thesis submitted in partial fulfillment of the requirements of the degree doctor rer nat

Table 5.2 Description of land cover and their appearance on SPOT-5 image 61 Table 5.4 Relevant factor classes indirectly affect forest status 74 Table 6.1 The regression between digital

CLASSIFICATION OF NATURAL BROAD-LEAVED EVERGREEN FORESTS BASED ON MULTI-DATA FOR FOREST INVENTORY IN THE CENTRAL

HIGHLANDS OF VIETNAM

Thesis submitted in partial fulfillment of the requirements of

the degree Doctor rer nat of the Faculty of Forest and Environmental Sciences,

Albert-Ludwigs-Universität Freiburg im Breisgau, Germany

by

Nguyen Thi Thanh Huong

Freiburg im Breisgau, Germany

2009

Name of Dean: Prof Dr Heinz Rennenberg

Name of Supervisor: Prof Dr Barbara Koch

Name of 2nd Reviewer: Prof Dr Dr h c Dieter R Pelz

Date of thesis defence: 2nd December 2009

The study was conducted at a site on the Central Highland of Vietnam where much natural forest remains The research used SPOT 5 satellite imagery captured during February

2006 Geomatic and topographic corrections were carried out using ground control points (GCPs) and Digital Terrain Model (DTM) which were obtained from Global Positioning System (GPS) and contour lines, respectively Three algorithms were tested for topographical normalization, which are cosine, minnaert and C-correction Satellite image corrected using C-correction algorithm was used as input data for further analysis Both unsupervised and supervised methods have been integrated in the classification process A set of sample plots and sample points were collected for the classification and validation during ground survey The study recognized four major distinguished natural forest classes and four non-forest land cover classes A total of one hundred and twenty one sample plots were collected from four forest classes for analysis Diameter breast height (DBH), tree height, crown diameter and distance to the nearest neighbor tree were measured by forest inventory tools, while the others such as canopy cover and vegetation cover were estimated visually The regression method was employed to simulate the relationship amongst forest variables, and between forest variables and spectral data Tree density, mean diameter and density of the trees having DBH ≥ 35cm were the important variables having a closed relation with spectral values Inverse J-shaped distribution Meyer N-DBH was selected as the basis to calculate mean characteristics of individual forest class Finally, an improved classification system of natural wooden forest was defined based on the result of image classification and mean value of forest conditions of quantitative criteria along with qualitative characteristics Four band SPOT5, three Principle Component (PCs), and a Normalized Difference Vegetation Index (NDVI) image, and the methods of regression, kNN and geostatistics were used to predict the forest stand volume The best result was obtained by applying regression kriging method on SPOT5 image In order to predict potential risk for the forest at the study area, the factors which related to accessibility in forest utilization were also analyzed These factors were divided into four different impact

iii

Zusammenfassung

Für die nachhaltige Bewirtschaftung von Wäldern sind forstliche Karten mit einer zweckmäßigen Waldeinteilung von großer praktischer Bedeutung Um diese wichtige Datengrundlage zu erstellen und fortzuführen müssen geeignete technische Lösungen gefunden werden, welche Kosten und zeitlichen Aufwand in vernünftigen Grenzen zu halten In dieser Arbeit wird nach einer optimierten Verknüpfung von Fernerkundungsdaten und terrestrisch erhobenen Daten für die flächige Klassifizierung von tropischen, naturnahen Wäldern gesucht Dabei wird auch der Frage nachgegangen, welche forstlichen Parameter eine enge Beziehung zu den spektralen Werten in den Satellitendaten haben Des Weiteren wurden Untersuchungen zur Schätzung der Holzvorräte durchgeführt

Das Untersuchungsgebiet liegt im zentralen Hochland von Vietnam Dort kommen noch verbreitet natürliche tropische Wälder vor Als Fernerkundungsmittel wurde eine SPOT 5 Szene vom 5 Februar 2006 verwendet Für die geometrischen und topographischen Korrekturen des Satellitenbildes wurden Passpunkte (Ground Control Points) aus GPS Messungen sowie ein Geländemodell aus digitalisierten Höhenschichtlinien eingesetzt Für die topographische Normalisierung wurden drei verschiedene Algorithmen getestet und zwar Cosine, Minnaert und C-correction Die besten Ergebnisse lieferte die C-correction, weshalb diese Ergebnisse in die weiteren Verarbeitungsschritte einflossen Die vorbearbeiteten Satellitendaten wurden sowohl mit unüberwachter Klassifizierung, als auch mit überwachten Klassifizierungsverfahren ausgewertet Für letzteres und auch für die Genauigkeitsüberprüfung der Ergebnisse wurden eine Reihe von Trainingsgebieten in dem Untersuchungsgebiet festgelegt Insgesamt 120 Plots wurden im Gelände angelegt, verteilt auf die vier unterschiedenen, natürlichen Waldklassen sowie für vier weiteren Landbedeckungsklassen In den Plots wurden forstliche Parameter wie BHD, Baumhöhe, Kronendurchmesser, Position der Bäume und deren relative Lage zu den Nachbarbäumen gemessen, Kronenschluß und Bedeckungsgrad wurden geschätzt Zur Berechnung der Zusammenhänge zwischen den erhobenen forstlichen Variablen selbst sowie den Variablen und den spektralen Fernerkundungsdaten wurden Regressionsanalysen angewandt Dabei wurden enge Zusammenhänge zwischen spektralen Eigenschaften mit der Dichte des Kronendaches, dem mittleren Durchmesser, sowie der Dichte aller Bäume über 35 cm Durchmesser gefunden Für die Berechnung von charakteristischen spektralen Eigenschaften der einzelnen Waldklassen wurde J-shaped Verteilung Meyer N-BHD zu Grunde gelegt Basierend auf den Ergebnissen der Satellitenbildklassifikation und den berechneten Kennziffern wird in einem letzten Schritt ein neues, verbessertes Klassifikationsschema für natürliche tropische Wälder in vorgeschlagen Für die Schätzung des Holzvorrates wurden die vier Kanäle der Spot 5 Szene, drei Principle Component Berechnungen sowie der Normalized Difference Vegetation Index (NDVI) herangezogen

iv

Dabei kamen verschiedene Methoden aus der Geostatistik, Regressionsanalyse und kNN vergleichend zur Anwendung Die besten Ergebnisse erzielte der geostatistische Ansatz des „Regression Krigings“ mit den SPOT 5 daten Als zusätzliche Information wurde das potentielle Risiko für ungeregelte Waldnutzungen in Abhängigkeit von der Zugänglichkeit der Waldflächen modelliert Vier abgestufte „Impact“ Klassen wurden für die zwei betroffenen Waldformationen mit GIS Techniken ausgeschieden Die daraus resultierenden thematischen Karten zeigen die gefährdeten Flächen und dienen für die Planung entsprechender Maßnahmen bei der Waldbewirtschaftung

