Modeling soil loss at plot scale with EUROSEM and RUSLE2 at stony soils of Khamesan watershed, Iran

The present study was carried out to evaluate the EUROSEM and RUSLE2 models for predicting soil loss on stony steep lands in Khamesan watershed using single event data obtained from natural rainfall. For this purpose, eighteen plots of 24 m long by 1.8 m wide were used to measure runoff and soil loss on six rangeland hillslopes (3 replicates) over three years. Twenty-four rainfall events occurred during the study period. The input data were continuously monitored for both models including rainfall, surface runoff, soil loss and soil properties. Sensitivity analysis and calibration results showed that saturated hydraulic conductivity had a large effect on EUROSEM simulation results, especially on runoff estimation. Validation of EUROSEM model showed that this model is applicable for predicting both runoff and soil loss for studied hillslopes (R2 = 0.7, R2 = 0.85, respectively). Efficiency tests indicated that the EUROSEM model outputs were affected by rock fragment cover in the study area. RUSLE2 model was able to predict soil loss in individual storms somewhat well (R2 = 0.53), but under the higher EI30 index, the over or under-prediction was higher. Including direct consideration of runoff in the event rainfall-runoff factor (average annual KUM associated with the QREI30 index) increased RUSLE2 efficiency in estimating event soil loss (R2 = 0.69). Calibration of the K-factor in RUSLE2 led to considerable improvement of model efficiency (R2 = 0.82). Comparison of the two models showed that EUROSEM revealed better model efficiency in estimation of event soil loss than calibrated RUSLE2, although in some hillslopes, performance of calibrated RUSLE2 model was more than EUROSEM model. With respect to spatially non-uniform canopy cover and ground cover as well as the stoniness of soils and direct and indirect impacts of surface and subsurface rock fragments on runoff and sediment yield, also existence of lime at low to high levels in the study area, more research is needed to examine the effect of these parameters on input and output of the models, especially for the RUSLE2. In addition to characteristics of studied hillslopes, it seems that storm pattern had a significant effect on estimations of two models. The results showed that effects of intra-storm variations in rainfall intensity on runoff and soil loss production have not been sufficiently considered in equations of two models, especially RUSLE2 model.