Abstract
With the aim of conducting efficient and accurate predictive mapping of actual vegetation, we set out to examine the data processing methods and to evaluate the accuracy of created maps. The study area contained natural vegetation and substitutional vegetation in a Fagetea-crenatae region and upper Camellietea-japonicae region distributed within a 6 km by 6 km square region on the border between Tottori and Hyogo Prefectures in Japan. A land cover map generated from satellite data and a predictive map of potential natural vegetation produced from GIS data - including elevation, landform, accumulated amount of solar radiation, and maximum snow depth data - were overlaid to create a predictive map of actual vegetation. The kappa coefficient accuracy was calculated in order to evaluate the accuracy of the predictive map of actual vegetation. Results indicated that 89% of the mapped target region had a kappa coefficient accuracy of over 0.21. Taken together, these results indicated a relatively good correspondence between the predicted results and actual onsite confirmation of distribution of vegetation in the mapped area. Within a certain degree of mapping accuracy, a predictive map of actual vegetation was successfully created.
