Analyzing the key drivers of tree planting from local people in cao phong district hoa binh province vietnam with bayesian networks

In this study, 100 households in Bac Phong and Xuan Phong communes, Hoa Binh province were surveyed on factors influenced by tree planting decision as well as the area of forest will be

MINISTRY OF AGRICULTURE AND RURAL DEVELOPMENT

VIETNAM FORESTRY UNIVERSITY

Faculty: Forest Resources and Environmental Management

Student: Tran Thi Mai Anh Student ID: 1153091133

Class: K56 Natural Resources Management Course: 2011 - 2015

Advanced Education Program Developed in collaboration with Colorado State University, USA

Supervisor: Dr Le Dinh Hai

Ha Noi, 10/2015

Abstract

Known as the climate and watershed protection roles, tree planting is an important reforestation activity in the area that involves planting seedling (United Nation Environment Program) In coping with significant deforestation and forest degradation currently in Cao Phong district, Hoa Binh province, a massive of reforestation projects have been implemented The typical forestry programs are the 5 million ha reforestation program, Project RENFODA and Afforestation and reforestation clean development mechanism (AR-CDM) project were contributed to a substantial increase in Hoa Binh‟s forest cover during several decades However, when remarkable attempts and investments have been made in reforestation, interaction of household characteristics, and socio-economic factors with tree planting decision are still little comprehension In this study, 100 households in Bac Phong and Xuan Phong communes, Hoa Binh province were surveyed on factors influenced by tree planting decision as well as the area of forest will be planted The research investigated and determined 6 success indicators including household characteristics and socio economic indicators which effect the decision of farm households to plant trees with Bayesian networks (BNs) BNs allow the incorporation of variances relationships in data analysis and can combine qualitative dependents with quantitative data Based on this there are 4 key factors directly correlated to tree planting decision and 5 other highly connected variances that influence to area of tree planting The two BNs also are developed with the child nodes to give suggestions and recommendations for increasing forest in research area We conclude that government forestry programs contribute great influences to forest planting activities and household characteristics as well as socioeconomic factors have returned effects on the success of reforestation programs

I wish to extend my thanks to express my heartfelt gratitude to Prof Dr Lee MacDonald, Department of Ecosystem Science and Sustainability, Colorado State University for his valuable and constructive suggestions during the planning of this research work His willingness to give his time so generously has been very much appreciated

Similarly, I would love to thank various people for their contribution to this project;

Mr Nguyen Huu Son, staff of Communal People Committee of Cao Phong district; the staffs

of Communal People Committee of Bac Phong and Xuan Phong Communes; Mr Bui Van Bu

in Re village, Bac Phong Commue who devote their time to help me during the field survey Special thanks to local people in Cao Phong district for providing me helpful information in this study

Finally, I own my gratefully thank to my grandmother, my parents, my sisters, my uncles and aunts for their endless love, supports, and encouragements to me throughout all my life Without their wonderful care, it would have been impossible for me to chase my dream

Table of Contents

Abstract 2

Acknowledgements 2

Table of Contents 3

List of photos 6

List of figures 7

List of tables 8

Chapter 1 Introduction 1

Chapter 2 Study Goals and Objectives 4

2.1 Study goal and objectives 4

2.2 Research questions 4

Chapter 3 Study Area and Research Methodology 5

3.1 Selection of research site 5

3.1.1 Hoa Binh Province 5

3.1.2 Cao Phong district 6

3.2 Research methodology 9

3.2.1 Framework of factors influencing tree planting development of smallholder 9

3.2.2 Data collection 16

3.2.3 Data analysis methods 19

3.2.4 Bayesian networks 22

Chapter 4 Results 27

4.1 Descriptive statistic on surveyed households in Cao Phong district 27

4.1.1 Land area 27

4.1.2 Education 28

4.1.3 Distance to market 28

4.1.4 Knowledge on silviculture 29

4.3 Key drivers affecting forest area will be planted of surveyed households 32

4.4 Relationships among independent variables 33

4.5 Bayesian networks (BNs) 36

4.5.1 Bayesian Network of tree planting decision model 36

4.5.2 Bayesian network of forest area will be planted 38

Chapter 5 Discussion 43

5.1 The difference of perspectives of households about planting trees 43

5.2 Potential actions which may increase planting forest in Cao Phong district 48

5.2.1 Suggestion based on attitude of tree planting 48

5.2.2 Suggestion based on investment capital for tree planting 49

5.2.3 Suggestion based on knowledge on silviculture of local people 50

5.2.4 Suggestion based on distance to market 51

5.2.5 Suggestion based on total land area of each household 53

5.2.6 Suggestion based on participation forest program 54

5.3 Limitations and suggestions for future research 55

Chapter 6 Conclusion 57

Chapter 7 References 60

Chapter 8 Appendices 65

List of photos

Photo 3.1 Land use change from timber tree to fruit tree farm 8

Photo 3.2 Photo of interviewing local household of tree planting decision 18

Photo 3.3 Acacia mangium is a popular species in most family 22

Photo 5.1 Tree planting for fuel wood in Bac Phong commune 46

Photo 5.2 Rich household invest on Aquilaria crassna Pierre species 50

List of figures

Figure 3.1.The map of Hoa Binh province 6

Figure 3.2 Map of communes in Cao Phong 7

Figure 3.3 Factors influence tree planting decision of smallholder 10

Figure 3.4 Bayesian of tree planting with the status of all variances 26

Figure 1 Frequency distribution of total land area 70

List of tables

Table 3.1 General Information on Bac Phong and Xuan Phong Commune (2015) 9

Table 3.2 Cost-benefit parameters and possible influencing cost benefit estimates 13

