Monthly runoff and soil loss simulated by WEPP (Water Erosion Prediction Project) were compared with field observations on a pineapple farm in south-east Queensland for a 3-year period. The soil at the site is sandy. Slope length and steepness are 36m and 5.5%, respectively. Three treatments, namely bare, farmers' conventional practice, and mulching of the furrows, were used. Infiltration and erodibility parameters were determined using WEPP-recommended equations and measurable soil properties. These parameters were also calibrated using the runoff and soil loss data for the bare plot only. Apart from the soil loss prediction for the mulching treatment, for which WEPP did not perform well, the average coefficient of efficiency in runoff and soil loss predictions was -0.02 using soil property-based parameter values and 0.66 using calibrated parameter values. The corresponding r 2 values are 0.57 and 0.81, respectively. On the whole, WEPP is able to reproduce the trend and variations in runoff and soil loss among different treatments for the site. Parameter values based on measurable soil properties would greatly under-estimate the runoff and soil loss for the site. Thus, use of WEPP outside its US database requires calibration with locally obtained data. It was also found that WEPP does not seem to model effectively the situation where there is considerable flow impediment with the furrows covered with mulch. We are unable to reject WEPP because the statistical performance indicators are reasonable for the site, and because the model is so complex that it is nearly impossible to pinpoint the source of discrepancy and articulate the model deficiency on physical grounds.