Detection of actual and assessment of potential plantations in Lao PDR using GIS and remote sensing technologies doc

In the study at hand these problems are addressed by an attempt to assess the current extent of the rubber plantations and by modelling the potential spatial distribution of rubber plant

Rubber in Laos Detection of actual and assessment of potential plantations in Lao

PDR using GIS and remote sensing technologies

Diplomarbeit der Philosophisch-naturwissenschaftlichen Fakultät

der Universität Bern

vorgelegt von:

Kaspar Hurni

2008

Leiter der Arbeit:

Prof Dr Urs Wiesmann Centre for Development and Environment,

University of Bern

Rubber in Laos Detection of actual and assessment of potential plantations in Lao

PDR using GIS and remote sensing technologies

Diplomarbeit der Philosophisch-naturwissenschaftlichen Fakultät

der Universität Bern

vorgelegt von Kaspar Hurni

2008

Leiter der Arbeit:

Prof Dr Urs Wiesmann Co-Leiter:

Dr Andreas Heinimann Centre for Development and Environment,

University of Bern

During my studies I had the chance to work as a tutorial assistant within the NCCR North-South at the Centre for Development and Environment of the Institute of Geog-raphy at the University of Bern It was a huge opportunity to gain insight into the re-search context of the NCCR North-South and obtain some practical experience Most

of my knowledge and skills regarding GIS and Remote Sensing technologies were tained during this employment

at-Besides the administrative work I could also participate in more research-related tasks where I got in touch for the first time (although only digitally) with Laos I performed a classification of satellite imagery in order to map land cover changes, which was part

of a study assessing land cover changes in the lower Mekong region When later on the decision on the topic of my diploma thesis came closer, I opted for a work in context with Laos, as it was a chance to visit a country for two months that I previously only knew through satellite imagery

Besides the opportunity to travel to Laos, the motivation for this thesis is mainly based

on giving a contribution to the current research on rubber in Laos Considering the amount of newly established rubber plantations in the recent years, research on the topic is indispensable regarding decision-making and elaboration of development strategies

Using my GIS and Remote Sensing competence and my fascination for the possibilities offered by spatial data and modelling approaches, I hope being able to add to the cur-rent research, or at least provide some thought-provoking impulses

Besides the institutional support acknowledged below many people contributed to this study in various ways First I would like to thank Professor Urs Wiesmann, the leader

of the thesis His professorship at the University of Bern and his functions within the CDE and the NCCR North-South provided the background for various studies and re-search projects in Southeast Asia and thus enabled my research in Laos

My thanks go also to Dr Andreas Heinimann, the co-leader of my thesis, who was, already during my time as tutorial assistant, my mentor regarding GIS and Remote Sensing technologies While working on the thesis his inputs on various levels and the motivational words were indispensable

I am also very thankful for the support I received from Dr Peter Messerli He works in the research project on “Contextuality of Development Interventions” in Laos, and was the contact during the time I spent in Laos He introduced me to several people and institutions in Laos and was very supportive in specifying the direction of my research

at the initial stage

My acknowledgements go to Mr Sangkhane who joined me for the field trip His tance during the interviews, knowledge on the Lao culture and the ability to contact people was very helpful Additionally I would like to mention Mr Mike Dwire and Mr Phaythoune Pilakone who were very obliging in terms of sharing information and re-sources

assis-My thanks go also to my parents, Marlies and Hans, who supported me throughout my studies and provided many inputs regarding the success of my research Your motiva-tional stimulus, thought-provoking impulses and corrections helped me a lot

Additionally I would like to thank Brigitte Portner for her skills and help regarding editing and the use of Microsoft Word, Daniel Loppacher for his advices on administra-tional processes and various discussions, and Adrian Weber with whom the work and leisure time in Laos was fun to spend with

Finally, my thanks go to all the co-workers and the student body for their advices and talks, and my friends where I could recharge during my off time However, you are too many to be named in person

Kaspar Hurni, August 2008

The cornerstones for my research activities in Laos are on the one hand the funding and network provided by the National Centre for Competence in Research North-South and, on the other hand, the support I received from various individuals The present study is performed within the context of research projects of the Centre for Develop-ment and Environment (CDE) at the Institute of Geography of the University of Berne The CDE is the executive agency of the NCCR North-South and involved in various research themes, projects and networks all over the globe One of these projects is lo-cated in Laos and performs research on the contextuality of development interventions The local institution this project is related to is the Lao National Mekong Commission Secretariat (LNMCS) Within the framework of these three institutions the research on rubber in Laos is performed During my stay in Laos the LNMCS was the main con-tact, while back in Switzerland the work was performed at the CDE while the NCCR North-South provided the funding and network

In Laos other Institutions contributed to this thesis, too, mainly by sharing information and data These are namely: The National Geographic Department (NGD), the National Land Management Authority (NLMA), the National Agriculture and Forestry Research Institute (NAFRI), Ecotourism Laos and the Ministry of Agriculture and Forestry (MAF)

Images xiii Figures xiv Tables xiv Equations xv

Laos 51

4 Conclusions and recommendations 83

With the crash of the Soviet bloc in the early 1990s Laos faced a cut off in foreign aid and needed to reorient itself Ties were bid up with China and Thailand, and an eco-nomic reform programme was pushed with focus on liberalizing the market This posed and still poses opportunities for the more developed surrounding countries, as Laos has

a lot of forest land that can be used for agricultural plantations The Laotian ment fosters this dynamic, as with the market integration of the smallholder, socio-economic development can be pushed But today’s dynamics can also have some det-rimental consequences for the economic development and the environment

govern-Prices on the world market for natural rubber in 2008 are at an all time high, mainly driven by the high oil price and the demand of China On the background of this global market development the regional disparities between Laos and China lead to a high dynamic Around 2001 the rubber boom in Laos set off, and up to now more and more land has been and still gets converted to rubber plantations An assessment and control-ling of this development is indispensable, as it can have detrimental effects regarding the environment and dependencies from the market and China In the study at hand these problems are addressed by an attempt to assess the current extent of the rubber plantations and by modelling the potential spatial distribution of rubber plantations according to different assumptions on policy implementations

These tasks are approached in three main steps: The first one consists of the fieldwork, where GPS points for the classification of the satellite image were collected, informal talks with farmers and officials conducted, and an impression of the landscape was gained The second step is a remote sensing approach, performed in order to gain some knowledge on the actual distribution of rubber plantations Information from the image classification is meant to provide inputs for the last step, the GIS-assisted modelling approach Through the combination of different datasets using fuzzy logic, models displaying the potential spatial distribution of rubber plantations according to different scenarios are elaborated

