Master Thesis submitted within the UNIGIS MSc. programme at the Interfaculty Department of Geoinformatics - Z_GIS University of Salzburg, Austria under the provisions of UNIGIS joint study programme with Goa University, India
GIS-Based Multi Criteria Decision Analysis for Promoting Teak Plantation in Bokeo Province, Lao PDR
by
Anja Nicolay - Grosse Hokamp (GIS_102905)
A thesis submitted in partial fulfilment of the requirements of the award of the degree of Master of Science (Geographical Information Science & Systems) - MSc (GISc)
Advisor (s): Dr. Shahnawaz Interfaculty Department of Geoinformatics - Z_GIS University of Salzburg, Austria
Bangkok, Thailand - May 2014
ACKNOWLEDGEMENT
I would like to thank the following people for their valuable support:
Employees of RECOFTC in Bokeo and Bangkok.
Frank Siegmund, Ministry of Planning and Investment, Department of Planning (MPI/DOP), Vientiane, LAO PDR
Dr Shahnawaz, Interfaculty Department of Geoinformatics - Z_GIS, University of Salzburg, Austria,
Last but not least I would like to thank my family, especially my husband for all the support and patience during my studies.
i
SCIENCE PLEDGE
By my signature below, I certify that my thesis is entirely the result of my own work. I have cited all sources I have used in my thesis and I have always indicated their origin.
Bangkok, Thailand – 22. May, 2014
Place and Date Signature
ii
ABSTRACT
GIS based multi criteria decision analysis is a process that combines geographic data with value judgement based on decision makers’ preferences. This analysis combines two different scientific methods: (1) the multi-criteria decision analysis, and (2) the GIS-based analysis.
Multi-criteria decision making (MCDM) is a set of tools to structure the process of decision making in order to enable the consideration and evaluation of a high number of often conflicting criteria, alternatives and opinions. In questions of spatial decisions where multiple criteria need to be considered, Multi-Criteria Decision Analysis (MCDA) is often combined with GIS to enable mutual benefits.
In this study GIS-based MCDA was used to find the most suitable locations for teak plantations in the province Bokeo in Lao PDR. Forest resources, mainly timber have played and still play a central role in Lao’s economy and development. The promotion of teak
(Tectona grandis) plantations is seen as one means to generate and increase the income for the local rural population in the nation’s strive to reduce poverty through sustainable resource management. In Bokeo Province teak is an important element in the rural economy, generating between 25 and 55% of the annual household income.
The Analytical Hierarchy Process (AHP) as a technique for organizing and analysing complex decisions based on a hierarchical tree structure with three levels. AHP was used to structure and organize the research problem and to define the necessary criteria used in the
GIS. The decided on criteria were transformed into spatial layers and translated into suitability raster to be used in the weighted overlay analysis.
The technique of pairwise comparison was used to determine inherent weights for each criterion. Four stakeholder groups determined four different sets of weights that were used in the final step of Weighted Overlay Analysis.
iii
During discussion with stakeholders it was decided on the following list of criteria to be used in the suitability analysis: altitude, slope, aspect, soil type, distance to and influence of transportation infrastructure, distance to sawmills, accessibility of villages, availability of manpower, landuse and poverty of villages. Protected areas and areas close to the hydrologic network were excluded.
Based on the four different set of weights four suitability maps in raster format were created with ArcGIS software. The results show areas possible for teak plantations and their calculated suitability based on the chosen criteria.
The experience of this study shows that the structured method inherent in the MCDA is very suitable for decisions with a multitude of criteria. Additionally the study shows that the method is very suitable in a problem solving environment where different stakeholders have different views about significance of factors and criteria, their measurement and their combination. The structured approach of MCDA helps in clarifying matters and guides the discussion and decision process. Easy replication and the visualisation of the results are two distinct advantages from using a GIS environment in location analysis.
Key Words: MCDA, GIS, Lao PDR, Analytical Hierarchy Process, Pairwise Comparison,
Location Analysis, Weighted Overlay Analysis
iv
CONTENTS
Acknowledgement ...... i Science Pledge ...... ii Abstract...... iii Contents ...... v List of tables ...... vii List of figures ...... viii List of maps ...... ix List of abbreviations ...... x 1 Introduction ...... 1 1.1 Background ...... 1 1.2 Area of focus ...... 4 1.3 Aims and objectives ...... 9 1.4 Literature review ...... 10 2 Methodology ...... 14 2.1 Methodology ...... 14 2.2 Data ...... 21 2.3 Software ...... 22 3 Processes and results ...... 22 3.1 MCDA procedure ...... 22 3.1.1 Altitude ...... 23 3.1.2 Slope ...... 24 3.1.3 Aspect ...... 27 3.1.4 Soil ...... 30 3.1.5 Transportation infrastructure ...... 32 3.1.6 Distance to sawmills ...... 35 3.1.7 Accessibility ...... 37 3.1.8 Availability of manpower ...... 38 3.1.9 Protected areas ...... 41 3.1.10 Sustainability ...... 42 3.1.11 Land use ...... 42 3.1.12 Poverty reduction ...... 45 3.2 Pairwise comparison ...... 46 3.3 Modelbuilder (steps within ArcGIS) ...... 48 3.3.1 Reference systems ...... 48
v
3.3.2 Resolution ...... 48 3.3.3 GIS Data Preparation ...... 49 3.4 Results of the location analysis ...... 59 3.5 Validation of Weighted Overlay Analysis...... 65 4 Discussion ...... 66 4.1 Interpretation of results ...... 66 4.2 Discussion on suitability of methods used ...... 70 4.2.1 Multi-criteria decision analysis ...... 70 4.2.2 GIS...... 72 4.3 Recommendations ...... 73 5 Conclusions ...... 75 5.1 Summary ...... 75 5.2 Outlook ...... 76 References ...... 77 Annex ...... 81
vi
LIST OF TABLES
Table 1: Steps and activities ...... 16
Table 2: Meaning and description of scales used in pairwise comparison ...... 20
Table 3: Suitability of aspect ...... 28
Table 4: Soil suitability ...... 32
Table 5: Length and suitability of different transportation routes ...... 34
Table 6: Availability of manpower ...... 40
Table 7: Area sizes of landuse types ...... 44
Table 8: Pairwise comparison matrix ...... 47
Table 9: Weights resulting from pairwise comparison...... 47
Table 10: Analysis result summary ...... 64
Table 11: Analysis criteria ...... 81
Table 12: Data sources ...... 83
vii
LIST OF FIGURES
Figure 1: Hierarchy tree ...... 17
Figure 2: Influence of radiation and precipitation on aspect ...... 29
Figure 3: Major soils in km^2 ...... 31
Figure 4: Histogram of population density 2009, Bokeo Province, Lao PDR ...... 40
Figure 5: Steps in ArcMap to create elevation, slope and aspect raster ...... 51
Figure 6: Steps in ArcMap to create soil raster ...... 52
Figure 7: Steps in ArcMap to create raster for infrastructure ...... 53
Figure 8: Steps in ArcMap to create raster of distance to processing facilities ...... 54
Figure 9: Steps in ArcMap to create raster: manpower, accessibility and poverty ...... 55
Figure 10: Steps in ArcMap to create the mask ...... 56
Figure 11: Weighted Overlay: final step in ArcMap ...... 59
Figure 12: Comparison of result 2 (left) and result 1 (right) ...... 68
viii
LIST OF MAPS
Map 1: Overview of Bokeo Province, Lao PDR ...... 5 Map 2: Physiography in Bokeo Province, Lao PDR ...... 6 Map 3: Mean annual temperature in Bokeo Province, Lao PDR ...... 7 Map 4: Mean annual precipitation in Bokeo Province, Lao PDR ...... 7 Map 5: Suitability of altitudes in Bokeo Province, Lao PDR ...... 24 Map 6: Slopes in Bokeo Province, Lao PDR ...... 25 Map 7: Suitability of slopes in Bokeo Province, Lao PDR ...... 26 Map 8: Aspect in Bokeo Province, Lao PDR ...... 27 Map 9: Suitability of aspect in Bokeo Province, Lao PDR ...... 29 Map 10: Major soils in Bokeo Province, Lao PDR ...... 30 Map 11: Transportation network in Bokeo Province, Lao PDR ...... 33 Map 12: Location of sawmills in Bokeo Province, Lao PDR ...... 35 Map 13: Distance to sawmills in Bokeo Province, Lao PDR ...... 36 Map 14: Accessibility of village areas in Bokeo Province, Lao PDR ...... 38 Map 15: Population density in Bokeo Province, Lao PDR ...... 39 Map 16: Suitability based on population density in Bokeo Province, Lao PDR ...... 41 Map 17: Protected areas and protected forests in Bokeo Province, Lao PDR ...... 42 Map 18: Land use in Bokeo Province, Lao PDR, 2008 ...... 43 Map 19: Villages classified as poor in Bokeo Province, Lao PDR ...... 45 Map 20: Mask showing suitable areas for teak plantations in Bokeo Province, Lao PDR .... 57 Map 21: Analysis result (1) using weights determined by the ministry of Planning and Investment ...... 60 Map 22: Analysis result (2) using weights determined by RECOFTC Bokeo ...... 61 Map 23: Analysis result (3) using weights determined by RECOFTC Bangkok ...... 62 Map 24: Analysis result (4) using weights determined by the international expert ...... 63 Map 25: Validation of Weighted Overlay Analysis ...... 65 Map 26: Influence of the transportation infrastructure on analysis results ...... 67
ix
LIST OF ABBREVIATIONS
AHP Analytical Hierarchy Process
DEM Digital Elevation Model
DSS Decision Support System
GIS Geographical Information Science
MADM Multi-attribute decision making
MAF Ministry of Agriculture and Forestry, Lao PDR
MCDA Multi Criteria Decision Analysis
MoNRE Ministry of Natural Resources and Environment, Lao PDR
NGD National Geographic Department Lao PDR
PDPI Provincial Department of Planning and Investment, Bokeo, Lao PDR
RECOFTC Regional Community Forest Training Center for Asia and the Pacific
UNDP United Nations Development Program
UTM Universal Transverse Mercator
x
1 INTRODUCTION
1.1 BACKGROUND
Location is one of the central concepts in any Geographic information system. The location determines a place or position where something is or happens. The science of GIS is concerned with the location of features, their spatial relationships and correlations. (Longley et al. 2011)
One of the central analysis topics in GIS is location analysis. It uses the common attribute of position to combine different values in order to compute a specific value at every single location. In its definition local analysis is linked with raster data and the computation of values in an output raster based on a function of the input values. (Environmental Systems
Research Institute, Inc. 1995)
Finding the most suitable location is a location analysis in which carefully chosen favourable attributes are selected, combined and analysed to determine the most suitable location for a specific investment or action. In this context the location analysis is a decision support system.
