Tree species identification is an important element in many forest resources applications such as wildlife habitat management, inventory, and forest damage assessment. Field data collection for large or mountainous areas is often cost prohibitive, and good estimates of the number and spatial arrangement of species or species groups cannot be obtained. Knowledge-based and neural network species classification models were constructed for remotely sensed data of conifer stands located in the lower mountain regions near McCall, Idaho, and compared to field data. Analyses for each modeling system were made based on multi-spectral sensor (MSS) data alone and MSS plus LiDAR (light detection and ranging) data. The neural network system produced models identifying five of six species with 41% to 88% producer accuracies and greater overall accuracies than the knowledge-based system. The neural network analysis that included a LiDAR derived elevation variable plus multi-spectral variables gave the best overall accuracy at 63%.
| Identifer | oai:union.ndltd.org:MSSTATE/oai:scholarsjunction.msstate.edu:td-1471 |
| Date | 09 August 2008 |
| Creators | Posadas, Benedict Kit A |
| Publisher | Scholars Junction |
| Source Sets | Mississippi State University |
| Detected Language | English |
| Type | text |
| Format | application/pdf |
| Source | Theses and Dissertations |
Page generated in 0.0023 seconds