Table 3.3 Sampling design in Cao Phong district 17

Table 3.4 Basic to define household ranking in Vietnam 18

Table 3.5 Basic to define moderate and rich households in Xuan Phong Commune and Bac Phong 18

Table 3.6 Discretization methods for continuous variables 20

Table 3.7 Tree species are planted in Cao Phong district 21

Table 4.1 Model summary for key drivers affecting tree planting decision 31

Table 4.2 Model summary for key drivers affecting area will be planted 32

Table 5.1 Farmer‟s reasons for planting trees and not planting trees 45

Table 5.2 Percentages of farmers mentioning specific disadvantages of tree planting 47

Table 1 Frequency of people participating in forestry program 70

Table 2 Distribution of distance to market of surveyed households 70

Table 3 Description of discrete variables 71

Table 4 Characteristics of surveyed households by communes with discrete variances 72

Table 5 Correlation between households‟ characteristic and tree planting decision 73

Table 6 Correlation between households‟ characteristic and forest area will be planted 73

Table 7 Summary of significant relationships among tree planting decision and indicators 74

Table 8 Summary of significant relationships among forest area will be planted drivers and indicators 75

Table 9 Difference of socio-economic and household characteristic between tree planter and non-tree planter in Cao Phong district 76

Chapter 1 Introduction

Tree planting is the process of transplanting tree seedlings, generally for forestry, land reclamation, or landscaping purposes In silviculture the activity is known as reforestation and afforestation, depending on whether the area being planted has or has not recently been forested (Ngugi Tirus Kamau 2003) Afforestation and reforestation both refer to establishment of trees on non-tree land Reforestation refers to establishment of forest on land that had recent tree cover, whereas afforestation refers to land that has been without forest for much longer (IPCC 1998) Tree planting not only can be used as a geoengineering technique

to remove CO2 from the atmosphere but also can be used as solution for economic forest development in comprehensive and sustainable way

Hoa Binh is a mountainous province of Vietnam covering an area of 460,870 ha Of those forest land area occupies 332,800 ha holding 72% of total area; the number of labors working in agroforestry is 391,500 people taking roughly 71% of all labors in the province (Nhan Sinh 2015) However, since 2009, 25.7% of hill land in Hoa Binh province has not been covered by forests yet (Đăng Ngọc Oanh 2009) These forests protect Da electric power plant, regulate stream flow, maintain water quality, minimize erosion, conserve ecosystems, and provide other benefits via their protection Therefore, planting forest in Hoa Binh were determined as the most importance of environmental protection and a necessary insurance of short term and long term earning for the rural communities

Besides that, deforestation had become a problem in Hoa Binh Province In the last decade, due to cutting forest and unplanned cultivating of the ethnic communities, the forest cover of the province dropped to an alarming rate with 27% (GIZ project, 2000) Plus with forest fire forecast is on the level 5/5 which is above average activity, the forests were decreased from 23,103 ha to 23,042 ha during 2004 – 2011 period (Phạm Quang Tùng 2013)

In 28-September 2005, People‟s committee of Hoa Binh Province was given an official

document number 1314/UB-NLB to eliminate forest destruction inside the area, yet deforestation is still an insolvable problem

Throughout 20 years, a lot of projects from government and international organizations were established such as Decree 327, Decree 661(1998), Project 747 (472), and Project RENFODA (cooperate with JICA technologies), to fulfil the objectives of raising the forest cover on national level to an average of 43% (Thủ tướng chính phủ 1998) Up to now, Hoa Binh has implemented 15 plans with the budget by 2005 is 40.1 billion VND and in 2008

is 27.4 billion VND Particularly, on 20 April 2008, a project of Afforestation and reforestation clean development mechanism (AR-CDM) has firstly implemented in Cao Phong district, Hoa Binh Province This is a project formulated with the assistance of JICA study on Capacity Development for AR-CDM Promotion in Vietnam with a total of 3.5 billion VND to be donated to Forest Development Fund which established in April 2008 according to Decree No.148/2007/ND-CP by Honda Vietnam Company The period for the project is 16 years covering by 308.5 ha located in Xuan Phong and Bac Phong communes, Cao Phong district and be voluntary participated by 320 households of the land using certificate of the project area would conduct planting, tending, protecting and managing of the plantations (Tạ Văn Chính 2007) Planting forest has provided job, contributed the income, eliminated hunger and reduced poverty for thousands of local people, especially the ethnic minorities

At the Hoa Binh hydropower plant watershed, the forests transferred for local people were 81,850 ha, occupied 94% of total 86,980 ha of forest land in protection forests in 2009 (Hoang Lien Son 2009) which were considered as one of 13 focal points of headwater forests

in Vietnam (Ministry of Agriculture and Rural Development 2006) Moreover, at the moment, for the market output of plantation forests and economic forests, there are 33 business manufacturing, wood and forest product factories at small and medium scales inside the

province are recently operative This is a good condition for local people to connect between tree planting and harvesting, processing timber products Therefore it might promotes socio-economic of households to plant trees

Meanwhile Hoa Binh province is a cradle of Muong ethnic people as well as home of

so many minority communities, most of them live based on natural resources Thus, household characteristics understanding may be the main factor that influences tree planting Accordingly, it leads to the questions that what are the differences in the socio-economic and perceptional characteristics of tree planter and non-tree planters What are farmers‟ motivations for planting or not planting trees, and what are the main disadvantages related to tree planting?

