Tree spatial patterns and stand attributes in temperate forests: The importance of plot size, sampling design, and null model

Detection of tree spatial patterns and structural attributes in a forest stand can provide critical information on occurring dynamics, and steer management decisions. However, since tree spatial distribution depends on factors that operate at different scales, including environmental heterogeneity and tree-to-tree interactions, both the extent to which measurements are taken and the choice of null model for spatial analysis (including site heterogeneity or not), can considerably influence investigation outcomes and related inferences. In this study, we aim to evaluate the effect of plot size, sampling design (single or combined plots), and null model for spatial analysis, on point pattern analysis and stand attribute assessment in temperate forests. Analyses were performed on 4-ha plots in two old-growth and two previously managed stands in central Europe. For each site, we calculated tree density, mean diameter, mean height and basal area, and performed point pattern analysis (pair-correlation function) under complete spatial randomness (CSR) and heterogeneous Poisson (HP) null models. We then assessed stand attributes and spatial patterns on subplots, and calculated their deviation from the 4-ha reference plot. As expected, accuracy of stand attribute assessment improved by increasing subplot size. However, while accuracy of small (0.25-ha) plots was quite high for basal area, it was rather low for tree density, especially in the old-growth stands. Outcomes of point pattern analysis in 0.25-ha plots were variable, generally presenting low agreement with the reference. Larger plots assured more consistent results, but deviations from the reference were still rather high when CSR null model was used. In all the sites, subplot agreement improved using HP model. Our investigation indicates that 0.25-ha plots are mostly reliable for assessing stand attributes in previously managed forests. However, tree distribution can be very variable both in these and in old-growth stands, therefore spatial patterns cannot be reliably detected with one small plot. Combining several small plots, and using null models accounting for site heterogeneity, are efficient strategies to detect small-scale spatial patterns, but plot larger than 1-ha should still be used to assess large-scale patterns in high-diversity forest stands.