The Lodgepole Pine Partnership: Managing for Value in Lodgepole Pine

Integration and Application: A Tale of Opportunities

Jim Stewart and Roger Whitehead

Focus on one species

Lodgepole Pine:

• a high-value resource

• a simple forest type

• ideal for proof-of-concept

Two heads better than one

• combined research groups in Victoria and Edmonton

• complementary skills and experience

• cover the range of LPP in Canada

Two heads better than one

• The Team: Roger Whitehead, Jim Stewart, John Vallentgoed, Jared Salvail, Dominique Lejour, Ross Koppenaal, Mingliang Wang, Mike Wulder, Gordon Frazer, Joanne White

• Collaborators: FPInnovations, Canadian Forest Service, University of British Columbia, BCMinistry of Forest and Range Stand Development Modelling Group, Foothills G&Y Association

3 Gifts of the Magi

• LTRI data from CFS, FGYA

• Foothills LiDAR coverage from ASRD

• IWQTA study from West Fraser

Four Links in the Chain

• Inventory tools

• Correlations

• Production

• Valuation

Objectives:• Develop Inventory Tools

– tree & stand description from LiDAR into GIS – add fibre attributes/biomass to GIS using

correlations– link outputs to Optimization Models

• FPSuite, FPInterface, BiOS, Optitek, WoodSim, etc.

• Develop Production/Decision Support Tools– TASS , SYLVER, GYPSY– effects of silviculture on value

• Validate & Extend Applications– Long Term Research Installations– sampling tools and techniques for correlations

Lodgepole Pine Partnership

A Fistful of Challenges

1. Communication with partners

2. Communication among collaborators

3. Data issues

4. Bridging the gaps

5. Industry capacity

½ Dozen Golden Opportunities

• LiDAR-enhanced inventory • Predicting fibre attributes from LiDAR• Modelling fibre attributes from tree

and stand variables • Non-destructive tools • Silviculture systems for desired

attributes • Integration with existing models

LiDAR-enhanced Inventory

Three ongoing activities:1. Capacity of ground and airborne LiDAR to

predict stand level attributes 2. Demonstration of potential of ground

based LIDAR to predict individual tree attributes

3. Review of capacity of terrestrial and airborne LiDAR to assess wood fibre quality attributes

LiDAR & Digital Imagery

Ground calibration • Allometric relationships by spp. from geo-referenced plots at same resolution (DBH, Vol, etc)• Fibre Attributes from Silviscan & FQA for trees in same plots (MoE, MFA, DEN, PER, MFL, CRS)

+

Canopy Metrics & DEM

GMV m3/haAverage GMV = 50.4 Average GMV = 50.4 ±± 4.8 4.8 mm33/ha/ha

Regressions

Enhanced Inventory Layers

1.LiDAR-enhanced Inventory

Study Area

• Hinton FMA ~ 988,870 ha

• AVI available for the entire study area

• ~ 324,273 inventory polygons

LIDAR Metrics

• FUSION/LDV software (USFS)• 35,954,401 25m cells• ~ 8 GB of data• 51 metrics generated for each

cell:

Total return count > 2 m

Minimum height

Maximum height

Mean height

Modal height

Std. dev. height

Height percentiles

Percent cover

LegendAVI polygons

Maximum height (m)High : 39.9996

Low : 2.0105

LegendAVI polygons

Percent coverHigh : 100

Low : 0.1031

Frazer, G., M. Wulder, and O. Niemann, 2005; Simulation and quantification of the fine-scale spatial pattern and heterogeneity of forest canopy structure: A lacunarity-based method designed for analysis of continuous canopy heights, Forest Ecology and Management, Vol. 214, pp. 65- 90

Stand structure- Canopy closure - Multi-layers

Predicting Tree & Stand Attributes from LiDAR

• Predicted attributes: knots, social status, vigour, growth rate, stem volume, stem biomass, aboveground biomass, site quality, crown length, vertical light profile, stocking, local climate, seasonal distribution of growth

• Measuring: branchiness, branch diameter, crown dimensions, stem diameter & growth, Tree height & growth, vertical distribution of foliage, gap fraction, stocking density, terrain slope & aspect, precipitation and temperature, site nutrient availability