For a sample area in north Laos around the town of Luang Namtha the current tion of rubber plantations was tried to be captured on ASTER satellite imagery using image segmentation and fuzzy classification Due to the recent rubber boom this task could not be fulfilled, as most of the rubber plantations were still too young to be cap-tured according to their spectral properties A multi-resolution image classification could possibly tackle this task by including texture analysis in order to classify the younger plantations On the ASTER imagery however, the land cover changes could still be mapped using a binary classification approach The land cover changes do not allow a quantification of the rubber dynamic, but nevertheless they include the rubber plantations Some indications regarding the plantations can therefore still be gained when analyzing the land cover changes

distribu-The land cover changes around Luang Namtha were combined with a set of factors as slope, elevation, accessibility and different types of protection areas In a logistic re-gression model the linear correlation of the factors with the land cover changes was checked Accessibility and slope showed a correlation, and we are thus able to explain

changes in land cover Elevation in fact is linked with the land cover changes but in a non-linear way and is thus not able to explain land cover changes For the protection areas the sample area was too small to determine any relations with the land cover changes Additional calculations including a land cover dataset of 2002 revealed rela-tions between deforestation and accessibility Deforestation is much more likely to occur in good accessible forest areas; especially dense forest is more likely to get cut down in accessible places In more remote areas dense forest is less affected, it is rather the secondary forest that still gets cut down Additionally, deforestation is much higher and more concentrated around bigger cities, while at rather remote villages it is lower and more disperse

Using this information, and including literature on rubber crop requirements, economic considerations (accessibility) and interpretations of Laotian forest protection policies, a modelling of the potential spatial distribution of rubber plantations in three scenarios could be performed For the modelling fuzzy logic was used, as this allows including uncertainties on the data and the effects of data combination Three different scenarios were calculated, each including different assumptions on the policy implementations regarding forest and conservation area protection The first scenario delineated the po-tential rubber area when no protection is applied It thus shows the maximum potential extent of rubber in Laos according to its biological suitability and accessibility of the landscape The second one was the opposite, considering maximum conservation, in-duced by all the protected or classified forest areas The scenario is rather fictional as it would exclude 74% of the area of Laos from being used agriculturally The third sce-nario showed the potential extent of rubber according to the current level of policy implementation Some forest and conservation areas are protected, but in accessible areas the plantation of rubber is still tolerated

Comparing the different scenarios allows considerations of the effects the current icy or policy changes have on the landscape of Laos and other economic activities as ecotourism (mainly in protected areas) and collection of non-timber forest products Accessibility has been found to be a major factor regarding the plantation of rubber Increasing accessibility to a region can foster socio-economic development, but can also have detrimental effects on the environment and thus other economic activities Another important factor are the protection policies In many villages the definitions of land-use types, conservation areas and management rights are unclear; population den-sity and accessibility rather determine whether land is agriculturally used or not In order to reduce one-sided dependencies and environmental degradation, and foster socio-economic development policies, the Lao Government needs to be clear about management rights and possible land use activities and delineate protected areas prop-erly Through investments in infrastructure and the promotion of different crops and economic activities one-sided dependencies can be reduced through diversification

pol-ASTER Advanced Spaceborne Thermal Emission and Reflection

Radiometer CDE Centre for Development and Environment

CPI Committee for Planning and Investment

DSCA Demining Defense Security Cooperation Agency

ESRI Environmental Systems Research Institute

FAO Food and Agriculture Organization

Landsat TM/ETM Landsat (Enhanced) Thematic Mapper

Lao PDR Lao People’s Democratic Republic

LNMCS Lao National Mekong Commission Secretariat

LSUAFRP Lao Swedish Upland Agriculture and Forestry Research

Project MAF Ministry of Agriculture and Forestry

MCTPC Ministry of construction, Transport, Posts, and

Communica-tions NAFRI National Agriculture and Forestry Research Institute

NBCA National Biodiversity Conservation Area

NCCR N-S National Centre for Competence in Research North-South NDVI Normalized Difference Vegetation Index

NSC National Statistic Centre

PAFO Provincial Agriculture and Forestry Office

SPOT Satellite Pour l'Observation de la Terre

UNESCO United Nations Educational, Scientific and Cultural

Organi-zation UTM Universal Transverse Mercator (coordinate system)

VNIR Visible Near Infrared

WSCP Watershed Classification Project

Image 1: Rubber plantations in Yunnan (around Mengman and Yun-Chin-Hung)

covering large areas along the hillsides of the valleys (source: Google

Image 2: Smallholder rubber plantations northwards of Luang Namtha and the land

cover difference between Laos (on the left) and China (source: Google

Image 3: GPS points collected during field trip in June 2007 in Luang Namtha

Province 14 Image 4: Different stages of smallholder rubber plantations in the North of Luang

Image 5: Rubber plantations in the north of Luang Namtha Characteristic pattern

comes from the contour bunds in the plantation (source: Google Earth,

Image 6: ASTER Satellite image North of Luang Namtha Comparison of spectral

reflectance and land cover: Depending on the age the rubber

Image 7: Segmentation dialogue window of ‘Definiens eCognintion’ software

Image 8: Example of a class description and a membership function in eCognition 31 Image 9: Areas of classification problems: Haze and image transition zones

Image 11: Potential spatial distribution of rubber plantations without conservation

policies 69 Image 12: Potential spatial distribution of rubber plantations according to crop

requirements 70 Image 13: Potential spatial distribution of rubber plantations without conservation

policies in the north of Laos around Luang Namtha displayed with

Image 14: Potential spatial distribution of rubber plantations with maximum

conservation 73 Image 15: Potential spatial distribution of rubber plantations according to current

policy 75 Image 16: Potential spatial distribution of rubber plantations according to current

policies in the north of Laos around Luang Namtha displayed with

NBCA’s 76 Image 17: Percentage of areas suitable for rubber within NBCA’s according to crop

requirements 77 Image 18: Percentage of areas suitable for rubber within districts according to crop

requirements 79 Image 19: Potential spatial distribution of rubber plantations according to

accessibility from surrounding countries without conservation policies applied 81

Figures

Figure 1: Yearly mean price of rubber in US cent per kilo (source: Malaysian Rubber

Figure 2: Comparison of rubber and oil prices in percent of year 2000 prices

Figure 3: Area of rubber plantations in Luang Namtha 1994-2006 Data is

unavailable for 1996, 2000 and 2002 (source: PAFO Luang Namtha

Figure 4: Comparison of Spectral Bands between ASTER and Landsat-7 Thematic

Figure 6: Value ranges of selected image layers used for the classification (source:

author) 31 Figure 7: Combination of land cover changes with slope classes in the study area

Figure 8: Combination of land cover changes with elevation classes (source: author) 40 Figure 9: Combination of land cover changes with province accessibility (source:

author) 41 Figure 10: Combination of land cover changes with village accessibility

