate systematically from their reference value. The results indicate that the soil loss data obtained from plots and quoted in current erosion research papers must be approached with caution. As a result of the large scatter in erosion data the samples for study have to be extremely large. This means that future erosion sampling will be both time-consuming and expensive.