For those reasons, I come up with a goal of analyzing the key drivers of tree

planting from local people in Cao Phong District, Hoa Binh Province, Vietnam with Bayesian Network to be my study in order to have a better understanding of community

members „motivation for reforestation and afforestation

Chapter 2 Study Goals and Objectives

2.1 Study goal and objectives

 Goal: Analyzing key drivers of tree planting from local people with Bayesian Network

in order to provide suggestion in increasing planting forest in Cao Phong District, Hoa Binh Province

 The main objectives of this study are:

(1) To determine the household factors associated with household decision of local people to plant trees and explain them

(2) To provide potential suggestion which may increase planting forest in Cao Phong district, Hoa Binh province, Vietnam

2.2 Research questions

(1) What are key drivers affecting tree planting decision of households?

(2) What are key drivers affecting area will be planted by households?

(3) What are farmers‟ motivations for planting or not planting trees?

(4) What are the main disadvantages related to tree planting?

Chapter 3 Study Area and Research Methodology

3.1 Selection of research site

3.1.1 Hoa Binh Province

The area of Hoa Binh province is the source of headwater and major tributaries that influence the lives of more than 808,200 people (GSO, 2013) inside the province Located in the North of Vietnam It borders Son La and Phu Tho provinces to the northwest, Ha Noi city

to the north and northeast, Ha Nam province to the southeast, Ninh Binh and Thanh Hoa provinces to the south Hoa Binh is a mountainous province located on the entrance of the Northwest region and is proud to be famous with “Hoabinhian Culture” where human life is proven to existed here since 10,000 – 2,000 BCE The topography is combined by mountains and narrow valleys results in the climate of this district is representative for tropical monsoon, which is pretty cold and less rain in winter but hot and rainy in summer The annual temperature varies between 15 to 29 Degrees Celsius, depending on season Hoa Binh is in the region has a high poverty rate and a low standard of living of the population The growth

of GDP amounts to 11.8% during 2000-2010 The poverty rate was 31.31% in 2005, and was 14% in 2010, but in 2011 the rate of poverty has jumped again to 37.68%, according to the new rate of poverty (Mai Lan Phuong 2011) They are a large variety of ethnic groups, which has 15 ethnic communities, and 63.4% is Muong ethnic group The variety of both culture and environment leads to diverse land-use systems

Cao Phong District, Hoa Binh Province was chose to be a case study because of the following reasons Firstly, Cao Phong is located at the center of Hoa Binh province which is representative for mountainous area, bounded with streams, rivers, valleys, and limestone mountains Secondly, this area also is a focal point of planting protection forest for headwater which plays an important role for protecting water resource of whole province There are 3 ethnic communities living together (Kinh, Dzao, Muong), mainly practice agriculture, largely

on sloping areas Deforestation and change in land use are critical challenges for protecting and managing forest in Cao Phong District, Hoa Binh Province

Source:http://investinvietnam.vn/data/image/HoaBinh.jpg

Figure 3.1.The map of Hoa Binh province

3.1.2 Cao Phong district

a) Geographic location

Cao Phong is a rural area in Hoa Binh Province bordered Kim Boi district to the east, Hoa Binh city to the north, Da Bac district to the northwest, Tan Lac district to the west, and Lac Son district to the southeast The district extends along with the Number 6 Highway from Hanoi to the provinces of Hoa Binh, Son La, Lai Chau with the total area of 25,460 ha

Bac Phong Commune is located in the north of Cao Phong District, Hoa Binh Province, on the inter-commune road running from Tay Phong Commune, Cao Phong town to Hoa Binh city and Binh Thanh Commune It has a total land area of 2,329 ha, including 1,613

ha of forest land, with the area of natural forest is 72 ha and plantation forest is 359 ha (Communal People Committee of Bac Phong in 2015)

Xuan Phong Commune is bordered with Kim Boi district along from the north east to the south east, the west side bordered with Thu Phong, Dong Phong, Tan Phong, Yen Lap from the northwest to the south west, respectively Total land area is 3,111 ha including 816

ha of forest land, with the area of protection forest is 428 ha, and 388 ha of plantation forest (Communal People Committee of Xuan Phong in 2015)

c) Soil and land use

Due to the topography is quite complex and various, Cao Phong has different types of soil On the hill and mountain area, yellow ferarit soil type and limestone are the most

popular Downhill and lowland has alluvial soil Therefore, soil in lowland of Cao Phong district has high fertility and be able to plant diverse of tree species

Nevertheless, according to land use state of Cao Phong district 2015, agriculture land just hold 14% while non-use land in forest land accounted very high percentage is 40% As the survey from CDM reforestation program in 2008, the organic layer in the hill slope of both Bac Phong and Xuan Phong Commune was accessed as poor and thin, the main reason was soil was degraded very seriously due to long time deforestation This problem is also recorded during my surveyed interview through the answers of respondents and low development of tree species in the area