Figure 11: Deforestation rate between 2001 and 2006 in relation to village and

Figure 12: Combination of land cover changes with land cover classes Change

Figure 13: Combination of areas without land cover change with the land cover

Figure 14: Combination of ”low to high biomass density changes” with land cover

Figure 15: Combination of ”high to low biomass density changes” with land cover

Figure 16: Deforestation rate in relation to forest classes and province accessibility

Figure 17: Fuzzification of the elevation (DTM) and the slope according to

Figure 19: Comparison of different scenario calculations on the potential spatial

Tables

Table 3: Result of the Logistic Regression Model on the occurrence or

Table 4: Collection of crop requirements for rubber (Hevea brasiliensis) from

rubber plantations 55 Table 7: Assignment of values between zero and one according to the suitability

Table 8: Assignment of values between zero and one according to the suitability for

rubber of the soil depth and soil drainage (1 = high suitability, 0 = no

Table 9: Considerations for the different model scenarios according to the

Table 10: Reclassification of forest and land cover layers to model different policy

implementation effects in the potential spatial distribution of rubber plantations 66 Equations

Equation 7: Probability values of the linear logistic model (source: Mertens &

1 Introduction

1.1 Rubber in Laos

Rubber cultivation has a rather young history in Laos In 1930 small areas around Pakse Town in southern Laos were planted with rubber trees that are still alive up to now The villagers tapped the trees rather for fun in order to trap small animals like insects and birds The economic potential of these trees was not considered and the cultivation of local species was preferred The basis for the rubber boom in Laos was laid in 1994 when smallholders established a 400 ha plantation around Ban Had Ngao,

a village in the north of Luang Namtha The villagers, mainly from the Hmong ity, were supported by resettled relatives from China, which had gained experience in cultivating rubber from their work in the Chinese agricultural collective Even though there was an increasing area planted with rubber it never caught the attention of policy-makers, traders or stakeholders This rapidly changed after the first two to three years when tapping the trees generated high income from latex sales The boom set off around 2003 with rubber being promoted by foreign traders and companies, the Lao Government and by the villagers themselves (Shi, 2008; Alton et al., 2005; Ketphanh et al., 2006) The situation on the global rubber market heavily supported, and is still sup-porting, the expansion of rubber plantations within Laos Figure 1 shows the develop-ment of the rubber price over the last eight years

ethnic-Yearly mean price of natural rubber in US cent per kilo

Figure 1: Yearly mean price of rubber in US cent per kilo (source: Malaysian Rubber Board, 2008)

For the enormous increase in the price over the last eight years (over 400%) a set of different parameters can be accounted But what may be more important is the fact that such a huge increase was not expected when looking at rubber price scenario calcula-tions performed after the crisis in Asia and for the new millennium In general, the pro-jections underestimated the price and the quantity of natural rubber that has been traded

in the past couple of years (ICNR, 2008) A reason for this can be found to some extent

in the nature of rubber Most importantly natural rubber can be substituted perfectly by synthetic rubber which is produced using oil Several factors affecting one or both types of rubber can thus influence the overall price The price level of natural rubber may be shifted due to natural hazards such as storms or floods, while the price of syn-thetic rubber rather depends on the dynamics on the oil market Other factors leading to

a price shift are the stocks on the supply and demand side and also effects of currency de- and appreciation However, a rise or decline in the price of one of both, natural or synthetic rubber, leads to a substitution and thus to an overall price increase The price increase displayed in Figure 1 is mainly due to an increased oil price and an increased demand for rubber in Asia, but especially in China Figure 2 shows the oil price and the price for natural rubber To be comparable both prices are indicated in percent of the year 2000 prices

Comparison of rubber and oil prices (percent of year 2000 price level)

Figure 2: Comparison of rubber and oil prices in percent of year 2000 prices (source: OPEC, 2008; Malaysian Rubber Board, 2008)

In general both prices evolve similarly – the increasing oil price lead to higher prices for synthetic rubber which in turn boosted the demand for natural rubber resulting in a higher price for both, synthetic and natural rubber Gaps between the two graphs in Figure 2 may be due to speculations The steady growing oil price causes expectations

of an increasing rubber price in the future and thus a higher demand which fuels the price increase This dynamic may not be foreseen in the pre-millennium scenario calcu-lations but between 2001 and 2003 the potential was recognized by both, smallholders and governmental institutions of Laos Within these first years of the boom the relation

to China seems to play a crucial role in northern Laos The first rubber plantations in Ban Hat Ngao have been established by resettled people from China or people with relatives in China providing the technical knowledge and the experience (Alton et al., 2005) Farmers from other villages also reported to have some connection to China, be

it relatives or just traders animating them to plant rubber Additionally Chinese, but also Thai (rather in central Laos) and Vietnamese (rather in south Laos) companies seek to establish rubber plantations on the basis of concession or contract farming (Ket-

Different types of concessions and contracts exist, but in general within concessions the farmers only provide labour and are paid monthly wages while in the contract farming system land and labour is supplied by the farmer and profits from latex and timber sales are shared among farmers and investors (Douangsavanh & Thammavong, 2006) This dynamic (regarding smallholders but also contracts and concessions) is heavily sup-ported by the Lao Government Rubber is one of the focus commodities for poverty alleviation and eradication of shifting cultivation In the sixth national socio-economic development plan goals and means for the promotion of rubber are formulated regard-ing the labour market, investment and processing industries (Lao People’s Democratic Republic, 2006)

Regarding the high or even increasing prices for rubber on the world market and the heavy promotion from the side of the government but also private investors an ongoing rubber boom in Laos can be expected resulting in vast areas of forests and other land cover types being converted to rubber plantations The effects of this dynamics are difficult to predict First, returns from the rubber trees can be expected after approxi-mately seven years, when the tree can be tapped for the first time Within this time-span a lot can happen considering market uncertainties, natural hazards, etc Addition-ally, the outcome of the environmental impacts of rubber monoculture (especially on large-scales) and the effects on other crops and income diversification of the farmers are mainly unknown Historically, heavy governmental promotion of rubber already occurred in the Chinese province Yunnan bordering with Laos Beginning in the late 1950s and in a second step in the late 1970s the expansion of rubber was fostered, which resulted in a successful natural rubber industry on steep slopes in southern Yun-nan (Chapman, 1990) Analyzing Chinese policy measures and environmental and socio-economic impacts may help to formulate predictions and recommendation for Laos

