| Changes in vegetation distribution due to climate change are a concern in alpine tundra ecosystems. Past vegetation change was assessed and a cellular automata-Markov (CA-Markov) model was used to predict future land cover scenarios in the Torngat Mountains National Park Reserve (Labrador, Canada). Post-classification image comparison was applied to classified, multi-temporal satellite imagery to detect changes in vegetation patterns since 1985. Deciduous shrubs (typically less than 3m in height) increased in areal coverage whereas heath (low-growing, woody vegetation) experienced a decrease in coverage. Transition matrices were developed from these observed changes, and were used in the Markov chain component of the model. Topographic variables were classified, and used as prior information to calculate Bayesian probabilities (B Prob). The BProb's describe suitable areas of growth based on known patterns and were used as a suitability map in the cellular automata component of the model. The CA-Markov model was initially used to predict a known vegetation pattern for 2008, using classified imagery from 1985 and 2001. The model predicted the 2008 land cover with 70.7% accuracy and data, recorded in 2008, was used to predict scenarios for 2018, 2028, and 2038. Results of the CA-Markov simulations show that deciduous shrubs will increase in area by 7.7% but heath will decrease by 14.4%. The results indicate that deciduous shrubs have a tendency to move into higher elevations over an extended period of time. |
