Canadian Forest Service Publications
Multidimensional scaling of first-return airborne laser echoes for prediction and model-assisted estimation of a distribution of tree stem diameters. 2016. Magnussen, S., Renaud, J-P. Annals of Forest Science, 73, 1089-1098.
Issued by: Pacific Forestry Centre
Catalog ID: 39005
Availability: PDF (request by e-mail)
Available from the Journal's Web site. †
† This site may require a fee
Key message - We demonstrate how multidimensional scaling can be used to combine forest inventory field data and airborne laser scanner data to obtain both predictions and model-assisted estimation of a tree stem diameter distribution. Context - The size distribution of forest trees is important both for management planning and analysis purposes. Yet field samples are rarely large enough to assuage a desired accuracy of a direct estimation in all areas of interest. Improvements in spatial coverage and accuracy are possible with a census—or a very large sample of one or more cost-effective auxiliary variables that can inform one about the tree size distribution. Aims - The objective of this study is to demonstrate how a relative frequency distribution of canopy heights from airborne laser scanner data can be used to improve direct estimates of a tree size distribution. Methods - Multidimensional scaling is used to link a relative frequency distribution of canopy heights to an observed plotlevel distribution of tree size. Results - A multivariate linear model can be used for both predictions and model-assisted estimation of a tree stem diameter distribution. Conclusion - Multidimensional scaling can provide a multivariate linear link between two relative frequency distributions and is therefore ideally suited for both stand-level predictions and design-based inference of tree size distributions.
Plain Language Summary
The size distribution of trees is an important source of information for forest management, reporting and analysis. Airborne laser scanner (ALS) data has improved our ability to predict a size distribution from ALS data, but only within a model-based framework with a non-trivial risk of model bias. Many inventories are required to report on design-unbiased estimators but an estimator of this type for a size distribution has been lacking. This study provides a model-assisted estimator (design-based) for a tree size distribution by linking field observations and ALS data through a multidimensional scaling. Testing of the estimator with data from four sites in eastern France was successful for 11 leading species strata.