Characterization of Spatial Structures and Autocorrelations in Forest Stands: Loblolly Pine Plantations in the Southeast US, Low Dipterocarp Natural Forests in Indonesia, and Eucalyptus Plantations in Brazil.

摘要

Development of a forest is influenced by various factors often resulting in spatially autocorrelated patterns. Depending on the underlying spatial process in a forest, a forest will exhibit aggregated patterns of similarities or dissimilarities among spatial attributes close in proximity. Understanding spatial structures and autocorrelations of forests provides key information on underlying processes determining spatial patterns of forests.;This dissertation is comprised of three chapters analyzing and characterizing spatial structures and autocorrelations in forest stands. The first study deals with characterization of spatial autocorrelations of detrended diameters of loblolly pine stands at age five in the southeast US. Loblolly pine stands of four different levels of genetic homogeneity including clones, full-sib families, half-sib families, and a seed orchard mix randomized into two spacings (3.0 x 6.1 m; 538 trees per hectare, and 1.5 x 6.1 m; 1077 trees per hectare) were analyzed. The results showed that 12.5% of the plots were positively autocorrelated and 5.0% of the plots were negatively autocorrelated at an alpha level of 0.10. It was found that spacing had an effect on both significant positive and negative spatial autocorrelations. A majority of positively and negatively autocorrelated plots were located in the wider spacing and closer spacing, respectively. No particular trend in terms of different levels of genetic homogeneity was noted except for in one specific clone when planted at a wider spacing.;The second study deals with applying geostatistical approach to quantify aboveground biomass (AGB) of Labanan Concession Forest in East Kalimantan, Indonesia. Diameters collected with transect sampling were converted to AGB data. Two approaches of estimating the spatial distributions, global and stratified, were compared for AGB mapping. The global approach does not take local varying structures into account, whereas the stratified approach accounts for the heterogeneity of land cover types. Thus, AGBs estimated from each land cover type were pooled for the stratified approach. Ordinary kriging was performed to predict AGB at unsampled locations. The total estimates of AGB and RMSEs for the global and stratified methods were 13,512,392.2 tons (161.92 ton/ha) and 13,607,205.5 tons (163.05 ton/ha), respectively, which the estimate of stratified method was larger than the global estimate by 94,813.3 tons, and 81.0 ton/ha and 81.2 ton/ha, respectively, suggesting the degree of accuracy for the two methods was similar.;The last study deals with change in spatial autocorrelations of diameters, heights, and biomass of clonal and seed-origin Eucalyptus plantations in Brazil according to different silvicultural treatments over the rotation. Spatial autocorrelations of conventional vs. enhanced levels of fertilization, non-irrigated vs. irrigated, and uniform vs. heterogeneous stand structures were evaluated and compared, and two distance-based neighborhood structures were applied to look at changes in the degree of spatial autocorrelations when different numbers of trees were accounted as neighbors. The results suggested that spatial autocorrelations of diameters, heights, and biomass of Eucalyptus plantations were mostly insignificant at an alpha level of 0.10 over the rotation, however values of spatial autocorrelations were maintained below zero in most plots after the middle of rotation indicating occurrence of size differences among neighboring trees. In particular for diameters and biomass, general declining patterns from positive to negative spatial autocorrelations, except for plots with heterogeneous stand structures, were noted at plots treated with enhanced levels of fertilization, suggesting the influence of fertilization on the realizations of microsite effects in the early stage of stand development. No particular effect was detected for the two neighborhood structures except for the increased frequencies of significances in spatial autocorrelations for diameters, heights, and biomass.