improving the accuracy of diameter distribution title
TRANSCRIPT
1
Title
Improving the accuracy of diameter distribution
predictions by using stand table projection
Meeting: Precision Forestry Symposium
Location: Stellenbosch
Prepared by: Heyns Kotze
Date: 3 March 2014
12 March 2014 Heyns Kotze - PFS PAGE 1
Location
2
12 March 2014 Heyns Kotze - PFS PAGE 2
Working Circles and Species
Working
Circle
Area
(%)
Species Rotation
(Years)
PSPH
Gum Pulp 72
Eucalyptus dunnii
E. grandis
E. grandis x urophylla
E. grandis x nitens
E. grandis x camaldulensis
...
±8 1389
1667
Pine Pulp 21
Pinus patula
P. elliottii
...
±15 1333
1667
Wattle Pulp 7 Acacia mearnsii ±10 2222
Unthinned, Even-aged, Single species stands
Plans • Annual Plan of Operations (APO) • 3 Year Tactical Plan • 30 Year Long-term Plan
12 March 2014 Heyns Kotze - PFS PAGE 3
Forest Planning
Growth & Yield Simulator
Compartment Database
Harvest Scheduling & Tactical Planning
Home-grown forest planning systems e.g. MicroForest Similar architecture for • Stand-level G&Y models • G&Y simulators
3
Improved accuracy of yield prediction
• Use accurate compartment description: Species, PSPH, Planting Date, etc.
• Use Effective compartment Area as verified from latest orthophotos
• Use a Volume% adjustment factor to cater for damages
• Use applicable Species-level models
• Use Log specification, stump height, log trim and Util top diameter
that reflects business practice
• Use accurate enumeration instead of SI-based defaults
• Calibrate models with stand-level parameters such as:
• Age, TPH, Hdom, BA, DbhStdDev, Dbhmin
• Calibrate the stand table, by using stand-table projection
Plans • Annual Plan of Operations (APO) • 3 Year Tactical Plan • 30 Year Long-term Plan
12 March 2014 Heyns Kotze - PFS PAGE 5
Forest Planning - Inventories
G&Y Simulator
Compartment Database
Harvest Scheduling
Pre-clearfell Enumeration • Sampling Intensity ~ 5% • ± 20% of area
SI-based defaults • Management level
4
• Trees per hectare (TPH) • Basal Area (m2/ha) • Dominant Height (m) • Dbhq (cm) • Util Volume (m3/ha) • MAI (m3/ha/yr)
Stand-level view of a compartment
Size-class variables • Dbh distribution: Number of trees in each Dbh class • Average Ht by Dbh relationship
Size-class view of a compartment
Dbh class (cm)
He
igh
t (m
)
TPH
Fre
qu
en
cy
5
Single tree Stand-level Process
Stand-level projection Growth = f(Age, Hd, TPH, BA)
Physiologically based • Photosynthesis • respiration e.g. 3-PG; Evaluate Climatic effects
Individual tree models Spatial or Non-spatial e.g. SILVA
• Management planning;
• Optmization of regimes.
Understanding
• Inter-tree competition
• Mixed species
Understanding growth
responses to:
• climate & site
Stand table projection
Size class
Projecting stand
structure
Growth modelling approaches
Commercial Planning Scenario Analysis
Group Component Function Graph
Growth Hdom HT2HF3:
TPH NS2CLJ:
BA BA1MR1:
Stand
Structure
Dbh distribution
Weibull MOM
- Dmin:
DMINBHWK:
- Dsdev SDBHWK:
Ave Ht by Dbh HQDBH_DD:
Products Voltaper & Bucking VOLTAPERMB:
Calibration
For SI-based
Assumptions &
BA calibration model
BA1MR1:
Projection Dbh-distribution
projection
Nepal-Summers method
Growth model architecture - unthinned
23
1
113
1
121
2
2
AGEAGE
AGEHD
AGEAGE
AGEHD
133
1
1
12212
100100
AGEAGETPHTPH
1
15
1
141312
1
101
lnlnlnlnexp
AGE
HD
AGE
TPHHDTPH
AGEBA
bRATb DmeanDminD
AGENRSDD RATbRAT 3210
BAAGEDbSd 210
2
2231
21
2
2
10
2 11 IXaIXaXXDBHd ib
1
15
1
141312
1
101
lnlnlnlnexp
AGE
HD
AGE
TPHHDTPH
AGEBA
Dq
DBHHmeanh i
i 210exp1
6
Growth modelling approach published in the 2012 South African Forestry Handbook
12 March 2014 Heyns Kotze - PFS PAGE 11
SI-based growth prediction
For longer term plans, we use SI-based growth prediction.
• Require the growth model to reflect ave behaviour over Age, SQ and Stand density.
Surv% SI Pred
Cal BA
Weibull
Log specification, Greedy bucking algorithm,
Defaults for Stump Ht, Log trim, Util top diameter
• Use stable defaults to calibrate the models.
