View
215
Download
1
Tags:
Embed Size (px)
Citation preview
Impact of plot size on the effect of competition in individual-tree models and their applications
Jari Hynynen & Risto OjansuuFinnish Forest Research Institute,
Vantaa, Finland
Introduction• Description of competition has major impact on model
behavior• Correlation between observed competition and
increment are affected by sampling > plot size effect– statistical effect – sampling error
• when plot size in the modelling data differs from plot size in the simulation data
• correction method (Stage and Wykoff 1998)– coefficients of the structural model are corrected in
simulation according the plot size in the simulation data
– biological effect• plot size vs. size of competition zone• affects predicted stand dynamics• affects comparisons between the simulated development of
different treatment regimes
Goal
To empirically study the effect of alternative sample plot size on the description of competition
1) in growth model (model parameters)
2) in simulation results
in Norway spruce dominated stands based on an extensive data from inventory growth plots
Description of Competition
Overall stand density: Relative density factor: RDF
• describes the relative distance with respect to self-thinning line (Hynynen 1993)
• expressed by tree species
• calculated treewise
Dg
N self-thinning line:RDF=1.0
a stand withRDF=0.8
Competitive status of a tree:RDF of larger than subject tree:
RDFL
Modelling data
• an objective sample of Norway spruce stands: Repeatedly measured inventory growth plots – 337 stands
– 802 sample plots
– 9300 tree measurements
– one 5-year growth period from each stand were used
• 4550 growth observations Norway Spruce stands in INKA data
tally tree plots
stand
Experimental design
• Three systematically located sample plots within a stand– circular tally tree plots
(36 trees/plot on the average) including
sample tree plots
– smaller concentric sample tree plots (10 trees/plot on the average)
• Models were developed for trees on sample tree plots
Alternative sampling applied in the estimation of stand density variables
• Sample 1: RDF and RDFL calculated separately for each sample plot– 10 trees/sample– 116 m2 sample area
model Variant 1
Alternative sampling applied in the estimation of stand density variables
• Sample 2: RDF and RDFL calculated separately for each tally tree plot – 36 trees/sample– 331 m2 sample area
model Variant 2
Alternative sampling applied in the estimation of stand density variables
• Sample 3: RDF and RDFL calculated from pooled data of sample tree plots– 29 trees/sample– 337 m2 sample area
model Variant 3
Alternative sampling applied in the estimation of stand density variables
• Sample 4: RDF and RDFL calculated from pooled data of tally tree plots – 103 trees/sample– 964 m2 sample area
model Variant 4
Model development
• Individual-tree, distance-independent models were developed for– tree basal area growth– tree crown ratio
of Norway spruce trees• Models were developed for trees of sample
tree plots• Four model variants (Variants 1 to 4) were
fitted to data with four alternative values of competition variables obtained from four different sampling (Samples 1 to 4)
Model for tree basal area growth
where d = tree diameter at breast height, cmcr = tree crown ratioTS = temperature sum, dd.RDFL = relative density factor of trees larger than subject treeRDFNs RDFSp, RDFbl = relative density factor of Scots pine, Norway spruce,
and broad-leaved tree speciesHdom = stand dominant height, m
SI = site index of Norway spruce (index age 50 years), m
SC1, SC2, SC4 = categorical variable referring site types
u = random stand effectv = random sample plot effect
e = random effect of a tree
ln(ig)= a0+a1ln(d)+a2d2+a3(ln(d))2 +a4cr+a5cr(TS/1000)
+a6RDFL2 +a7(ln(RDFNs+1)+a8(ln(RDFSp+1)+a9(ln(RDFbl+1)
+a10(1/ln(Hdom)) + a11(1/Hdom2)
+ a12ln(SI) +a13SC1+a13SC1+a14SC2+a15SC4
+ u + v +e
tree dimensions
competition
stage of stand development
site
random parameters
Tree basal area growth model variants
Variant 1 Variant 2 Variant 3 Variant 4 Variable Parameter values Intercept -4.259 -3.411 -5.946 -5.699
ln(d) 0.665 0.720 0.672 0.738 d2 -0.0005 -0.0004 -0.0005 -0.0004
(ln(d)) 2 0.183 0.143 0.175 0.133 cr -0.678 -1.036 -0.696 -0.716
cr .(TS/1000) 1.916 2.101 1.923 1.911 (RDFL)2 -0.441 -0.938 -0.632 -1.095
ln(RDFNs +1) -0.434 -0.809 -0.991 -1.018 ln(RDFSp +1) -0.704 -0.555 -1.041 -0.972 ln(RDFDt +1) -0.444 -0.357 -0.916 -0.425 1/(ln(Hdom )) 6.622 5.718 7.682 6.184
1/Hdom2 -23.107 -20.922 -27.859 -22.904
ln(SI) 0.476 0.453 1.029 1.222 SC1 0.196 0.268 0.270 0.356 SC2 0.144 0.187 0.171 0.179 SC4 -0.029 -0.104 -0.135 -0.123
var(u) 0.09697 0.09380 0.09271 0.08976 var(v) 0.02573 0.02333 0.02163 0.02227 var(e) 0.2076 0.2048 0.2079 0.2055
The effect of relative stand density (RDF) on tree basal area growth of largest tree
in a stand (RDFL=0)
0
10
20
30
40
50
60
70
80
90
100
0 0.2 0.4 0.6 0.8 1RDF
Variant 1
Variant 2
Variant 3
Variant 4