Inherent in location analysis is the consideration of a multitude of different criteria. GIS has the capability of managing and analyse a high volume of diverse data. (Baban 2004)
Multi-criteria decision making (MCDM) is a set of tools to structure the process of decision making in order to enable the consideration and evaluation of a high number of often conflicting criteria, alternatives and opinions. As mentioned by Pomerol & Adam decision is always “a matter of compromise”. No straightforward decision exists. Instead there are always a number of more or less contradictory objectives involved. (Pomerol & Adam 2004)
Herbert Simon was the pioneer in the study of decision making. His work promoted the decision making process from a simple matter of choice into the field of Information
1
Systems. Today it is widely agreed that decision making is a scientific process with information and the consideration of possible alternatives at its centre.
One frequently used class of methods in the science of decision making are the multi-criteria decision making (MCDM) methods. They involve the evaluation of a set of alternatives on the basis of conflicting and often incommensurate criteria. Conventionally MCDM techniques have been aspatial and assumed a homogenous environment. (Malczewski 1999; Longley et al. 2011)
Another inherent characteristic of decision making is the multicriteria aspect. This is either based on the fact that several objectives are present or that more than one individual is involved in the decision making process. (Pomerol & Adam 2004)
Belton (2002) speaks of a multi-criteria framework in which there is no ‘right answer’ but only a consideration of different choices. The decision is based on criteria as a means or standard of judgement. Every decision requires the balancing of multiple criteria. Multi- criteria decision analysis (MCDA) is therefore an aid to decision making. (Belton 2002)
Malczewski (2006) defines criterion as a “standard of judgement or rule on the basis of which alternative decisions can be evaluated and ordered according to their desirability”.
Criteria are therefore factors that are important and influence the outcome of the decision.
In MCDA the term ‘criteria’ includes both ‘objectives’ and ‘attributes’ resulting in two different methods: multi-objective decision analysis (MODA) and a multi-attribute decision analysis
(MADA). It is important to distinguish between ‘objective’ and ‘attribute’. Hwang and Yoon
(1981) define objectives as a desired direction of change while attributes are defined as synonym to performance parameters, characteristics, or properties. An attribute or a set of attributes enables the evaluation of an objective. These definitions explain the difference between MODA and MADA. (Hwang & Yoon 1981)
2
Simon (in Malczewski 2006) structures the method of MCDA in three main stages: intelligence, design and choice. This places an emphasis on the identification and evaluation of relevant components, the organization, structuring and understanding the different components and on the communication of participants and stakeholders. The result is a deeper understanding of the nature of the decision. The decision itself is therefore a result of the process but not the main activity. (Malczewski 2006)
In questions of spatial decisions where multiple criteria need to be considered, MCDA is often combined with GIS to enable mutual benefits. Cowen describes GIS as “a decision support system involving the integration of spatially referenced data in a problem solving environment (Cowen in (Malczewski 2006)). Malczewski (2006) shows in his literature review that the application of MCDA combined with GIS is a frequently and increasingly applied method. Additionally the myriad of application domains confirm the success of this methodology. Urban planning, site selection for economic or environmental decisions, land use resource planning, and hazard and risk maps are only a few domains mentioned.
(Malczewski 2006; Effat & Hegazy 2013; Al-Hanbali 2011, p.-; Nyeko 2012; Onunkwo-
Akunne et al. 2012; Peng et al. 2012; van Westen & Damen 2013)
Huang et al. also confirm in their study that GIS-based MCDA was successfully used in an increasing number of environmental applications and problems solving analyses during the considered decade up to 2009. (Huang et al. 2011) Recent successful examples are described by Uribe et al. (2014), Arianoutsou et al. (2011) or Nas et al. (2009) and many others. (Nas et al. 2009; Arianoutsou et al. 2011; Uribe et al. 2014)
Additionally MCDM allows for weighting the criteria according to their importance. (Uribe et al. 2014)
3
1.2 AREA OF FOCUS
Lao People’s Democratic Republic (PDR) is a landlocked country in South-East Asia, the local heart of the Indo-Chinese Peninsula. The country is situated between 13°50’ and
22°30’ N and 100°4’ and 107°49’ E. The total area is 236,800 km 2. Bordering countries are
China and Myanmar in the North, Vietnam in the East, Cambodia in the South and Thailand in the West. (Lao PDR 2010)
Within the country four major geographic regions can be found: Upper Mekong, Upper
Annamite, Central Plain, Lower Mekong Basin. The Mekong River runs for a length of 1,898 km through the country, draining 80% of Lao PDR’s land area. Approximately 80% of the country is classified as mountainous or hilly. More than 30% of the country’s area has slopes steeper than 30%. (PAD Partnership 2003)
The nation is divided into seventeen provinces. Vientiane situated in the municipal province of Vientiane is the capital. In 2005 the census counted a population of 5.8 million people. In
2012 the population was estimated at 6.5 million people. On average the population density is 24 persons per km 2 which is the lowest density within South-East Asia.
Lao PDR belongs to the group of least developed countries. Together with Cambodia, Lao
PDR is ranked 138 by UNDP in the Human Development Index, placing the country in the
last ranks of medium human development. (Malik & United Nations Development
Programme 2013) With 33.9 % of the population living below the poverty line of 1.25$ per
day (Ravaillon et al. 2008), poverty reduction is one of the main development challenges
within the country. (UNDP 2013; Lao PDR 2010)
The Lao PDR is rich in forests and forest resources. Forest resources, mainly timber have
played and still play a central role in Lao’s economy. In 1998 for example forest products
accounted for 42% of the country’s foreign earning. (Lao PDR 2010) Next to timber water for
hydropower is the principal national resource. (Keonakhone 2006)
4
Bokeo Province is with an area of 6968 km 2 the smallest province of Lao PDR. It is situated
at the north-western border of the nation, between N 19° 47’ and N 20° 50’ latitude and E
100° 5’ and 101° 15’ longitude. The province is subdivided into five districts: Houayxay,
Tonpheung, Meung, Pha-oudom and Paktha with the capital Houayxay. The main economic
activity is mining and maize production. (IMF 2008)
MAP 1: OVERVIEW OF BOKEO PROVINCE, LAO PDR
Bokeo province is situated in the physiographic unit of the northern Highlands. Apart from the Mekong floodplains along the western border, the entire province has a rugged mountainous topography with an elevation between 500 and 2000m. Map 2 shows the general physiography of the region including the main rivers to give an impression of the area.
5
MAP 2: PHYSIOGRAPHY IN BOKEO PROVINCE, LAO PDR
Bokeo province has a dry subtropical climate with less than 2000mm annual precipitation.
Map 3 displays the mean annual temperature in Bokeo ranging from 18.2° to 25.1° Celsius.
Coldest months are December and January with lowest mean temperatures of 13° Celsius, hottest month is May with a maximum mean temperature above 28° Celsius. (Hijmans et al.
1965)
Map 4 shows the mean annual precipitation in Bokeo ranging from 1398 to 1686mm annually. The climate is strongly influenced by the annual south west monsoon cycle with a distinct wet season from April to October when up to 90% of the precipitation occurs.
