Theses and Dissertations


Tianyu Li

Issuing Body

Mississippi State University


Meng, Qingmin

Committee Member

Rodgers, John C., III

Committee Member

Du, Qian

Committee Member

Ambinakudige, Shrinidhi

Committee Member

Cooke, William H., III

Date of Degree


Document Type

Dissertation - Open Access


Earth and Atmospheric Sciences

Degree Name

Doctor of Philosophy


College of Arts and Sciences


Department of Geosciences


The forest ecosystem is a dominant landscape in the Gulf of Mexico (GOM) coastal region. Currently, many studies have been carried out to identify factors that drive forest dynamics. Changes in meteorological conditions have been considered as the main factors affecting the forest dynamics. For this study, the statistical regression analysis was used for modeling forest dynamics. Meteorological impact analysis was driven by observed data from PRISM (parameter-elevation regressions on independent slopes model) climate dataset. The forest dynamics was characterized by an indicator, the normalized difference vegetation index (NDVI). The objectives of this study are to 1) to specify and estimate statistical regression models that account for forest dynamics in the Golf of Mexico coastal region, 2) to assess which model used to capture the relationship between forest dynamics and its explanatory variables with the best explanatory power, and 3) to use the best fitted regression model to explain forest dynamics. By using fixed-effects regression methods: ordinary least squares (OLS) and geographically weighted regression (GWR), the sample-point-based regression analysis showed that meteorological factors could generally explain more than half of variation in forest dynamics. In respect of the unexplained variation of forest dynamics, the necessity of using soil to explain forest dynamics was then discussed. The result suggested that the forest dynamics could be explained by both meteorological parameters and soil texture. One of the basic considerations in this study is to include the spatiotemporal heterogeneity caused by seasonality and forest types. The model explanatory power was found differ among forest types (spatially) and seasons (temporally). By constructing regression models with randomly varying intercepts and varying slopes, the linear mixed-effects model (LMM) was fitted on composite county-based data (e.g., precipitation, temperature and NDVI). The use of LMMs was proved to be appropriate for describing forest dynamics to mixed-effects induced by meteorological changes. Based on this finding, I concluded that meteorological changes could play a significant role in forest dynamics through both fixed-effects and random-effects.



meteorology||GWR||OLS||regression model||NDVI||Gulf of Mexico||forest dynamics