Sowohl für Beobachtung der Waldentwicklung als auch für die nachhaltige Bewirtschaftung der Wälder Vietnams werden durch die Kombination von Fernerkundungsdaten und Feldaufnahmen sehr brauchbare Forstkarten zur Verfügung gestellt In Verbindung mit der Abschätzung potentieller Risiken sind diese ein Planungsinstrument von großer praktischer Bedeutung für die Forstwirtschaft in Vietnam

v

Acknowledgements

First of all, I would like to express my deep appreciation to Professor Dr Barbara Koch and Dr Claus-Peter Gross of the Depatement of the Remote Sensing Lanscape Information Systems (Felis) of the Albert Luwig University Freiburg for their kind supervision, guidance and support I also thank Prof Dr Dr h c Dieter R Pelz of Department of Forest Biometry for his valuable comments and for taking up the role of co-referent of this study

The warmest thanks go to Assoc Prof Dr Bao Huy, who has been giving me valuable suggestions and kind assistance during my research I wish to thank Mr Sandeep Gupta,

Mr Johannes Heinzel and Dr Fillip Langa for their technical support

My special gratitude goes to forest staff in Quang Tan forest enterprise, farmers, lecturers and forestry students from Tay Nguyen University for their kind assistance during field work Without their support it would not have been possible to conduct such an extensive forest inventory

I would like to give my thanks to the Government of Vietnam for granting me the scholarship and research subsidy support by the DAAD (German Academic Exchange Service) I am also greatly indebted to the OASIS for providing remote sensing images

I would like to express my sincere appreciation to all of the friends and colleagues of Felis, Freiburg University, Germany and Department of Forest Resource and Environment Management, Tay Nguyen University, Vietnam for their kind assistance during my research

Last but not least, my deepest thanks go to my family for their patience, support and encouragement throughout my many years of higher education My parents have been an endless source of love and understanding throughout my life My husband, daughter, brothers and sisters have always given me all their infinite love and best wishes To all I