Precipitation in these months is up to 400mm. The dry season from November to March has months with no rainfall at all. (PAD Partnership 2003; Hijmans et al. 1965)
6
MAP 3: MEAN ANNUAL TEMPERATURE IN BOKEO PROVINCE, LAO PDR
MAP 4: MEAN ANNUAL PRECIPITATION IN BOKEO PROVINCE, LAO PDR
7
In 2009 the population in Bokeo province was 156,173 people resulting in an average population density of 22.4 people per km 2. The rural population depend strongly on forests and forest products for their livelihood. The impacts of forest loss and degradation and the resulting reduced availability of forest products and environmental services like soil and water protection affect these poor rural communities most severely. (Senyavong 2010;
Roder et al. 1995)
Teak (Tectona grandis) plantations are one means to generate income for the local population in the nations strive to reduce poverty through sustainable resource management. Farmer-owned teak plantations were introduced by the French colonial regime as early as 1950 and expanded rapidly since 1998. The north of Laos with Luang Prabang as the commercial centre is the focus area of this development. In this area teak is an important element in the rural economy, generating between 25 and 55% of the annual household income. (Mohns & Laity 2010)
Teak has many advantages that make the plant ideal for plantations in Bokeo province. Teak is an indigenous plant in the northern Highlands of Lao PDR (Fogdestam & Galnander
2003). The northern highlands have a moist to dry sub-tropical climate with annual rainfalls between 1500-2000mm. Precipitation mainly occurs during the monsoon months April to
October and leads to a seasonal climate with a distinct dry season (PAD Partnership 2003).
This corresponds with the ideal conditions for the production of high quality wood as described by Kaosa-ard (Kaosa-ard in Regional Seminar on Teak et al. 1998).
Additionally teak is easy to manage, grows quickly and is fire tolerant. The timber is valued for its durability, strength, resistance to fungus and termites, little risk of splitting and warping during drying and its carving capability. This is reflected in the high market price that can be achieved by qualitative high teak.
8
1.3 AIMS AND OBJECTIVES
In the province Bokeo many communities rely on forests and forest products for their daily needs (Roder et al. 1995). The development aim of the government is to get away from the unsustainable slash and burn practices and to enable communities to use forests sustainably and efficiently to increase and stabilize their livelihoods (Mohns & Laity 2010).
To achieve this sustainable forest based production of teak timber and accompanying non- timber goods must be enabled and improved in such a way that smallholder can compete in markets locally as well as on the global level. (Hansen et al. 2005; Newby et al. 2010) Site suitability is seen as a basic question that has not yet been considered scientifically
(Manivong & Sophathilath 2007) in Bokeo province.
The aim of this study is to prove the suitability of the method of using a GIS-based
Multicriteria Decision Analysis to find the most suitable locations for teak plantations. The results of the study will additionally contribute to the development of the study area and will form the basis of decisions concerning the promotion of locations for teak plantations by the provincial government in Bokeo.
The main objective of this study is to identify areas most suitable for promoting teak plantations using a geographic information system. In order to achieve this, the following minor objectives were defined:
Identify, define and evaluate ecologic criteria for teak plantations based on
physiography and climatic conditions. Possible factors include:
o Slope,
o Aspect,
o Altitude, and
o Soil.
Identify, define and evaluate criteria related to the pursued policy of economic
development and poverty reduction
9
Identify, define and evaluate economic criteria like
o Transportation infrastructure
o Distances, and
o Accessibility
Identify, define and evaluate socio-economic criteria like
o Existing land use, and
o Manpower based on population.
Identify, define and evaluate criteria based on the National Planning Framework like
o Protected areas.
Transform and transfer the identified criteria into spatial parameters and attributes to
integrate them into the GIS environment.
1.4 LITERATURE REVIEW
Herbert Simon is regarded as the pioneer in decision making processes. Since the 1940’s he is seen as the most influential and important contemporary author in terms of organisational theory. Simon changed the viewpoint in organisations and management from the hierarchy in management to the concept of information flow within organisations and the manager as decision maker and the decision making and the resulting action as central part.
Hwang and Yoon define 1980 the term Multi-attribute decision making (MADM) as “making decisions in the presence of multiple, usually conflicting, criteria” (Hwang & Yoon 1981). In their book, they focus on a summary of MADM methods. They also describe decision making as a central concept present in common problems in everyday life. This includes spatial problems like the communal water resources development plan they mention as example.
Pomerol and Adam examined Simons legacy in 2004 and illustrated his impact on research carried out in the decision making area. In their text they cite Crozier, who describes Simon as the “father of the sciences of decisions” and Simon’s work as revolutionary for social
10 sciences and the topic of decision making (Crozier in (Pomerol & Adam 2004)). Pomerol and
Adam link Simon’s work to the current topic of Decision Support Systems (DSS) in which decisions and the resulting actions can-not be separated (Pomerol & Adam 2004).
In 2002 Belton summarized the current most common schools of thoughts within Multicriteria
Decision Analysis (MCDA) and in that created a comprehensive summary of the topics and aspects of decision making (Belton 2002). This summary aims at providing background and information to enable decision maker, students and researchers to use MCDA in an informed manner.
Fundamental concepts of the approach in MCDA are also described and taught in social sciences. Additionally supporting tools exists. An example is the workbook of Mabin &
Beattie that forms a practical guide to employ MCDA practices with the help of supporting software (Mabin & Beattie 2006).
For approximately 25 years the integration of GIS and MCDA started with increasing interest over the years since many spatial decisions involve a large number of different and often conflicting criteria. Joining GIS with the MCDA tools and procedures proved to be a beneficial combination for tackling decision problems. The synergetic effect of combining both techniques is described in several research studies and led to advancement in theoretical and applied research on combining GIS and MCDA (Malczewski 1999).
The success of combining MCDA with GIS is also explained by the fact, that MCDA provides a structured framework that helps in handling and understanding complex problems, and the relationships between different criteria. The same principal benefit to “facilitate decision makers’ learning about and understanding the problem faced” as stated by Belton (2002) is regarded as the main aim of MCDA.
Malczewski also reasons that the increase of research in GIS-MCDA is based on the increased recognition and perception of the decision analysis and support element within
11
GIS, the availability of low-cost and user friendly MCDA software and an integration of
MCDA techniques in GIS software products (Malczewski 2006).
Within MCDA Malczewski differentiates between two main schemes: multi-attribute decision analysis (MADA) and multi-objective decision analysis (MODA). MADA unite problems with a focus on attributes. They have per definition a limited number of alternatives and this is therefore considered as a selection process. MODA are described as a continuous design process where the best solution can be found within a range of possible solutions
(Malczewski 2006).
A further classification of MADA and MODA is concerned with the combination rule used in the analysis. One of the most common combination rules within MADA is the weighted summation that is also the combination rule used in weighted overlay analysis of many GIS software packages like ArcGIS (Malczewski 2006).
Another very common method within GIS-based MCDA is the analytical hierarchy process
(AHP) developed by Saaty. Due to the clear methodology AHP is very useful in complex decision analysis that contain a large number of attributes or alternatives (Saaty 1995).
GIS-based MCDA with slightly different methods was used in a large number of researches during the last decade. Location analysis and suitability analysis are the two very common research problems in which GIS-based MCDA was applied.
Nas et al. (2010) performed a multi-criteria evaluation combined with GIS. In their study a number of proposed land fill sites were evaluated concerning their suitability based on a limited number of evaluation criteria. Criteria were represented as layers within the GIS framework (Nas et al. 2009).
Similarly Babalola used GIS and Multi-criteria decision methods in his study of land suitability for land fill sites in Malaysia. Multiple data sets containing environmental and
12 policy factors were used to determine a risk free and environmentally friendly waste disposal site. (Babalola 2011)
Al Hanbali (2011) used the method of weighted linear combination to select the most suitable solid waste disposal site in Jordan. Similarly to Nas et al. and Babalola several criteria were considered in the research (Al-Hanbali 2011).
A number of diverse criteria were combined in the analysis of Arianoutsou (2011). Bio- indicators and geo-indicators were synthesized in a GIS to determine the post-fire resilience of forests in Greece. The different criteria included for example forest cover and relative species richness as well as fire history, parent soil material, and slope. (Arianoutsou et al.
2011)
Meng (2011) used in his study the analytical hierarchy process (AHP) to determine the relative importance of criteria. Meng emphasises that AHP is very helpful in understanding and solving complex problems. Within the AHP the complex problem gets decomposed into a hierarchy of elements. This helps in understanding the problem and simplifies the decision analysis by structuring. (Meng 2011)
Also Peng (2011) used AHP to determine the relative weights of criteria in the creation of potential hazard maps to be used in disaster prevention in Taiwan. Peng also shows the practical implementation of AHP in a questionnaire. The determined weights are then used in what Peng calls a map overlaying analysis. The description shows that the method is identical to the weighted overlay analysis, where weights are assigned to each criterion and the sum of these weighted criteria are calculated. (Peng et al. 2012)
Also in the agricultural context site suitability analysis are performed with the method of GIS- based MCDA. Das et al. (2014) performed a research on site suitability for pineapple and oranges in Meghalaya, India. Basis for the analysis were several layers consisting of evaluated soil criteria. For both fruits different suitability criteria were assigned to the attributes. An overlay analysis resulted in separate suitability maps. (Das & Sudhakar 2014)
13
Uribe et al. (2014) included stakeholder preferences in their research of land suitability.
Including stakeholders is getting more and more common in environmental decisions where
participation of stakeholders is a necessary or even compulsory precondition. Uribe et al.’s
approach is based on the definition and weighting of multiple criteria for evaluating land
suitability. The weights are defined according to preferences of stakeholders. (Uribe et al.
2014)
Teak is a very common species used in plantations to increase the household income in
rural Lao PDR. In 2005 Hansen (2005) described the current situation and best practices
concerning teak plantations in the area of Luang Prabang province in central Lao PDR
(Hansen et al. 2005). Also Keonakhone (2006) focuses in his assessment of the use of teak
plantations at landscape level on the area around Luang Prabang (Keonakhone 2006).
