Thạc Sĩ Modelling Growth and Yield of Dipterocarp Forests in Central Highlands of Vietnam

iii
Abstract

Dipterocarp forests in Vietnam are distinct ecosystems with specific characteristics
which are different from other forest types such as evergreen forests, semi-deciduous
forests and conifer forests. According to inventory results of the Forest Inventory and
Planning Institute of Vietnam in 2005, the area of the Dipterocarp forests is approximately
680.000 ha, accounting about 5.4% the total forest area of the country and concentrates
mainly in the Central Highlands of Vietnam.
The main objective of this study is to develop a size class model based on systems
of differential equations for supporting sustainable management of the Dipterocarp forests
in Vietnam. Two data sets collected in the Dipterocarp forest in YokDon National Park
were used in this study to construct the growth model and calculate the main stand level
characteristics. They include plot group A consisting of twelve one-hectare permanent plots
with two measurements of a 5-year growth interval, and plot group B of 21 one fourth
hectare plots with a single measurement. For calibrating the growth model, only data set of
group A plots was used. In addition to be used to calculate the main stand level parameters,
the group B plots will supply reliable data sources to recalibrate the model in the future.
The study area was classified into three site quality levels based on mean height of the 20
largest trees in each plot. The measurements on these permanent plots recorded a total of
4,975 trees belonging to 64 species with diameter at breast height (dbh) from 6 cm and
above. Based on biological characteristics, trees on these plots were grouped into three
species groups: Dipterocarp species, evergreen tall species, and small-sized, lower species.
The diameter distribution of the average stands follows the form of negative exponential
distribution for all three species groups in accordance with the distribution rule of natural
uneven-aged forests. The number of trees per hectare has a tendency to decrease when
diameter increases. Stand basal area ranges from 10.15 to 26.9 m 2 ha -1 and the range of
basal area increment is between 0.27 and 0.48 m 2 ha -1 yr -1 . Standing volume ranges from
53.8 to 208.8 m 3 ha -1 and the range of standing volume increment between 1.5 and 3.86
m 3 ha -1 yr -1 . The number of tree per hectare ranges from 223 to 1.156 trees ha -1 .
The four major components of the growth model are diameter increment, mortality,
recruitment and harvesting. The first three models were developed separately for each
species group and site quality level. Multiple linear regression, non-linear regression and
logistic regression were used to estimate the parameters of diameter increment, recruitment
and mortality functions. Significant stand level variables included stand basal area, basal
area in larger trees, tree number, site quality, and significant individual-tree level variables
were diameter, diameter squared and reciprocal of diameter. Selecting the model equations iv
was based on the following criteria: suitability of biological interpretation and goodness-of-
fit statistics. The results indicated that diameter growth level of three species groups on
different site quality levels was significantly different with the exception of species group 3
on good and medium site quality. Trees grow more quickly on good sites than on poor
ones. However, the effect of site quality on mortality rate was not obvious in this study.
These major components were then embedded to the final growth model which is a size
class management-oriented model. The model was implemented in the framework of the
modelling software Vensim DSS 5.7a. It consists of 76 one-cm diameter classes ranging
from 6 to 81cm dbh for three species groups, the last class gathering all trees with diameter
above 80.5cm. Time interval for each simulation step of the model was set one year and
diameter class width was one cm.
A thorough evaluation of the growth model showed that the models were fitted very
well with the empirical data. Simulation results with the models showed that the difference
between observed and predicted values of basal areas and tree number distribution by
diameter class for a growth period of five years was small. The long-term performances of
the simulation proved plausible states of the stand evolution which is consistent with
general knowledge of stand growth over long time. This indicates that the model can be
applied in practice.
The example applications of the growth model in determining appropriate
silvicultural regimes based on the method of scenario analysis. Given the initial condition
of the stand, the model estimated the state of the stand after given years with the alternative
assumed prescriptions. The simulation results indicated that, with a selection harvesting
cycle of 10 years, different initial stand distributions will produce different sustainable
yields. The q-factor method was applied to determine the target diameter distributions that
produce maximum sustainable yields on three site qualities. The maximum diameters for
each species group were selected based on management purpose and diameter growth level
as follows: for species group 1 and 2, maximum diameters are 70, 60 and 50 cm for good,
medium and poor site quality, respectively. For species group 3, maximum diameter is 35
cm for all site qualities. From the simulation results of the model, the following target
distributions have been defined: on good site quality with following parameters: basal area
equal to 20 m 2 ha -1 , q-quotient (slope of the stem number-diameter distribution of 5 cm
classes) equal to 1.4, with the sustainable yield of 3.91 m 3 ha -1 yr -1 . For medium site quality:
basal area equal to 18 m 2 ha -1 , q-quotient equal to 1.5, sustainable yield of 3.22 m 3 ha -1 yr -1 .
And on poor site quality: basal area equal to 16 m 2 ha -1 , q-quotient equal to 1.6, sustainable
yield of 2.75 m 3 ha -1 yr -1 . In addition, the model was also used to estimate the return time
that regulates a given stand towards the target distribution stand for the twelve plots of
group A and to assess effect of wildfires on long-term yields of the Dipterocarp forests. v
The example applications presented in this study provide valuable information to
the forest managers for supporting decision making in sustainable management of
Dipterocarp forests. Other applications of the model need to be further explored in specific
contexts of the production practice.
Although there were several studies on growth and yield of natural uneven-aged
forests in Vietnam before, those studies modeled only important species in the forests and
produced yield tables dependent on the age of trees that provide less information for forest
management. In comparison to those studies, this growth model was constructed
incorporating competition effects as well as mortality and recruitment so that it has the
advantage of being able to estimate the growth of forests dynamically and independent on
the tree age for long time spans with reliable results.
However, due to the comparably small amount of data available in this study, all
data was used to calibrate the model, there was no data set aside for validating the model.
So, it is necessary to obtain more data from permanent plots and when it is available the
model should be recalibrated in order to expand the geographic research area and achieve
more accurate results. Although the growth model in this study was developed for
Dipterocarp forests that are uneven-aged, multi-species deciduous forests, the approach can
be applied to develop models for other forest types such as evergreen, semi-evergreen
forests or plantation forests. vi
Zusammenfassung