am grateful for their roles in my life

Nguyễn Thị Thanh Hương

vi

Abstract i 

Zusammenfassung iii 

Acknowledgements v 

List of Tables xi 

Illustrations xiii 

List of Abbreviation xv 

1  INTRODUCTION……….1 

1.1  Background 1 

1.2  Problem analysis 3 

1.3  Objectives 7 

1.4  Hypothesis 8 

1.5  Outline 8 

2  FOREST CLASSIFICATION SYSTEM……….10 

2.1  The forest classification systems 10 

2.2  The classification systems of forest in Vietnam 13 

2.2.1   Several forest classification systems before 1975 13  

2.2.2   The developed classifications in the post-war period 14  

3  MULTI-DATA SOURCES IN FOREST INVENTORY……….19 

3.1  Remote sensing techniques 19 

3.1.1   Pre-processing remote sensing images 19  

3.1.2   Image processing 22  

3.1.3   Normalized Difference Vegetation Index (NDVI) 27  

vii

3.1.4   Principle component analysis (PCA) 28  

3.2  Forest parameter considerations 29 

3.2.1   Definiton of some forest characteristics 29  

3.2.2   Regression analysis considered as the major double sampling among forest parameters 30  

3.3  Combining of remote sensing and terrestrial data 31 

3.3.1   Multiphase sampling 31  

3.3.2   Determining the sample size 33  

3.3.3   Estimation of forest variable using remote sensing 33  

3.4  GIS technique 36 

3.5  Relation between relevant factors and forest status 37 

3.6  Literature review 37 

3.6.1   Classification of forest using remote sensing images 37  

3.6.2   Prediction of forest parameters using remotely sensed data 38  

4  DESCRIPTION OF RESEARCH AREA……….43 

4.1  Location 43 

4.2  Topography 44 

4.3  Geology and soil 44 

4.4  Climate 44 

4.5  Description of forest types 45 

4.6  Forest jurisdiction 46 

5  METHODOLOGY……… 48 

5.1  Data, Software and equipment 49 

5.1.1   Data availability 49  

5.1.2   Software 51  

viii

5.1.3   Equipment and tools 51  

5.2  Development of a set of field data 52 

5.2.1   Pre-field preparation 52  

5.2.2   Field measurement 53  

5.2.3   Statistical approach to field data survey 54  

5.3  Pre-processing remote sensing data 56 

5.3.1   Ortho-rectification 56  

5.3.2   Topographical correction 57  

5.4  Multispectral classification 60 

5.4.1   Unsupervised classification 60  

5.4.2   Supervised classification 60  

5.4.3   Accuracy assessment 63  

5.5  Statistical description of forest classes 65 

5.5.1   Transformation of data 65  

5.5.2   Regression analysis technique 66  

5.6  Estimation of stand volume by different methods 67 

5.6.1   Using empirical regression 67  

5.6.2   The k-NN algorithm 68  

5.6.3   Geostatistics with regression kriging 69  

5.6.4   Model evaluation 72  

5.7  Assessing risk potential of forest status by indirect factors for forest management ……… 73 

5.7.1   Basis of assessment 73  

5.7.2   An analysis based on database and forest map 74  

6  RESULTS……….77 

ix

6.1  Topographic correction 77 

6.2  Statistical approach with stand forest variables 78 

6.2.1   Height – Diameter Relationship (H - DBH) 78  

6.2.2   Volume equation for forest stands 80  

6.3  Proposal for a new forest classification system based on both quantitative and qualitative criteria of the stand 80 

6.3.1   Average characteristics of each forest status 80  

6.3.2   Estimates of forest stand parameters 81  

6.3.3   Qualitative description as a definition of the forest class 85  

6.4  Forest/land cover classification 86 

6.4.1   Forest/Land cover map 86  

6.4.2   Natural broad-leaved evergreen forest 88  

6.4.3   Plantation forest 90  

6.4.4   Bamboo 90  

6.4.5   Shrub/Grass land 90  

6.4.6   Other lands 91  

6.5  Accuracy assessment 91 

6.6  Relation of directly surveyed data with spectral images 92 

6.6.1   Correlation Analysis 92  

6.6.2   Relationship model 96  

6.7  Estimation of stand volume from spectral image and field data survey 96 

6.7.1   Regression method 96  

6.7.2   KNN method 101  

6.7.3   Geostatistical method with regression kriging 102  

6.7.4   Accuracy assessment of the models 105  

x

6.8  Involvement of relevant factors in forest management 106 

6.8.1   Risk potential of poor forest status from disturbance 106  

6.8.2   Potential risk from clearing forest for agricultural land 109  

7  DISCUSSION………113 

7.1  Improvement of the existing classification system based on satellite images 113 

7.2  Relation of stand volume and image data 116 

7.3  Relevant factors in the categorization of forest for better management and planning 121 

8  CONCLUSION……… 122 

8.1  Important research findings 122 

8.2  Limitations of the study 123 

8.3  Recommendations and future outlook 124 

References 127 

Table 5.2 Description of land cover and their appearance on SPOT-5 image 61 

Table 5.4 Relevant factor classes indirectly affect forest status 74  Table 6.1 The regression between digital value with incidence angle 77 

Table 6.3 The mean forest characteristics of forest status 82 

Table 6.6 The land cover area in ha and percentage 87  Table 6.7 Confusion matrix of the Maximum Likelihood classification 92  Table 6.8 Pearson’s correlation pairs of variables 93  Table 6.9 Pearson correlation matrix for forest variables and classified forest classes 95  Table 6.10 Pearson correlation matrix for the variables analyzed for stand volume

Table 6.11 Results of simple regression modeling for stand volume 98  Table 6.12 Results of simple regression modeling for logarithmic stand volume 98  Table 6.13 Error of the volume estimates using three different methods 105  Table 6.14 Risk levels based on relevant factors from poor forest 109 

xii Table 6.15 Risk levels based on relevant factors from agricultural land 112  Table 7.1 Correlation of spectral data and stand volume characteristics 117 

xiii

Illustrations

Figure 2.1 Vietnamese classification of evergreen natural forest 15 

Figure 5.2 Equipment for stem diameter and tree height measurement 52 

Figure 5.4 A schematic example of regression-kriging: fitting a vertical cross-section with assumed distribution of an environmental variable in horizontal space 72 

Figure 6.2 Scattergram of the relationship between tree height and breast height diameter

79  Figure 6.3 Inverse J-shaped distribution Meyer N-DBH 84 

Figure 6.6 Volume map using SPOT 5 image and regression estimator 100  Figure 6.7 Volume map using SPOT 5 image and kNN method 101  Figure 6.8 Histogram of original and log transformed volume 102  Figure 6.9 Experimental variogram with fitted model, log (volume) 103  Figure 6.10 Volume map using SPOT 5 image and regression-kriging method 104  Figure 6.11 Accuracy assessment of standing volume estimations 106 

Figure 8.1 Applicability of the study in practice 126 

CPU Central processing Unit

DBH Diameter at Breast Height

DEM Digital Elevation Model

Dis to NN tree Distance to nearest neighbor tree

FAO Food Agricultural Organization

GoVN Government of Vietnam

GPS Global Position System

HRS High Resolution Stereoscopic

ISODATA Iterative Self – Organization Data Analysis

MARD Ministry of Agricultural Rural Development

MLC Maximum likelihood classifier

MSS Multi Spectral Scanner

NDBH35 Density of tree having DBH equal to 35cm and over

NDVI Normalized Difference Vegetation Index

PCA Principal Component Analysis

xvi

RMSE Root Mean Squared Error

SPOT Satellite Pour l’Observation de la Terre

SPOT XS Satellite Pour l’Observation de la Terre

UNESCO United Nations Educational, Scientific and Cultural Organization USGS United State Geological Survey

WGS 84 World Geodetic System 1984

1

Chapter I

1 INTRODUCTION 1.1 Background

Forest mapping is one of the most important factors for organization of forest resource inventory and monitoring With forest mapping, a classification of the forest in homogenious strata is carried out as a basis for sustainable forest management plans

The main purpose of the classification of forest type is to support forest management planning The classification should be based on needs of potential resources, forest characteristics with up-to-date information using the minimum time necessary and at a reasonable cost Therefore, the definition of a forest ecosystem and the relevant characteristics vary with the resource managed and the issue under consideration (Wulder and Franklin, 2007) The role of classification is to provide a set of criteria that bring a certain degree of order to ecological community patterns A classification system has been developed to address a wide variety of spatial scale and purpose, therefore, a suitable classification system is not only important for management planning but also for forestry development strategy The system can be directly interpreted from imagery and can be associated with the ground information as well as the ancillary information with an acceptable accuracy According to Franklin (2001), the use of remote sensing in classification is based on the fact that the differences on the ground between vegetation types can be isolated or separated as differences in the image characteristics When different vegetation structures define the classes, and the latter correspond with recognizable vegetation types on the ground, there is a good reason to believe that the types can then be mapped with digital remote sensing data and methods (Merchant, 1981) The goal of remote sensing in forestry is the provision, based on available or purposely acquired remote sensing data, of information that foresters need to accomplish the various activities that comprise sustainable forest management The objective of classification is to generate spatially explicit generalizations that show individual classes selected to represent different scales of forest organization (Franklin, 2001) Of all the available ways of extracting information from remote sensing data, image classification in particular can be considered as a prime candidate for a standard methodology with the potential to be distilled into a protocol that can be extended spatially and temporally (Lillesand, 1996)

2

In forestry, as in many other disciplines involving land management, there exists a definite need for timely, reliable information on which to base resource management decisions (Fleming and Hoffer, 1979) Hence, effective policy decisions aimed at protection and sustainable use of forest lands need to be based on reliable, accurate and relevant

information (Stibig et al., 2003)

