This paper describes the application of a two-layer stochastic model of rainfall interception in modelling the interception loss from tropical forest (Kandyan Forest Garden), in Sri Lanka. A small data set collected at a Kandyan Forest Garden site near Kandy has been used to illustrate the performance of the model. The model simulates the asymptotic wetting of tree canopies and accounts for the effects of changing drop size, resulting from changing rainfall intensity and drop modification by the canopy, on the canopy storage capacity. The net rainfall predicted by the model is compared with the measured net rainfall at 5 min intervals as well as with the predictions of the widely used Rutter model. The two-layer stochastic model performed better in that it accounted for 76% of the variance, compared with 21% accounted for by the Rutter model using parameters optimised to minimise the sums of squares of the differences between the measured and predicted cumulative net rainfall. The improved performance of the stochastic model is due largely to the improved prediction of the net rainfall during the initial wetting phase of the storm.