Pesticide degradation models are compared which simulate the response of biofilters for treatment of pesticide-contaminated waste water to time-irregular pesticide supply in which the pesticide is used for growth and mineralized. Biofilter microcosms containing a mixture of straw, peat and soil and harboring micropopulations which uses the herbicide linuron for growth, were irrigated with linuron for 28 weeks with a stop in its supply between week 12 and 17. Matrix samples were regularly taken to assay linuron mineralization. A first-order approximation of the Monod model was used to simulate the observed mineralization data, while an inverse modeling framework combining a sensitivity analysis (Morris Sensitivity Analysis) with an inverse modeling approach (Shuffled Complex Evolution Metropolis) adopted to parameterize the model. Lag times in linuron mineralization decreased during the initial weeks of linuron irrigation but increased after supply of linuron ceased. The model well-simulated the lag time dynamics which were related to the dynamics of the predicted linuron-degrading population size in the microcosms. it was predicted that the population size decreased at a rate of 0.031 d(-1) after pesticide supply ceased to reach its initial population size after 25 weeks. We conclude that modeling pesticide degradation in biofilters should incorporate biomass dynamics in case the pesticide is used as C-source. First-order approaches without incorporating biomass dynamics could lead to underestimation of the risk of pesticide leaching. (c) 2009 Elsevier Ltd. All rights reserved.