A suitable classification system will aid in the large management context while assessments of forest structure and status are of fundamental importance to forest management The purpose of traditional forest inventory is to provide unbiased and reliable forest resource information Typically these inventories lack fine spatial resolution and even performed teresstrially only In recent decades, however, remote sensing, the Global Positioning System (GPS), and Geographic Information Systems (GIS) have provided new opportunities for such forest inventories Integrating these techniques and tererrestrial data as an analysis of multi-source information is therefore of particular value According to Franklin (2001), remote sensing and geographical information systems (GISs) are more than tools to satisfy increasing information needs of resource managers

They represent essentially new approaches to forest disturbance and spatial pattern

mapping and analysis because they enable new ways of viewing disturbances and landscapes, which in turn influence our understanding and management practices Consideration of the fragmented status of forests is also necessary There are several reasons causing fragmentation, but largely forest fragmentation is caused by human activities Disturbances can be defined as relatively discrete events that disrupt ecosystem, community or population structure, and change the availability of resources or the physical environment (Pickett and White, 1985) Analysis of the disturbance regime of a forest can

be of great value for understanding patterns of structure and composition, as well as being important for defining appropriate management interventions Characterizing the disturbance regime typically involves assessing the severity, timing, and spatial distribution of the different types of disturbance affecting the forest (Newton, 2007) Consequently, an understanding of the natural and societal dynamics also has great implications for forest management Knowing the factors which are related to forest accessibility e.g topography, accessibility to forest area and so on, and how they impact natural resources in indirect manner may contribute to better forest management planning Although the great progress in remote sensing and GIS in recent decades has provided considerable opportunities for improving the effectiveness of forest management, an application of remote sensing in forestry in Vietnam is limited due to several reasons,

3

namely in technique and resources as well Meanwhile, characteristics of natural rainforests are specific and complex, and always be put under change by forest land utilization and social pressures Correspondingly, the traditional method in forest inventory cannot satisfy increasing information requirements in a sustainable forest management system For this reason, a multi-source approach of combining remote sensing, GIS and field data may improve effectiveness of forest management planning at the location

1.2 Problem analysis

The forestry issues in Continental Southeast Asia are probably the most complex in the world (Stibig, 2003) As in many Southeast Asian countries, Vietnam has experienced the highest rates of net forest cover decline Estimates of the change in forest cover in Vietnam

in the last half century vary greatly in measurement The forests have dramatically decreased overall during the last 60 years, despite having slightly increased recently The forest cover declined from 43% of the country’s area to 30% in 1985, and 28% in 1995 (Lung, 2001; Maurand, 1943) Especially in the war time, large areas of mangrove and rainforests of Vietnam were eradicated by toxic chemicals (Nakamura, 2007) The summary tally is that 104,909 ha of mangrove forests and about 3,000,000 ha of inland forests were eliminated The total defoliated area amounted to 3,104,000 ha Some 82,830,000 cubic meters of wood were lost Inland forests were converted into the fields of tall grass with a strong, thick reticulate (Nakamura, 1995) The forests have continued to decrease in area and degrade in quality during recent years The forests have been destroyed for many reasons, such as overexploitation, shifting cultivation, and agricultural extension (Lung, 2001) However, Lang (2001) expresses that the key causal forces come from the role of government and state forest enterprise such as resettlement programs, migration and logging by state forest enterprises Nevertheless, illegal logging is a serious problem as it has annually destroyed around 30,000 ha of forests in recent decades (GoVN, 1994) In oder to raise forest cover, the 5 Million ha Reforestation Program was initiated in

1998 with a target to plant 5 million ha of forests by 2010, restoring the forest cover to 43% (MARD, 1998) The program aims not only to reforest, but also to protect existing natural forests As a result, the forest cover of Vietnam has gradually increased In 2003, the forested area of Vietnam was 12,094,518 ha, of which 10,004,709 ha were natural forests and 2,089,809 ha were plantation forests, resulting in a forest cover of 36.1% (Hung, 2004) However, the quality of forests is still low since most of the forests are poor

in timber volume and lack valuable species as a result of a long time of overexploitation

and random logging (Dang et al., 2001; Nguyen, 1997) It is reported that natural forests

of stand structure are also given for natural forests Lack of modern techniques and multi data sources and poor logistical support in forest resource assessment in Vietnam are the main causes that lead to deceptive forest data creation Consequently, there is a need for reliable sources of data for natural forest resources

Effective policy decisions aimed at protection and sustainable use of forest lands are indispensable, therefore, accurate and relevant information to quantify and assess their

quality in terms of biophysical variables (i.e diameter at breast height, stand height,

species composition, basal area, volume, tree density and age) are important Knowing just

what forests are left, and where these are is a good starting point (Stibig et al., 2003) Thus,

a suitable classification of forests and mapping of forest cover to provide a comprehensive and reliable view of the region’s forest cover is important to help forestry managers make appropriate decisions in forest management and planning

Several ecological classification systems have been developed for Vietnamese forests during the past century in North and South Vietnam separately In forest planning and management, a common system, which came into force from the Vietnamese Government, has been used since 1975 (see chapter II) This system was established for the whole country without a region specific focus The classification criteria of this system was developed from basis of forest condition in the North region, then applied to the whole country As a consequence, this leads to not only difficulties in inventory but also in forest management strategies

Vietnam basically has a tropical climate as it lies inside the Northern Tropical Zone However, due to a varying monsoon season and complex terrain, the climate of Vietnam differs with latitude and altitude (Lap, 1999; Thai, 1998) In other words, the climate differs from the South to the North Meanwhile, climate – hydrology is the decisive factor group to physiognomy and structure of vegetation type Forest types vary in general as a function of environmental and climatic factors including temperature, rainfall, humidity, seasonality - often governed by latitude and topography (Thai, 1978) As a result, characteristics of vegetation vary not only in physiognomy but in structures from the North

5

to the South Thus, to apply the classification system which was based on the forest condition of the Northern Vietnam to the whole country is unsuitable, especially for the forests distributed in the Central Highlands, because they have far different ecological conditions from the North In practice, this system has caused some issues while used in inventory in the Central Highlands, namely:

