359 – Nonstationary Models for Spatial Data
Quantifying Spatial Variability in Forest Plots
Gavin Corral
USDA/NASS
A paucity of information exists on spatial patterns of tree attributes in regularly spaced plantations. Most tree models assume independence among individual sampling units (trees) to meet the assumptions of Ordinary Least Squares. Moreover, spatial variability is often confounded with other non-spatial variables, such as genetic factors. The objectives of this research are to test the feasibility of redundancy analysis to partition total variability into pure spatial and pure environmental components and to test the efficacy of cluster analysis to map areas of contrasting productivity. Furthermore, we will use the results of this study to identify primary sources of tree attribute variability and to discuss the consequences of those findings on forest management. To achieve these objectives, we utilize simulated and real data sets. The results of this study indicate that redundancy analysis and spatially constrained cluster analysis are useful forest management tools. Furthermore, it is apparent from these results that the predominant source of tree attribute variability in planted stands is spatial and not genetic