Biofilters are a potential treatment option for removing pollutants from feedlot runoff but little research has been done on their use and design. In this study, two mechanism-based models were developed to simulate biofilter processes: A first-order model and a logistic model. The two models were calibrated and evaluated using nitrogen (N) and phosphorus (F) data collected from rainfall events for an experimental biofilter at Melrose, Minnesota, USA. The first-order model predicted removal efficiencies better than the logistic model. The sensitivity analysis suggested that the predictions of the first-order model are more sensitive to parameter. In addition, the uncertainty analysis suggested that the range in predictive errors could be a consequence of uncertainty of estimating parameter from the limited data set for the first-order model. In contrast, the uncertainty analysis for the logistic model of N suggested that reasons other than the uncertainty in parameter estimation are needed to explain predictive errors. Overall, the study provides a useful tool for assessing biofilter performance that can easily be improved with larger observed data sets. The biofilter model has been implemented in the most recent version of the Minnesota feedlot annualized runoff model (MinnFARM).