- When the status is used for forest classification the factor of basal area or volume is not homogeneous

- In general, the forests in the Central Highlands are distributed on more propitious land than that of the north Hence, applying specific criteria of DBH, crown closure, and species of this system as indicative parameters in classification for Central Highlands forests is seemingly inappropriate

- The classification criteria are mainly descriptive, making it difficult to specify and distinguish in practice

As presented alrealdy, one of the vital causes of ineffective management of the forests is lack of spatial information resources Thus, the combination of remote sensing data in forest inventories has not been done, the forest data have not been frequently updated and the forest maps are coarse As a result, out-of-date forest information is the basis of forest planning management Consequently, erroneous decision-making is unavoidable

Natural forests in the Central Highlands account for 50% of its land, in which areas of natural broadleaved evergreen forests are dominant However, a dynamic view of forests in the Central Highland of Vietnam is necessary because primary forests disappeared almost completely and were replaced by a mosaic of forest formations in varying degrees of several kinds of degradations in the past Daknong is one of the similar provinces that has the most abundant forest resource in Vietnam But, as in other provinces in the Central Highlands, the forests in Daknong appear as a mosaic of different communities and characteristics, distinct from others with respect to their structural parameters However, with the inflexible system for the aforementioned reasons, the forest management in the Central Highlands in general, but in Daknong in particular, has encountered more difficulty

in forest resource assessment than other provinces as well as in forest management planning Therefore, adjustment of the countrywide classification system to the conditions and requirements of specific area e.g the Central Highlands, is necessary in forest management and forest development strategy

In addition, although the forests of tropical countries are very diverse in species and complex in structures, they are being lost at an alarming rate The most crucial underlying

6

cause of deforestation is due to highest population (people) growth rates in the world (Laurance, 1999; Wright and Muller-Landau, 2006) Therefore, a successful forest strategy requires not only a good understanding of the natural but also social conditions and their relations to the strategy’s objectives For this reason, assessment of accessibility to forest resources is also one of the considerations under study

The main problems as mentioned above are summarised in a diagram of cause and effect relations, illustrated in Figure 1.1 The unsuitable forest classification system along with drawback techniques in forest inventory are major causes leading to ineffective forest management planning

Figure 1.1 Problem Analysis Tree

Difference of natural forests of the Central Highlands from other parts

in the Country

Mechanism for application and updating

Lack of human

resource Lack of finance

Limitation of education/

training in RS and GIS

Rapid change in forest status

in overtime

Uncontinuous update, unconfident maps and

database

Linkage between RS + GIS and Inventory

Time consuming, expensive

Be difficult classify forest status in the field

Invaluable forest management

8

i Classify forest status based on satellite images and terrestrial data

ii Suggest an improved forest classification system for forest management

iii Discover which field inventory data is in relation with parameter of SPOT 5 data and to estimate stand volume in relationship with spectral data

iv Combine database and GIS modelling to evaluate the relative roles of relevant factors in driving force forested area modification

1.4 Hypothesis

- Remote sensing is indispensable for an up-to-date forest map Digital classification of SPOT images can improve existing forest maps which are too coarse for effective planning management Therefore it can meet the requirements for sustainable forest management planning with reasonable costs and within a rational time period

- Combination of spectral and forest field data can demonstrate the relationship between them The variance of forest characteristics can influence the spectral parameters of the satellite image

- The forest at the research site is of the same origin, but the different status of existing forest is due to human activities Factors such as rivers and topography can contribute in this change

1.5 Outline

Chapter 2 provides an overview of the general basis of tropical forest classification Some classication systems of Vietnamese forests are also mentioned in this chapter The multi-data sources in forest inventory are revealed in Chapter 3 This chapter contains an overview of processing techniques of satellite image, field data analysis, the combination

of remote sensing data and terrestrial data, and the GIS technique Additionally, a literature review of forest classification and prediction of forest variables using spectral value associated with field data are presented in the last part of the chapter

In Chapter 4 there is an introduction to the research area, general information about the location, topography, geology and soil, climate, and a description of the forest situation is given Methodology is introduced by Chapter 5 This chapter specifies data, software, and equipment used The methods used to process remote sensing, to develop a set of field data, to estimate forest variables using satellite data, and to produce risk potential maps are also provided by Chapter 5 Chapter 6 presents results from the combination of field

9

data and satellite image data in forest inventory and the creation of potential risk maps The first section aims to demonstrate the use of SPOT 5 data and terrestrial data for forest classification The relationship between spectral data, forest variables and volume estimations are given thereafter Risk potential maps follow in the latter section of this chapter Chapter 7 discusses the results obtained and Chapter 8 gives conclusions of the research and contains recommendations for further research

Jensen (1996) defined classification as an abstract representation of the situation in the field using well-defined diagnostic criteria A classification describes the systematic framework with the names of the class and the criteria used to distinguish among them Classification of tropical forest can be important in determining management plans Attempts to characterize and differentiate tropical vegetation ecosystems have generally involved grouping observations into classes, the members of which share common characteristics This grouping may be followed by attempts to relate classes to underlying environmental conditions However, because of the great variety of physiognomic and environmental characteristics used to classify vegetation for various purposes, no classification has yet been proven to be the most useful in all circumstances (USGS) During the last century, many classification systems of forest have been described and developed by authors such as Shimper (1903) Champion (1936), Burt-Davy (1938), Bear (1944, 1949), Richard (1952)

Classification systems of forest are based on forest structure and/or on function The classification of forests into different communities or types provides an important basis for forest management and conservation planning According to Jordan (1993), successful forest management depends more on forest function than on structure Management for forests based upon structure is satisfactory when structure reflects function, as in the case

of moisture which is a critical factor in classification, for example However, structure does not always reflect function, especially in the case of nutrient status While studying ecosystem characteristics along a nutrient gradient on different types of forest, Jordan (1985) claimed that nutrient-conserving mechanisms can compensate for low nutrient availability, with the result that the biomass of a forest on nutrient poor soils differs only

11

slightly from that of a forest on rich soils, despite the fact that forests on the two sites differ greatly According to Montagnini and Jordan (2005), for lack of understanding of how tropical forests function, functional characteristics are rarely used as a basis for classification of tropical forests More commonly, they are classified by the climate in which they occur, by species, by successional state, or by the soil type on which they occur