Nowadays, many households in Cao Phong district are changing their land use from agriculture land (rice cultivate, sugar cane, maize crops) into fruit trees land (orange, lemon, pomelon) with the purpose of increasing their income As noted by a village leader, around 70% of household in Mon Village, Bac Phong Commune is selling and clear cutting protection forests, plantation forests to plant fruit trees It leads to so many problems occurring in the region such as lack of water in the head stream which used to crop cultivating, polluted water in the upstream and soil degradation are more and more frequent

d) Socio-economic characteristics

 Population and ethnic composition

According to Communal People of Bac Phong and Xuan Phong Communes by 2015, the total of population in Bac Phong is 1,095 households comprising 4,167 inhabitants, with the natural growth rate is 1.2% Xuan Phong is considered sparsely populated with 785 households comprising 3,485 inhabitants, with the natural growth rate is 1% which slightly decreased (0.2%) from 2006 to 2015 Muong ethic is accounted for 99% of communities in whole district

Table 3.1 General Information on Bac Phong and Xuan Phong Commune (2015)

Source: Communal People Committee of Bac Phong and Xuan Phong (2015)

 Labor

The percentage of people in labor age (from 15 to 59 years old) in Bac Phong Commune is 60% and 58% in Xuan Phong Commune Average GDP of people in Cao Phong district is estimated around 8.5 million VND/person/year (2015) Nearly all inhabitants in Cao Phong are working in agroforestry Hence their income are depended on tree planting This also is one of the reason for me to choose Cao Phong to do my study

3.2 Research methodology

3.2.1 Framework of factors influencing tree planting development of smallholder

A wide range of factors that influences smallholder tree planting development are identified in the literature review These factors are grouped into those that relate to characteristics specific to household and farming factors, socio-economic aspects, institutional policy aspects and biophysical characteristics that are discussed below

Figure 3.3 Factors influence tree planting decision of smallholder

1/ Household characteristic factors

Socio-demographic variables such as the age, gender, and educational level of framers might influence whether they engage in tree planting activities (Mercer and Pattanayak 2003) Older farmers are often viewed as less flexible, more risk averse and less willing to engage in

a new, innovative activity Thatcher et al (1996) and Zhang and Flick (2001) report that age has no influence on planting while Romm et al (1987) state that older age reduces the

probability of Silvicultural investment Education has been reported to influence significantly tree planting and conservation by famers because it is a medium of learning about a resource (Thacher, Lee et al 1996, Glendinning, Mahapatra et al 2001, Owubah, Le Master et al 2001) Nsiah and Pretzsch (2005) indicate that where household with forest plantation had higher education and also showed positive attitude towards forest plantation development than household without forest plantation However, educational level of household head was found not to impact tree planting in studies in Sumatra and the Philippines (Otsuka, Suyanto et al

2000, Mercer and Pattanayak 2003) Access to information and forestry technical assistance

improves the quality of household labor, and their willingness to engage in forestry activities including tree planting (Amacher, Hyde et al 1993)

Reasons for planting and managing trees on land also influence smallholder tree planting These reasons include the provision of a legacy for children and grandchildren; timber production for sales; provision of construction materials; increase of farm income; and

other environmental improvement However, according to Sevare et al (2007), the main

purpose of famers for tree planting in Leyte Island, the Philippines is to provide a legacy for their children and grandchildren as a source for their income in the future, to provide for their own use and in order for them to perceive the forest and appreciate nature

2/ Farm characteristics/resource endowments

Important farm characteristics/resource endowments include level of intensification, size of landholding and land availability, total household income, livestock, accessing to existing forest resources, and distance to market The level of diversification of the production system is an important factors influencing farm tree management decisions (Mahapatra and Mitchell 2001)

The asset holdings or resource endowments (e.g land, labor, and wealth) that a farming household possesses are a measure of resources available to the household for implementing a new agroforestry practice such as tree planting Household wealth has a positive influence on agroforestry adoption and conservation investments (Glendinning, Mahapatra et al 2001, Pattanayak, Evan Mercer et al 2003) Nsiah and Pretzsch (2005) also note that household income has significant influence on smallholder forest plantation development Larger farm size often play a role in farming household‟s participation in tree programs (Thacher, Lee et al 1996, Salam, Noguchi et al 2000) However, Nsiah and

Pretzsch (2005) and Otsuka et al.(2000) report that size of household land did not influence

tree planting

Resource constraints influence landholder investment in tree planting Harrison et al.(2000) show that the small-scale landholder often lacks the inputs such as land, capital,

planting material and silvicultural knowledge to grow trees Lack of finance can lead to failure to apply fertilizer or to adopt other recommended practices Peak labor requirements may clash with labor demands for other crops Extension services may be limited, and species recommended are not always the most suitable for the area