Newby et al. (2010) concentrates on northern Lao PDR in his study, while his main focus is
on Luang Prabang Province (Newby et al. 2010).
As far as the literature suggest Bokeo province was never focus of a suitability study for teak
plantations or any other agricultural or silvicultural usage. Nevertheless many statements
concerning teak plantations, especially the ecologic conditions and statements concerning
community based forest management or general policy can be transferred to the situation in
Bokeo province.
2 METHODOLOGY
2.1 METHODOLOGY
Typically multiple, conflicting and incommensurable evaluation criteria influence the decision and give rise to a large set of alternatives. In order to structure the decision making process to evaluate and prioritize alternative decisions a MCDA will be applied.
14
Location analysis is based on a multitude of criteria. The basic influencing factors comprise ecologic conditions, economic factors, socio-economic factors and the political and planning framework. Ecologic criteria are the topography with slope, aspect, altitude and soil conditions. Economic criteria are concerned with the transportation infrastructure and the relative location of sites to existing sawmills. Socio-economic criteria are related to population and land use. The political framework comprises existing conditions or limitations
(protected areas) and policy decisions.
In order to find the most sustainable and suitable locations for the promotion of teak plantations the ecological, economic, social and political criteria resulted from the MCDA will be combined in a weighted GIS location analysis.
Wade and Sommer (2006) define analysis as a “systematic examination of a problem or complex entity in order to provide new information from what is already known” (Wade &
Sommer 2006). Important is the systematic methodology that can be replicated. Equally important is the creation of new knowledge from available information.
GIS based multi criteria decision analysis is a process that combines geographic data with value judgement based on decision makers’ preferences. The analysis combines two different scientific methods: (1) the multi-criteria decision analysis, and (2) the GIS-based analysis.
Herbert Simon, the pioneer in studies of decision making (in (Uribe et al. 2014)) divides the process of decision making into three main stages: Intelligence, design and choice. These three stages correspond to the three key phases of MCDA: (1) Problem identification and structuring, (2) Model building and use and (3) Development of action plans (here: interpretation of results) (Belton 2002). Table 1 shows the sequence of steps in this study assigned to these three stages.
15
TABLE 1: STEPS AND ACTIVITIES
Sequences of steps and activities performed in this study. Stage Steps and Activities Tool Problem Define problem and main objective Analytical Hierarchy Process identification and Identify and evaluate relevant criteria Discussion with stakeholder and structuring literature work Define relevant criteria Analytical Hierarchy Process Assign attribute values of importance Analytical Hierarchy Process Assign weights to the criteria Pairwise Comparison Model building Create criteria layers GIS – ArcMap and use Create different constraint layers GIS – ArcMap Combine constraint layer = masks GIS – ArcMap Mask criteria layer GIS – ArcMap Transform every criteria layer to raster GIS – ArcMap Weighted analysis of raster layers GIS – ArcMap Interpretation of Interpretation of resulting raster layer results
Problem identification and structuring is one of the central themes in MCDA. Saaty’s
Analytical Hierarchy Process (AHP) is a structured technique for organizing and analysing complex decisions based on a hierarchical tree structure with three levels. The first step is to decompose the problem into a hierarchy that consists of all essential elements of the problem in question. At the top level the ultimate goal is places. The lowest level consists the specific elements of the problem, the criteria and attributes (Meng 2011). Saaty’s AHP results in a sound structure and understanding of the problem and enables decision making in an organised way (Saaty 2008).
Figure 1 show the hierarchical value tree of this study with the main objective at the top level, the criteria class at the second level and the actual criteria at the lowest level. This value tree is the result of literature work as well as discussions with consultants of the
Regional Community Forest Training Center for Asia and the Pacific (RECOFTC), GIS and forestry experts from the Ministry of Planning and Investment, Vientiane, Lao PDR.
16
FIGURE 1: HIERARCHY TREE
The first level contains the overall goal, namely the location analysis of finding the most suitable locations for farmer owned teak plantations in Bokeo, Lao PDR. The second level contains the categories of criteria. The relevant criteria that were identified can be sorted in the following four groups: ecological criteria, economic criteria, socio-economic criteria and the policy and national planning framework. Ecological criteria are the ecologic circumstances, like physiography and soil properties that are favourable for the commercial planting of teak. Economic criteria are factors related to the economic efficiency of teak plantations like distances to market facilities. Socio-economic criteria are related to the population and the current land use. Policy and national planning framework comprises mainly restrictions due to protected areas or considerations in order to promote a desired development.
17
Each criterion is discussed and assessed according to its attributes that influences teak plantations. From this assessment a scale of suitability is worked out. Afterwards the criteria are transformed into spatial attributes to be used in the GIS analysis process.
Ecologic criteria are based on the topographic conditions. The basic data set is a digital elevation model (DEM) that contains continuous elevation values over a topographic surface
(Wade & Sommer 2006). From this DEM the topographic criteria of slope and aspect can be derived. A second necessary data set contains major soils.
Economic criteria are based on the existing infrastructure. Layers of roads, rivers and other transportation networks are the basic source of information. Socio-economic criteria are based on the census data and on land use data.
Criteria within the political framework and policy are normally not directly represented by a single data set. Restrictions of protected area can be represented by one layer but policy decisions must be translated into a measurable spatial criterion. In this study the main political aim is poverty reduction. Areas classified as poor must therefore get preferential treatment. Poverty is measurable within the census data combined with location of the counted villages.
To enable the comparison of the different criteria and attributes, each criterion needs to be scored or standardized. This involves the conversion of original values into degrees of suitability. Scores do not necessarily have to be linear but depend on the criterion itself.
Sometimes reversing scales might be necessary as well. Standardization was done separately for each criterion explicitly in a subsequent step.
In the model building phase a layer was created for each criterion. The spatial attributes were reclassified into measures of suitability. During standardization it is important to distinguish between factors and constraints.
18
Factors are attributes that contribute to the location analysis and that possess a degree of suitability. The used standardized output values of factors range from 1 (least suitable) to 10
(most suitable). These standardized values are on a rational scale. The ratio scale has a reference point and the numbers within the scale are comparable. This means that for example 10 is twice as suitable as 5, and 4 is twice as suitable as 2. To comply with the program and to ensure comparability this evaluation system is used in reclassifying the different criterion layers. (Environmental Systems Research Institute, Inc. 2013)
Constraints are restrictions that exclude locations from the analysis. Output values are in a binary format either NoData (a “value” that excludes cells from the analysis) or 1 (suitable).
In GIS constraints are pre-emptive criteria that screen out location alternatives before they are evaluated. They correspond to masks that exclude locations. (van Westen & Damen
2013; Mabin & Beattie 2006)
The aim of standardization is to enable comparability of different criteria. The scale of the criterion must be transformed into the suitability scale. This transformation describes the relationship between criterion and suitability for the specific question, here the suitability of areas for teak plantation promotion. This relationship is not necessarily straight forward and depends on the criterion itself.
With each criterion the following question must be asked: how does the change of a criterion impact on suitability? For some criteria it is a linear relationship. An increase in the original value increases the suitability for the specific aim. An example is the availability of manpower: the denser the population, the more workers are available. Other criteria have an exponential relationship. And some call for an assignment of suitability values based on a general judgement of experts. Additionally a mix of these relationships is possible. The relationship between criterion and suitability is discussed for each criterion separately below.
It is important to consider one criterion at a time. Mixing criteria distorts the analysis and might result in an emphasis on a single criterion. For example stakeholders suggested to
19
assign a high value to sparsely populated areas in the criterion “availability of manpower” to
support these areas. Instead of integrating this policy measure into the existing criterion this
called for an additional criterion and therefore an additional layer describing the policy of
supporting poor villages.
The last step before the calculation is the assignment of weights to the different criterion
based on the relative importance for the location analysis. To determine the relative
importance of each criterion a pairwise comparison was performed.
In the process of pairwise comparison all unique pairs are compared directly with each other
with the help of a matrix. Saathy’s weights as described in words in table 2 are assigned to
each combination in the matrix. The weighted values are then obtained by adding the values
in each row for each criterion; this sum is then divided by the total sum of all rows. (Saaty
2008)
TABLE 2: MEANING AND DESCRIPTION OF SCALES USED IN PAIRWISE COMPARISON
Scale Definition Description 1 Equally important The contribution of the two factors are equally important 3 Slightly important Experiences and judgement slightly tend to certain factor 5 Quite important Experiences and judgement strongly tend to certain factor 7 Extremely important Experiences and judgement extremely strongly tend to certain factor 9 Absolutely important There is sufficient evidence for absolutely tending to certain factor 2, 4, 6, 8 The median between In between two judgements two neighbouring scales (Saaty in Peng et al. 2012)
For the weighted overlay analysis process, the vector data was generally transformed into raster data by using the ‘polygon to raster’ - tool in ArcMap. The different raster were first reclassified according to the standardised suitability values for each criterion. In case of pre- emptive criteria a mask was created. The detailed processes for the single criterion are described in the chapter “Processes and Results”.
20
The final suitability map was created by using the raster calculation tool “weighted overlay”.
Resulting output values range from 1 to 10 similar to the input values. This is a function of
the calculation method that multiplied the suitability values with each corresponding weight
and then added the values of the different raster layers.