[german] Das Ziel dieser Arbeit ist die Entwicklung eines differentialgleichungsbasierten
Durchmesserklassen-Wachstumsmodells für nachhaltige Bewirtschaftung von
Dipterocarpaceenwäldern in Vietnam. Die Daten wurden im YokDon Nationalpark
erhoben. Das Programm besteht aus vier Modulen zur Abschätzung des
Durchmesserzuwachses, der Mortalität, der Verjüngung und einem Durchforstungsmodell.
Als Simulationssoftware wurde Vensim DSS 5.7a verwendet. Das Modell wurde eingesetzt,
um über Szenarioanalysen geeignete Behandlungsstrategien zu finden. vii
Table of contents


Preface and Acknowledgements i
Abstract iii
Zusammenfassung vi
Table of contents .vii
List of figures . xi
List of tables xiii
Chapter 1 Introduction .01
1.1 General Introduction 01
1.2 Research Questions and Objective of the Study 05
1.3 Outline of the Dissertation 06
Chapter 2 Literature Review .08
2.1 Studies About Forest Structure and Growth in Vietnam in General .08
2.1.1 Studies about Forest Growth and Yield 08
2.1.2 Studies about Diameter Distribution Rules .10
2.2 Studies about Dipterocarp Forests .11
2.2.1 Studies about the Dipterocarp Forests in the World .11
2.2.2 Studies about the Dipterocarp Forests in Vietnam 12
2.3 Historical Development and Classification of Forest Growth and Yield
Models 18
2.3.1 Stand Growth Models Based on Mean Stand Variables .18
2.3.2 Stem Number Frequency Models .19
2.3.3 Single-Tree Orientated Management Models .21
2.3.4 Gap and Hybrid Models 22
2.3.5 Matter Balance Models .22
2.3.6 Landscape Models 23
2.3.7 Selection of the Model Approach to be Used in This Study 24 viii

