Reducing the effects of disturbance on tree-ring data using intervention detection

In order to analyse the radial growth reactions of trees caused by a specific forcing factor, the effects of other disturbing signals have to be eliminated from the tree-ring chronology. The possibility of using detrending methods and time series analysis to determine and eliminate growth reactions caused by thinnings was studied on material obtained from a long-term growth experiment on Norway spruce (Picea abies (L.) Karst.) in southern Finland. Special attention was paid to intervention analysis. The low-pass filter used eliminated long-term growth changes but it did not filter out rapid growth reactions caused by thinnings. Short-term changes were modelled and eliminated from tree-ring chronology by means of intervention analysis combined with time series analysis. Significant autocorrelation and interventions were found at lower stem heights but not in the upper part of the stem. The pulse intervention whose effect died out within 3 yrs after thinning had the best fit to the mean chronology. Tree-ring series from long-term growth experiments were shown to be well suited to intervention analysis because the time and intensity of disturbing silvicultural events are known.