Scarcity of wood supply from existing resources is cited as an important factor influencing tree management and wood production by small-scale landholders Gilmour‟s (1995) models on the likely responses of Nepalese farmers to tree planting and protection indicate that the farmers‟ interest in devoting land and labor for growing tree declines with proximity to available existing tree and forest resources Where wood supply from nearby forest resources is scarce, high collection time and purchasing cost of fuel wood are expected

to have a positive influence on farmers‟ tree planting development

3/ Biophysical factors

Biophysical characteristics, especially soil quality and slop of farmland are related to the production process of food crop and tree farming In general, poorer biophysical production conditions (plots with poor soil fertility for food crop production or highly exposed to erosion) are likely to be positively related to tree planting (Pattanayak, Evan Mercer et al 2003) Tree growing is further positively related to fine-textured clayey soils and negatively related to coarse-textured soils (Schuren and Snelder 2008) However, there is threshold quality below which any productive investment may not be useful

In addition, the distance between a field and farmer‟s house is negatively related to tree growing Trees are preferably grown close to the house where farmers can more easily inspect them and prevent damage or losses by fire, astray animal and theft (Schuren and Snelder 2008)

Moreover, climate also influences tree planting For example, the increasing political interest in climate change and the ability of forests to sequester carbon is expected to encourage more state led efforts to expand plantations, while also providing supplemental income for the rural poor via payments for carbon sequestration (Maarit Helena KALLIO 2013) In this study, local people in Cao Phong district are supported by reforestation project was registered as small-scale project under the Afforestation/Reforestation Clean Development Mechanism within the UN Framework Convention on Climate Change (UNFCCC), part of the Kyoto Protocol and the 661 national reforestation project

4/ Socio-economic factors

In addition to other factors, socio-economic factors also strongly influence smallholder land use decisions Households determine their land use portfolios based upon potential net benefits given their environmental and economic resource endowments, and taking into consideration the time frame in which outputs will be profitable These factors, including opportunity cost of the various factors of production, access to markets for inputs and outputs, transaction costs associated with institutions, risk and access to credit, and the discount rates

of economic decision making units, need to be taken into account to identify locations economically as well as ecologically suited to tree growing Table 3.5 outline the socio-economic parameters we employ in our ex ante cost benefit estimates and summarizes how each of the various factors may influence economic returns to tree planting

Table 3.2 Cost-benefit parameters and possible influencing cost benefit estimates Cost-benefit

High potential land with high opportunity cost

Few negative environmental Significant negative environmental

externalities and/or positive externalities

externalities

Labour Low or medium wage rates (may be

associated with high population density distant from off-farm labour markets)

High wage rates (may be associated with low population density relatively close to off-farm labour markets)

High wage rates may also favour tree planting when compared with other labour intensive activities

Material

inputs

Low cost material inputs related to good access to input markets

High cost of material inputs related

to remoteness from input markets Presence of NGOs and other

organisations that offer inputs for free

or at subsidised prices

Absence of NGOs and other subsidising agencies

Output prices Good output market access (may be

partially related to proximity to major road, town and high population density

Poor market access (characteristics

of poor access to roads and towns)

Markets with sufficiently elastic demand

Inefficient markets and/or inelastic demand

Discounted

rates

Good access to credit markets Poor access to credit markets High potential lands yield high MAI

providing returns in shorter time horizon

Low potential yields yield low MAI providing returns in medium

to long run Institutional

factors

Access to benefits not restricted Restricted access to benefits

Source: Jagger and Pender (2000)

5/ Institutional and policy factors

Institutional and policy factors include land and tree tenure security, the organization

of overall rural development and forestry system, extension services, and information sources

Land and tree property rights security

Tenure insecurity is defined here as the perceived probability of losing ownership of a part or the whole one‟s land without his or her consent (Sjaastad and Bromley 1997) Land

ownership in agrarian societies is not only the main means of generating a livelihood, but it is often also the primary means for accumulating wealth and transferring it to future generations Right of access and use to the products that are produced from woodlots is an institutional factors that may influence long-term investments such as tree planting If smallholders perceive that they may not be given approval to harvest products from woodlots that they have invested in, they are less likely to get involved in community tree planting because of the uncertainty of when or if they will benefit from their investment In addition, smallholders who have weak property right (e.g insecure land tenure) may not undertake a socially efficient level of production (Harrison 2005) If land and tree tenure are uncertain, this will discourage the investment in tree planting

For many smallholders and communities in developing countries the question of tenure security is major constraint to tree planting and management (Herbohn, Emtage et al 2005) Lack of secured land and tree tenure have been documented as a significant constraint

to tree cultivation as it impacts the potential benefit accrued to famers (Zhang and Pearse

1997, Place and Otsuka 2000) Basically, other things being equal, farmers who hold a secured title to land are more likely to participate in tree planting activities than those who do not The forms of tenure that have longer terms, are more clearly defined, provide more of the economic benefits to their holders, are likely to stimulate planting (Zhang and Pearse 1997) The duration of tenure determines whether farmers plant trees or short-term crops (Sellers 1988) Within the same village, farmers respond differently to tree planting programs depending on the tenure they hold to farmlands, while the same famers respond differently to tree planting programs on lands he or she cultivate under different types of tenure (Pasicolan, Udo de Haes et al 1997) Farmers rarely participate in tree planting programs when they have

no rights to trees (Fortmann and Bruce 1988) Therefore, the strongest incentives for promoting timber production are to give farmers full rights over tree they cultivate (Treue 2001)