Chapter 3 Study Area and Establishment of Research Plots .26
3.1 General Information about the Study Area 26
3.1.1 Geographic Position and Boundary of the YokDon National Park 26
3.1.2 Forest types in the Park 27
3.1.3 Topography and Hydrography 28
3.1.4 Climate 30
3.1.5 Flore and Fauna Resources 31
3.1.6 Social Economic Conditions 32
3.2 Establishment of Research Plots as an Empirical Data Base for
Modelling Growth and Yield in Dipterocarp Forests 33
Chapter 4 Data and Description of Stand Characteristics 38
4.1 Ecological Classification of the Research Plots by Species Composition 38
4.2 Establishment of Stand Height Curves and Site Quality Classification 43
4.2.1 Selecting Height Curve Functions .43
4.2.2 Categorizing Species Groups .44
4.2.3 The Results of Height Curve Fitting 46
4.2.4 Site Quality Classification .47
4.3 Data sets 48
4.3.1 Data for Calculating Stand Characteristics .48
4.3.2 Data Used to Calibrate the Growth Model 49
4.4 Stand Variables 53
4.4.1 The Method of Calculating Stand Variables 53
4.4.2 Calculation of Stand Variables 53
4.4.3 Relationships between Stand Variables 57
Chapter 5 Model Conception and Parameterization 60
5.1 Model Conception 60
5.1.1 The Concept of System Dynamics Diagrams 60
5.1.2 Model Structure and Implementation 62 ix
5.2 Development of the Major Components of the Growth Model .71
5.2.1 Diameter Increment Model 71
5.2.2 Mortality Model 74
5.2.3 Recruitment Model 76
5.3 Results of Model Parameterization 77
5.3.1 Diameter Increment Model 77
5.3.2 Mortality Model 81
5.3.3 Recruitment Model 84
Chapter 6 Model Evaluation .88
6.1 Evaluation of the Model Approach .89
6.2 Validation of the Growth Model 90
6.2.1 Short-Term Prediction of a 5-Year Period .91
6.2.2 Long-Term Validation of Steady States 94
6.3 Evaluation of the Growth Simulator 99
Chapter 7 Applications of the Growth Model DIGROW . 101
7.1 Estimation of the Growth and Yield of Forest Stands and Determination
of the Target Diameter Distributions 102
7.2 Estimation of Time to Regulate a Given Stand to Target Stand . 110
7.3 Evaluation of Effects of Wildfires on Long-Term Sustainable Forest Yield 114
Chapter 8 Discussion . 118
8.1 Growth Model Approach and Parameterization . 118
8.2 Simulation Results of the Growth Model 121
8.3 Effects of Wildfire . 123
Chapter 9 Conclusion and Perspective 124
9.1 General Conclusion . 124
9.1.1 The growth Model Approach and Development 124
9.1.2 Model Applications 125
9.1.3 Data Assessment 126
9.2 Perspective of the Study 127 x
9.2.1 Recalibration of the Growth Model and Extention of its Applications .127
9.2.2 Development of Growth Models for Other Forest Types in Vietnam 128
9.2.3 Technical Development 128
Literatures . 130
Appendix 149



















