Mortality processes are a critical component of forest stand spatial pattern formation and maintenance; understanding these processes can inform management decisions and improve future modeling efforts. Old growth forest gives us an opportunity for deepen understanding the stand dynamic and mortality process in longleaf pine stand. Mortality in small trees was primarily due to the suppression while in the mature trees they were contagious. Competition based on the nearest neighbors is an important deriving factor of the tree mortality; however, most of the mortality models deal with the density of a whole stand or plot to incorporate competition on their models. We have analyzed historical data gathered from the two large permanent plots, Caffey Hill (1.5 ha) and Red-Tail Ridge (1.8 ha), of the Mountain Longleaf Pine National Wildlife Refuge, Alabama. The data had information such as spatial location, Age, DBH, Status of each tree from 1999 to 2014. Live and dead trees were subdivided into four different classes based on the age structure in 1999 for modeling separately. Multiple spatial and statistical methods were applied to explore and analyze the data. Finally, mortality models were developed using Geographically Weighted Generalized Regressions for each dead trees classes.
Results/Conclusions
Each tree was plotted and competition index (Hegyi competition index) was calculated based on the nearest ten neighbor’s trees. Competition index was higher for those smaller trees which were surrounded by larger trees. Each class of live and dead trees was plotted separately and examined their pattern and association. We found that most of the classes were in cluster patterns. A strong positive association was seen between small dead trees and live trees on both sites. Intensity map of the competition index was made and dead trees were plotted on the competition intensity map. We have examined that small trees were highly concentrated where the competition intensity was higher on both sites. Geographically weighted generalized regression models were used to predict the probability of mortality. Backward selection approach was used to find the best model using the smallest AIC values. Local R2 and percent deviation explained by the models were plotted and shown on the map. Parameter estimate varied depending on the location. Furthermore, models suggested that competition index was one of the most important parameters to predict mortality in both sites for the small trees. However, larger trees mortality models were not as good as small trees mortality.