Training Workshop on Forest Landscape Models and Land-use by wuzhenguang


									          START FELLOWSHIP REPORT

      Training Workshop on
  Forest Landscape Models and
  Land-use Change in Sumatra

                       Reported by:
                    Dr Budi Suharjo
                  Mr. Agus Eka Putera
               Mr. Desi Ariyadhi Suyamto

Department of Environmental Sciences, University of Virginia
           21 October 1997 - 6 December 1997
START Fellowship Grantees

              Dr Budi Suharjo.          Lecturer     on applied
              mathematics      for   environmental      sciences,
              Department     of    Mathematics,      Faculty   of
              Mathematics and Natural Sciences, Bogor
              Agricultural University (IPB), Jl Raya Pajajaran,
              Bogor 16143, Phone: 62-251-338488, 313384,
              Fax:     62-251-312708,        E-mail:    ppm-mat-

              Mr. Agus Eka Putera. Research assistant, Remote
              Sensing and Ecology Laboratory, BIOTROP, Jl
              Raya Tajur Km 6, P.O. Box 116, Bogor 16000,
              Indonesia. Phone: 62-251-323848 ext. 160, Fax:
              62-251-371656,                            E-mail:

              Mr. Desi Ariyadhi Suyamto. Technical Assistant
              for    Ecosystem      Modelling,     BIOTROP-
              GCTE/Impacts Centre for Southeast Asia (IC-SEA),
              Jl Raya Tajur Km 6, P.O. Box 116, Bogor 16000,
              Indonesia. Phone: 62-251-371655, Fax: 62-251-
              371-656, E-mail:
A mosaic of colorful patches in better understanding landscape

1. Introduction
Sumatra is one of regions in Indonesia that is experiencing very rapid change in land-
cover and land-use. The high rate of deforestation on this island amounting to 367.7
thousand hectares per year in 1981-1991 period (FAO 1990) well illustrates the
magnitude of the change. Such a rapid change could modify environmental quality,
both in short-term and long-term, which in turn would affect the capacity of the region to
support livelihood of communities especially whose life depends directly to the
environment and also to support the development at large.

The prominent feature of the changes in land-use/-cover in Sumatra has been the
conversion of humid tropical rainforest into other land-uses. There is a world wide
concern over the reduction in natural forests in this region since it is a home of millions
of plant and animals species including those classified as endangered species.
Degradation of ecosystem services in maintaining vital processes such as nutrient,
hydrological, carbon cycle is another concern. In other words, the overall impact of the
changes in land-use/-cover on biophysical components, such as soil-C stock, SOM (Soil
Organic Material) dynamics, vegetation (structure, function, biodiversity) and
atmosphere (GHG concentration etc.) will not only be felt at local and regional level,
but also at global level.

It is widely accepted that the changes in land-use/-cover are a result of complex
interactions between biophysical, social, economic and political factors (Turner, Skole
et al. 1995). In Sumatra, population increase, extensive and intensive mode of
agricultural practices, the transmigration project, forest harvesting activities as well as
natural disturbances such as fire are among the forces that play an important role in
driving the changes. It is noticeable, therefore, that those changes are mostly human-
driven changes.

The influence of humans in driving the changes in land-use/-cover can take place at
various spatial and temporal scales. For instance, farmers' decision on ‘what to grow this
year’ determines the land-use pattern at the smallest scale (i.e. field). On the other hand,
government policy, such as forest utilisation policy (e.g. setting aside an area for
logging, non-timber forest product harvesting, transmigration project and conservation)
operates at higher scales (e.g. sub-district) which in this case plays an important role in
determining the ‘allowable’ land-use change for a part of the region.

