Using rainfall simulation and site measurements to predict annual interrill erodibility and phosphorus generation rates from unsealed forest roads: Validation against in-situ erosion measurements

Abstract

The measurement of soil erosion rates under natural rainfall conditions is costly and time consuming. Data provided by rainfall simulation and static site measurements can be used to predict erosion rates under natural conditions, however the accuracy of this method is largely untested. This is especially true for erosion rates from unsealed forest roads. In this study, the values for a range of erodibility indices calculated from rainfall simulation experiments are compared to observed erodibility index values from 1 year of detailed in-situ erosion monitoring of seven different forest road types. The prediction of phosphorus generation rates from rainfall simulation and/or soil sampling was also evaluated. The results showed that a series of commonly used erodibility indices such as sediment per unit rainfall, sediment per unit runoff, sediment per unit EI30, and sediment per unit rainfall energy were poorly predicted from the rainfall simulation experiments. Five of the six indices tested substantially overpredicted the observed erodibility at one site, a gravel road subjected to minimal traffic. For the other six road sites predictions were poor and highly variable, the coefficient of efficiency ranging from − 13.32 to 0.17 for these erodibility indices. A modified index, the ratio of sediment per unit rainfall energy to the mean rate of rainfall energy input, was able to predict annual erosion rates from six different road surfaces using rainfall simulation data with a coefficient of efficiency of 0.9. The results indicate that existing erodibility indices are not suitable for predicting observed erosion rates on forest roads using rainfall simulation data as collected in this study. It is argued that the modified index is more suited to sites (such as compacted roads) where interrill processes dominate, and erosion rates are less sensitive to peak flows. With respect to nutrient generation rates, rainfall simulation was able to accurately predict the observed proportion by mass of total phosphorus (TP) in runoff with a coefficient of efficiency of 0.96. Direct soil sampling of the road surface could also be used to predict the proportion by mass of TP in runoff. Concentrations of total nitrogen in forest road materials were found to be at the lower detection limit of the laboratory instruments.