2. Terrestrial LIDAR and Individual Tree Attributes

Echidna Validation Instrument (EVI) ©

Full waveform

1.5M points per scan

Co-registration of five scans per plot

CSIRO Echidna Validation Instrument, photo taken by Martin van Leeuwen

T-LiDAR

Tree detection

Tree reconstruction

Structural analysis

Raw Terrestrial LiDAR point cloud

MacKay plot D1

Edge detection

Medial axis Transform

Filtering

1521m

10

T-LiDAR• Discrete return

image showing a combination of terrestrial and airborne LiDAR

Full waveform image from the echidna

Predicting Fibre Attributes from LiDAR

• Modelling Fibre Attributes directly from LiDAR

Predicting Fibre Attributes from LiDAR

PC1 – 53.2 % MFA, MOE, DEN

PC2 – 34.4 % CRS, PER, MFLH1

H2

H3

H4

H5

H6H7

H8

H9

H10

H11

H12

H13

H14

H15

H16

H17

H18

H19

H20

H21

H22

H23

H24

H25

H26

H27

H28

H29 H30

H31

H32

H33

H34H35

H36

S101

S102

S103

S104

S105

S106

S107S108

S109

S110

S111

S112

S113

S114

S115

S116

S117

S118

DEN

T_PER

T_CRS

T_MFA

MOE

T_MFL

PCA - Wood Quality

Axis 1

Axi

s 2 Sp

PlSbSw

Exploratory analyses of the IWQTA wood fibre dataset using principal components analysis

Predicting Fibre Attributes from LiDAR

PC1: DBH Incr, Branch diam. Slenderness coef

PC2 – 34.4 % Height, Ht to Live Crown

H1

H2

H3

H4

H5

H6H7

H8

H9

H10

H11

H12

H13

H14

H15

H16

H17

H18

H19

H20

H21

H22

H23

H24

H25

H26

H27

H28

H29 H30

H31

H32

H33

H34H35

H36

S101

S102

S103

S104

S105

S106

S107S108

S109

S110

S111

S112

S113

S114

S115

S116

S117

S118

T_STEMS

T_PLC

SC

T_DBH

HT

T_AGE

T_DBHI

SI

T_BD

T_CR

HTLC

PCA - Wood Quality

Axis 1

Axi

s 2 Sp

PlSbSw

Correlation analysis with stand variables

Predicting Fibre Attributes from LiDAR

Ground- based Airborne LiDAR

attribute Predictors adj. R2 adj. R2 Predictors

Perimeter DBH, SI 0.781 0.500 Lh100

Coarseness Ht 0.611 0.525 Lh90

Fibre length SI 0.592 0.428 Lh100

Density Br. Diam., SpH

0.587 0.317 TBZ

MoE Br. Diam. 0.485 0.394 Canopy cover, TBZ

MFA DBH incr. 0.272 0.158 TBZ

Ground-based Fibre Attribute Prediction

• Evaluating IWQTA models

• Developing improved models with LTRI data

Evaluating IWQTA models• IWQTA models for within-tree (radial) variation

well for ring width, MFA and MOE (EF>0.15), MFA somewhat over-estimated. Other variables were poorly estimated (EF<0).

• better prediction for juvenile than for mature wood 2

432

210 PHaPHaRaRaaP

Evaluating IWQTA models• BH-averaged between-tree models worked best for

Density, MFA and MOE, but not when split into separate Juvenile Wood and Mature Wood models.

Fibre phase n mean Bias RMSE R2

DEN Juvenile 244 499.7503 -8.7824 42.1115 0.0819

Mature 233 513.7682 -8.5921 62.0114 0.0791

PER Juvenile 244 113.4229 13.0168 13.9962 -5.0401

Mature 233 115.6814 13.6861 14.7177 -5.4000

CRS Juvenile 244 383.0726 54.1408 62.7640 -2.0607

Mature 233 409.7597 59.0901 73.5273 -1.0810

MFA Juvenile 244 14.1050 -0.2373 2.5409 -0.0087

Mature 233 10.6557 -0.7968 2.1685 -0.1568

MOE Juvenile 244 12.6777 -0.2091 1.8235 0.1081

Mature 233 15.6914 -0.3397 2.5016 0.0553

KSCaDBHIaSCaHTaDBHaaPFH 543210

Evaluating IWQTA models

• BH-averaged between-tree models

• Site-specific

Fibre region n mean Bias RMSE R2DEN CR 118 527.3699 11.7548 48.4509 0.0498

MC 113 477.0029 -39.3912 56.8807 -0.5716MK 89 526.9228 5.2138 39.7338 0.1415PA 40 547.3564 23.8942 60.7223 -0.2124TF 59 482.1844 -28.5361 63.3780 -0.1465TN 58 487.5313 -14.0840 52.7015 -0.1338