We should also think that there is a set of biophysical, socio-economic and political
context driving those human decisions. For example, the decision of a farmer to replace
unproductive jungle rubber with oil palm plantation can be driven by the promising
benefits of the oil palm since the demand of palm oil in the world market is high. The
same economical consideration may also drive the government policy to allocate some
forest area under the conversion forest category to be used for large-scale oil palm
plantation. The point we would like to make is that, to understand the dynamics of the
changes in land-use/-cover in Sumatra, it is important to recognize the complexity of the
problem and the significant role of social, economic and political factors in driving the

Modeling exercise is one tool that can be used to study the dynamics of land-use/-cover
change. Modeling allows us to explore possible changes in landscape compositions and
its environmental consequences for a given set of assumption of the driving forces. This
kind of ‘what-if ’ exercise could ultimately be valuable as a tool to assist environmental
management of the region.

            START FELLOWSHIP REPORT                                                               1

However, developing a land-use/-cover change model is not an easy task considering
the complexity of factors driving the changes. It is essential therefore that such an effort
take a multidisciplinary approach. A wide range of expertise such as forestry, economics,
statistics/mathematics, biology and computer science is required in achieving a
comprehensive understanding of the nature of the changes and in formulating those
understanding into the models. There is an opportunity to undertake a land-use/-cover
change modeling exercise for Sumatra since there has been a number of studies
conducted by various research institutions (regional, national and international
institutions) in this region that could provide a valuable information to start the model

“The Training Workshop on Forest Landscape Models and Land-use Change in Sumatra”
held by Department of Environmental Sciences, University of Virginia (UVA) is one step
towards that direction. This course, which was coordinated by Prof. H.H. Shugart,
involved UVA students from a wide range of fields, i.e. environmental sciences,
anthropology, environmental planning as well as an Indonesian team consisting of
persons with forestry, ecology and mathematics background. During the workshop, the
UVA students and the Indonesian team interacted to exchange information and ideas to
build an understanding of the problems related to the changes in land-use/-cover in

            Figure 1. Landscape ecology of Sumatra.

As a part of the training, both the UVA students and the Indonesian team developed
land-use change models with a slightly different approach. In this exercise, the
Indonesia team took a benchmark areas of Alternative to Slash-and-Burn (ASB) Project in
Jambi province as a target area (i.e. Muaro Tebo). The description of the modeling
approach will be elaborated later. It is expected that the result of this exercise can be
refined and developed in the future since there is a substantial gap of knowledge and
lack of data in our present model.

            START FELLOWSHIP REPORT                                                               2

2. Activities
Period : 21 October 1997 - 6 December 1997 (at University of Virginia)

       Activities during this period includes:

       n   Meeting with the class on Tuesday, 06:00-09:00 p.m. In this meeting, we
           exchanged information with UVA students regarding the land-
           use/forestry issues in Jambi/Sumatra. We helped the UVA student to
           choose a target area to be modeled and also provided them with some
           information from ASB Report and other sources. In addition to the formal
           meeting in the class, we also had some discussion outside the class with
           some groups when they needed additional information/clarification.

       n   Attending Prof. H. H. Shugart’s lecture on “Ecological Issues on Global
           Change”. This lecture was held on Tuesday and Thursday, 09:30-10:30

       n   Attending Environmental Sciences Seminar, Thursday 04:00-05:00 p.m.

       n   Outside the scheduled activities above, the Indonesian team worked in-
           group to develop a model of land-use change for Sumatra.

       n   Literature review at UVA library

       Figure 2. Developing land-use change model.

       START FELLOWSHIP REPORT                                                               3

Period : 7 December 1997 - 10 December 1997 (in San Francisco)

       The main activity during this period is attending American Geophysical
       Union (AGU) Fall Meeting at Moscone Center, San Francisco.

       Figure 3. Attending AGU 1997 Fall Meeting, San Francisco, California.