Laying on the ecological margin for Hevea brasiliensis the establishment of rubber

plantations in southern Yunnan faced environmental difficulties with heavy losses of even mature rubber trees Major constraints were the occasionally very severe cold events causing damage to the bark, and sometimes leading to death in the rather long dry season of four to five months After primary failures the establishment of a success-ful Chinese rubber industry can mainly be accounted for in the planning skills of the Bureau of State Farm Managements and China’s determination to increase domestic production By the end of the 1980s, rubber plantation in Yunnan covered approxi-mately 100’000 ha of which 46’000 ha were owned by smallholders Planning skills and a protected domestic market take a big share in the success story In order to mini-mize losses from environmental impacts plantations were restricted to areas below 900 – 1000 m asl In narrow valleys on the bottom and on lower slopes no plantations were established to avoid areas of cold air drainage Additionally, before planting, site as-sessments had to be performed including three years of winter temperature observation (Chapman, 1990)

Up to the mid-1980s most of the Chinese rubber was coming from state farms, as ily farming was prohibited until 1978 The state farms had good funding, technical efficiency and sometimes even provided housing Wages of rubber tappers could thus easily compare with those of urban workers China’s main concern was laid on techni-

fam-cal efficiency and increasing the production Economic efficiency had a secondary function as rubber enjoyed a protected market with domestic prices well above the world market price Not until 1978, with family farming being allowed again, some of the state farms were reorganized into smaller production units at village level and many villagers became in effect smallholders (Chapman, 1990) Due to the favorable domes-tic prices rubber became even a boom crop in Yunnan Up to now, large areas of Yun-nan are covered with rubber plantations being planted in state farms or by smallholders

on smaller patches, as can be seen in Image 1

Image 1: Rubber plantations in Yunnan (around Mengman and Yun-Chin-Hung) covering large areas along the hillsides of the valleys (source: Google Earth, accessed in July 2008)

A set of different factors can thus be accounted for in the successful establishment of a rubber industry in Yunnan, some of which may be adapted to the context of the Laotian rubber boom The Chinese achievement can be bound to the following favourable con-ditions: The promotion of rubber was well planned and technically supervised Initially, the rubber was only planted in state farms according to regulations concerning altitude,

the smallholders received training in planting, maintaining and tapping the trees, which they could adopt later when planting rubber on their own On national level there was the determination to increase domestic production in order to reduce dependence (Chapman, 1990) This plays a crucial role for the successful rubber industry and espe-cially for the vast expansion of rubber plantations within Yunnan With the protected market and favourable prices a long term investment like rubber (seven years until tap-ping can be started and first returns expected) faces fewer risks and uncertainties Combined with the knowhow the smallholders were able to gain on the state farms they are more likely to successfully plant rubber on their own

The setting of the rubber boom in Laos shows some analogies but also major ences especially in terms of dependencies on the market and overseas During the late 1980s and 1990s Laos was in a wind of change Due to the crash of the Soviet bloc it faced a cut-off in foreign aid The reorientation of the country resulted in bidding up ties with China and Thailand and an economic reform programme called the New Eco-nomic Mechanism, involving decentralization of economic decision-making, more accountability for public enterprises, fiscal and financial reforms, deregulation of prices, removal of trade barriers, the implementation of new investment codes and an overall greater reliance on market forces Agriculture is still considered the engine of growth, but production has been shifted back to a family-based form opposed to the prior work within collectives (Gunn, 1991)

differ-Regarding the actual dynamic in the field of rubber plantations the economic reform played a crucial role: Both domestic and foreign ownership and capital investments have constitutional guarantees of protection and private ownership of production is guaranteed along with state and collective forms of ownership, transfer and inheritance

of property The legal basis for investments on the side of smallholders, but also eign companies, is thus given While the implementation of the New Economic Mechanism basically opened Laos economically, the political situation was not much altered The one-party system was not touched, and any antiregime associations or calls for a multiparty system are against the constitution (Gunn, 1991; Inoue & Hyakumura, 1999) The challenge Laos faces is to cover the span between the reliance on the market forces on the economic side, while political decision-making and planning is limited to one party

for-The economic reform and the reorientation towards China led to agreements on trade and cooperation as well as an agreement on technical cooperation between authorities

of China’s Yunnan province On this background today’s high dynamics in this region become explicable Due to the prior economic closure Laos, compared with its sur-rounding countries, still has large areas under forest cover, and within rural areas most people rely on subsistence with some local market integration (Foppes & Ketphanh, 2000) These disparities are a potential to foster socio-economic development and in-crease the income of the rural population The promotion of certain industrial crops based on the potential of each region should boost export volumes (Lao People’s De-mocratic Republic, 2006) Nevertheless, the vacuum Laos poses within Southeast Asia needs good political guidance The economic potential for both, the rural smallholders and the foreign investors can create a dynamic that calls for control measures in order

to reduce possible negative environmental effects and one-sided dependencies

The example of coffee in Vietnam enlightens how local land use decisions are enced by central policy-making combined with global economics The “de moin” pro-gramme in 1975 changed the Vietnamese centrally planned command economy into a market economy with socialist direction The land use in the Vietnamese highlands could be influenced by the global economy, and with high global coffee prices in the mid-1990s the area under coffee doubled within a few years only, resulting in a major change in the land use and land cover characteristics (Heinimann, 2006) Temporary changes on the global market can thus have an impact on regional level with uncertain future effects on environment and socio-economic development

influ-The case of rubber in Laos in fact faces a similar starting position as the coffee in nam The price increase of rubber on the global market made its cultivation more valu-able Disparities between Laos with vast areas of agriculturally unused forest land and its more developed neighbouring countries are expected to result in a fast expansion of rubber fuelled by foreign investors and smallholders Controlling these dynamics is crucial as they can have negative environmental effects and easily result in dependen-cies from the global market and other countries The currently high demand on the Chinese side for natural rubber, which increases the rubber area in Laos, results in an increased dependence from the global market and from China Up to now the rubber plantations in northern Laos have not yet reached the scale and the extent of those in Yunnan, as can be seen in Image 2

Viet-Image 2: Smallholder rubber plantations northwards of Luang Namtha and the land cover ence between Laos (on the left) and China (source: Google Earth, accessed in July 2008)

differ-Nevertheless, large areas have already been converted to rubber plantations, and ready many of the smallholders rely on prospects of future sales of latex and thus de-pend on the dynamics of the global market The success story of Yunnan and the high demand of China fuels expectations of future gains and motivates many to participate

al-in the latex production

Although there are differences between Laos and the Chinese example, especially cerning the influence of the market The challenge Laos is facing at the moment is the balancing of the different potentials in order to reduce one-sided dependencies, prevent environmental degradation and foster socio-economic development As the Chinese example shows, rubber is a good commodity to increase the income of the rural popula-tion However, there are differences regarding the rubber production between the two countries The Chinese rubber industry was established in order to achieve self-sufficiency They solely produce for the domestic market and enjoy fixed prices well above the world market price Production is utterly independent from foreign influ-ences with the major concern laid on achieving technical efficiency and increasing

con-productivity Technical assistance was indispensable as plantation was first limited to state farms With a policy change, the latex production was later opened to small-holders, which could profit from the infrastructure and the diffused knowhow (Chap-man, 1990)