Tree harvest and log transport rights

Slow and costly approval processes for tree registration, tree harvesting and log transport, frequently changes to these processes and regulations all serve to increase the perceived level of transaction costs and sovereign risk associated with forestry development

in general and with tree planting in particular (Harrison 2005) In the Philippines, although the DENR has sought to reduce the complexity of these requirements over the past 10 years, the log harvest and transport permit system is still inappropriate for tree planting smallholders (Donoghue 1999, Harrison 2005) The lack of stability of the regulations relating to commercial tree planting is likely to reduce incentives of farmers to invest in tree planting

Forestry support programs

The frequency with which the farmers have contact with extension agents is important

in the acquisition of informal education (Salam, Noguchi et al 2000, Adesina and Chianu 2002) Hence, the efficiency of the agroforestry extension services and dissemination of information is essential in improving the farmers‟ human capital and their decision making and farm management activities

Information for growers

Tree growers would benefit from greater silvicultural, marketing and regulation information (Steve Harrison 2005)

household characteristics, factors influencing tree planting decision of household A copy of the questionnaire is available on request

Table 3.3 Sampling design in Cao Phong district

100 households, the household ranking was equally separated as 33 rich households, 33 moderate households and 34 poor households The interview design was followed by a stratified random sampling approach to obtain representative strata by participation in forestry programs and estimated wealth

The questionnaire was administered face-to-face, usually the head of households This method allows researchers the opportunity to ask more questions, longer questions, more detailed questions, more open-ended questions, and more complicated or technical questions

Moreover, face-to-face surveys also offer advantages in terms of data quality (Manurung et al

2008) The survey was conducted from 1st August 2015 to 20th August 2015

The criteria of rich, moderate, and poor households are classified as the table below

Table 3.4 Basic to define household ranking in Vietnam

Poor household

Rural areas: < 4,800,000 VND Urban areas: < 6,000,000 VND

Source: General Statistic Office of Vietnam (2015)

Near poor household

Rural areas: 4,812,000 – 6,240,000 VND Urban area: 6,012,000 – 7,800,000 VND

Sources: General Statistic Office of Vietnam (2015)

household

Income per capita per year: > 10,000,000 VND

Land area: > 2 ha

Assets owned by household

Table 3.5 Basic to define moderate and rich households in Xuan Phong Commune and

Bac Phong Commune (2015)

Photo 3.2 Photo of interviewing local household of tree planting decision

Secondary data collection

Secondary data collection is used in this study if the key informant did not have information for a particular question, then data for that question was obtained from official government records, academic publications of different agencies, such as Communal People Committee of Bac Phong and Communal People Committee of Xuan Phong, Ministry of Agriculture and Rural Development, and Vietnam Bureau of Statistics

3.2.3 Data analysis methods

IBM SPSS Statistics 20 (2011) was used for data analysis Bivariate analysis was used

to identify association between indicators (dependent variables) and drivers (dependent variables) (see Table 3 - Appendices and Table 3.6 for a full list of drivers included in the analysis) For binary indicators „Tree planting decision by household‟ (0 or 1), the Student‟s t test was used to explore associations with continuous drivers and the Pearson‟s χ2 test was used to explore associations with categorical drivers For continuous dependent variable „Area will be planted‟, linear regression was used to explore associations with continuous drivers and the Student‟s t test was used to explore associations with categorical drivers Drivers found to be significantly associated with indicators in the bivariate analyses (p<0.05) were considered as candidates in stepwise multiple regressions with indicators

Before conducting stepwise multiple regressions, preliminary analyses were conducted

to ensure no violation of the assumptions of normality, linearity, multicollinearity and homoscedasticity among the variables For continuous dependent variable, standard stepwise multiple linear regression was used For binary dependent variable, forward stepwise binary logistic regression was used Drivers were entered into the stepwise regressions if the significance of their relationship with an indicator was p<0.05 and removed from the stepwise regressions if the significance of their relationship with a dependent variable became p≥0.10 Drivers were entered into the stepwise regressions in order of their correlation with an

dependent variable , from most strongly (lowest p value) to least strongly correlated (highest p value) (Brace, Kemp et al 2006, Ho 2006)

The end result was a set of significant driver-dependent variable, dependent-dependent and driver-driver relationships that was used to identify a system of relationships that affect tree planting decision and area will be planted

A discrete variable is one with a well-defined, finite set of states; A continuous variable can take a value between any other tow values (Jeremy Cain 2001)

Based on (Kanellopoulos 2006), the ability of BN to deal with continuous data is limited, so we often discretized The discretization of continuous variables requires a decision

on the number of discrete bins, and the corresponding cut-off values Thus, there are infinite ways to discretize continuous variables A good discretization method should minimize information loss and maintain or maximize the interdependence among variables Heuristic methods such as equal width (or equal interval), equal frequency methods and entropy minimization are widely used (Marcot 2006) suggested using the least possible number of intervals to find a balance between parsimony and precision In addition, when deciding on cut-off values, model objectives and domain knowledge on certain variables should be considered in order to make the binning more sensible and logical (Frayeer 2014)

Following these principals, we binned the area-related and the income-related

variables into same intervals to facilitate comparison, reduced number of labor to three

categories and kept the remaining categorical variables in the original measurement scale

Table 3.6 Discretization methods for continuous variables

1 Age of household head

(years old)

