Thinning intensity studies and growth modeling of Montana mixed conifer

forests at the University of Montana’s Lubrecht Experimental Forest

Thomas PerryResearch Forester

Applied Forest Management ProgramCollege of Forestry and Conservation

University of MontanaMissoula, MT

Applied Forest Management Program

Developing and promoting silvicultural tools and techniques for the restoration and renewal of western

forests.

http://www.cfc.umt.edu/AFMP/default.php

Lubrecht Experimental Forest▪ Timber ▪ Education ▪ Research ▪ Recreation ▪

The LandbasePre-acquisition period: pre-1937.• Owned by Anaconda Timber Company.• Explotive harvesting; stand re-generating disturbance.

Early Lubrecht Years: 1938-1960’s.• Focus on managing uncontrolled grazing• Small thinning studies established

Timber Management Era Begins: 1960’s.• Road building increases• Clearcutting implemented; Greenough Ridge, Stinkwater Creek, Old

Coloma Road.

Transition to Stand Tending: 1970’s.• Timber sales primarily salvage, thinning and some overstory

removal.

Stand Tending Period: 1980’s-2000’s• Diameter in many stands is large enough for viable commercial

thinning. Large scale thinning program implemented.• Viable pulp markets encourage continued thinning through 1980’s

and 1990’s.

Pine Beetle Salvage: 2000’s to present• MPB salvage operations account for more and more harvest

volume.

1100m-1900m (3630ft-6270ft)

8500ha (21,000 acres)

Douglas fir (Psme)Ponderosa pine (Pipo)Western larch (Laoc)Lodgepole pine (Pico)

Overstory

DF

PP

WL

LP

HW

OSW

Understory

DF

PP

WL

LP

HW

OSW

4-8 8-12 12-16 16-20 20-24 24-28 28-320

10

20

30

40

50

60

70

OSWHWLPWLPPDF

Diameter ClassTr

ees p

er A

cre

Overstory UnderstoryTPA 135 280BA (ft2/ac) 86.6 7.3DF (%) 53 71LP (%) 7 10PP (%) 23 9

WL (%) 14 7

The Levels of Growing Stock Thinning Network (LOGS)

• History– Established in 1983, measured at 5 year intervals

until 2003, then six years elapsed until the 6th measurement

• Intent– Establish permanent growth and yield plots for a

range of sites, species, and stand densities.– Compare several alternative stand density

measures computed for the same stands.– Evaluate multi-resource productivity in side by

side comparison (timber, range, wildlife, watershed, recreation).

• Implementation– 6 sites– 4 thinning levels (treatment) per site– 3-7 plots per treatment

3 Age Groups3 Habitat Types5 Composition classes

LOGSSite Name Code Stand Age Habitat Type Species Composition

Baker Road M1 120 PSME/SYAL, CARU Ponderosa pine, Douglas fir, Western larch

Coyote Park WL 70 PSME/LIBO, VAGL Western larch

Gate of Many Locks M2 120 PSME/SYAL, CARU Douglas fir, Ponderosa pine

Section 12 LP 80 PSME/VACA Lodgepole pine

Shoestring M3 120 PSME/SYAL, CARU Douglas fir, Ponderosa pine

Upper Section 16 PP 120 PSME/SYAL, CARU Ponderosa pine

LP M1 M2 M3 PP WL0

5

10

15

20

25

30

WLPPLPDF

Basa

l Are

a (m

2/he

ctar

e)

No Thin 10x10 14x14 20x200

500

1000

1500

2000

2500

3000

LPM1M2M3PPWL

Tree

s per

Hec

tare

No Thin 10x10 14x14 20x200

5

10

15

20

25

30

35

40

45

50

LPM1M2M3PPWL

Basa

l Are

a (m

2/he

ctar

e)

No Thin 10x10 14x14 20x200

5

10

15

20

25

30

LPM1M2M3PPWLQ

MD

(cm

)

Study Design Summary

• 6 Installations• Varied Site Conditions

– Age– Site– Composition

• No Replication• No Randomization• Design will not facillitate

statistically robust comparisons between treatments.

70 80 120

PSME/SYAL, CARU

M1, M2, M3, PP

PSME/LIBO, VAGL WL

PSME/VACA

LP

Data Set

3137 individual trees, measured 2-6 times since 1983, 12548 records.

