Linking spatial patterns to forest ecological processes by using spatial statistical methods

While spatial patterns have long been recognized as important aspects of forest vegetation and site factors, the task of explaining patterns in relation to underlying processes has been more difficult. Advances in spatial statistical analysis are offering ecologists new strategies for examining spatial patterns that have the potential to better link observed patterns to processes. This dissertation project applied multiple methods of spatial statistical analysis to four field studies conducted in Montana (USA) and Chile (South America), including some novel approaches for examining causal factors and extending the utility of plot-based data. In the first study of natural stands of Nothofagus glauca in south-central Chile, analysis using Ripley's L-function indicated that the ecological status of N. glauca may vary from pioneer to gap strategist depending on site and stand conditions. On harsh sites, N. glauca seedlings displayed positive spatial associations with overstory trees, suggesting facilitation. To further examine bivariate associations observed between regeneration and overstory trees I studied ponderosa pine/Douglas-fir forests of western Montana. A new index was developed that quantifies the strength of association between canopy layers. This index considers the ratio between the value (Lˆ 1.2(t)) of the bivariate L-function and the corresponding confidence envelope at a specified distance (t). Larger index values indicate greater departures from the hypothesis of no spatial association, prompting further ANOVA comparisons among groups of plots differing in moisture availability. Seedlings of both ponderosa pine and Douglas-fir tend to be positively associated with overstory trees in drier sites, and negatively associated on moister sites. In a third study examining pine plantations in Patagonia, an effort to further discriminate between multiple factors influencing spatial patterns at different scales utilized a semivariogram approach to formulate stand development models for even-aged populations. In the fourth study, data from ponderosa pine restoration treatments in a randomized block design were weighted using a spatial ANOVA model to control for spatial auto-correlation, allowing expansion of the original design to examine age classes separately. This case study indicated that old trees responded positively to release from competition via harvesting, but that spring broadcast burning may reduce both growth and vigor. These various studies emphasize the importance of both spatial pattern description and the utility of statistical strategies for examining potential causal factors.