The primary goal in Laos, on the other hand, is to foster socio-economic development

in the rural areas through the promotion of industrial crops However, the current ket price of rubber may be detrimental to the primary goal of socio-economic develop-ment Prospects of future gains caused a high dynamic, and many areas have been con-verted to rubber Some villages in north Laos have even converted most of their agricultural land into rubber plantations Subsistence food production has been heavily diminished, and with the income from latex sales food is meant to be bought in future

mar-In the time until the tapping is going to start, farmers reported to “not have an easy living”

As it seems, many smallholders put all their eggs in one basket, making their future dependent from the development on the rubber market On the background of the actual dynamic this can be understood The proximity of north Laos to China and the large forest areas that could be used economically more productive lead to regional dispari-ties that result in a high dynamic regarding the current market price of rubber How-ever, increasing the area of rubber plantations means increasing the dependence on the demand of China and on the rubber price, while income diversification and a greater degree of independence is being reduced

In order to push socio-economic development and secure future gains diversification is indispensable Different agricultural crops need to be promoted where the plantation of rubber is bound to risks, income generation through other activities (i.e ecotourism) need to be considered to avert conflicts with other use functions and where necessary the environment needs protection in order to reduce natural hazards and environmental degradation To support the decision-making of the responsible authorities, research in different fields has been performed Most prominent may be the Para Rubber Study of Alton et al (2005) It studies relevant aspects of the cultivation of rubber as well as its production and marketing in Luang Namtha and gives recommendations on issues like rubber extension, research, and the environment Subsequent to this publication a lot of follow-up research was performed, resulting in papers and conferences as for example the Workshop on Rubber Development in Lao PDR organized by the NAFRI Major topics are ecotourism, environment, economics and spatial rubber suitability assess-ments, in general aiming at bringing light into the dynamics, providing technical assis-tance and determining policy actions

On this background, the study at hand has the overall aim to fill gaps and to contribute

to the current discourse on rubber Within a remote sensing approach the possibilities

of mapping the present extent of rubber plantations using medium-resolution satellite imagery on a sample area around Luang Namtha shall be explored By analyzing the actual distribution of rubber and considering the rubber crop requirements, an attempt

is made to model the potential spatial distribution of rubber plantations in Laos Through the inclusion of economic considerations and assumptions on policy imple-mentations different scenarios are then calculated in order to display the effects of envi-

ronmental protection and accessibility on the potential spatial distribution of rubber in Laos

In order to add to the current research on rubber, its dynamic in Laos is addressed along two different objectives The first objective is to present in a cost effective way the actual rubber plantations using medium resolution remote sensing data Regarding the satellite imagery (medium resolution images) and the processing methods (image seg-mentation and fuzzy classification) the potential for the assessment of a specific land cover (rubber) shall be examined in order to give recommendations for future assess-ments covering the whole area of Laos Concerning the rubber dynamic the classifica-tion of the satellite imagery is meant to provide area estimates of the land covered with rubber, its spatial distribution, size of plantations, etc Knowing about these facts can give some information on the pushing factors of the rubber dynamic, provide some inputs regarding the modelling of the potential rubber areas and maybe even allow formulating first policy recommendations

The second objective is to present a model of the potential spatial distribution of rubber plantations according to different scenarios, indicating what effects different policy implementations can have on the spatial distribution of rubber Also here the major goal is to obtain information on the pushing factors and to formulate policy recommen-dations However, in difference to the satellite image classification, the factors in the model affecting the distribution of rubber plantations can be adjusted and thus the ef-fects of for example different policy implementations on the potential spatial distribu-tion of rubber are meant to be displayed

Regional disparities between Laos and its neighbouring countries regarding the tial area for agricultural plantations pose an opportunity for socio-economic develop-ment and increasing the income of the rural population However, the chances can also result in a deterioration of conditions if the dynamics are not handled properly The combination of regional disparities and favourable conditions on the market side can initiate fast and large-scale land cover changes Environmental effects of vast areas of monocultures can reduce income from other activities, as for example the collection of non-timber forest products and ecotourism Focusing on only one main crop can result

poten-in a dependency from the market with detrimental effects of price downturns cal assistance from the state and regulations concerning conservation areas and forest lands are indispensable As the Chinese example showed, technical assistance is a cen-tral element in fostering industrial plantations Crop suitability assessments and consid-erations on other use functions help increasing diversification, dispersing dependencies and minimizing environmental degradation In the study at hand, issues of concern are the actual distribution of rubber and the possibility to model its potential spatial distri-bution Knowing the present area of rubber and its distribution in Laos enables to get

Techni-information on the dynamics in the different regions, and allows choosing appropriate policies

The first hypothesis thus states:

“Actual rubber areas in Laos can be detected with medium-resolution satellite imagery available at present.”

The reason for the use of medium-resolution imagery is cost efficiency In general, such images cost less, cover larger areas and need less processing time Using remote sensing technologies allows for a quick assessment and mapping of the dynamics in a region, easing decisions on policy measures Conflicts within areas of other use func-tions, such as conservation or ecotourism, are also well detectable and can be assessed With the mapping of the present rubber plantations already a good amount of the dy-namics can be targeted with policies, and conflicts tackled But using this strategy pro-vokes a lag between the dynamics and the politics Policies, meant to be the steering element, are actually following the dynamics and are hence less efficient A model of the potential spatial distribution of rubber plantations has the potential to solve this problem

This leads to the second hypothesis:

“Actual rubber plantations can be correlated with a set of factors that allow for potential rubber mapping.”

With the remote sensing approach the extent of the rubber plantations will be mapped

A correlation of these plantation areas with factors like elevation, slope, accessibility, etc allows determining which factors (or ranges of the factors) are able to describe the occurrence or absence of rubber plantations The combination of these factors results in

a map showing the potential spatial distribution of rubber plantations The planning and the adaption of policies can be performed by considering both – the actual and the po-tential extent of rubber plantations It can help in delineating regions with different foci, as for example, to rather increase technical assistance or renegotiate different use functions

But what about the different policy measures? Is it also possible to model their effects

on the potential spatial distribution of rubber? Being able to include the policy ures in the model would allow for better technical assistance and planning, as the ef-fects could be predicted to some extent from the model

meas-The third hypothesis therefore states:

“The potential spatial distribution of rubber can be modelled according to ent policy measures.”

differ-For example, the effect of policies on the distribution of rubber can be seen in Yunnan The dedication of the Chinese to increase domestic rubber production resulted in a suc-cessful rubber industry In the case of Laos the setting is different, but also here, the policies play an important role In order to foster socio-economic development, policies need to control the rubber dynamic and balance between protection of the environment, diversification and poverty alleviation, while still considering the dynamics on the market The model may not play a role in the final decision-making, but can certainly