Climatic classifications are based on the fact that climate is one of the major factors

influencing vegetation distribution Climate types can be defined in terms of temperature and precipitation, which are associated with particular types of vegetation, for example, tropical rainy and warm temperate climates are associated with tropical rain forest and temperate rain forest, respectively (Newton, 2007) Champion (1936) and Holdridge (1967) proposed a classification based solely on climate They considered temperature and rainfall to prevail over other environmental factors in determining vegetation Beard (1949) classified forests on mountains of small tropical islands according to rainfall and elevation Elevation of each forest type varies, depending on the size of island and the direction of wind relative to the mountains

Although some aspects of forest function are reflected by climatic classification, others such as tightness of nutrient cycling are not, therefore, climatic classifications alone are often inadequate as a system on which to base forest management plans A useful classification for the plants of lowland evergreen rainforests was provided by Richards (1952) The characterization of types is based mainly on the occurrence of economically important species The community type found at any given location depends on not only local climate, but also on the physical and chemical properties of the soils, topography and elevation, and previous site history Classification based upon community or dominant species is useful in designing management strategies because managers are usually interested primarily in species, since markets demand trees of given species Community classification is also useful because such classification reflects function (Mongtanini and Jordan, 2005)

The functional classification of forests can be based on the successional stages or on the nutrient status (Jordan, 1993) The successional stage is a common forest classification system that is based on function Woody vegetation communities that occur on larger areas

of abandoned agricultural land, pasture or forest gaps are usually called “secondary successional forests”, “secondary forests” or “pioneer forests” They have several

a silvicultural system, such as enrichment planting, improvement felling or selective harvesting is often based on the forest successional stage (Camirand and Everlyn, 2003) Forest classification based upon soil nutrient status is considered by Jordan (1985) To characterize tropical forest according to their degree of adaptation to low soil nutrients, Jordan (1985) classified forest along a continuum from those adapted to extremely low nutrient supplies This classification also suggests impactions for management Species diversity, productivity, decomposition process and nutrient cycle depend mainly on soil nutrients These help to make suitable decisions for forest management

In 1978, a committee from UNESCO attempted to standardize the classification of vegetation worldwide The UNESCO system includes many categories for tropical and subtropical forest formations The UNESCO system establishes physiognomically and environmentally separated vegetation types in a multi-tiered classification hierarchy A major problem with the UNESCO system is that it must be used in conjunction with accurate maps of forest communities Due to the time and expense of preparing them over large areas such maps are often unavailable (Montaginii and Jordan, 2005)

A classification, however, must deal with grouping of the same things, and also with their relationship and development The primary categories must be as broad as is consistent with the actual facts, and be capable of division and subdivision as necessity demands; that

is, the divisions must proceed from the general to the particular The classes in the scheme must be defined by their ground characteristics for the classification products to be useful Regardless of the types of classes selected, experience has shown that the classification scheme must be developed with full involvement of at least two collaborative interests: (1) those generating the classification products and (2) those using the classification products (Franklin, 2001) Furthermore, the classification must provide the basis for an information-storage and -retrieval system that is convenient for use The classification is an organized index to information that is stored elsewhere Finally, a good classification is an effective

a particular forest type Selection of forest types used in a map will depend on the map’s purpose, and whether the classes can be accurately and efficiently delimited When a classification system has been selected, it is important to document the details of each class and apply these definitions in an objective and consistent manner (Franklin, 2001)

Montagnini and Jordan (2005) concluded that there is no single classifcation scheme that is general enough to be of use to everyone yet particular enough to take local conditions into account in a way that is meaningful for management Correlation between species and forest function based on local knowledge are still the most useful bases for small-scale forest management Remote sensing techniques along with geographic information systems offer the possibility of defining forest communities on a scale useful for management over larger areas

2.2 The classification systems of forest in Vietnam

Several classification systems have been applied to Vietnamese forest during the past century The vegetation of Vietnam has been studied and classified from the beginning of the last century

2.2.1 Several forest classification systems before 1975

In 1918, Chavalier performed the first classification system for the forests in the North of Vietnam based on vegetation types By 1943, Maurand divided vegetation in Indochina into three vegetation zones including North Indochina vegetation, South Indochina vegetation and an intermediate one

In the south of Vietnam, a particular classification system of forest was proclaimed by Maurand in 1953

Using the classification system of UNESCO (1973), Vietnamese vegetation exists in four formations of which two are related to forest They are dense forest and open forest Each

of these two is divided into several sub-formations, and some different classes are divided from the sub-formations

14

2.2.2 The developed classifications in the post-war period

The above classification systems were performed separately in two parts of Vietnam Hence there had not been any common system for the whole country until after 1975 when Thai (1978) developed the classification of forests based on site condition for the whole country

2.2.2.1 The forest classification system developed by Thai Van Trung

Thai (1978, 1999) classified the vegetation of Vietnam based on the viewpoint of ecological factors of phytocoenosis of vegetation According to this view, in a specific ecological environment they can appear only in a given intact vegetation type, in which the various forest characteristics such as species composition, physiognomy, and structure are decisively affected by five ecological groups e.g physio-geographic, climate, edaphic, floristic and bio-anthropogenesis Accordingly, the forests of the entire country are divided into five main types with 14 subtypes as table below:

Table 2.1 Classification system of forest developed by Thai Van Trung (1978, 1999) Types of lowland closed rainforest:

1 Humid tropical evergreen closed rainforest

2 Humid tropical semi-deciduous closed forest

3 Deciduous tropical forest

4 Hard leaf, dry closed tropical forest

Types of open forest

5 Broadleaved open tropical forest

6 Needle-leaved evergreen slightly open dry tropical rainforest

7 Needle-leaved evergreen open slightly dry semitropical rainforest in lower montane

Steppe

8 Bushes, dry tropical tall-grass

9 Thorn scrubs

Closed montane rainforest

10 Humid evergreen semi-tropical rainforest in lower montane

11 Humid mixed broad-needle leaved lower montane evergreen rainforest

12 Closed needle-leaved temperate forest

Cold-dry montane formation

13 Montane dry formation

14 Montane cold formation

15

2.2.2.2 The forest classification according to forest status

In order to put the suitably silvicultural treatments into practice for different forest stands, the Department of Forest Planning Inventory applied hierarchical classification to categorize the forests Accordingly, the forest is divided into 4 main classes with several subclasses each These classes are distinguished by level of disturbance, potential regeneration, species, maximum diameter class and so forth The hierarchy of the system is presented in Figure 2.1 and their criteria are detailed in Table 2.2 This system is simple, easy to recognize in the field, hence it has been broadly used by forest rangers However, according to Thai (1978), the basical weakness of the system is although the three criteria used for classification are: tree composition, ecological characteristics and structure, in reality the forests are categorized based mainly on forest status with different disturbances Therefore it is not possible to distinguish among the pristine forest and secondary forest and their complex successional periods