20-35; 35-50; >50 Expert knowledge (According

to Law on marriage and family

of Vietnam in 2015; man from

20 can get married, and normally be head of household

Thus 20 is the first cut, and 51 will be the last cut We have 3 equal width of classification)

2 Number of labor

(person)

Low (1-2); Medium 4); High ( >5)

(3-Logical, Equal frequency

3 Total land area (ha) Little (0ha – 0.5 ha);

Medium (0.5 ha – 2 ha);

A lot ( > 2 ha)

Equal grouped frequency distribution (Criteria of household ranking in Cao Phong district with poor, medium, and rich household)

4 Forest land area (ha) Little (0ha – 0.5 ha);

Medium (0.5 ha – 2 ha);

A lot (> 2 ha)

Expert knowledge, equal width

5 Total household income

(million VND)

Low (0-50); Medium (50-100); High (>100)

Equal grouped frequency distribution

6 Distance to market

(km)

Near (0-6); Medium 12); Far (>12)

(6-Equal width and frequency

7 Distance to the field

(m)

Near (0-3000); Far (>3000)

Equal grouped frequency distribution

8 Area of tree will be

To deal with defining long time tree and short time tree type, the list name of tree species is classified as the table bellows

Table 3.7 Tree species are planted in Cao Phong district

1 Bạch đàn (Eucalyptus camaldulensis Dehn) Short time

3 Luồng (Dendrocalamus membranceus Munro) Short time

6 Xoan (Melia azedarach Linn) Long time

8 Sưa Bắc bộ (Dalbergia tonkinensis Prain) Long time

Source: Surveyed household interviews in 2015

Photo 3.3 Acacia mangium is a popular species in most family

3.2.4 Bayesian networks

Bayesian networks (BNs) are used for the analysis of tree planting behavior BNs are non-parametric statistical tools that rely on Bayesian inference to deduce the influence of explanatory variables on the outcomes of interest A Bayesian networks consists of two parts: the first is a directed acyclic graph (DAG), also known as the structure of a BN This depicts interdependence among variables with directed arrows connecting nodes (corresponding to variables) The second component are conditional probability tables (CPTs), also known as parameters of a BN, which define the probability distributions of nodes conditioned upon the values of their parent nodes (where an arrow originates) The CPT of a child node (where an arrow ends) contains the conditional probability of being in a specific state, given the states of its parent node When a node has no parent, the CPT is simply its prior probability

distribution The conditional probability of a variable is also known as the belief for this state

of the variable.(Frayeer 2014)

Probabilistic inference following Bayes‟ theorem is used to quantify influences in the network with conditional dependencies Given observations of some variables (i.e., evidence), BNs can in principal, regardless of the directions of arrows connecting them deduce the posterior probability of any other variable Thus, BNs can support not only forward inference

or predictive analysis (from causes to effects), but also backward inference or diagnostic analysis (from effects to causes) (Judea Pearl 2000)

One important advantage of BNs is their ability to incorporate qualitative stakeholder knowledge, and quantitative and spatially explicit data This flexibility can account for subjective decision-making and qualitative reasoning Moreover, the representation of effects

as probability distributions implicitly incorporates an uncertainty component by going beyond

a mere point estimate, such as in traditional regression analysis The capability of BNs to combine causal stakeholder knowledge and empirical, evidence-based data explains their growing importance in environmental analysis Recent applications include environmental management, water management, forestry, wildlife management, and land-use change BNs also enable the assessment of different scenarios on the outcome variable Scenarios can be implemented either by incorporating new variables into the network or by changing the probability distributions, which are the CPTs of existing variables The resulting variation in the target variable‟s probability distribution then corresponds to the potential developments under the scenario conditions Altering the CPT of a node to examine corresponding changes

in outcome probabilities also facilitates insights into the sensitivity of individual variables (Frayeer 2014)

Construction of the Bayesian network

The construction of a BNs entails two steps: building the structure (the DAG) and learning the CPTs The DAG is developed by structural learning from the survey data Structural learning is conducted via data mining that automatically searches for statistical relationships among variables However, purely statistical interdependence may not reflect realistic cause-effect relationships among variables because the direction of the influences between variables cannot be determined by machine learning Given these limitations, we opted for a supervised learning strategy that draws on our prior knowledge derived from domain experts and our qualitative interviews with local people and local officials During supervised learning, we imposed constraints that prohibit some links while enforcing others

After the structure of the BN is defined, the CPTs for each node will be learned from questionnaire data using the expectation-maximization (EM) algorithm implemented in Netica software (Norsys software Corp 1995-2015) The EM algorithm is selected due to its robustness In sum, the CPT learning process can be viewed as a probabilistic classification to estimate the conditional probabilities between a node and its parent nodes.(David Heckerman March 1995)

Variables are kept in the model if their inclusion improve the predictive accuracy in form of a reduced ratio of incorrectly predicted cases to the total number of cases (lower error rate) of the target variable The influence of an input variable on the target variable is measured by the magnitude of probability changes of the target variable when the input variable is altered The influence of a variable corresponds to how much explanatory information a variable imposes on the target variable Consequently, a number of variables, which are a priori hypothesized to being important, is excluded because they carry redundant information and have relatively low explanatory power Variables are kept in the model if the error rate stayed stable or decreased and the influence on the other variables is above 0.05% After removing the variables with low influence from the network the desired compromise

between simplicity and high model accuracy, measured by the error rate, is achieved The resulting model will be updated with prior and conditional probabilities that represent the distribution in the sample population The final BN illustrates the variables that influence tree planting decision making and the interactions between variables From the chosen variables in Part 2, we can visualize the Bayesian Network as following

