Altering finite Markov chains to more accurately model forest succession
Document Type Dissertation/Thesis
A finite Markov chain model of forest succession was modified to add considerations of: 1) the different death rates of canopy species; 2) the probability a given forest species has of reaching the canopy based on microhabitat factors; 3) the canopy-composition dependent death rates of understory species. Two datasets were collected, one at Mt. Pisgah in Winthrop, Maine and the other at the Alonzo Garcelon Wildlife Management Area in Vassalboro, Maine. The standard finite Markov chain model and the canopy death rate adjusted model were both applied to the datasets and their results compared. The adjusted model yielded more ecologically realistic predictions of forest succession than the standard model for the Mt. Pisgah data set, while the predictions based on the Alonzo Garcelon data set were relatively similar for both models. For each canopy species present over the two study areas, a logistic regression model was used to determine a species' probability of reaching the forest canopy using four microhabitat predictor variables: soil depth, soil texture, slope, and aspect. The logistic regression model was fined using US Forest Service Forest Inventory and Analysis Program (FIA) data colleted in Maine in 2003. Due to a lack of variability in the FIA data used, none of the four microhabitat variables were found to be significant predictors of the success or failure for any of the species present in the two datasets with the exception of the slope variable, which was found to be a significant predictor for the success or failure of yellow birch (Betula allghaniensis). The finite Markov chain model adjusted to account for each species' probability of reaching the canopy and the finite Markov chain model adjusted to account for canopy-composition-dependent death rates of understory species were both applied to hypothetical forest systems. The predictions of both altered models were compared to the predictions of the standard finite Markov chain model. Again, the altered models produced more ecologically realistic predictions of forest succession than the base model. Thus, results indicate that increasing the number of parameters considered in a finite Markov chain model of forest succession can increase the ecological realism of the model's predictions.