       START FELLOWSHIP REPORT                                                               4

3. Result
Modeling Land-use Change for Sumatra with Muaro Tebo,
Jambi as a Case Study                   -

3.1. General Approach

         The area to be modeled in this exercise was sub-district Muaro Tebo, Jambi
         Province of Sumatra. Muaro Tebo is a benchmark area of Alternatives to
         Slash and Burn (ASB) project, which is now still underway in Sumatra (van
         Noordwijk, Tomich et al. 1995). This area is an interesting spot to study,
         since there has been a massive land-use change in the last 20 years
         involving the role of many parties with many interests. It also provides a good
         case study of land-use change study where factors like human population,
         government policy in forestry, agriculture and transmigration sector and
         other socio-economic factors interact to influence the dynamics of the
         landscape. The general approach for developing a model of land-use

         change for this area is briefly described bellow.

         Figure 4. Study sites of Alternatives to Slash and Burn (ASB) Project in

             This is a result of team work consisting of Dr Budi Suhardjo, Ms Titiek Setyawati , Ms Endah Sulistyawati,
              Mr Agus Eka Putera, and Mr Desi Ariyadhi Suyamto.

         START FELLOWSHIP REPORT                                                                                          5

We started the model development by obtaining a general picture of what happened in
the landscape. The purpose was to recognize observable land-use components and to
identify the magnitude of the change. In this regard, remote sensing technique is a
powerful tool which can be use to achieve that objectives (Murdiyarso and Wasrin 1995).
It is a particularly useful tool in the process of determining the level of detail of land-use
components to be included into the model. In this exercise, we used remotely sensed
data available for Muaro Tebo (BIOTROP, unpublished data) to identify the land-use
types and the magnitude of the changes among those types over 1988-1996 period. Also
included in this step also was combining several land-use types together into a new type
in order to simplify the land-cover representation in the model. After that, to quantify the
magnitude of the change, a parameter describing the annual rate of conversion among
the land-use types was derived. In this case, an annual conversion rate refers to the
proportion of a land-use type that is converted into another type annually.

The second step was using the information on annual conversion rate to construct a
simple land-use change model in the STELLA environment. The basic structure of the
model consists of land-use type compartments and annual conversion rates between the
compartments illustrating the flow of land- use in the landscape. At this stage, the
annual conversion rate was the only parameter determining the dynamics of the system
without considering the mechanism of the change. The model simulates the distribution
of the land- use overtime.

Although the model seems somewhat unrealistic because of the constant annual
conversion rate, there are some points that can be highlighted from this simple model.
Firstly, it allows ones to evaluate the performance of the model in the STELLA
environment. This is done by setting the model with the initial condition as 1988 data
and running it for eight years (1988-1996), during which the rate conversion data is
derived. The performance of the model is assessed from its ability to predict the
distribution of land-use type in 1996. Secondly, in this model, we basically make an
assumption that the rate of change is constant. Taking this assumption, we can use the
model to explore questions such as: ‘If the present conversion rate continues, what
would be the land-use distribution after 100 years?' or ‘In what year would the secondary
forest be completely gone?'

Furthermore, since the actual land-use dynamics is a result of complex interaction of
biophysical, social, economic and political factors, any ideal form of a land-use change
model should take those factors into account. Therefore, in the third step in the model
development, we tried to identify driving forces affecting land-use conversion. Then, the
model is expanded to make it more ‘mechanistic’ by replacing the constant conversion
rates with sub-models simulating the mechanisms by which a set of driving forces affects
a particular land-use conversion. The following sections will describe each step of the
model development in more detail.

3.2. Pattern of Land-use Change in Muaro Tebo 1988-1996

To identify the pattern of land-use change in Muaro Tebo, we used a map derived from
LANDSAT TM with the scale of 1:250000 (BIOTROP, unpublished data) as shown in
Figure 5. There are seven land-use types suggested in the map. We then reduced the
number of types in order to simplify the land-use representation in the model by merging
some of them. The new land-use classification is presented in Table 1.