Figure 3.4 Bayesian of tree planting with the status of all variances

Table 4.1 Characteristics of surveyed households by communes in Cao Phong district

Parameters

Xuan Phong Commune (n=50)

Bac Phong Commune

(n=50)

T-test

P value (2 tailed)

Mean Std Dev Mean Std Dev

Total household income 61.78 47.134 128.32 213.024 0.28

Distance to the field 2587.76 2950.257 1643.87 1636.754 0.57

Source: Research data (2015)

The size of the standard deviation for the mean of the communities and for all respondents indicates the substantial variantion within communities as well as between them

4.1.2 Education

Each of respondents were asked to indicate the education achieved by household head

In general, the formal education level in the average household is secondary school which holds 35% (Figure 4.1), by most of households surveyed is Muong ethnicity (89%) Beside that, 43% of total households is high school or college above, 20% is primary school, and only 2% of total household is illiterate The result indicates that, in overall, households are attaited the primary level of education since Cao Phong district is still a poor rural area Different of education level beteen 2 communes is presented by Table 4 - Appendices

Figure 4.1 Education level of household heads in Cao Phong district

4.1.3 Distance to market

The results (Table 2 – Appendices) show that 33% of households have difficulties in accessing to market due to long distance (>6km), however most of households just need to travel with the distance of smaller than <6km

Student t test indicated statistically significant differences (P=.000) in terms of mean

of distance to market between two communes by presenting the distance to market per each household in Xuan Phong is 2 times higher than in Bac Phong (Table 4.1)

For most of households, the distance to market is comprehended by the distance which they have to travel to buy seedling or fertilizers for tree planting In general, as respondents

cited that they often order the seedling sources from the surrounding districts such as Kim Boi, Luong Son, or nursery trees in Xuan Mai, Chuong My, Hanoi

4.1.4 Knowledge on silviculture

Source: Research data (2015)

Figure 4.2 Knowledge on Silviculture of household on tree planting

The results in Figure 4.2 show that 52% of the respondents having good knowledge on Silviculture and around haft of total households admit that they have little or even no knowledge on this field In addition, most of interviewees cited that an extension officer from government forestry program is very important in training and educating communities on tree planting practices The more the farmers interact with them, the more likely it is for them to gain knowledge on Silvilcuture

The fact that, 77% of the respondents have said their knowledge on Silviculture was taken from forestry program training, only 23% is learnt by themselves and from the other tree planters

Table 4.2 Distribution of silvicutural knowledge sources of households

Where do you take the

silviculture from?

By myself From forestry program Total

Significant differences were found between communites in terms of the proportion of knowledge on silviculture (P=0.000) and knowing about forestry program (P=0.013) As the surveyed results, households in Xuan Phong showed higher experience of planting tree and knowledge on forest field than households in Bac Phong The reason might be people in Xuan Phong are belonged in remote area, thus they can get a greater amout of incentives from government for planting tree such as practical training, or materials and seedling than people

in other communes in Cao Phong district

4.2 Key drivers affecting tree planting decision of surveyed households

To clear understanding of the intended tree planting activities of households is required to estimate the likely impact of planting decision on various types of households In general, the households that intend to plant trees on their own land or their management have higher levels of resources, or higher awareness of tree regulations than households that do not intend to plant trees in the future Statistically tests between the households characteristics and their current planting tree will help to answer this question The results of correlation between household characteristic and tree planting decision is showed in Table 5 - Appendices

As the analysis method, I choose binary logistic regression with forward LR options, because LR options utilize the likelihood ratio test (Chi-square difference), inaddition forward selection will start with the constant-only model and adding variables one at a time in the order they are best until some cutoff level is reached which variables not in the model have a significance higher than 0.05

Input 11 drivers correlated with tree planting decision (Table 4.0.4) The overal percent correct before using logistic stepwise method is 53.6% and it has improved to 84.5%

We have summary model for key drivers affecting tree planting decision as table 4.3 below

Table 4.0.1 Model summary for key drivers affecting tree planting decision of surveyed

households

(P-value)

Participation in Forest Programs 4.779 1.177 119.038 000***

4.3 Key drivers affecting forest area will be planted of surveyed households

Similarly, the socioeconomic and household factors are analyzed by Pearson correlation to find the relationship between each variables with factor of forest area will be planted Based on Sig (2-tailed) of each correlation, we pick the smallest P value and get the table of correlation between households‟ characteristic and forest area will be planted as Table

Coefficients (B)

Standardised Coefficients (Beta)

t-value Significa

nt value)

(P-VIF

Participation on Forest

Dependent variable: Are will be planted by households

0,336 0,299 Note: *** p<0,001, ** p<0,05, *p<0,10 (two-sided)

There are 5 main variables are statistically significant in predicting the forest area will

be planted in the future The beta weights (Table 4.4) reveal that the attitude of farmer on tree