Tree Records by Species

DF LP PP WL

3068 3276 2144 1572

Tree Records by Thinning Intensity

No Thin Level 1 Level 2 Level 3

5556 3276 2144 1572

Analysis - Data Set Goals

• Diameter growth model• H:D model• Volume growth model• Compare with FVS

growth predictions for local stands.

Diameter Growth Model

Modeling Process- Overview

• Stepwise process• Predicting diameter – Previous diameter– Density measures– Species effects

• Species specific models• Linear modeling in R

DBH =

DBH t-1

DBH t-1 + TPH t-1

DBH t-1 + BA t-1

DBH t-1 + BA t-1 + Sp

Time series of basal area; level 1

Time series of basal area; level 3 Time series of basal area; level 4

Time series of basal area; level 2

Competition and Growth

Competition (Basal Area/hectare) Growth (Annual Increment [cm])

Thinning Intensity Thinning Intensity

Treatment Treatment

Variables-Why Drop Treatment ?

• Treatment tried to create 4 levels of thinning intensity and residual density.

• Thinning intensity, residual density, and species composition varied too much for distinctions by treatment to be meaningful.

• A better option was to use actual density per plot to describe competition for individual trees.

• Use a measured variable rather than a categorical variable that did not adequately reflect stand conditions.

Variables-Density

• Trees per Hectare versus Basal Area– Expected stronger

correlation using BA– Better measure of

competition than TPH since same levels of TPH could have wide ranges of competitive stress based on QMD

Model Iterations - DetailStep Formula Intercept Coeff.1 Coeff.2 Coeff.3 R-squared F-statistic p-value

1 DBH~DBHt-1 -0.0162596 1.047564 0.9954 2.91E+06 2.20E-16

2 DBH~DBHt-1+TPHt-1 0.6245 1.032 -3.17E-04 0.9959 1.62E+06 2.20E-16

3 DBH~DBH.t-1+BA.t-1 0.77384 1.046783 -2.73E-02 0.9963 1.81E+06 2.20E-16

4 DBH~DBH.t-1+TPH.t-1+BA.t-1 0.9053 1.041 -1.21E-04 -2.34E-02 0.9963 1.22E+06 2.20E-16

5 DBH~DBH.t-1+BA.t-1+Sp 0.90108 1.04418 2.30E-02 *** 0.9964 7.35E+05 2.20E-16

6 DF -- DBH~DBH.t-1+BA.t-1 0.6926 1.03808 -1.89E-02 0.9961 4.26E+05 2.20E-16

6 LP -- DBH~DBH.t-1+BA.t-1 1.410207 1.024712 -4.19E-02 0.9914 1.20E+05 2.20E-16

6 PP -- DBH~DBH.t-1+BA.t-1 0.767952 1.04891 -2.62E-02 0.9961 5.72E+05 2.20E-16

6 WL -- DBH~DBH.t-1+BA.t-1 1.0805 1.0509 -4.41E-02 0.9973 6.68E+05 2.20E-16

Growth Increment

Formula Intercept Coeff.1 Coeff.2 Coeff.3 (Species) R-squared F-statistic p-value

Inc~Inc.t-1 + BA.t-1 + Sp 9.76E-02 0.8166 -1.09E-03 0 DF 0.7339 5.59E+03 2.20E-16 -4.19E-02 LP 2.20E-16 -1.62E-02 PP 2.20E-16 -3.47E-02 WL 2.20E-16

Wrap Up• Good fit with diameter based

model.• Utilizes 80% of data set.• Strong autocorrelation.

• Increment model is less autocorrelated.

• Utilizes 100% of data set.• Weak fit without good data

describing environmental and morphological parameters.

How useful is a diameter based model predicting a fixed growth period?

While not biologically valid, will it perform across a local landscape?

For the increment model – What could be done to account for more of the variability in the model?

Will increased site and stand factors limit the portability of this model?

Is the dataset powerful but not useful or is it a diamond in the rough?

What would you do with this data?

• Acknowledgements– Dr. David Affleck: University of Montana– Dr. Aaron Weiskittel: Universisty of Maine– Dr. Chris Keyes: University of Montana– Kevin Barnett: University of Montana– Woongsoon Jang: University of Montana