2.2 DATA
Lao PDR is collecting, preparing and supplying digital geographic data on a national scale.
Most of the data on national and province level was provided by the National Geographic
Department (NGD), Lao PDR. Data on district level was mainly provided by the Ministry of
Agriculture and Forestry (MAF) and their agricultural census of 2011. Data concerning the protected areas was provided by the Ministry of Natural Resources and Environment
(MoNRE), Lao PDR. Detailed census data was provided by the Provincial Department of
Planning and Investment (PDPI) of Bokeo province. The Center for Development and
Environment (CDE), SWISS Development, Lao PDR provided the data about main roads on all levels.
In general the timeliness of the data used was very good. Most of the provided shapefiles were from 2011. Road data was more recent with shapefiles from 2013. Unfortunately land use data was only available of 2008 and census data in English of 2009.
Despite the availability of GIS data and the mission of the government of Lao PDR to cover the entire country with GIS data (shapefiles), this collected data is up to now rarely used in any sophisticated spatial analysis. The use of GIS-based decision making tools to generate a map of suitable areas has so far not been tried in this region.
During the preparation of the data several small inconsistencies in census data like double allocation to the same village code were discovered during joining the census data with the shapefile of the villages in the 5 districts. Automatic cross-checking with the name of the villages was impossible due to the phonetic character of the Lao language. The English
21
names have different spellings in the different data sets. Correcting these inconsistencies by
hand was necessary.
The digital elevation model (DEM) was provided by the National Geographic Department
(NGD), Lao PDR. The DEM was developed in 2003 within a development project carried out
and funded by a Japanese organisation. Digitized contour lines from topographic maps with
a scale of 1:100,000 form the basis of the developed DEM.
A concise list of the different data sets used including their source and description is
provided in the Annex.
2.3 SOFTWARE
For all GIS operations ArcGIS 10.1 including the Spatial Analyst Extension was used.
3 PROCESSES AND RESULTS
3.1 MCDA PROCEDURE
The main partner during the procedure of finding and deciding on the criteria for a suitability analysis for teak plantations in Bokeo province were the employees of the Regional
Community Forest Training Center for Asia and the Pacific (RECOFTC). In discussions the hierarchy tree as displayed in figure 1 was worked out.
Literature review and further discussions led to the assessment of criteria and the assignment of suitability values for each criterion as shown in table 10 of the Annex.
Standardized suitability values range from 1 to 10 with 1 indicating a low suitability and 10 a high. The scale is a rational scale. As a general procedure several suitability values were assigned, the in between values were then calculated.
22
3.1.1 ALTITUDE
Altitude is defined as “the height or vertical elevation of a point above a reference surface”
(Wade & Sommer 2006). In other words altitude is the height of an area above sea level.
Map 2 shows the topography of Bokeo province in a natural colour scheme combined with hillshading to give an impression of the topography and altitudes in the study area.
Altitude influences the plant community. The limiting factor concerning altitude is the temperature. Teak is neither frost-resistant nor frost-tolerant. Therefore the lowest temperature limit is +2° Celsius for teak.
The most suitable altitude for teak plants in the relevant latitudes of 12 to 23 degrees North is according to Hansen below 700 to 900m (Hansen et al. 2005). The inability of teak plants, especially of seedlings and saplings to survive temperatures below +2° Celsius is the main limiting factor concerning altitude (Regional Seminar on Teak et al. 1998). Experiences shared during the AHP discussion process show that the altitude limit for teak is approximately 900m and the most suitable range lies between 400 and 700m. Above 700m suitability reduces linear with higher altitude. Below 400 m the suitability decreases at a lesser rate than above 700m. This was incorporated into the assignment of standardized suitability values.
Based on the discussion the following reclassification was decided:
Altitude 0 - 200 - suitability value 8
Altitude 200 – 400 - suitability value 9
Altitude 400 – 700 - suitability value 10
Altitude 700 – 800 - suitability value 8
Altitude 800 – 900 - suitability value 5
Altitude > 900m - suitability value ‘NoData’
Map 5 shows these altitude classes.
23
MAP 5: SUITABILITY OF ALTITUDES IN BOKEO PROVINCE, LAO PDR
3.1.2 SLOPE
ESRI defines slope as the incline or steepness of a surface. (Environmental Systems
Research Institute, Inc. 1995). ESRI uses two different units of measurement: degree and
percent. Degree slope is the angle between the surface and the horizontal plane. Values
range from 0 to 90 degree. Percent slope is the ratio between change in elevation (rise) to
the horizontal distance travelled (run) multiplied by 100. Values range from 0 to infinite.
Longley et al. (Longley et al. 2011) define a third unit of measurement, the ratio between the
elevation and the actual distance travelled with resulting slope values between 0 and 1.
Therefore it is important to know which one is used.
In ArcGIS software slope is defined and calculated as the rate of maximum change in z-
values from each cell. Important are the units in the data set. If the unit of the z-values is
different to the ground units (x and y) a z-factor needs to be defined.
24
The unit for x and y of the DEM in this study is similar to the unit of the elevation value:
meter. No z-value needs to be considered. Map 6 shows the slopes in Bokeo province.
MAP 6: SLOPES IN BOKEO PROVINCE, LAO PDR
The slope map was created with the spatial analysis tool “Slope”. Slope values range from
0° to 88.08°. Most authors in forestry literature and practice use the unit of percent to describe slopes. The description of the discussion of this criterion below follows this practice.
To calculate and create the suitability map from slopes the percentages were transformed into degree.
In timber harvesting slope is one of the main limiting factors. Greulich et al. (Greulich et al.
1985) define common slope classes in timber harvesting with < 30%, 30 – 70% and >70%.
However this number is given for the North-American context. During discussions it was discovered that the maximum usable slope is 50% and that slopes below 20% are the most
25
easy to cultivate. Additionally there was agreement that the relationship between slope and
suitability is not linear but that suitability decreases faster in steeper slopes.
Another aspect with slope is, that the lesser the slope, the higher the moisture content of the
slope. Therefore lesser slopes are more suitable.
Based on the above discussion the following classification is chosen:
Slope < 20% (< 11.31°) - very suitable - suitability value 10
Slope 20 – 30 % (11.31 – 16.70°) - suitable - suitability value 9
Slope 30 – 40% (16.7 – 21.8°) - moderately suitable - suitability value 7
Slope 40 – 50 % (21.8 – 26.57°) - less suitable - suitability value 5
Slope > 50% (> 26.57°) - unsuitable - suitability value ‘NoData’
Map 7 shows the resulting suitability classes based on the above discussed classification of
slopes.
MAP 7: SUITABILITY OF SLOPES IN BOKEO PROVINCE, LAO PDR
26
3.1.3 ASPECT
Aspect is defined as “the compass direction that a topographic slope faces” (Environmental
Systems Research Institute, Inc. 1995). The measurement unit is ‘degrees from North’. Map
8 displays the aspect map of the study area.
MAP 8: ASPECT IN BOKEO PROVINCE, LAO PDR
Micro climate conditions are influenced by the aspect of a surface. The direction of the slope influences the microclimate of an area. A very conspicuous example is viniculture in northern
Europe where the preferred hillside location for growing wine is south facing. As visible in this example, the latitude of the area of interest is important.
Bokeo province is situated in the northern hemisphere between N 19° 47’ and N 20° 50’ latitude. The sun is predominantly shining from the South throughout most of the year apart from one to two months around the summer solstice in June. This results in a sunnier but also drier microclimate on the southern facing slopes. Teak is a pioneer species and needs
27
light conditions. However like in all plantations there is an increased danger to the
unprotected soil. These two factors were considered in the assessment of the aspect
conditions related to radiation.
Additionally the south-west monsoon influences the microclimate. South-west facing slopes
are directly exposed to the heavy rainfalls while north-east facing slopes are situated in the
precipitation shadow and receive less rainfall. The slopes directly exposed to sun and
monsoon rain are extremely prone to erosion and degradation.
Resulting from the position towards the sun, the east and west facing slopes are the most
suitable with less danger of drought and a much better soil moisture condition. Concerning
precipitation, the south-west facing slopes are less suitable due to high erosion susceptibility
and the north-west facing slopes due to their position in the rain shadow. The influence of
these two factors is displayed in figure 2.
The two factors were considered separately using a simple classification with three classes:
0 = less suitable, 1 = moderately suitable, 2 = suitable; Locations in between received mean
values. The values of the two factors were combined resulting in the combined suitability.
Finally this was transformed into the standardized value scheme.
Resulting from this discussion it was decided on the following classification:
TABLE 3: SUITABILITY OF ASPECT
Compass Aspect in Suitability Suitability Combined Standardized direction degree resulting from resulting from suitability value radiation precipitation North 0 – 22.5 and 1 1.5 2.5 9 337.5 - 360 North-East 22.5 – 67.5 1.5 2 3.5 10 East 67.5 – 112.5 2 1.5 3.5 10 South-East 112.5 – 157.5 1 1 2 8 South 157.5 – 202.5 0 0.5 0.5 6 South-West 202.5 – 247.5 1 0 1 7 West 247.5 – 292.5 2 0.5 2.5 9 North-West 292.5 – 337.5 1.5 1 2.5 9 Flat areas -1 10
28
FIGURE 2: INFLUENCE OF RADIATION AND PRECIPITATION ON ASPECT
MAP 9: SUITABILITY OF ASPECT IN BOKEO PROVINCE, LAO PDR
Map 9 shows the suitability categories resulting from the aspect as described above.
