Evaluation of eight different methods to predict hillslope runoff rates for a grazing catchment in Australia

Unlike the USLE/RUSLE models, which require only rainfall intensity data to quantify climatic effects on soil erosion, physically based erosion models require data on runoff rates as their input. However, runoff rate data are rarely measured in the field. This study evaluates eight models in terms of their performance in predicting peak ( Q p) and effective ( Q e) runoff rates required by erosion models. The eight models are: (1) a multiple regression model (MR), (2) a power function model (PF), (3) a scaling technique (ST), (4) a constant infiltration model (CI), (5) a constant runoff coefficient model (RC), (6) a spatially variable infiltration model (VI), (7) the CREAMS peak runoff rate equation ( Q p_CREAMS), and (8) an empirical peak runoff rate equation ( Q P_SAL). Rainfall and runoff data from experimental plots in a grazing catchment in central Queensland (Australia) were used. A commonly used model efficiency statistic ( E) was used to compare the performance of these models. Models resulting in high E values are said to perform better than models resulting in low values of E. Hence, with E values of 0.85 and 0.81 in predicting Q p and Q e, respectively, the PF model ranked first. On the other hand, with an E value of −12.7, the Q p_CREAMS performed the worst in predicting peak runoff rates. On the basis of input data requirements and number of free parameters involved in each model, however, the VI model, with E values of 0.82 and 0.79 for Q p and Q e, respectively, is found to be the best choice when breakpoint rainfall is available for an event. If only peak rainfall intensity is available, the ST with E values of 0.80 and 0.63 for Q p and Q e, respectively, would be the best model to use to predict these two runoff rate characteristics for the site.