Submitted to: Global Change and Terrestrial Ecosystems Working Group on Soil Erosion
Publication Type: Abstract Only
Publication Acceptance Date: September 15, 1997
Publication Date: N/A
Technical Abstract: Erosion plot data contains a high degree of variability which is often not recognized or considered in evaluating the results of field experiments. The variability has significance for erosion monitoring and for the evaluation of soil erosion models. In this study replicated erosion plot data from several hundreds of storm events were analyzed. The results showed that large differences between measured values of erosion on replicated plots were the norm. Of 2802 storm data pairs, the 80 percent confidence interval for replicates covered the range from 81 percent underpredicted to 110 percent overpredicted. Of 1498 annual data pairs, the 80 percent confidence interval covered the range from 49 percent underpredicted to 92 percent overpredicted. Relative differences in replicates was greater for smaller values of erosion for both storm- by-storm and annual data. In other words, the tendency was that the greater the erosion values, the less the relative difference between replicates. The basic conclusions of the study are: 1) that replicated data are necessary in order to evaluate erosional differences between experimental treatments, and 2) large differences between predicted and measured erosion may be acceptable for evaluating erosion model results.