xi
List of Figures

Fig. 1.1 Geographic position of the Central Highlands in Vietnam .03
Fig. 3.1 Geographic position of the YokDon National Park in the Dak Lak
province 27
Fig. 3.2 Hydrography system in the YokDon National Park 29
Fig. 3.3 Average air temperature in the period 2001-2006 in the study area 30
Fig. 3.4 Average atmosphere humidity in the period 2001-2006 in the study
Area 30
Fig. 3.5 Average rainfall in the period 2001-2006 in the study area .31
Fig. 3.6 Forest state map of YokDon National Park .35
Fig. 4.1 Association type 1: Dipterocarpus tuberculatus as dominating species 39
Fig. 4.2 Association type 2: Dipterocarpus tuberculatus forest with
Shorea obtusa .40
Fig. 4.3 Diameter-height curves of plot A1 and plot A4 for three species groups .46
Fig. 4.4 Diameter-height curves for three species groups of the twelve group A
plots 47
Fig. 4.5 Average tree number by diameter class distribution per hectare of
the two plot groups 49
Fig. 4.6 Average number of trees per hectare for three species groups over
twelve plots of group plot A .50
Fig. 4.7 Relationships between important stand variables .58
Fig. 5.1 System Dynamic Diagram notation .61
Fig. 5.2 Stock- and Flow-structure of a diameter class .63
Fig. 5.3 Principle of tree transition from one class to the successive higher class .64
Fig. 5.4 Diagram of recruitment to the smallest diameter class of 6cm 65
Fig. 5.5 Structure of the mortality model for each diameter class .66
Fig. 5.6 Structure of harvesting method of diameter limit cut and proportion xii
harvesting rule 67
Fig. 5.7 Structure of the harvesting model for q-factor guide 68
Fig. 5.8 Complete SD Diagram of the simulation model DIPGROW 71
Fig. 5.9 Partial effect of variables on diameter increment .80
Fig. 5.10 Plots of residuals against the fitted values of individual-tree diameter
increment model for three species groups 81
Fig. 5.11 Partial effect of variables on mortality rate of three species groups 84
Fig. 5.12 Partial effect of variables on recruitment .86
Fig. 5.13 Annual predicted vs. observed recruitment of 12 plots for the three
species groups .87
Fig. 6.1 Observed vs. predicted values after a simulation period of five years for all
group A plots 94
Fig. 6.2 Simulated basal area evolutions over one thousand years in total and
seprated species roup of an undisturbed stand on three site qualities .96
Fig. 6.3 Predited long-term diameter distribution evolutions of an undisburbed
forest stand for three site qualities 98
Fig. 7.1 Results of a scenario simulation of a stand .104
Fig. 7.2 Simulation results of mean annual volume increment obtained by the
stands with different basal areas and q-values .107
Fig. 7.3 The target diameter distributions of three species groups for three site
qualities 109
Fig. 7.4 Simulation results of the growth model for plot A8 following the harvesting
method of q-factor guide .111
Fig. 7.5 Simulation results of the growth model for plot A8 following the harvesting
method of q-factor guide with slight modification .112
Fig. 7.6 Diameter distribution of example plots .113
Fig. 7.7. Simulated effects of wildfires with different frequencies and intensities on
long-term yields 115
Fig. 7.8 Diameter distribution of the stands with different wildfire frequencies at the
time of 200 years 116 xiii
List of tables


Table 3.1 Areas of different forest types in the YokDon National Park .28
Table 4.1 Species association on the research plots .39
Table 4.2 Diversity of species composition for plot group A .41
Table 4.3 Diversity of species composition of group B 42
Table 4.4 Summary statistics for individual trees data on 12 plots .50
Table 4.5 Data for developing the recruitment function 51
Table 4.6 Summary of the data of mortality status in the plots used to develop
mortality functions .52
Table 4.7 Summary statistics of the mortality data used for the model
development .52
Table 4.8 Growth and yield characteristics of plot group A 54
Table 4.9 Growth and yield characteristics of plot group B 56
Table 4.10 Range of mean diameter and mean height in the stands of group A 57
Table 4.11 Range of mean diameter and mean height in the stands of group B 57
Table 5.1 The estimated parameters and fit statistics of individual tree diameter
increment models by species group .78
Table 5.2 The estimated parameters and fit statistics of mortality rate models by
species group .83
Table 5.3 The estimated parameters and fit statistics of recruitment models .85
Table 6.1 Predicted vs. observed basal areas for each plot of group A for the
three species groups .91
Table 6.2 Predicted vs. observed average number of trees per hectare for each
site quality by 5-cm diameter classes 93
Table 7.1 Mean annual volume increments produced by various initial stands 105
Table 7.2 The target diameter distribution for three site quality levels 108
Table 7.3 Return time of the group A plots .113
Table 7.4 Average annual long-term yields with different intensities and
frequencies of wildfire .116