Local variability of stand structural features in beech dominated natural forests of Central Europe: Implications for sampling

摘要

a- Local variability of natural stands and appropriate sampling is often neglected. a- Our analyses are based on extensive datasets (in total about 29,000 trees on 107ha). a- Variability of estimates in relation to plot sizes follows function y = a A- x b . a- Min. numbers of plots and total sampling area for proper estimates were calculated. a- Range of historical and recent studies have been based on insufficient datasets. Current knowledge of the within-site variability of major stand structural features in beech dominated natural forests is limited. Numerous studies have used just several small plots for characterizations of natural stands, but this may lead to generalizations based on unreliable results. This study shows how major stand structural features vary at the local scale, and how suitable sampling may reflect this variability. Stem position maps of three natural forests in the Czech Republic (Zofin 71ha, Salajka 19ha and Zakova hora 17ha) were used. Each vector stem position map representing all live and dead trees with DBHa[control][yen10cm was intensively analyzed using square sample plots of different sizes (10A-10m; 20A-20m; 30A-30m; 50A-50m; 100A-100m; 140A-140m and 200A-200m). Basic statistics (mean, standard deviation, coefficient of variation, min., and max.) were calculated for every plot size and each of six major stand features: density, basal area and volume of living trees, volume of course woody debris, total volume and proportion of course woody debris in total volume. As expected, the local relative variability of major stand structural features decreased with increasing plot size. For particular stand features, the observed trend was expressed by significant regression models (y = a A- x b ) with high coefficients of determination across all study sites. The results further show that even hectare-to-hectare values are still highly variable; e.g. the volume stock of living trees can range from 474 to 1049m3/ha within one study site. Hence, single samples of 1ha can be poorly representative. We also calculated the minimal number of plots and total sampled area required for estimations of major stand features to within 20% (A plus or minus 10%) of the mean with 95% confidence for particular plot sizes. It appears that plot sizes between 0.01 and 0.09ha would be the most efficient for sampling the above-mentioned variables in European beech-dominated natural forests.