29
3.1.4 SOIL
Teak can grow on a variety of soils. The quality of the timber however depends on different soil properties. Keonakhone describes suitable soils as deep, well-drained and fertile. The range of soil pH in teak forests is with 5.0 to 8.0 very wide although the optimal soil pH is between 6.5 and 7.5 (Regional Seminar on Teak et al. 1998). Another important factor is high calcium content since Calcium deficiency may result in stunted growth. (Keonakhone
2006)
The soil map of Bokeo province is displayed in Map 10. Figure 3 shows a summary of existing major soil types. Acrisols is the most common major soil type in Bokeo province and cover 7686 km 2. Cambisols are the second most common covering an area of 2339 km 2.
The existing soils in the region were assessed concerning their suitability for teak
plantations. The following paragraphs outline this assessment.
MAP 10: MAJOR SOILS IN BOKEO PROVINCE, LAO PDR
30
The most common soils in Bokeo province are Acrisols. Acrisols are strongly weathered acid
soils with a low level of plant nutrients. They are not very productive and perform best under
acidity-tolerant crops such as pineapple, cashew, oil palm or rubber. Additionally they are
prone to erosion. These limited soil resources are the main problem for agricultural use. It is
suggested that the best way to use them is the very common slash and burn agriculture with
long fallow periods (Driessen et al. 2001)
With the general acidity, the low nutrient content and the susceptibility to erosion, these soils
are only moderately suitable for teak plantations.
FIGURE 3: MAJOR SOILS IN KM^2
Cambisols are soils with incipient soil formation. They can be found on a variety of base material. In the humid tropics and also in Bokeo province dystric and ferallic subtypes are predominant. Despite being poor in nutrients Cambisols are still richer than Acrisols or
Ferrasols. Suitability for teak is not very good but better than Agrisols or Ferrasols.
Luvisols are soils with favourable physical properties. Their good internal drainage, moderate state of weathering and high base saturation makes Luvisols potentially suitable
31
for a wide range of cultural usage. Gleyic and Ferric subtypes are less fertile but in general
Luvisols are very suitable for teak plantations (and therefore most probably in competition
with other agricultural uses).
Lixisols are strongly weathered soils with low level of available nutrients and low nutrient
reserves. Their moisture holding capacity is slightly better than with Acrisols. Additionally
their higher soil pH is more favourable for teak plantations. Since Lixisols are prone to
erosion perennial crops are preferred on these soils.
Fluvisols are young soils on alluvial deposits. Since Fluvisols are found in the floodplain of
major rivers, they are flooded periodically. This makes them unsuitable for teak plantations.
Leptosols are either shallow soils on hard rock or deeper soils on gravelly base material.
They are highly calcareous which makes them in general suitable for teak plantations.
Leptosols are prone to erosion and are best kept under forest. With a sound management
these soils are the most suitable for teak plantations in Bokeo province.
Table 4 shows the resulting classification of existing soils based on the discussion above.
Included is the classification “Water” that is present in the soil layer but obviously not suitable
for teak plantations.
TABLE 4: SOIL SUITABILITY
Major soil Suitability for teak Suitability value Acrisol Moderately suitable 6 Cambisol Moderately suitable 7 Luvisol Highly suitable 10 Lixisol Moderately suitable 8 Fluvisol Unsuitable NoData Leptosol Highly suitable 9 Water Unsuitable NoData
3.1.5 TRANSPORTATION INFRASTRUCTURE
There are three possible transportation paths for timber: roads, ferry routes and waterways.
The criterion ‘Transportation infrastructure’ encloses the distance to any of these
32
transportation paths. Distances to roads, to ferry routes and to waterways were combined
into one layer.
Timber transportation in Bokeo province on water is limited to the major rivers. Map 11
shows the transportation network in Bokeo.
MAP 11: TRANSPORTATION NETWORK IN BOKEO PROVINCE, LAO PDR
The road network in the road layer is classified into 7 different road types. Table 5 summarises the road types present:
33
TABLE 5: LENGTH AND SUITABILITY OF DIFFERENT TRANSPORTATION ROUTES
Transportation Suitability value of areas within 3000m distance of routes Length in km these routes Unpaved Road 1288.675 9 Trail 306.724 8 Collector Road 906.258 10 Interstate 806.18 10 Residential Road 58.085 10 Alley or Driveway 26.932 10 Arterial Road 3.914 10
In the discussion it was decided to assign a lesser suitability to areas close to roads that have an unpaved surface (unpaved road and trail), since they are more difficult or even impossible to travel on during the rainy season. Additionally ferry routes and major rivers received a suitability value of 10.
According to Mohns & Laity studies show that manual extraction of teak harvest is the predominant practice in farmer-owned teak plantations. This limits the financial and economic viability to a maximum distance of 500m to transportation infrastructure (Mohns &
Laity 2010).
At the moment there is an intensive research and educational campaign happening in Bokeo province with the aim to increase the transportation distance of teak. This increases the economically viable distance of teak plantations to transportation infrastructure to a maximum of up to 3000m.
Areas with a greater distance than 3000m to the transportation network are masked out with a suitability value of ‘NoData’. Areas within the 3000m buffer are divided into three classes according to the surface and type of road. Suitable areas are situated around paved roads like ‘Collector Road’, ‘Interstate’, ‘Residential Road’, ‘Alley or Driveway’ or ‘Arterial Road’. A lesser suitability is assigned to ‘unpaved road’ and to ‘trails’. Suitability values are displayed in table 5 above.
34
3.1.6 DISTANCE TO SAWMILLS
The main customers for teak timber are sawmills that prepare timber for the export market.
Within Bokeo province there is only a limited number of sawmills, most of them located in
Houay Xai district as displayed in map 12. Additionally timber is processed on the opposite side of the Mekong River in Thailand. The distance to these markets are one factor that needs to be considered.
MAP 12: LOCATION OF SAWMILLS IN BOKEO PROVINCE, LAO PDR
The distance to sawmills is closely linked to the transportation network. Distances should be measured along the transportation network to determine the suitability of locations. However during the analysis it was discovered that the road network is of poor quality. Gaps in the connectivity of the roads make this layer unusable for network analysis. To approximate the distance to the sawmills, an Euclidean distance layer was created
35
Euclidean distance or ‘distance as the crow flies’ is defined as the straight-line distance
between two points (Wade & Sommer 2006). ArcGIS calculates the Euclidean distance
within a raster based on a source layer. Map 13 displays the calculated distance layer. The
distances are classified into 10 classes with regular intervals according to the 10 suitability
classes.
For the distance to suitability a linear relationship was assumed. The calculated values of 0
to 70 km were divided equally on the suitability value range resulting in the reclassification
scheme displayed in the maps legend.
MAP 13: DISTANCE TO SAWMILLS IN BOKEO PROVINCE, LAO PDR
36
3.1.7 ACCESSIBILITY
The criterion of accessibility to the villages and therefore to manpower came up during discussing the criterion infrastructure. Some villages are difficult or even impossible to access during the rainy season due to bad road conditions. These villages are from an economic viewpoint less suitable for teak plantations.
The necessary data about accessibility is found in the census data of Bokeo province from
2009. Each village was assessed according the state of the access road. Resulting categories are “no road access”, “accessible during dry season only” and “accessible all year round during both seasons”.
Joining the census data tables with the village point layer enables a display of accessibility to villages. Since in Bokeo province the land area is continuously assigned to one village, the suitability was chosen to represent the village land. Map 14 shows the accessibility based on the census data.
It was decided on a suitability value of 10 for all-year accessible villages and a suitability value of 9 for possible disruptions of accessibility in the rainy season. Villages with no road access were classified with a suitability value of 5.
37
MAP 14: ACCESSIBILITY OF VILLAGE AREAS IN BOKEO PROVINCE, LAO PDR
3.1.8 AVAILABILITY OF MANPOWER
Manpower is defined as the labour force available for a specific task (Anon 2014). Manpower is therefore closely linked to the overall population. In forestry in Lao PDR the criterion of availability of manpower is influenced by several factors including total population, adult population and male population. The discussion of this criterion revealed that within the decades of growing teak coming up tending work like planting, weeding and cutting are performed by available workers within families disregarding age or sex. Only harvesting and the final cutting of timber is performed by groups of male population. It was therefore decided to use the total population as reference value for the criterion of availability of manpower.
38
Map 15 shows the population density within Bokeo province. The polygons display the
current borders of each village territory. The census data makes use of these boundaries as
basis for their data collection.
MAP 15: POPULATION DENSITY IN BOKEO PROVINCE, LAO PDR
In general a dense population suggests a high availability of manpower and a sparse population lower availability of manpower. Another factor that influences the availability of manpower is the available time of farmers. Since there is no data available on this aspect it was decided to translate population density directly into a scale of suitability.
The histogram of population density in Bokeo province is displayed in figure 4. The graph shows the population density on the x-axis and the number of villages related to the densities on the y-axis. The graph is extremely skewed to the left with several outlier values on the right. Only one village has a population density higher than 2000 ppl per km 2. And
39
only three villages have densities higher than 1000 ppl per km 2. More than 80 villages have
densities below 15 ppl per km 2.