“Policy-relevant science often fosters public understanding and support for major litical decisions It is thus essential to provide early warning of emerging issues.”

po-2 Approaches and methods

2.1 Approaches to assess the rubber dynamics

In order to assess the current rubber dynamics in Laos, detect the major pushing tors, perform different model calculations and give policy recommendations, the topic

fac-is addressed in three different approaches: Fieldwork, a remote sensing approach sification of satellite imagery) and a GIS approach (mapping of potential rubber areas) There is some ranking between the three, each providing some input for the next ap-proach Still they have the ability to be performed as stand-alone activity, in case one of the prior approaches fails

(clas-The first approach is the fieldwork and goes mainly hand in hand with the classification

of the satellite imagery and the modelling First of all it should give the author an pression of the landscape and land cover, land use practices, distribution of settlements, etc With the collection of GPS points the classification of the satellite imagery and its verification should be eased Informal talks with farmers and officials are meant to provide some information for the modeling approach

im-The second approach using remote sensing attempts displaying the actual distribution

of rubber plantations within a sample area in the north of Laos, around the town of Luang Namtha Knowing the actual distribution of rubber allows for a quantification, maybe revealing some of the pushing factors and enabling the detection of conflicts with other use functions Correlating the actual rubber areas with other data (as eleva-tion, slope, accessibility, etc.) may provide inputs for the modelling and even allows formulating firs policy recommendations

The third for assessing and giving recommendations on the rubber dynamic is a assisted step The idea is to model the potential spatial distribution of rubber plantations using the information provided by the two prior approaches and additional information from the literature as for example the crop requirements of rubber Including different assumptions on accessibility and environmental protection is meant to display the ef-fects different policy implementations have on the distribution of rubber plantations Comparing the different scenarios allows for estimating the effects different policy implementations have on the spatial distribution of rubber and thus on the environment, dependencies from neighbouring countries and socio-economic development

From 9-19 of June 2007 the northern part of Laos was visited as part of a two-month stay in Laos The aim was to get an impression of the landscape and the people, collect GPS points of various land cover for ground truth data, and get information from local people and officials on rubber by conducting informal talks The route travelled was chosen to cover the extent of the satellite images as good as possible and to get an im-pression of the landscape to ease the task of classifying the images later on Approxi-mately 80 GPS points of different land cover types were collected, of which 46 lie within rubber plantations Image 3 shows the spatial distribution of the collected infor-

mation on land cover As can be seen for the image in the south, only a couple of GPS points were collected in the upper left corner of the image Due to road works, bad weather conditions and time constraints it was not possible to follow the road further southwards

Image 3: GPS points collected during field trip in June 2007 in Luang Namtha Province

Informal talks were conducted at province, district and smallholder level On provincial level the Committee for Planning and Investment (CPI) and the Provincial Agriculture and Forestry Office (PAFO) were visited while on district level the Department of For-estry (DoF) provided some information The smallholders were talked with while visit-ing the rubber plantations

2.3 Classification of satellite images

The high dynamic in the field of rubber plantations within Laos calls for a tion of the recent developments For some districts and provinces there is some data on rubber plantations available, but mostly it contains information on rather large-scale rubber plantations on the basis of concessions or contract farming Generally, it gives some information about the size, ownership, etc of the plantation, but does not include any spatial information Area estimates for the spread of smallholder rubber plantations are mostly unavailable This calls for a cost-effective way to capture the ongoing dy-namics

quantifica-2.3.1 Choice of Remote Sensing data

A classification of medium-resolution satellite images seems to be an appropriate proach, as the cost for these images is low or nil (Landsat); they cover areas of reason-able sizes for an efficient classification of bigger parts of Laos, and the image resolu-tion still enables to detect rather small fields as planted by the smallholders The usage

ap-of Landsat ETM satellite images seems promising, as the detection ap-of rubber tions with Landsat has already been conducted in Indonesia (Ekadinata et al., 2004), and the Landsat images can be downloaded freely from the Global Land Cover Facility (GLCF).1 Unfortunately, Landsat 7 images of high quality are only available until May

planta-2003 when a hardware component failure left wedge-shaped spaces of missing data on either side of the Landsat 7 images The satellite is still able to record images, but is only able to capture approximately 75% of the data for any given scene (NASA, 2008)

In Laos, the plantation of rubber has only started recently Already in the years 1994 and 1995 in some villages in the Northern Province of Luang Namtha a couple of rub-ber plantations were established But it took a couple of years until the rubber boom set off and other villages followed the example As Figure 3 shows, an increase in the plantation of rubber started around 2001, but did really set off after 2003 only, when the Landsat ETM satellite was defect already (Shi, 2008)

Figure 3: Area of rubber plantations in Luang Namtha 1994-2006 Data is unavailable for 1996,

2000 and 2002 (source: PAFO Luang Namtha in Shi, 2008)

The classification of the satellite imagery is done by adapting the approach of nata et al (2004) using ASTER imagery instead of the defect Landsat images The ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) satellite was launched in 1999 and provides images with similar properties as the pre-2003 Landsat:

Ekadi-“ASTER covers a wide spectral region with 14 bands from the visible to the thermal infrared with high spatial, spectral and radiometric resolution (…) The spatial resolu-

tion varies with wavelength: 15m in the visible and near-infrared (VNIR), 30m in the short wave infrared (SWIR) and 90m in the thermal infrared (TIR) Each ASTER scene covers an area of 60 x 60 km,” (Abrams, 2002: 8)

Figure 4: Comparison of Spectral Bands between ASTER and Landsat-7 Thematic Mapper (% Ref is reflectance percent) (source: Abrams, 2002: 10)

Figure 4 shows a comparison between the spectral bands of the ASTER sensor with those of the Landsat 7 Both cover a similar range of wavelengths, with the ASTER instrument being more accurate for long wavelengths, while the Landsat 7 instrument reaches further into the visible infrared Especially ASTER’s VNIR instrument is of interest for the study at hand, as the visible to near infrared can be used for vegetation detection and the calculation of vegetation indices

For selected regions in Northern Laos, ASTER satellite images for two time-cuts (2001 and 2006) were acquired Using multi-temporal image analysis is meant to ease the task

of detecting the rubber plantations by including the interpretation of land cover change trajectories Adverse to the Landsat satellite, the ASTER does not capture the images using a path and row system It has a sensor that can be adjusted and is able to provide images as demanded by the customer The images purchased for the study at hand were available from the archive and therefore the two time-cuts do not match spatially To cover the area of the two 2006 images three images from 2001 had to be bought

