1. Updating Forest Stand Inventories: Integration of Photo-Interpreted and Airborne Laser Scanning Forest Attributes Using Generic Region Merging Segmentation and kNN Imputation
- Author
-
Ethan E. Berman, Nicholas C. Coops, Geordie Robere-McGugan, Ian Sinclair, Grant McCartney, and Alexis Achim
- Subjects
Environmental sciences ,GE1-350 ,Technology - Abstract
Integrating airborne laser scanning (ALS) forest attributes with photo-interpreted forest stand age and species attributes can provide managers with the best information to drive estate planning, growth and yield projections, and forest operations. Photo-interpreted forest inventories provide certain forest attributes that are difficult to measure with ALS, yet are subjective and irregularly updated. ALS is objective and provides detailed estimates of forest structure attributes, but poorly estimates age and species composition. We used Generic Region Merging segmentation and k-nearest neighbor imputation to integrate photo-interpreted and ALS-derived forest attributes into a contemporary stand-based forest inventory. We first segmented gridded ALS attributes into forest stand polygons across a ∼630,000 ha managed forest in Ontario, Canada. We next applied imputation to a photo-interpreted inventory, assessing the influence of model parameters and imputed vs. observed values of age and species using leave-one-out cross-validation. Compared to the photo-interpreted inventory, the optimal imputation model estimated age with a mean absolute and mean bias difference of 16.06 and −0.14 years, and classified leading species with 65.46% accuracy. We lastly integrated imputed age and species attributes into the automatically segmented forest stand polygons, finding similar age and species distributions across the landscape when compared to the photo-interpreted inventory.
- Published
- 2024
- Full Text
- View/download PDF