Figure 2.1 Vietnamese classification of evergreen natural forest (Source: MARD)

Table 2.2 Class definition for Vietnamese Classification of Evergreen Natural Wooden

Forest (Source: MARD)

Group I: Non-forest Only grasses, bushes with very few trees, scattered bamboos; coverage index is under 0.3 (full coverage is 1.0)

This group has 3 sub-groups:

• IA: Characterized by grasses, bushes or wild bananas

• IB: Characterized by bushes, scattered wooden trees and bamboos

IIIA

I

Non - Forest

II Regeneration Forest

III Degraded Forest

IV Primary Forest

IIIA2 IIIA3 IIIA1

16

• IC: Characterized by high density regenerating wooden trees Trees of over 1 meter height and with more than 1000 stems per hectare

Group II: Regenerating forest with pioneer species that have a smaller diameter

Based on status and origin, there are 2 sub-groups:

• IIA: Regenerating forest after shifting cultivation, characterized by pioneer species that are fast growing and prefer light Trees are of similar age and with only 1 layer

• IIB: Regenerating forest following heavy exploitation for timber Young

community with species preferring light; diverse species composition; trees of

different ages; unclear dominance There may be some big trees remaining, but the numbers are not relevant with bad quality Forest is only classified into this group if the community with the most common diameter equal to or less than 20cm

Group III: Degraded forest Communities have been exploited, and changed the

stand structure of the forest Depending on exploitation levels and potential stand volume, two sub-groups are recognized:

IIIA: Communities with heavy exploitation; present potential for exploitation of timber

is limited The structure of the forest is significantly changed There are 3 sub-groups:

- IIIA1 (Poor Forest): Most heavily exploited forest The upper storey may have some large trees, but generally the forest is of low quality with numerous vines, bushes, and bamboos, and has a defragmented canopy

- IIIA2 (Medium Forest): Heavily exploited, but significant time for regeneration Characterized by middle storey that becomes dominant with the majority of trees of 20-30cm diameter The forest has at least 2 stories; the upper storey coverage is not continuous, and mostly previously established by the trees from the lower storey; there are a few large trees

- IIIA3 (Rich Forest): After selecting cutting with low intensity or forest developing from IIIA2 The communities have a relatively closed coverage, with at least 2 stories The main difference from type IIIA2 is that the number

of trees are higher with some trees with having diameter of more than 35cm IIIB: Characterized by communities that have low levels of selective logging, with few valuable wooden species exploited The stable structure of the forest has not changed;

biomass is high with a high percentage of large trees

Group IV: Primary forest, stable forest Pristine forest or matured secondary forest that has not yet been exploited The forest has a stable structure, multi-storey, diverse diameter sizes, but sometimes lacks a lower storey

There are 2 sub-groups:

17

- IVA: Primary forest

- IVB: Regenerating secondary forest

In addition, several classification systems of forest have been used for different purposes such as:

• Classification based on forest functions (forestland utilization purposes)

This classification system was specifically regulated upon Decision 08/2001/QD-TTg dated 11 January 2001 by the Prime Minister Under this decision, natural forest has been classified in three categories for the following utilization purposes:

- Production forest: this kind of forest is used for commercial purposes It supplies timber and non-timber forest products (NTFP) In addition, it provides environmental protection

- Protection forest: the forests used to protect water resources, soil, and to prevent from land erosion, desertifcation, or to regulate climate and so on

- Special use forest: the forests used mainly for natural conservation, genes, historical/cultural monument protection, scientific research, and ecologically environmental protection

This classification is employed for long term forestry strategy in the entire country

• Classification based on stand volume: the forests are distinguished by their volume Accordingly, volume categories are

- Little affected natural forest with three-storey tree structure, unclear vestige of ravaged forest, and timber stock of 300 - 400 m3/ha;

- Medium-level affected natural forest with three-storied tree structure and timber stock of 200 – 300 m3/ha;

- Exhausted forest with timber stock of 120 - 200 m3/ha;

- Exhausted forest with timber stock of 80 - 120 m3/ha; and

- Exhausted forest with timber stock of 50 - 80 m3/ha

However, such classifications only provide generally indicative information and criteria description in the forest without using remotely sensed data for interpretation during forest inventory Whereas, forest inventories, particularly in the rainforest, are both time consuming and expensive if they are conducted using the pure terrestrial approach only Monitoring and management of tropical rainforest is of paramount importance to many countries in the tropical regions because it is essential for economic reasons to establish a

18

sustainable process of exploitation of the timber resources in the forests There is a pressing need for low cost and effective inventory and monitoring Universally, this has meant the use of remotely sensed imagery from satellite (Rennolls, 1998), and classification of forest types from image data is an inference process that reduces raw data

to useful information (Moffett, 1996)

In forest management and planning in Vietnam, the classification of forest status with regard to the level of disturbance is the most practical and suitable method of all Therefore, this system was considered for improvement in this study

19

Chapter III

3 MULTI-DATA SOURCES IN FOREST INVENTORY

Inventories have produced large area forest resource information in some countries since the beginning of the 1920s The first world-wide forest resource assessment was compiled from national statistics by the FAO in 1947 The requirements for forest inventories have increased during the last couple of decades New information sources and new methods make it possible to increase the cost-efficiency of inventory and change them from a field measurement-based system into a multi-source monitoring of the whole forest ecosystem, thereby, providing information about the structure of forests, their health and their biodiversity status for small and large areas Multi-source based forest inventory and management including remote sensing and a geographical information system (GIS) become increasingly necessary in most forest inventory and monitoring Some main subjects considered in this chapter involve remote sensing techniques, the definition of some forest variables of interest, using remote sensing in forest inventory (e.g combining remote sensing and field data for classification or extracting forest variables from remote sensing images) An overview of the GIS technique in forest management is given in the last part of the chapter