Different levels of pre-processed ASTER data are available The ASTER Level-1B data acquired for this study comes with the necessary geo-referencing information When taking a closer look at the two time-cuts, shifts of different sizes appeared be-tween the images In the rugged landscape of Northern Laos the geo-referencing infor-mation is not accurate enough to make the satellite images fit with the actual landscape Therefore, a rectification of the images had to be conducted An ortho-rectification

within the study area did not suffice The ASTER data was rectified to an rectified Landsat image using the ‘spline’ transformation in ArcGIS with an average number of displacement links of 35

ortho-The ASTER imagery was used to calculate two indices that serve as an addition for the image classification: The Normalized Difference Vegetation Index (NDVI), and the Tasseled Cap Index The NDVI is a calculation between the visible and the near-infrared bands (bands 2 and 3 for ASTER images)

NDVI = (NIR – VIS) / (NIR + VIS)

Equation 1: Calculation of the NDVI (source: Weier & Herring, year unknown)

The visible and the near-infrared is reflected or absorbed by the plants in a different manner The chlorophyll (green leaves) strongly absorbs visible light for the use in photosynthesis, while the cell structure of the leaves reflect near-infrared light If there

is more reflected radiation in the near-infrared wavelengths than in the visible, the vegetation is more likely to be dense and indicates some kind of forest cover, and vice versa The calculation of the NDVI returns values between -1 and +1 Negative values close to -1 mean no vegetation at all; around zero means no green leafs and values near +1 indicate a high density of green leafs (Weier & Herring, year unknown)

The Tasseled Cap was developed for an application with Landsat images Basically, it

is a transformation of the six available TM bands into six new bands by applying ferent linear functions to each band Three of the six bands are often used: Brightness (measure of soil), Greenness (measure of vegetation), and Wetness (interrelationship of soil and canopy moisture) The algorithm for the Tasseled Cap transformation could be downloaded from the homepage of the Geoinformatics Center of the University of Mis-sissippi (Yarbrough et al., 2004; NCRSTE, 2008)

The extraction of information from Earth Observation data has been the same for many years: The spectral reflectance of the earth’s surface is stored in rectangular format and measured in its ground level resolution, the pixel, which serves as smallest unit for classification Interpretation and analysis are conducted on single or a combination of pixel values without taking contextual information into account (Baaz et al., 2004) Problems arise especially when pixels display transition areas from one land cover to another and thus show mixed values The class assignment of these pixels is much lower than for any given pixel lying within a certain land cover

The pixel-based classification, however, ignores a substantial proportion of tion, as only the spectral properties are considered, without looking at spatial associa-tions like context and neighbourhood information, as well as the geometrical shape of the land cover units Especially, when working with medium to high resolution data, it

informa-is more likely that a certain land cover informa-is represented by a group of pixels and maybe a certain shape Considering a pixel within a certain land cover (i.e water), it is more probable that the surrounding pixels belong to the same land cover class than the pixels

further apart The pixel as the smallest unit of classification, therefore, does no longer meet the requirements (Benz et al., 2003; Yu et al., 2006)

Recent developments in photogrammetry software have taken these facts into account Generally, two different approaches to image segmentation are applied: edge-based segmentation, and region-based segmentation The first algorithm aims at detecting the boundaries (edges) between the image objects, while the region-growing algorithm clusters pixels starting at a certain seed point until a defined threshold is reached The threshold usually is a homogeneity criterion or a combination of homogeneity and size

(Blaschke, 2000) According to Benz et al (2003: 240) the “advantages of

object-oriented analysis are meaningful statistic and texture calculation, an increased related feature space using shape (e.g length, number of edges, etc.) and topological features (neighbour, super-object, etc.), and the close relation between real-world ob- jects and image objects This relation improves the value of the final classification and cannot be fulfilled by common, pixel-based approaches.”

uncor-2.4 Input data used to model the potential spatial

distribu-tion of rubber plantadistribu-tions

Modelling those areas where potentially rubber plantations can be established means balancing between data needs and data availability, but it also implies also considera-tions about the importance of an included dataset Generally, the input data for model-ling can be split up into the biophysical factors (according to the crop requirements of the plant) and the socioeconomic and political factors (rather influencing the distribu-tion of the plant within suitable areas)2 The determination of the biophysical factors was done by reviewing literature on crop requirements and species information, while the identification of the relevant socioeconomic and political factors is based on litera-ture review, but mainly on informal talks with farmers and officials, economic consid-erations and the laws and policies of Laos The following datasets were available and found to be useful for modelling the spread of the rubber trees: Elevation (digital ter-rain model), slope and different soil properties as biophysical input data, and accessibil-ity, forest and land cover of 2002, and protection areas as socioeconomic inputs

Elevation (DTM)

The Digital Terrain Model (DTM) was provided by the Mekong River Commission Secretariat It originally covered the area of the Lower Mekong Basin (LMB), and in-cluding the parts of Laos and Cambodia lying outside the LMB The source maps for the DTM are 1:50’000 American topo maps and partially 1:100’000 Russian topo maps; the raster cell size is 50m The dataset was created by the Watershed Classifica-tion Project (WSCP) and last updated in October 2000

opportuni-Slope

The slope was calculated using the Digital Terrain Model Slope values are in percent Source maps are, as for the DTM, the 1:50’000 American and the 1:100’000 Russian topo maps; the raster cell size is 50m The dataset was created by the Watershed Classi-fication Project (WSCP)

Forest / Land cover

The land cover data was provided by the Ministry of Agriculture and Forestry (MAF) and represents the land cover in 2002 The dataset is part of a forest cover and land use changes assessment during the period from 1992 to 2002 conducted in Laos The data-set is produced by interpreting SPOT satellite images at 1:50’000 and 1:100’000 scales and by conducting forest and land use mapping and field verification It distinguishes

24 different types of land cover, of which ten represent different forest classes (MAF, Department of Forestry, 2005) Unfortunately, no metadata for the dataset could be obtained There exists a report on the assessment of forest cover and land use during

1992 to 2002, but there is little to no information on how the land cover data was rated

elabo-Production Forest, National Biodiversity Conservation Area and Protection Forest

The data delineating the different forest areas according to forest classification type is provided by the Ministry of Agriculture and Forestry The delineations for the Produc-tion Forests and the National Biodiversity Conservation Areas (NBCA) are in a dataset for each type, while the Protection Forest Areas are divided into the following sub-categories: Watershed Protection, Reservoir Protection, and Tributary Protection Alto-gether, theses forest areas account for approximately 138’256 km2 or a 58% share of the country A description about the laws and responsibilities within the different areas

is given in section 3.2

Accessibility

The circumstances how a certain landscape at a specific site can be reached are seen as

a key determinant of the characteristics of this landscape Accessibility is thus one of the most important driving factors of land use and land cover change (Kaimowitz & Angelsen, 1998; Chomitz & Gray, 1996; Geist & Lambin, 2002 in Heinimann, 2006; Verburg et al., 2004) The theoretical basis for this approach reaches back to von Thünen’s bid rent model describing that land is normally devoted to the use that gener-ates the highest returns Roads, and thus accessibility, play a crucial role in determining this rent and consequently influence land use and cover (Chomitz & Gray, 1996; Nel-son, 2001; Verburg et al., 2004; Heinimann, 2006)

