Predicting Shoot height of coppiced poplars : A comparison between a simple and a mixed effect linear height model

Abstract

Poplar (Populus sp.) is a fast-growing tree species of increasing interest to establish as short rotation coppice plantations for bioenergy in Sweden (10-15 year cutting cycles). After harvest, most poplar clones produce a large number of self-generated shootings. Volume or biomass models are usually based on total height (TH) and diameter at breast height (DBH). A function to predict shoot heights will therefore be useful. The objective of this study was to investigate if a linear mixed height model can improve height predictions compared to a simple model. Data of fifty shoots with observed DBH and TH in each of five harvested poplar stands in Skåne, South Sweden, was used.The simple linear model was expressed as TH = b0 + b1DBH, with transformation log (TH) = b0 + b1* log(DBH). For calculations of the transformed models, the retransformation equation TH=exp(b0)*DBHb1 was used. “Site” was the group level (random intercept and “slope”) in the mixed effect model. The transformed simple model reduced the absolute bias with 3.2 %, yet with a small tendency to underestimate height (0.9 dm). However, the main effect was observed when selecting a linear mixed effect model, with absolute bias reduction from 9 dm down to 6 dm resulting in a remarkable lower span between Min and Max values. A linear mixed model will give a reduction of absolute bias up to almost 30 % compared to a simple linear model (Table 1.). Poplar (Populus sp.) is a fast-growing tree species of increasing interest to establish as short rotation coppice plantations for bioenergy in Sweden (10-15 year cutting cycles). After harvest, most poplar clones produce a large number of self-generated shootings. Volume or biomass models are usually based on total height (TH) and diameter at breast height (DBH). A function to predict shoot heights will therefore be useful. The objective of this study was to investigate if a linear mixed height model can improve height predictions compared to a simple model. Data of fifty shoots with observed DBH and TH in each of five harvested poplar stands in Skåne, South Sweden, was used. The simple linear model was expressed as TH = b0 + b1DBH, with transformation log (TH) = b0 + b1* log(DBH). For calculations of the transformed models, the retransformation equation TH=exp(b0)*DBHb1 was used. “Site” was the group level (random intercept and “slope”) in the mixed effect model. The transformed simple model reduced the absolute bias with 3.2 %, yet with a small tendency to underestimate height (0.9 dm). However, the main effect was observed when selecting a linear mixed effect model, with absolute bias reduction from 9 dm down to 6 dm resulting in a remarkable lower span between Min and Max values. A linear mixed model will give a reduction of absolute bias up to almost 30 % compared to a simple linear model.

Published in

Title: Forest Biomass Conference. Book of Abstracts
Publisher: Poznań University of Life Sciences, ORNATUS

Conference

Forest Biomass Conference

SLU Authors

• Hjelm, Birger

  • Department of Crop Production Ecology, Swedish University of Agricultural Sciences
    

UKÄ Subject classification

Forest Science

Publication identifier

• ISBN: 978-83-921460-9-4

Permanent link to this page (URI)

https://res.slu.se/id/publ/53588