FIGURE 4: HISTOGRAM OF POPULATION DENSITY 2009, BOKEO PROVINCE, LAO PDR
Based on these observations it was decided that natural breaks represent the inherent structure of the population densities most accurate. According to the range of the suitability values 10 classes were created using the classification along natural breaks within the dataset. Table 6 shows the resulting classes and the assigned suitability values used. Map
16 shows the distribution of the assigned suitability value based on this criterion.
TABLE 6: AVAILABILITY OF MANPOWER
Population density in people per km^2 Suitability value 0 – 14.90 1 14.90 – 28.94 2 28.94 – 53.23 3 53.23 – 92.44 4 92.44 -156.5 5 156.5 – 289.5 6 289.5 – 491.7 7 491.7 – 774.1 8 774.1 – 1173 9 1173 – 2574 10
40
MAP 16: SUITABILITY BASED ON POPULATION DENSITY IN BOKEO PROVINCE, LAO PDR
3.1.9 PROTECTED AREAS
Lao PDR is one of the few least developed countries that established an extensive set of protected areas as an integrated system. Protected areas cover approximately 21 % of the land area of the entire country.
In Bokeo province two types of protection area exist: (1) the national protection area Nam
Kan in the north-east and (2) protected forest areas on national and district level. Together they cover an area of 2885.7 km 2 or 41.41% of the province’s area.
Concerning the use in GIS, the protection areas form a pre-emptive criterion since plantation is a restricted use in these protected landscapes. Accordingly a value of “NoData” was assigned to the protected areas while all other areas received a value of 1.
41
MAP 17: PROTECTED AREAS AND PROTECTED FORESTS IN BOKEO PROVINCE, LAO PDR
3.1.10 SUSTAINABILITY
During the discussion of the different criteria experts suggested that to ensure sustainability
and protection of the environment the area around rivers and creeks should be restricted.
This pre-emptive criterion was incorporated into the GIS system by creating a buffer of 100m
from the sides of a river and 50m from the sides of a creek (Food and Agriculture
Organization of the United Nations (FAO) 1998).
3.1.11 LAND USE
Bokeo province is in general a rural province characterised by self-reliant subsistence
farming. Even district centres have populations well below 5000 people per km 2. Larger
settlements are mainly a conglomeration of neighbouring villages (Ireson 1995).
42
In Bokeo province the land use type is closely related to forests and agricultural activities.
Map 18 shows a map of the land use classification found in Bokeo province. The data
displays the land use classes of 2008. Conspicuous is the absence of settlements and
roads. The data set was provided from the National Geographic Department (NGD), Lao
PDR in Vientiane. After inquiry the exclusion of settlements and roads was explained with
the availability of separate datasets on these topics that can be overlayed on the land use
data layer. For this analysis the exclusion of roads and settlements was discussed during the
processes and judged as unproblematic. The low population density and the rural character
of the entire area were two of the most important reasons in this decision.
MAP 18: LAND USE IN BOKEO PROVINCE, LAO PDR, 2008
43
TABLE 7: AREA SIZES OF LANDUSE TYPES
Land use type Area in km^2 Total Current Forest Dry Evergreen Forest 46.93 Mixed Deciduous Forest 3223.33 Forest Plantation 1.28 3271.54 Potential Forest Bamboo 773.43 Un-stocked Forest Area 2471.53 Ray 208.53 3453.49 Permanent agricultural land Rice Paddy 164.88 Agricultural Plantation 7 171.88 Other non-forest land Grassland 3.85 3.86 Other Areas Water Bodies 78.41 78.41
As visible in table 7 and in map 18 most of the area in Bokeo province is covered by forests.
Bokeo is a rural province with no large settlements. Only 171 km 2 are covered by permanent
agricultural land. The predominant slash and burn agriculture is practiced in all areas around
settlements. There is quite an amount of change in landuse around villages. This has no
influence on the suitability for teak plantations.
Table 7 suggests a classification concerning the suitability for teak plantations. Water bodies
and Grassland will be excluded from the analysis. Grassland areas possess an inherent
unsuitability based on ecological conditions for teak plantations. This could be shallow soils,
water saturation or any other limiting factor that prevents the growth of trees.
Similarly it was decided to exclude permanent agricultural land. Bokeo is a rural province
with subsistence farming as the most common form or agriculture. Decreasing the
agricultural area would put unnecessary stress on the rural population. All other areas are
classified as suitable for teak plantations. A mask is created from the suitable landuse
categories.
44
3.1.12 POVERTY REDUCTION
The measure to promote teak plantations in Bokeo province is an attempt to reduce poverty
by creating income opportunities for the local communities and farmers. This fact leads to
the incorporation of another criterion: the poverty of the villages.
The census of Lao PDR evaluated the poverty of the villages on the basis of the following
criteria:
• Number of poor households in the village is higher than 51%,
• The village does not have health service
• The village does not have education service
• The village does not have water/gravity use
• The village does not have road access.
MAP 19: VILLAGES CLASSIFIED AS POOR IN BOKEO PROVINCE, LAO PDR
45
Map 19 shows the villages classified as poor in the census of 2009. To include this criterion
a suitability value of 10 was assigned to poor villages and a suitability value of 5 was
assigned to non-poor villages.
3.2 PAIRWISE COMPARISON
The next step after deciding on the criteria and their attributes was to assign weights to each criterion according to its relative importance. Weight is defined as a value that indicates the relative importance value of each criterion for a particular calculation. The larger the weight, the higher is the influence of this particular variable on the outcome (Environmental Systems
Research Institute, Inc. 1995). Weights are always numbers between 0 and 1. The sum of weights within a group equals 1. In the GIS analysis these weights will be used in the weighted overlay analysis. (Uribe et al. 2014)
It is very difficult to judge several factors simultaneously, especially with a number of varying criteria as discussed above. For this purpose Saaty developed the method of pairwise comparison within the AHP. The general process was described above. Table 2 shows
Saaty’s scale that was used in the comparison.
The matrix used in the pairwise comparison process was developed on basis of the discussed and decided on criteria. Table 8 shows this matrix. Together with Saaty’s scale and introductions this matrix was distributed to the groups of stakeholders who discussed and decided on the ranking of criteria. The weights for the analysis were calculated from the completed matrix.
46
TABLE 8: PAIRWISE COMPARISON MATRIX
Pairwise Comparison Matrix Distance Availability Transportation to Accessibility Poverty Elevation Slope Aspect Soil of network processing of villages reduction manpower facilities Elevation 1 Slope 1 Aspect 1 Soil 1
Transportation 1 network
Distance to processing 1 facilities
Accessibility of 1 villages Availability of 1 manpower Poverty 1 reduction
This matrix including a letter with instructions and explanations about the procedure was distributed to 4 groups of stakeholder. The pairwise comparison delivered no indisputable result. Table 9 shows the 4 different weight schemes calculated from the results of the pairwise comparison that was performed by the different groups of stakeholders.
TABLE 9: WEIGHTS RESULTING FROM PAIRWISE COMPARISON
Criteria Weights Ministry of RECOFTC Bokeo RECOFTC Independent Planning and Bangkok expert Investment, Vientiane Elevation 22 5 3 4 Slope 23 13 10 11 Aspect 14 2 1 2 Soil 19 26 16 18 Transportation 10 25 18 14 infrastructure Distance to 4 12 8 4 processing facilities Accessibility of 2 5 19 14 villages Availability of 4 5 4 9 manpower Poverty reduction 2 7 21 24
47
3.3 MODELBUILDER (STEPS WITHIN ARC GIS)
Esri’s ArcGIS software was used in the entire analysis. ArcMap 10.1 and the extension
“Spatial Analyst” formed the basis of the GIS steps and analysis described below.
3.3.1 REFERENCE SYSTEMS
The datum is defined by ESRI as ”The reference specifications of a measurement system, usually a system of coordinate positions on a surface (a horizontal datum) or heights above or below a surface (a vertical datum).” (Environmental Systems Research Institute, Inc.
1995) The provided data sets possess two different reference systems:
Lao97_UTM_zone_48 and WGS_1984_UTM_Zone_48N. UTM is the acronym for the projected coordinate system Universal Transverse Mercator that divides the world into 60 north and south zones 6 degrees wide. Both systems are based on UTM projection and the same zone 48.
In ArcMap the environment settings include the definition of a reference system. The projection of the digital elevation model (DEM) Lao97_UTM_zone48 was used. Similar this reference system was used as the output coordinate system. Setting the output coordinate system automatically transforms the results of tools used into this specified projection.
3.3.2 RESOLUTION
Resolution is defined as the detail with which a map depicts the location and shape of geographic features (Environmental Systems Research Institute, Inc. 1995). In a raster the resolution corresponds to the dimension represented by each cell.
The specifications of the DEM were used as basis raster for any raster analysis performed.
This DEM was made available by the National Geographic Department (NGD), Lao PDR. It consists of 3851 rows and 3967 columns. The cells in this raster have a size of
30.79634901m x 30.79634901m. One cell covers therefore an area of 948.4151 m2. The
projection is a Transverse Mercator projection with a national datum of ‘D Lao National
48
Datum 1997’. Resampling the raster to a smaller cell size does not increase the detail of the map area.