            START FELLOWSHIP REPORT                                                               6

            Table 1. A new land-use classification used in the model.
                     Types used in this model                           Types as in the maps
            Primary forest                       Lowland primary forest
            Secondary forest                     Lowland logged-over forest
                                                 Old secondary forest
            Jungle rubber                        Mosaic of smallholder rubber and young secondary forest
                                                 Mosaic of small holder and thicket or shrub
            Mosaic of crop, homegarden,          Mosaic of Crops, fruit trees, alang-alang and settlement
            settlement and grassland (Mosaics)
                                                 Crops (ladang)
            Rubber plantation                    Rubber plantation
            Oil palm plantation                  Oil palm plantation

            Figure 5. Land-use/-cover change in Muaro Tebo, Sumatra, from 1988 to

Furthermore, the maps were processed to quantify the pattern and the magnitude of the
change. Since we did not have the data in a digital form, the maps were manually
divided into 0.5 X 0.5-cm2 cells and a coordinate was given for each cell. Then, a land-
use type --according to the new classification-- was assigned for each cell. In situations
where there are more than one land-use types within a cell, the assignment were based
on the dominant land-use type. We were fully aware that this was a very coarse method,
as there might be some degree of subjectivity and errors when reclassifying the map.
However, with limitations of the data we had this method should be enough to as a
mean to depict the general feature of land-use changes.

            START FELLOWSHIP REPORT                                                                         7

Pairs of cells of the same coordinate from 1988 and 1996 data were compared to
identify whether there had been a conversion in land-use for each pair. After
determining the land-use conversion types (e.g. secondary forest to jungle rubber) for all
pairs, the number of cells associated with each conversion type were calculated and
then converted into area (ha), as presented in Table 2.

            Table 2. Summary of the land-use change during 1988-1996
                  Initial area(ha), 1988                                Converted into, 1996 (ha)
                                               Prim.        Sec.         Jungle      Mosaic       Rubber      Oil palm
                                               forest       forest       rubber                   plant.
              Prim. forest           4531.25            0        4375            0       156.25           0          0
              Sec. forest             75000             0     47812.5        13750         3750        2500     7187.5
              Jungle rubber           34375             0     5156.25        28125          625       312.5     156.25
              Mosaic                27656.25            0     2656.25       2187.5        18750     3906.25     156.25
              Rubber plant.            1875             0           0            0            0        1875          0
              Oil palm                     0            0           0            0            0           0          0

Several points can be made regarding the general feature of land-use change in Muaro
Tebo. All primary forest were gone by 1996 and converted into secondary forests. In this
case, the logging operation seems the most likely factor driving this conversion. During
1988-1996, the conversion of primary forest was only one direction with no inflow back to
primary forest establishment. On the contrary, the change in area of secondary forest,
jungle rubber and mosaic was a result of two opposite processes i.e. inflow from and
outflow to a particular land-use type. These two processes can be both human-driven
and nature-driven phenomena. Conversion from secondary forest to mosaic is likely to be
human-driven, but conversion from jungle rubber to secondary forest could be as a result
of a succession change. For rubber plantation and oil palm, one-direction conversion
also took place with no outflow to other land-use type during the eight-year period.

Furthermore, using the data in Table 2, the proportion of total cells of each land-use
type in 1988 that has changed to other land-use types in 1996 were calculated. Then,
those proportion were divided by eight to determine annual conversion rate variables
associated with each land-use conversion type. To summarize the pattern of land-use
change, those two variables were combined into a diagram shown in Figure 6.

            Figure 6. Annual land-use conversion rate of Muaro Tebo, 1988-1996

            START FELLOWSHIP REPORT                                                                                      8
3.3. A Simple Model of Land-use Change for Muaro Tebo

Using the information from the diagram above, a simple model of land-use change was
constructed in the STELLA environment. The structure of the model is very simple
consisting of land-use types as state variables and annual conversion rates as
parameters. The structure of the model is presented in Figure 7.

