Sparse Bayesian Estimation of Forest Stand Characteristics from Airborne Laser Scanning
In this article, a new method is applied to modeling forest stand characteristics from airborne laser scanning measurements. The method is an alternative to the cross-validation procedure of variable selection used in the ordinary least-squares (OLS) method and seemingly unrelated regression (SUR) with automatic selection of the features used in the model. This method is called the sparse Bayesian method. It does not suffer from overfitting thanks to the Bayesian formulation of the problem. The proposed method is applied to sample plot data obtained from inventory by compartments. The results show that the sparse Bayesian method performs as well as OLS and SUR methods, in terms of accuracy of total stand characteristics. The methods are comparable also in their demand for sample plot data. None of the methods lose much of their accuracy, even when just a few dozen sample plots are available. A Bayesian approach makes it possible to automate model formulation and sample plot selection processes. It is therefore possible to automatically generate a different model for intrastand strata, thus addressing intrastand variability. The proposed method automatically maintains a balance between the number of forest parameters and the rank of the model used to estimate them.
No Reference information available - sign in for access.
No Citation information available - sign in for access.
No Supplementary Data.
Document Type: Research Article
Publication date: 2008-10-01
More about this publication?
- Forest Science is a peer-reviewed journal publishing fundamental and applied research that explores all aspects of natural and social sciences as they apply to the function and management of the forested ecosystems of the world. Topics include silviculture, forest management, biometrics, economics, entomology & pathology, fire & fuels management, forest ecology, genetics & tree improvement, geospatial technologies, harvesting & utilization, landscape ecology, operations research, forest policy, physiology, recreation, social sciences, soils & hydrology, and wildlife management.
Forest Science is published bimonthly in February, April, June, August, October, and December.
2016 Impact Factor: 1.782 (Rank 17/64 in forestry)
Average time from submission to first decision: 62.5 days*
June 1, 2016 to Feb. 28, 2017
Also published by SAF:
Journal of Forestry
Other SAF Publications
- Submit a Paper
- Membership Information
- Author Guidelines
- Ingenta Connect is not responsible for the content or availability of external websites