The aim of this study was to analyse the relation of horizontal and vertical canopy structure to tree functional diversity of a highly diverse subtropical broad-leaved slope forest, stratified for different successional stages. This is of particular interest because many key ecosystem processes and functions are related to the arrangement of forest canopies. We assessed the effect of stand-related functional diversity (FDQ, measured as Rao’s Quadratic Entropy of leaf traits), together with other environmental variables on horizontal (measured as relative crown projection areas CPAr) and vertical (relative crown overlap) structure of the upper canopy at the local neighbourhood level. The analyses with mixed effects models revealed a negative relation (p = 0.025; estimate: -0.07) between FDQ and CPAr. No significant effect of FDQ on vertical canopy structure has been found (p > 0.05). The findings are discussed with regard to resource partitioning and niche differentiation of canopy and subcanopy species. This dataset is obtained from the raw data of the dataset: "Tree neighbour competitive traits in the comparative study plots - tree individualss"

Our response variables were based on measurements of horizontal and vertical crown parameters of canopy tree individuals. For these measurements we focused on 70 groups of upper canopy trees (formed by a target tree and its local neighbours; in the following referred to as target groups) that were spread over all study plots. All trees growing within a circular plot with a radius equal to half of the target tree’s height (Ammer et al. 2005) and fulfilling the criterion of minimum diameter at breast height (dbh, measured at 1.3 m above ground; see criterion (ii) below) were considered as neighbours. The number of neighbours ranged from three to 38, resulting in a total number of 996 surveyed individuals. To allow for greater generalisability of tree individual-based measurements, we selected four highly abundant tree species (Yu et al. 2001) as target species: S. superba, C. eyrei, Q. serrata var. brevipetiolata and Castanea henryi (Skan) Rehd. et Wills.. Target trees were chosen randomly within the plots from all target species individuals complying with the following criteria: (i) single stemmed; (ii) dbh > 10 cm for intermediate and old successional stages 3-5 or dbh > 3 cm for young successional stages 1 and 2; (iii) crown position in the upper canopy layer; (iv) no serious damage due to recent snow break (less than 30% crown loss, no newly formed gap in close vicinity of the target tree).

CSPs, Gutianshan National Nature Reserve 29°08'-29°17'N 118°02'-118°11'E For every individual tree belonging to the studied target groups, the dbh (measured with a diameter measurement tape) and total height as well as the crown-base height (defined as the height at which the lowest living crown branch of the tree branches off, excluding epicormics or sprigs) were recorded. The relative position of the neighbours to the target tree was measured as the horizontal distance from stem base to stem base. All height and distance measurements were conducted using a Forester Vertex Hypsometer (Haglöf, Sweden). Crown length was computed as total height minus crown-base height and mean crown height as crown-base height added to half of the crown length. To characterise the sampling trees (of our target groups) within the different successional stages more thoroughly, we analysed relative frequency distributions of height classes. Crown radii in the eight subcardinal directions (N, NE, …., NW) were determined by means of a densiometer (a gimbal mirror positioned in an angle of 45°, allowing the precise observation of objects in orthogonality to the earths magnetic field; this is especially important in sloping terrain; Röhle and Huber 1985). In cases of extreme crown displacement – the crown projection did not include the stem base – the distances to the proximal and distal edge of the crown were measured in all possible directions. Whenever this was only possible for one direction, four crown radii were measured as follows: the distances to the proximal and distal edge of the crown were determined and, starting at the centre of this crown diameter, on the axis perpendicular to this diameter. Calculation of response variables To convey the individual-based crown projections to the target group-level response variables for horizontal and vertical canopy structure, the crown radii were used to construct crown projection maps for each target group in the geoinformation system ArcGIS (Version 9.0, ESRI). Crown projection maps were created by connecting the crown edge points to a polygon. The sum of crown projection area per target group (CPA) was calculated as the sum of the area of the polygons. In addition, the area of crown overlap (O) was calculated for each target group from the crown projection maps. In order to account for the different successional stages that are reflected in systematically higher CPAs of older target groups, relative CPA (CPAr) was defined as the ratio of CPA and the circular plot area for the selection of neighbours (C). For the same reason, the relative crown overlap Or was calculated as the ratio of O and C. Increasing CPAr describes a higher horizontal canopy space use of canopy trees leading to a more closed canopy, while an increase in Or reflects a higher vertical overlap of canopy trees. Both response variables are considered to be important factors determining canopy structure. Environmental plot-related variable To quantify the severity of disturbance by an unusual snow break event in February 2008, we used mean values of crown loss of all measured individuals per plot. Crown loss of all observed tree individuals was estimated as the percentage loss of the formerly undisturbed crown volume. Categories were defined as: undamaged: 0% loss = d0, 1-25% loss = d1 etc., crown completely broken 100% = d5.

07/2008-05/2009

Tree species in the Comparative study plots

Statistical analyses were performed using R 2.10. (R Development Core Team 2010). For the analyses of mixed effects models the package “nlme” was used (Pinheiro et al. 2009). Multiple comparisons were performed by means of the package “multcomp” (Hothorn et al. 2008). The influence of a set of explanatory variables on the response variables CPAr and Or was tested by means of mixed effects models.