basal area growth, cm²
Model
The effect of relative tree size (RDFL) on tree basal area growth in a stand with
high relative density
0
10
20
30
40
50
60
70
0 0.2 0.4 0.6 0.8 1RDFL
Variant 1
Variant 2
Variant 3
Variant 4
basal area growth, cm²
RDF=1 Model
The effect of relative tree size (RDFL) on tree basal area growth in a stand with
moderate relative density
0
10
20
30
40
50
60
70
0 0.2 0.4 0.6 0.8 1RDFL
Variant 1
Variant 2
Variant 3
Variant 4
basal area growth, cm²
RDF=0.5 Model
Model for tree crown ratio
where d =tree diameter at breast height, cmcr = tree crown ratioTS = temperature sum, dd.RDF = relative stand density factor (incl. all tree species)
Hdom = stand dominant height, m
SI = site index of Norway spruce, m
TH0-5 = categorical variable referring recent thinning (< 5 years ago)
cr = 1-e-f(x),
in which
f(x)=(a1-a11TH0-5).( Hdom)-a2 .da3 .exp(-a4RDF).TSa5 .SIa6
Variants of the model for tree crown ratio
Variant 1 Variant 2 Variant 3 Variant 4 Parameter Value
a1 a2 a3 a4 a5 a6 a11
1.426 0.589 0.269 0.491 -0.195 0.417 0.016
1.413 0.531 0.239 0.698 -0.336 0.454 0.063
0.550 0.564 0.272 0.648 -0.205 0.743 0.021
0.527 0.539 0.254 0.649 -0.237 0.770 0.028
RMSE 0.0999 0.0948 0.0992 0.0980
Predicted, mean
0.753 0.753 0.753 0.753
The effect of relative stand density on the predicted tree crown ratio
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.2 0.4 0.6 0.8 1RDF
Variant 1
Variant 2
Variant 3
Variant 4
crown ratio
Model
Simulation study
1. Model variants were added to Motti-simulator– stand simulator based on individual tree growth models
– stand-level analysis tool for assessing the effects of alternative management practices
2. The development of sample plots of a thinning trial were predicted with the model variants
3. The simulation results were analyzed by – comparing the results of model variants
– comparing the simulation results with measured stand development
in stands with different management schedule
Simulation data
• Repeatedly measured spacing trial for Norway spruce located in southern Finland– independent– subjectively chosen– treatments:
• four thinning intensities: 0,10,25,40 % of stand basal area removed
• three replicates
– plot size 1000 m2
– established in 1961– 37-year observation period (1961- 1988)– 8 measurements
• One unthinned and one repeatedly thinned sample plot were chosen for simulation study
Simulated and observed development of stand basal area in unthinned sample plot of Norway spruce
0
10
20
30
40
50
60
40 45 50 55 60 65 70 75 80stand age, years
Variant 1Variant 2Variant 3Variant 4Observed
stand basal area,
m2ha-1
Model
Predicted and observed yield of stand basal area in unthinned sample plot of
Norway spruce
0
10
20
30
40
50
60
70
Variant 1 Variant 2 Variant 3 Variant 4 Observed
MortalityThinning removalGrowing stock
basal area,
m2ha-1
Simulated and observed development of stand basal area in repeatedly
thinned sample plot of Norway spruce
0
5
10
15
20
25
30
35
40
40 45 50 55 60 65 70 75 80stand age, years
Variant 1Variant 2Variant 3Variant 4Observed
stand basal area,
m2ha-1
Model
Predicted and observed yield of stand basal area in thinned sample plot of
Norway spruce
0
10
20
30
40
50
60
70
Variant 1 Variant 2 Variant 3 Variant 4 Observed
MortalityThining removalGrowing stock
basal area,
m2ha-1
Relative difference between the predicted yields of thinned and
unthinned sample plots
-5
0
5
10
15
Variant 1 Variant 2 Variant 3 Variant 4 Observed
%
Conclusions (1/4)
1) Model parameters
• Competition effect is clearly affected by sample (and plot) size
• Overall stand density:– increase in sample size increased the the effect of
overall stand density
• Competitive status of a tree: – the effect increased with increasing plot size
• sampling area cannot compensate the small plot size: Variants 2 and 4 (large plots) more sensitive to RDFL than Variants 1 and 3 (small plots)
Conclusions (2/4)
2) Simulation results• in unthinned stand:
– notable differences in the predicted total yield between model variants
• highest level of mortality predicted with model variant 1
– largest overprediction of total yield obtained with model variant 1 (based on small plots)
• in thinned stand– no major differences in the predictions between model
variants
– all the models ended up in a slight overprediction
Conclusions (3/4)
2) Simulation results (... continued)
• Differences between the predicted development of different treatment regimes affected by model variant– include biological and statistical effects
– affect the conclusions that are drawn from the comparison of alternative management regimes
• referring growth and yield
• the value of harvestable wood
• the profitability of forest management
Conclusions (4/4)
3) Biological plot size effect
• Has major impact on the description of stand dynamics
• Sample tree plots of this study were too small for reliable description of the effective competition zone
• The impact remains small if spacing and size distribution is controlled by thinnings