3.1 Remote sensing techniques

3.1.1 Pre-processing remote sensing images

Image data received from imaging sensors mounted on satellite platforms generally contains flaws or deficient data The correction of deficiencies and the removal of flaws present in the data is part of pre-processing because, quite logically, such operations are carried out before the data is used for a particular purpose Despite the fact that some corrections are performed at the ground receiving station, there is often still a need on the user’s part for some further pre-processing Correction for geometric, radiometric, and atmospheric deficiencies and the removal of data errors or flaws is part of this process However, what should be included in pre-processing is dependent to a considerable extent

on the use to which data is to be put Therefore, all or some of these corrections may apply

to a given analysis or application, and the selective use of each is dictated by the type of sensor data and by the objectives of the work (Mather, 2004)

20

Nowadays, this process of information extraction and translation of radiance into a useful object identity or object property is mainly computer-based This process of translating reflected or emitted electromagnetic radiation into useful information is normally referred

to as image processing Image processing comprises a number of preparatory steps i.e geometric correction and radiometric processing of images, image enhancement or image improvement steps such as contrast stretch, image transformation and noise reduction Among them, ortho-rectification and topographic correction were considered under the current study

3.1.1.1 Ortho-rectification

Remotely sensed data displays usually in varying degrees of geometric and location distortion There are two broad types of spatial distortion: systematic and nonsystematic Systematic distortion occurs when coordinates are consistently wrong by a certain amount across the whole image while nonsystematic distortion occurs when random factors cause local variations in image scale and coordinate location These errors need to be reduced

before intepreting

Ortho-rectification is a form of geometric correction that takes into account the relief of the terrain To locate ground features on imagery, or to compare a series of images, a geometric correction procedure is used to register each pixel to real world coordinates

In ortho-rectification, the distortions induced by the imaging platform, film and dimensional shape of the earth are digitally removed from the space imagery The final result is an image that has a precise geometry of a map which is a very popular product due

three-to its diversity of use, particularly in its use as base information for Geographic Information Systems (GISs) Applications in many disciplines integrate the existing line data with digital images A useful and common integration is that of ortho-rectified images input directly into a GIS This is quite advantageous for providing a basis for a new data set or for updating existing databases Another remarkable advantage of a digital ortho-rectified image is that they provide more readily usable data sources for a GIS when compared with conventional data sources (Jensen, 1996)

21

3.1.1.2 Topographic correction

The topographic effect is defined as the variation in radiance from surfaces, compared with radiance from a horizontal surface, as a function of the orientation surface relative to the light source and sensor position (Holben and Justice, 1980) Topography does not only affect the geometric properties of an image but has an impact on the illumination and the reflection of the scanned area This effect is caused by the local variations of view and illumination angles due to mountainous terrain Therefore, identical land-cover might be represented by totally different intensity values depending on its orientation and on the position of the sun at the time of data acquisition

If the atmospheric influence and adjacency effects are ignored, it can be believed that in the visible and near infrared bands, the direct sun radiation is the only illuminating factor

If the terrain were completely flat and all objects had a Lambertian reflection characteristic, the reflected energy measured by a sensor (radiance) would only depend on the direct irradiance and the reflectance of the objects on the surface However, most

objects including forest have a non-Lambertian reflectance characteristic (Meyer et al.,

1993)

Methods for topographic correction can be grouped into two categories, one is based on band ratios e.g NDVI, and the other requires DEM (digital elevation model)

Some algorithms have been pursued and developed to remove these effects Teillet et al

(1982) described four topographic correction methods including a simple cosine correction, a statistic-empirical correction and two semi- empirical methods which are the Minaert method and the C-correction Civco (1989) identified some important considerations e.g the digital elevation model of the study area has a spatial resolution comparable to the spatial resolution of the digital remote sensor data; when correcting for topographic effects using a Lambertian surface assumption, remotely sensed data is often overcorrected Slopes facing away from the sun appear brighter than sun-facing slopes and

so on

Corrections for this effect have been developed, together with attempts at building methods

of incorporating the topographic effect into image analysis to better extract the required forestry or vegetation information from the imagery Neither of these two ideas - correcting for topography, or using topographic information to help make decisions - has attained the status of an accepted standard method in remote sensing image analysis (Franklin, 2001)

al., 1993; Riano et al.; 2003; and Tokola et al., 2001)

3.1.2 Image processing

The process of translating reflected or emitted electromagnetic radiation into useful information is normally referred to as image processing An image processing system is a key component of the infrastructure required to support remote sensing applications (Franklin, 2001) After the pre-processing of the image as mentioned above, image analysis is normally continued by a number of information extraction steps These information extraction steps refer to supervised or unsupervised image classification methods, the use of spectral vegetation indices or other types of image rationing, pattern

recognition and/or visual interpretation (Jong et al., 2004)

3.1.2.1 Remote sensing image visualization and interpretation

In general, classification approaches aim at the production of a thematic map from multispectral satellite imagery They are based on a principle that different objects or land cover/land use types show typical reflectance properties which are at best clearly distinguishable in the multispectral feature space Thus, the actual level of detail of the constructed thematic map depends on the particular ability to distinguish between the relevant categories which, in this study, are forest status and other land cover

Traditionally, an image analyst carried out the extraction of information from imagery by visual interpretation This visual interpretation is based on image properties like colour, pattern, shape, shadow, size, texture and tone of objects (Lillesand and Kiefer, 1994, Janssen, 2001) Based upon the possibilities given by the human eye–brain system, the interpretation of images by an analyst starts at a primary level by observing contrasts of tone and colour At a secondary level, the size, shape and texture are compared At the third level, the pattern, height difference and shadow aids are interpreted At a fourth level, the association with adjacent objects plays a role (Konecny, 2003)

Interpretation keys or guidelines are required to instruct the image interpreter In such guidelines, the (seven) interpretation elements can be used to describe how to recognize