Contribution of climate, soil, and MODIS predictors when modeling forest inventory invasive species distribution using forest inventory data
Forest inventory and analysis data are used to monitor the presence and extent of certain non-native invasive species. Effective control of its spread requires quality spatial distribution information. There is no clear consensus why some ecosystems are more favorable to non-native species. The objective of this study is to evaluate the reelative contribution of geo-spatial predictor variables, individually and groups, to the overall classification accuracy of the model. The three selected major gropus of geo-spatial data are MODIS satellite imagery, soil properties, and climate information. We combined predictor variables with forest inventory information, to model/classify the spatial distribution of privet invasive species. Models were separately developed for each group (MODIS group, soil group, climate group), and all data (269 layers) together. Forest inventory plot information and predictor variable information from each group were used to model spatial distribution of forested areas exhibiting the potential to contain privet. Classification results are based on a ten percent set aside dataset. Overall classification accuracy showed that the all data together model performs better than the stand-alone gorup predictive mode. The climate predictive model persorms better in identifying forest with privet among the stand alone (MODIS and soil) predictive models. There is no significant difference between the overall classification accuracy obtained from soil predictive model and the climate predictive mode. However, there is a significant difference on overall classification accuracy at 90% and 95% significance level between the MODIS predictive model and both the soil and climate predictive models.