PER CR 118 117.5522 14.9810 15.7209 -8.1660MC 113 115.5063 15.4336 16.1311 -9.3883MK 89 111.0494 10.3607 11.5234 -3.0322PA 40 115.3086 11.7036 12.7746 -5.1757TF 59 111.5412 11.4887 12.5704 -3.9204TN 58 114.2918 13.5365 14.4031 -6.1021

CRS CR 118 433.1360 80.6769 88.3442 -3.6802MC 113 380.0317 50.2538 59.7309 -1.9411MK 89 386.1173 47.2340 59.2053 -1.2673PA 40 430.2524 60.8689 66.9641 -4.0622TF 59 361.0944 35.2979 57.8211 -0.7855TN 58 379.4998 52.7353 59.9064 -3.0000

MFA CR 118 12.1829 -0.6853 2.1869 0.3469MC 113 12.8982 -0.1128 2.2249 0.3976MK 89 12.0586 -0.8421 2.3331 0.2213PA 40 11.2699 -1.6203 2.4828 -0.2290TF 59 13.1071 0.1830 2.8492 0.3198TN 58 12.6206 -0.3621 2.4116 0.2974

MOE CR 118 15.0352 0.5690 2.1087 0.4667MC 113 13.0883 -1.3266 2.1523 0.1290MK 89 14.7092 0.0107 1.9127 0.2984PA 40 16.0863 1.3019 2.4303 0.1591TF 59 13.1035 -1.1438 2.6939 0.0164TN 58 13.2871 -0.5681 1.9969 0.1274

Evaluating IWQTA models

• Prediction error shows different limitations of separate JW & MW BH-average models for different attributes

Modelling Fibre Attributes

• New models based on CFS and West Fraser data, and using new statistical methods

• Fibre Length at BH can be well predicted from tree height and ring age

0

1

2

3

4

10 30 50 70 90Ring age (yr)

Fibr

e Le

ngth

(mm

)

ObservedPredicted

FL=a+(b0+b1H) ln(ring) R2=0.8668

Modelling JW/MW transition • simplifying fibre attribute

prediction models by segregating mature from juvenile wood

• tested 2-segment, 3- segment and empirical (rate of change) models

• e.g., MFA data was best fit by empirical model, with 2.5% change marking the JW/MW transition

• other attributes show less pronounced transition

5

15

25

35

45

55

0 20 40 60 80Ring Number (CR)

MFA

(deg

rees

)

Empirical2-seg3-segEXPTransition point

cRingbeaMFAEXP :

0

10

20

30

40

50

60

70

80

90

0 10 20 30 40 50 60 70 80Ring Number

MFA

(de

gree

s)

Empirical

EXP

Percentage 2.5

CRPA+10TN+20TF+30MK+40

MC+50

cRingbeaMFAEXP :

Non-destructive tools

Evaluating acoustic tools for segregating standing trees and logs by stiffness (AV) for increased value recovery and grade outturns of structural lumber and LVL productsModel calibration?

• Acoustic tools measure acoustic velocity (AV) of a stress wave in a tree or log.

• AV increases with wood stiffness.

… for segregating stands

…for sorting logs

Acoustic tools

Acoustic Velocity Tools

Some early results

Resonance (Hitman on BH-LB) MoEest. vs. SilviScan MoEMacKay and Cranbrook sites

R2 = 0.5132

8

10

12

14

16

18

20

8 9 10 11 12 13 14

Resonance (Hitman) MoEest. (GPa)

Silv

iSca

n M

oE (G

pa)

MacKay

Cranbrook

Linear (MacKay and Cranbrook)

Predicted Modulus of Elasticity (stiffness) calculated from acoustic velocity and wood density, shows a good relationship to MOE from SilviScan analysis.

Acoustic Velocity Tools Tree (ST300) vs. log (HM200 on SH-MT and BH-LB) acoustic velocity

(Cranbrook and Mackay sites combined)

R2 = 0.67

R2 = 0.4614

3.0

3.2

3.4

3.6

3.8

4.0

4.2

4.4

3.8 4.0 4.2 4.4 4.6 4.8 5.0 5.2 5.4 5.6

ST300 (km/sec)

HM

200

(km

/sec

)

BH-LB

SH-MT

Linear (BH-LB)

Linear (SH-MT) Relationship between acoustic velocity in standing trees and in logs

Acoustic Velocity Tools Cumulative frequency of tree acoustic velocity on three lodgepole pine sites

(MSR grade thresholds overlayed)

0

20

40

60

80

100

120

3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6 4.8 5.0 5.2 5.4 5.6 5.8 6.0

Acoustic velocity (km/sec.)

