Annualized diameter and height growth equations for plantation grown Douglasfir, western hemlock, an

1
Annualized diameter and height growth equations
for plantation grown Douglas-fir, western
hemlock, and red alder
2
Introduction

• Most regional individual tree growth yield
  models operate on a 5-10 year time step
• It is commonly assumed that increasing the
  temporal resolution of the model will decrease
  overall precision
• Plot data are typically collected on a 2-10 year
  interval
• makes estimating annual growth difficult and
  imprecise

3
Current state of regional models1

• (Cubic volume (ft3/acre) after 20 years of
  simulation)

1Johnson, G. 2005. Growth model runoff II. Growth
Model User Group Meeting. Vancouver, WA. 15 Dec.
2005. Available online http//www.growthmodel.org
/
4
Current state of regional models1(response to
200 lbs N/acre fertilization)
1Johnson, G. 2005. Growth model runoff II. Growth
Model User Group Meeting. Vancouver, WA. 15 Dec.
2005. Available online http//www.growthmodel.org
/
5
Current state of regional models1

• There is a wide range of responses to thinning,
  fertilization, and the combination of treatments
  for 6 commonly used models
• No one model adhered to all the general research
  findings on these treatments
• Results suggest that long model time steps may be
  inadequate for capturing growth dynamics
  following silvicultural treatment

6
Objectives/Justification

• Use the iterative method of Cao (2002 CJFR 32
  2051-2059) to estimate annualized growth
  equations
• Diameter and height for now, but crown recession
  and mortality in the future
• Fit equations with maximum likelihood and
  multi-level mixed-effects
• random effects were then correlated with
  installation physiographic features
• Estimate parameters for 3 plantation species in
  western OR and WA (Douglas-fir, western hemlock,
  red alder)

7
Methods

• Plantation data obtained from the Stand
  Management Cooperative, Swiss Needle Cast
  Cooperative, and Hardwood Silviculture
  Cooperative
• Only control (untreated) plots used
• Hann et al. (2003 OSU FRL Res. Contrib. 40)
  model forms used
• Site indices used
• DF, Bruce (1981 For Sci 4 711-725)
• WH, Bonner et al. (1995 Can. For. Serv. Info
  Report BC-X-353)
• RA, Nigh Courtin (1998 New Forest 16 59-70)

8
Methods Model fitting technique

• Caos approach
• Requires no modification of the growth data (i.e.
  no interpolation to a common remeasurement
  length)
• Constrains predicted periodic growth, which
  reduces the error associated with annually
  updating a tree list
• Uses a simple do loop combined with a
  minimization function
• Automatically weights longer remeasurement
  intervals more than short intervals.

9
Results

• Models fit the data well (r2 0.5 0.9) and
  were consistent with biological expectations
• Multi-level mixed effects indicated significant
  installation and plot variation
• Diameter growth peaked at 30, 25, and 15 cm DBH
  for DF, WH, and RA respectively
• Hann et al. (2003) height growth equation worked
  well for DF, but modifications are required for
  WH and RA

10
Results

• Installation random effects provided a few
  interesting relationships for DF and RA, but fits
  were generally poor (r2 lt 0.35)
• WH showed no relationship with any physiographic
  variable

11
Results
12
Simulation

• 5 SMC control plots with varying site indices and
  the longest period of observation (gt15 years)
  were selected
• Growth was simulated using the annualized growth
  equations combined with a previously fit annual
  mortality function and a static crown recession
  model
• Predictions were compared with SMC-variant of
  ORGANON v8

13
Simulation

• After 15 years of simulation, the annualized
  equations were comparable or in some cases,
  better than ORGANON predictions

14
Simulation
15
Simulation LOGS plots

• Similar degree of bias observed after 25-32 years
  of simulation on 6 LOGS control plots
• Height growth overpredicted on intermediate and
  suppressed individuals
• Mortality significantly overpredicted
• Degree of bias similar to a model with a much
  longer time step

16
Discussion

• We found systematic variation in growth across
  the landscape for DF and RA
• south aspects were the poorest
• DF growth increased with greater slopes, while
  RA decreased
• The multi-level mixed effects model fits were
  poorer predictors than those obtained with
  maximum likelihood when applied to new locations,
  but the technique is useful for
• partitioning variation
• updating tree lists on locations with previous
  measurements

17
Future plans

• Modify the WH and RA height growth equations
• WH needs to be simplified to provide more stable
  parameter estimates
• RA shows a bias across stand density
• Fit modifiers for thinning and fertilization
• Preliminary simulation code for R/SPLUS is
  available online (www.holoros.com/goab.htm) and
  an EXCEL/ACCESS interface is currently being
  developed

18
Conclusion

• Annualized equations offer an opportunity to
  improve the precision of growth projections,
  while providing several additional benefits
• Not restricted to a predetermined time interval
  (useful for updating inventories to the present)
• Biologically justified (i.e. trees grow on an
  annual basis so should our models)
• Improved chance of capturing the growth dynamics
  following intensive management
• Opportunity to connect empirical equations with a
  process-based model (focus of my dissertation)

19
Acknowledgements

• USDA PNW Research Station for funding this work
• Stand Management Cooperative, Swiss Needle Cast
  Cooperative, Hardwood Silviculture Cooperative,
  and their supporting members for access to the
  data and maintenance of the plots
• Andy Bluhm, Randol Collier, David Hann, David
  Marshall, and Doug Mainwaring for assistance on
  creating the growth database