Delineating forest stands from grid data
Abstract Background Forest inventories are increasingly based on airborne laser scanning (ALS). In Finland, the results of these inventories are calculated for small grid cells, 16 m by 16 m in size. Use of grid data in forest planning results in the additional requirement of aggregating management...
| Published in: | Forest Ecosystems |
|---|---|
| Main Author: | |
| Format: | Article in Journal/Newspaper |
| Language: | English |
| Published: |
SpringerOpen
2020
|
| Subjects: | |
| Online Access: | https://doi.org/10.1186/s40663-020-00221-8 https://doaj.org/article/3e9e703269664f2d8955dba72d285850 |
| Summary: | Abstract Background Forest inventories are increasingly based on airborne laser scanning (ALS). In Finland, the results of these inventories are calculated for small grid cells, 16 m by 16 m in size. Use of grid data in forest planning results in the additional requirement of aggregating management prescriptions into large enough continuous treatment units. This can be done before the planning calculations, using various segmentation techniques, or during the planning calculations, using spatial optimization. Forestry practice usually prefers reasonably permanent segments created before planning. These segments are expected to be homogeneous in terms of site properties, growing stock characteristics and treatments. Recent research has developed methods for partitioning grids of ALS inventory results into segments that are homogeneous in terms of site and growing stock characteristics. The current study extended previous methods so that also the similarity of treatments was considered in the segmentation process. The study also proposed methods to deal with biases that are likely to appear in the results when grid data are aggregated into large segments. Methods The analyses were conducted for two datasets, one from southern and the other from northern Finland. Cellular automaton (CA) was used to aggregate the grid cells into segments using site characteristics with (1) growing stock attributes interpreted from ALS data, (2) predicted cutting prescriptions and (3) both stand attributes cutting prescriptions. The CA was optimized for each segmentation task. A method based on virtual stands was used to correct systematic errors in variable estimates calculated for segments. Results The segmentation was rather similar in all cases. The result is not surprising since treatment prescriptions depend on stand attributes. The use of virtual stands decreased biases in growth prediction and in the areas of different fertility classes. Conclusions Automated stand delineation was not sensitive to the type of variables that ... |
|---|