Cum

mul

ativ

e pe

rcen

tage

of t

rees

(%)

MacKay, ABParson, BCCranbrook, BC

g3 g2 g1 Distribution of tree acoustic velocity (stiffness) in stands relative to MSR grade

Silviculture systems

• 7 different LTRIs testing a range of silviculture systems – Commercial Thinning – Precommercial Thinning – Fertilization

Commercial Thinning, Fertilization, & Fibre Quality

• CT from below to uniform 4m or 5m spacing (~50% BA removal). Fertilizer applied 3 years later.

• Captured 125 – 175 m3/ha in small sawlogs, while leaving the larger diameter trees.

• After 15 years, significantly increased tree-level volume increment; fertilization had a smaller effect.

Commercial Thinning, Fertilization, & Fibre Quality• CT gave higher % of

larger logs 15 years later • lower harvest cost/m3 • higher potential value

recovery/m3 (wider & longer boards)

• No significant effects of either thinning or fertilization on MoE, microfibril angle, cell dimensions, density, mature fibre length or fibre coarseness.

0%

20%

40%

60%

80%

100%

Thinned (4

m)Thinn

ed (4m)+F

ertUnth

inned+Fert

Control

Thinned (5

m)Thinn

ed (5m)+F

ert

Control

>25cm20-25cm15-20cm10-15cm

Size Distribution of 5 metre logs* as a % of total within each treatment

Pre-Commercial Thinning & Fibre Quality• PCT at 22-27 y.o. in

3 sites in Alberta • At the heavier

thinning levels tested, tree-level volume increment was sharply increased

Pre-Commercial Thinning & Fibre Quality

Size Distribution of 5 metre logs* as a % of total within each treatment

0%

20%

40%

60%

80%

100%

Thinned (750sph)

Thinned (2600sph)

Control

Thinned (1000sph)

Thinned (4000sph)

Thinned (8000sph)

Control

Thinned (1000sph)

Thinned (4000sph)

Thinned (8000sph)

Control

>22cm17-22cm12-17cm7-12cm

MacKay TP Pole Flat TP Pole North• The heavier thinning levels gave larger potential log- size distribution after 40+ years

• But, all fibre attributes were affected, except for MFA

• Operationally significant?

Integration

• SYLVER • GYPSY • FPSuite• Wood Fibre Value Simulator

TASS Tree And Stand Simulator

• BC MFR Stand Dev Modelling Team• Tree-level, accumulated to stand• Used in FM planning and AAC allocation in BC

(some species) • Predicts stand volume by piece-size, knots &

early/late wood, juv./mature, wood density (pith-bark and bottom to top) and is linked to DSS software

Updating “SYLVER”…“SYLVER”: Stand Yield, Lumber Value & Economic Return

Tree File

Harvest & Bucking Module

Log Grading Module

Log File

Log Valuation Module

Lumber & Chip Grading Module

Lumber & Chip Valuation Module

G&Y Module

(TASS)

Silviculture Costs

Lumber & Chip File

Sawmill Simulation

Module

GYPSY Growth & Yield Projection SYstem

• AB Sustainable Resource Development• Stand level G&Y model• Basis for FM planning and AAC allocation

in Alberta • Appears to be compatible with some

quality predictors in IWQTA study

FPInnovations Tools

• FPSuite• FPInterface• BiOS• Optitek• WoodSim• Wood Fibre Value Simulator

½ Dozen Golden Opportunities

• LiDAR-enhanced inventory • Predicting fibre attributes from LiDAR• Modelling fibre attributes from tree

and stand variables • Non-destructive tools • Silviculture systems for desired

attributes • Integration with existing models

Objectives:• Develop Inventory Tools

– tree & stand description from LiDAR into GIS – add fibre attributes/biomass to GIS using

correlations– link outputs to Optimization Models

• FPSuite, FPInterface, BiOS, Optitek, WoodSim, etc.

• Develop Production/Decision Support Tools– TASS , SYLVER, GYPSY– effects of silviculture on value

• Validate & Extend Applications– Long Term Research Installations– sampling tools and techniques for correlations

Lodgepole Pine Partnership