However simple approaches relating distance to the road with deforestation mate the causality, as roads can be built after an area was cleared or vice versa (Kai-mowitz & Angelsen, 1998; Verburg et al., 2004) Additionally, distance seems to be an inappropriate measure, as the quality of the transportation network and the costs of transportation need to be taken into account Measuring the cost or time of travel can provide more realistic results (Mertens et al., 2002; Verburg et al., 2004) Land use and land cover change, however, are complex processes that are affected by different bio-physical, political and socio-economic conditions and spheres of influence (Nagendra

overesti-et al., 2004) According to Heinimann (2006), different relations of accessibility and land use and land cover change can be expected because of the heterogeneous devel-opment status within Southeast Asia In the study at hand, accessibility is therefore calculated on different levels, as accessibility from villages presumably has a direct influence on land use and land cover change, while accessibility from cities or neighbouring countries can have an underlying effect and result in contrasting patterns

On the background of modelling the potential spatial distribution of rubber plantations, three representations of accessibility are calculated: village accessibility, accessibility

to provincial capitals, and accessibility to riparian countries

Development of accessibility surfaces

The calculation of the accessibility is done using the cost-distance model included in the software package of ESRI’s ArcGIS 9.2 It calculates the shortest weighted distance

or travel cost from a source dataset (e.g villages) across the friction surface The input factors for creating a friction surface are roads, slope, land cover, or even the cost for public transport.3

Three different accessibility representations are calculated: village level, provincial capital level, and access to neighbouring countries The friction surface remains the same for all three representations; only the source for the calculation of cost is altered First, a glance at the elaboration and assumptions made to construct the friction surface

is taken Six variables serve as input for estimating the travel time to cross each cell (in

this case 50 m x 50 m): road network, slope, land cover, main rivers and lakes, country boundaries and official boarder check points

• Road network: The road network can be considered as backbone for

estimat-ing how much time is needed to cross each cell To each road class an average speed is related The assignment is based on the experience of A Heinimann and his personal contacts; average speed on flat ground is assumed: National roads 70 km/h, provincial roads 40 km/h, car tracks 25 km/h The influence of slope on the speed is included at a later stage.4

• Slope: The inclusion of slope is done by a simple multiplication of the time

re-quired to cross a cell with a factor according to the road class or walking speed

if no road is available The factor ranges from 1 for flat areas to 6 for very steep slopes The factors for the road classes were kept smaller, as cars are assumed

to have more power reserves to compensate the slope.5

• Land cover: Based on Heinimann’s experience, walking speed (average on and

off trail) within forest areas is considered as 3 km/h, in all other areas (except waterways) 4 km/h The influence of the slope on walking speed was already integrated in the calculation above.6

• Main rivers and lakes: Within the landscape rivers and lakes have two

differ-ent functions On one hand they are natural barriers separating landscapes, and

on the other hand they can serve as transport routes connecting different gions In the work at hand, rivers and lakes are only integrated as obstacles, as the inclusion as transport routes is complex (different values for up- and down-stream movements) and thus beyond the framework of present thesis

re-• Country boundaries: Considering country boundaries in terms of travel time

as an obstacle, they are integrated in the friction surface, too Borders are sidered as a major obstacle if there is no official check point or road crossing A value of three hours is therefore set to cross the border making the crossing pos-sible, but time consuming as it is illegal

con-• Border check points: The location of the different official border check points

are provided by various contacts and through online investigations Border check points are considered as ‘holes’ in the barrier the country boundary nor-mally represents To include the custom formalities when crossing the border in the friction surface, the speed on the roads in the border section is limited to 3 km/h

assistance and the Demining Defense Security Cooperation Agency (DSCA) of the U.S Department of Defense, in cooperation with the Lao Ministry of Construction, Transport, Posts, and Communications (MCTPC) The data was produced based on high resolution SPOT imagery (5-10m) The scale of the road data is approximately 1:25’000

Commis-sion Secretariat

cover in 2002 The dataset is produced by interpreting SPOT satellite images at 1:50’000 and 1:100’000 scales and by conducting forest and land use mapping and field verification

The combination of these indicators results in the friction surface, one of the two inputs

in the cost-distance function For running the function, one has to define the source points from where accessibility is to be delineated In the study at hand three different accessibility representations are considered, thus three different sources are defined: villages, provincial capitals, and riparian countries All three datasets could be obtained from the database of the Socioeconomic Atlas of Laos, a joint project of the NCCR North-South, the National Statistics Centre (NSC), and the Lao National Mekong Commission Secretariat (LNMCS)

• Village points: Village points were collected in 2005 in a GPS survey of the

NSC during the population census

• Provincial capitals: The layer indicating the provincial capitals of Laos is

pro-vided by the Mekong River Commission (MRC)

• Riparian countries: The accessibility of Laos from the neighbouring countries

is calculated using the Laotian country boundary as source The boundary is provided by the National Geographic Department (NGD) and represents the lat-est boundary delineation A 5 km buffer is applied, thus the roads in the sur-rounding countries and the border points play a crucial role in determining the accessibility of Laos from the surrounding countries

For the classification of the satellite images and for the model of the potential spatial distribution of rubber plantations, thresholds need to be defined indicating the occur-rence and absence of a phenomenon In general, real world phenomena show a degree

of complexity that can hardly be captured and modelled A process (i.e land use and land cover change) underlies a multitude of factors influencing each other in different ways, depending for example on the spatial, socio-economic, and political setting Un-certainties play a crucial role in describing a process, especially the more complex it gets Within this context, the ability to formulate significant statements is diminished with increasing complexity At a certain point the precision and the relevance start to exclude each other (Zimmermann, 1993)

Classical approaches using hard classifiers are not able to take this uncertainty into account Human thinking and language are able to express this vagueness When de-scribing a problem verbally, we create a model of the reality that reduces the complex-ity to a manageable degree – leaving space for incertitude An example is, when a per-son is considered tall “tall” then generally is not an expression that includes all the people taller than 2 metres, but it is rather a range, i.e starting from “quite tall” at 1.90 metres to “really tall” at 2.10 metres This fuzziness leaves space for uncertainties that cannot be expressed using hard classifiers

The mathematical approach to quantify uncertain statements is the fuzzy logic The two strictly logical statements “yes” and “no” are replaced by the continuous range of [0…1], where 0 means “exactly no” and 1 means “exactly yes” The process of assign-ing these values is called fuzzification – a membership degree between 0 and 1 is as-