Dieters (2014) gives the number of 1.4ha for woodlot sizes in farmer owned teak plantations in Luang Prabang (Dieters 2014). But he also concedes that this number is positively skewed due to a number of large plantations. RECOFTC in Bokeo calculated the average size of small-scale woodlots with 0.3 ha but also mentions a positive skew with 60% of plantations less than 0.3ha (Bianchi 2014). This number justifies the resolution which is 1/3 of this average size.
3.3.3 GIS DATA PREPARATION
The basis of this GIS-based suitability analysis is a weighted overlay analysis. Overlay analysis is a “technique for combining multiple raster by applying a common measurement scale of values to each raster, weighting each according to its importance, and adding them together to create an integrated analysis.” (Environmental Systems Research Institute, Inc.
1995) The precondition is that there exists one raster for each criterion discussed and decided on. The preparation of each raster is described below.
The step of creating a raster from the shapefiles to be used in the weighted overlay was performed with the tool “polygon to raster” and the environmental settings above. Each shapefile was projected to the Lao97-coordinate system before the “polygon to raster” tool was run to guarantee congruence of the different raster. Additionally the spatial references and the resolution of the digital elevation model bokeo_dem_utm_wgs84_z48n.tif that was provided by the NGD of Lao PDR, were used transforming polygon shapefiles into raster with a cell size of 30.8m x 30.8m based on the following coordinate system:
Lao97_UTM_zone_48.
The Spatial Analyst tool “Reclassify” was run on each raster to assign the agreed on suitability value to the attribute values. This was also necessary since the weighted overlay
49 tool only allows integer values in the input value field. The settings within the weighted overlay tool are simplified with this preceding reclassification.
In the following paragraphs the preparation of each raster layer in ArcMap will be described.
Each description is accompanied with a diagram of the steps performed in ArcMap. The symbolisation is similar to the one used in ArcGIS’ ModelBuilder. Tools are represented as yellow rectangles. Existing data is displayed in blue ovals and created data in green ovals.
3.3.3.1 ALTITUDE , SLOPE AND ASPECT
The DEM provided from the NGD of Lao PDR consists of a continuous surface of z-values representing the elevation in m as z-value at each cell. The specification of this DEM forms the basis of all the raster layers used in the analysis. The area of interest was extracted by masking the area with the shapefile of the province boundary using the tool “Extract by mask”. This new raster was used as source for creating raster layers related to the topography of the area.
To receive the slope raster the spatial analysis tool “slope” was performed. The tool extracts the maximum rate of change in z-values of each cell. The slope was calculated in percent as explained above.
Similarly the aspect layer was created by using the spatial analyst tool “aspect”. This tool identifies the direction of maximum change in z-values from each cell to its neighbour.
50
FIGURE 5: STEPS IN ARCMAP TO CREATE ELEVATION, SLOPE AND ASPECT RASTER
3.3.3.2 SOIL
The shapefile of the soil covers entire Lao PDR. As informed by the Ministry of Agriculture and Forestry (MAF), the data was digitized from the Atlas of Physical, Economic and Social
Resources of the lower Mekong Basin.
Similar to the masking of the DEM, the area of interest was retrieved by combining the soil shape file with the province boundary shapefile. The analysis tool “union” was used to create a geometric union of the two input features. To generate a raster from the vector layer, the tool “polygon to raster’ was run. The attribute of interest “MAJOR_SOIL” was specified as the fields used to assign values to the output raster. Since the input field contains string values, the output raster has an integer value field as well as a string field. Reclassification to the suitability value is not necessary but was done to simplify the settings in the weighted overlay tool.
51
FIGURE 6: STEPS IN ARCMAP TO CREATE SOIL RASTER
3.3.3.3 TRANSPORTATION INFRASTRUCTURE
The transportation infrastructure consists of roads, including ferry routes and waterways. To generate one road layer of the entire province, the 5 layers of the main roads of each district were combined. Since teak transportation on waterways is limited to major rivers the layer bko_main_river_pg represents the possible transportation routes on waterways.
The area suitable for teak plantations are areas within 3 km distance to these transportation routes. The suitability for teak plantation of the different buffer depends on the type of routes as described above. The analysis tool “buffer” with the specification of dissolving features according to the field “LAYER“ that contains the type of roads that is related to the type of transportation route created a polygon layer with overlapping polygons. These polygons were separated according to the route type as specified in the field “LAYER” by selecting each type of polygon and making a new layer from this selection. Additionally a 3 km buffer around the main rivers was created.
To remove the overlapping buffer areas the tool “erase” was used on the layers “unpaved roads”, “trails” and “main_rivers”. Then the 4 output layers were combined using the tool
“union”.
Before the final transformation to a raster layer was performed a field was added that contains the suitability value of each polygon. This field was used to as value field in the polygon to raster tool.
52
FIGURE 7: STEPS IN ARCMAP TO CREATE RASTER FOR INFRASTRUCTURE
3.3.3.4 DISTANCE TO SAWMILLS
The basis for this criterion layer formed a newly created point feature layer that contains the
existing sawmills in Bokeo province. To incorporate the entire area of the province two newly
created raster were combined: one from the point feature layer of sawmills and one from the
province boundaries. The new layer was reclassified to having a value of 1 at sawmill
locations and no data value at all other locations. This formed the input raster for the tool
“Euclidean distance” that calculated for each cell the distance to the closest source feature.
Additionally the province boundary was set as mask to exclude cells outside Bokeo province.
53
FIGURE 8: STEPS IN ARCMAP TO CREATE RASTER OF DISTANCE TO PROCESSING FACILITIES
3.3.3.5 ACCESSIBILITY , AVAILABILITY OF MANPOWER , POVERTY REDUCTION
All three of these criteria are related to the census data of the villages in Bokeo province.
The census data must be joined to a polygon shapefile of the village areas in Bokeo province. In ArcMap the command “join” appends attributes from one to the other based on a field common to both. The common field in both data sets is the village code that functions also as a unique identifier.
Before the table of the census data could be joined with the polygon shapefile of the villages of each district, the table was separated into one table for each district. Additionally the tables were simplified in simple columns for each attribute.
After joining a new column for the population density in peoples per km 2 “Pop_density” was
created in the attribute table of each layer. With the tool “Field calculator” the population
density in peoples per km 2 was calculated and filled into the respective field. The used formula divided the total population of each village area with the area size divided by
1,000,000.
The shapefiles of the 5 districts were merged into one dataset that covers the entire province.
54
To get the population density in raster format the merged shapefile was converted to a raster
specifying the newly created column as value field. Similarly the accessibility of each village
region needed specifying the column of accessibility as value field. A third raster showing the
poor villages was created by specifying the respective column as value field.
FIGURE 9: STEPS IN ARCMAP TO CREATE RASTER: MANPOWER, ACCESSIBILITY AND POVERTY
3.3.3.6 PROTECTED AREAS , SUSTAINABILITY AND LANDUSE
The remaining three criteria “Protected areas”, “Sustainability” and “Landuse” are pre-
emptive criteria that exclude areas from the analysis without classifying areas according to
their suitability. For these three criteria a mask layer was created that show the suitable
area.
To create a mask for the protected areas the two shapefiles
bko_national_protected_area_pg.shp and bko_national_protection_forest_pg.shp were
55
combined using the tool “union”. The resulting polygon layer shows the areas that needed to
be excluded from the analysis. A mask however shows the areas included in the analysis.
The tool “Erase” in combination with the polygon of the province boundary was used to
receive the areas to be used in the mask.
FIGURE 10: STEPS IN ARCMAP TO CREATE THE MASK
The mask for sustainability is created by buffering the layer of hydrology lines with a buffer of
50m and the layer of hydrology polygons with a buffer of 100m. These two polygon layers were merged. Again the tool “Erase” in combination with the province boundary was used to receive the areas to be used in the mask.
From the landuse layer all suitable areas were selected by attribute. A new layer was created from the selected features to result in a layer that masks the wanted areas.
56
The three created polygon layers were combined to form a final mask by using the tool
“intersect”. Map 20 shows the layer containing the mask. The purple area shows the area
that is suitable for teak plantations. Using a mask in ArcGIS excludes all areas not contained
in the polygon layer defined as mask.
MAP 20: MASK SHOWING SUITABLE AREAS FOR TEAK PLANTATIONS IN BOKEO PROVINCE, LAO PDR
3.3.3.7 WEIGHTED OVERLAY
The nine raster layers were used in the weighted overlay analysis. The tool “Weighted
Overlay” demands input raster with integer values. The reclassification of each raster data set to a suitability value according to the discussion above guaranteed a compliance with this condition.
The tool’s syntax is a weighted overlay table that lists all input raster with the chosen evaluation field. A value according to an evaluation scale is assigned within the tool to each attribute in the chosen field. These values correspond to the standardized suitability values
57
described above. The assignment of these values can be done directly in the “Weighted
Overlay” tool or each layer can be reclassified with the “Reclassify” tool to assign the
standardized suitability value directly.
The weights calculated in the pairwise comparison are assigned to each layer within the
weighted overlay table. All weights must sum up to 100% obviously. The weighted overlay
table was saved to enable a reuse in the weighted overlay tool while only the weights
needed to be changed.
The algorithm used by the tool multiplies the cell by their percentage influence, and then
adds the results together to create the output raster. The formula is