Mapping tree species in a boreal forest area using RapidEye and lidar data

Tree species composition is one of the criteria required for assessing forest reclamation in the province of Alberta in Canada. This information is also very important for forest management and conservation purposes. In this paper the performances of RapidEye data alone and in combination with the Light Detection And Ranging data is assessed for mapping tree species in a boreal forest area in Alberta. Both the random forest and support vector machine classification techniques were evaluated. A significant improvement in the classification outputs was observed when using both data types. Random forest outperformed the support vector machine classifier. Overall, the difference in acquisition time between the RapidEye and Light Detection And Ranging data did not seem to affect significantly the classification results. Using random forest, six input variables were identified as the most important for the classification process including digital elevation model, terrain slope, canopy height, the red-edge normalized difference vegetation index, and the red-edge and near-infrared bands. © 2014 SPIE.