The factors influencing the interception process such as the vegetation structure and the meteorological conditions, especially rainfall characteristics, are diverse and highly variable. Therefore, derivation of canopy parameters for interception modelling requires data from field experiments over a long period. In this study, we analyse long-term changes in canopy parameters and rainfall interception for an old-growth spruce stand (composed mainly of Picea abies at a continuous flux site started within EUROFLUX in 1996) within the Tharandter Wald, Southwest of Dresden, Germany. The 10-minute data from the continuous measurements of gross precipitation (P) and throughfall (TF) from 1997 to 2018 were used as follows: Rainfall events were separated by a minimum time of three hours without rainfall and divided into rainfall classes according to their total amount and intensity. Due to vegetation changes in the measuring period caused by two thinning operations, the complete study period was divided into three periods. Canopy parameters S (storage capacity), p (free throughfall coefficient) and Ps (canopy saturation point) were then generated applying a regression-based method using the relationship between P and TF. We generated parameter sets for the complete period, for the three periods reflecting the vegetation changes and for single years. These parameter sets were further tested using a regression model. The statistical analysis of the long-term data set showed that stand parameters and interception are largely influenced by vegetation changes. S and Ps decreased with each thinning. Interestingly, interception and rainfall only changed after the first thinning (increase in p, decrease in interception), indicating counter-acting factors after the second thinning, such as changes in meteorological conditions, rainfall patterns or thinning induced changes in the microclimate of the canopy. Generally, the reliability of the derived canopy parameters depends on the length of the dataset used for the regression analysis. However, the parameter sets that considered three periods produced the best model fit compared to parameter sets for the complete period or for single years. This clearly shows that model performance of statistical interception models depend on a reliable parameterisation that can be improved when taking into account changes in stand characteristics.