Temporal development of crown condition of Picea abies: Two-step approach using statistical and geostatistical methods

The aim of the presented Internal Report 2001 is to analyse the temporal development of defoliation of Picea abies in Europe by means of the transnational Level I data set. Additional external deposition and meteorological data are used to describe the influence of single stress factors. Evaluations are conducted using multivariate linear models in a two step analysis. Step 1 evaluates variables that were available in annual or higher time resolution in relation to annual mean plot defoliation for the years 1994 to 2000. The regression coefficient for the interaction term 'year*plotid' is interpreted as plot-wise time trend. It is presented in maps and after a geostatistical analysis detecting spatial autocorrelation interpolated by the geostatistical method 'kriging'. Rooted mean squared errors (RMSE) are calculated for all plots. They give a sensitive measure for the discontinuity of defoliation development and serve as a conservative estimate for the accuracy of the linear models offering a tool for quality control when interpreting the time trends. Results of step 1 show significant influences of insects, age and country as well as their interaction terms on annual mean defoliation. Time trends were only significant for plots and countries, which shows that for Picea abies there is no significant mean European wide trend. Fungi gave implausible results, whereas summer precipitation showed insignificant but mostly plausible effects. Larger regions with deteriorating defoliation are observed in southern Sweden and Finland, in Estonia as well as in alpine regions of Switzerland, Austria, and Slovenia. Improving crown condition is observed in northern Scandinavia, Lithuania, southern Poland, and in Slovakia. The analysis of the RMSE reveals single plots with extreme defoliation values, these could however only be partly explained by information available from the Level I data base. The interpretation of time trends at these plots has to be conducted with care. Step 2 of the analysis is the explanation of plot-wise time trends and RMSE values by predictors of low temporal resolution. Preliminary results show that trends of defoliation are correlated with the difference in SOx deposition between 1998 and 1997, with water availability and base saturation. The promising results for sulphur deposition underline the necessity to include annual deposition values into step 1 of the analysis in the future. Site characteristics may show closer correlation when combined with meteorological data of higher time resolution. After discussion of the results in an editorial group the evaluations are foreseen to be extended and applied to other main tree species for a presentation in the Forest Condition Report of UN/ECE and EU in 2002.