Simulation of the increase of sorption in time is one of the options in higher tiers of pesticide regulatory leaching assessments to obtain more realistic leaching estimates. Therefore, accurate estimates of non-equilibrium sorption parameters are required as input for the pesticide leaching scenarios. Usually, non-equilibrium sorption is described with a two-site equilibrium/non-equilibrium model in which the non-equilibrium sorption is described with two parameters (i.e. the desorption rate coefficient of the non-equilibrium site and the Freundlich sorption coefficient of this site). Estimates of these parameters can be obtained with inverse modelling techniques. At the moment, there is little understanding about whether the confidence intervals provided by inverse modelling can be used as measure of the likely accuracy (i.e. how close the estimated value is to the true value) of these estimates. We set up a semi-global inverse modelling exercise for a large number of parameter sets (Le. different pesticides) using simulated datasets. Inverse modelling of non-equilibrium parameters demonstrated decreasing accuracy of the estimates for decreasing values of the non-equilibrium sorption parameters and the equilibrium sorption coefficient. Furthermore, we found a relationship between the accuracy of a parameter estimate and its CV (coefficient of variation) provided by the inverse modelling technique. Using this relationship we calculated the likelihood of rightly or wrongly accepting or rejecting a parameter estimate as a function of this CV. We recommend to use this likelihood as the basis of communication with decision makers on how to decide on accepting or rejecting parameter estimates. (C) 2013 Elsevier Ltd. All rights reserved.