Canadian Forest Service Publications

Model-dependent forest stand-level inference with and without estimates of stand-effects. 2017. Magnussen, S., Breidenbach, J. Forestry 90(5):675-685.

Year: 2017

Issued by: Pacific Forestry Centre

Catalog ID: 38880

Language: English

Availability: PDF (request by e-mail)

Available from the Journal's Web site.
DOI: 10.1093/forestry/cpx023

† This site may require a fee

Mark record


Forest stands are important units of management. A stand-by-stand estimation of the mean and variance of an attribute of interest (Y) remains a priority in forest enterprise inventories. The advent of powerful and cost effective remotely sensed auxiliary variables (X) correlated with Y means that a census of X in the forest enterprise is increasingly available. In combination with a probability sample of Y, the census affords a modeldependent stand-level inference. It is important, however, that the sampling design affords an estimation of possible stand-effects in the model linking X to Y. We demonstrate, with simulated data, that failing to quantify non-zero stand-effects in the intercept of a linear population-level model can lead to a serious underestimation of the uncertainty in a model-dependent estimate of a stand mean, and by extension a confidence interval with poor coverage.We also provide an approximation to the variance of stand-effects in an intercept for the case when a sampling design does not afford estimation. Furthermore, we propose a method to correct a potential negative bias in an estimate of the variance of stand-effects when a sampling design prescribes few stands with small within-stand sample sizes.

Plain Language Summary

The very success of remotely sensed data in forest inventories have promoted sampling designs for field data that do not foresee replicated sampling within forest stands. As a consequence stand effects cannot be quantified and the precision of stand-level estimates of inventory variables (e.g. biomass and volume) can be seriously underestimated when stand effects are present in the study population. This study describes a process for a re-evaluation of existing design and way forward towards suitable designs that affords an estimation of stand effects with a minimum loss of design efficiency.