Accuracy and equivalence testing of crown ratio models and assessment of their impact on diameter growth and basal area increment predictions of two variants of the forest vegetation simulator

Laura P. Leites, Andrew P. Robinson, Nicholas L. Crookston

Research output: Contribution to journalArticlepeer-review

39 Scopus citations

Abstract

Diameter growth (DG) equations in many existing forest growth and yield models use tree crown ratio (CR) as a predictor variable. Where CR is not measured, it is estimated from other measured variables. We evaluated CR estima- tion accuracy for the models in two Forest Vegetation Simulator variants: the exponential and the logistic CR models used in the North Idaho (NI) variant, and the Weibull model used in the South Central Oregon and Northeast California (SO) variant. We also assessed the effects of using measured (CRm) versus predicted (CRP) crown ratio for predicting 10 year DG and 30 year basal area increment (BAI). Evaluation criteria included equivalence tests, bias, root mean square error, and Spearman's coefficient of rank correlation. Inventory data from the Winema and the Colville National Forests were used. Results showed that the NI variant models overpredicted CR when CRm was below 40% and underpredicted CR when it was above 60%, whereas the SO variant model overpredicted CR when CRm was smaller than 60%. Differences between CRm and CRP were positively correlated with differences in DG predictions. Using CRm versus CRP resulted in 30 year BAI absolute percent differences of 10% or less for more than 50% of the plots.

Original languageEnglish (US)
Pages (from-to)655-665
Number of pages11
JournalCanadian Journal of Forest Research
Volume39
Issue number3
DOIs
StatePublished - Mar 1 2009

All Science Journal Classification (ASJC) codes

  • Forestry
  • Ecology
  • Global and Planetary Change

Fingerprint Dive into the research topics of 'Accuracy and equivalence testing of crown ratio models and assessment of their impact on diameter growth and basal area increment predictions of two variants of the forest vegetation simulator'. Together they form a unique fingerprint.

Cite this