7
12 March 2014 Heyns Kotze - PFS PAGE 12
Enumeration-based growth projection
For shorter-term plans (APO), we require improved accuracy for:
• Budgeting purposes • Rates for contractor payment • Balancing of man-machine in Tactical Plans
Therefore we use Enumeration based growth projection
Enumeration
• Age
• Dominant Height
• Trees per Hectare
• BA
• Dbh distribution • Minimum Dbh • Standard deviation of Dbh • Unique shape
12 March 2014 Heyns Kotze - PFS PAGE 13
Enumeration-based growth projection
TPH Hdom BA
Future Stand table
Dbh StdDev
Dbhmin
Observed Stand table
Project
8
12 March 2014 Heyns Kotze - PFS PAGE 14
Dbh distribution ~ Stand table
Dbh-class Midpoint Frequency
3 5.8
4 0.0
5 5.8
6 0.0
7 0.0
8 0.0
9 0.0
10 0.0
11 0.0
12 0.0
13 5.8
14 5.8
15 0.0
16 5.8
17 5.8
18 0.0
19 5.8
20 0.0
21 0.0
22 11.6
23 0.0
24 11.6
25 11.6
....
Total 296
Dbh distribution Stand table
12 March 2014 Heyns Kotze - PFS PAGE 15
Reconstructing Dbh distribution - Weibull method
To estimate Weibull parameters:
• Method of Moments (Garcia, 1981)
• Input:
• Dbh_min
• Dbh_mean
• Dbh_StdDev
• Estimate these from stand-level
parameters, such as Age, TPH, BA.
Weibull pdf:
• Parameters:
• a ~ location
• b ~ scale
• c ~ shape
• Provides smooth unimodal distributions
9
12 March 2014 Heyns Kotze - PFS PAGE 16
Generalized approach to Stand Table Projection
• Algorithm defined by Nepal and Somers (1992).
• Method requires: a current stand table,
and BA and TPH at a future age
• General diameter growth equation, as defined by Bailey (1980).
• Uses the Weibull parameters of the current age (a1, b1, c1)
and future age (a2, b2, c2).
• Assumes that trees remain in their relative positions to others.
12 March 2014 Heyns Kotze - PFS PAGE 17
Generalized approach to Stand Table Projection
• The diameter growth equation is used to project the observed stand
table to a future age.
• The future stand table is algorithmically adjusted so that it is
consistent with future BA and TPH.
• Complex and difficult to program
• Illustrated detail with South African data
• (Corral-Rivas etal. 2009, Southern Forests Journal)
• Demo program (STPUtility.exe)
• Implemented in SA Growth & Yield simulators:
• FORSAT (KLF)
• MicroForest Planning System (Syndicate Database Solutions)
• PSAT (Mondi)
10
12 March 2014 Heyns Kotze - PFS PAGE 18
Predicting future Dbh distributions - comparison
Weibull method Stand table projection
Both approaches are consistent with the stand-level growth model,
but the stand-table projection method maintains the structure of the
observed Dbh distribution
12 March 2014 Heyns Kotze - PFS PAGE 19
Demo software: STPUtility.exe (Input)
11
12 March 2014 Heyns Kotze - PFS PAGE 20
Demo software: STPUtility.exe (Results)
12 March 2014 Heyns Kotze - PFS PAGE 21
Conclusion
The generalized Stand Table Projection method:
• alternative method to predict future Dbh distributions
• consistent with stand-level growth models
• should provide for more accurate estimates of log products
• because it maintains the shape of observed Dbh distribution
• especially useful for multi-modal or irregular Dbh distributions
Valuable addition to the general stand-level modelling approach
for even-aged single species stands as used in South Africa.
12
12 March 2014 Heyns Kotze - PFS PAGE 22
FORWARD - LOOKING STATEMENTS
It should be noted that certain statements herein which are not historical facts, including, without limitation those regarding expectations of market growth and developments; expectations of growth and profitability; and statements preceded by “believes”, “expects”, “anticipates”, “foresees”, “may” or similar expressions, are forward-looking statements. Since these statements are based on current knowledge, plans, estimates and projections, they involve risks and uncertainties which may cause actual results to materially differ from those expressed in such forward-looking statements. Various factors could cause actual future results, performance or events to differ materially from those described in these statements. Such factors include in particular but without any limitation: (1) operating factors such as continued success of manufacturing activities and the achievement of efficiencies therein, continued success of product development plans and targets, changes in the degree of protection created by Group’s patents and other intellectual property rights, the availability of capital on acceptable terms; (2) industry conditions, such as strength of product demand, intensity of competition, prevailing and future global market prices for the Group’s products and raw materials and the pricing pressures thereto, financial condition of the customers, suppliers and the competitors of the Group, potential introduction of competing products and technologies by competitors; and (3) general economic conditions, such as rates of economic growth in the Group’s principal geographical markets or fluctuations of exchange rates and interest rates. Mondi does not a) assume any warranty or liability as to accuracy or completeness of the information provided herein b) undertake to review or confirm analysts’ expectations or estimates or to update any forward-looking statements to reflect events that occur or circumstances that arise after the date of making any forward-looking statements.