To evaluate the performance of the model in the STELLA environment, we run the
model for eight years --period during which the annual conversion rate data is derived.
The result as shown in Figure 8. Indicates that the land-use change simulated during
1988-1996 follows a similar trend with the actual data. However, there are slight
differences in area per land-use types between the model prediction and the actual
data. This small discrepancy could be due to the non-linear nature of a model
constructed in this way, since the rate of conversion is expressed as a proportion of a
state variable rather than an absolute number.

           Figure 7. A simple representation of a land-use change model for Muaro

               Area (ha)

                           50,000.00                                                              Initial (1988)
                           40,000.00                                                              Actual(1996)
                           30,000.00                                                              Predicted(1996)
                                                                           Rub. plan

                                                                                       Oil palm


                                                      Land use type


           Figure 8. Predicted and actual land-use distribution for eight year simulation.

           START FELLOWSHIP REPORT                                                                                  9
The result of the simulation for 200 years is presented in Figure 9. It should be kept in
mind that, in this case, we make an assumption of a constant rate of change. This could
be an unrealistic assumption, especially when the model is used to predict long term
change. According to the figure, primary forest is completely gone after 25 years. The
areas of secondary forest, jungle rubber and mosaic decrease overtime. In contrast, area
of oil palm and rubber plantation keeps increasing and dominates the landscape by the
end of the simulation. The domination of oil palm and jungle rubber is a consequence
of the model construction. If we look at the pattern of land-use changes shown in Figure
6, it is clear that these two land-use types did not have outflows. It means that there was
no conversion from these land-use types to other land-use types. Therefore, the land
under oil palm and rubber plantation keeps accumulating over time, which could also
be an unrealistic situation. In summary, the simple land-use change model described
above provides a general overview of the land-use flow over time. However, the constant
conversion rate might not be appropriate in this case, since there are many factors that
can modify the magnitude of conversion over time.

                                                                                Oil palm
                   Area (ha)

                        60000      Secondary

                                                                                Rubber plantation


                                               Mosaic                            Jungle rubber
                                   2000            2050        2100             2150

            Figure 9. Land-use distribution for 200 year simulated using the simple model.

3.4. Identification of Driving Factors of Land-use Change and a Preliminary
     Construction of ANDALAS Model (ANalysis of Driving factors
     Affecting the LAnd-use change on Sumatera)

After having the general pattern of the land-use change, the model was expanded to
incorporate factors that drive the change. Incorporating driving factors in a land-use
change model would allow ones to evaluate the role of different factors affecting the
dynamics of a landscape. Considering the complexity of the problem in the field,
inclusion of such driving forces is ideally based on a deep and comprehensive study
involving expertise from many related fields. However, at this occasion, we mostly based
the driving forces assessment on our knowledge and available data from various sources
that we could get during the training such as reports, scientific papers and newspapers,
which we believe is still limited. Most of the information used were based on the
following publication: van Noordwijk, Tomich et al. (1995), Gouyon, De Foresta et al.
(1993), Zaini, Basa et al. (1996), Ginting, Sumarhani et al. (1996), Lusiana, Suyanto et
al. (1997) and newspapers.
As a consequence, the model construction is still in a preliminary stage with more focus
on identifying possible driving factors and establishing the framework rather than
obtaining a right parameter value. Due to time and data limitation, not all possible
            START FELLOWSHIP REPORT                                                               10

driving factors could be identified and implemented in the model construction.
Therefore, in this report I will only describe the conceptual framework of how a set of
possible driving factor affect land-use conversions. No simulation result will be presented
in this report, as the model construction has not yet completed.
In analyzing the driving factors, I will discuss the possible causes/mechanisms behind
land-use change using the following context: human population dynamics, forest
utilization, crops cultivation, jungle rubber cultivation, monoculture rubber cultivation,
oil palm cultivation and fire disturbance. In terms of the structure of the model, the main
part of the model consists of land-use type compartments --as in the simple model -- and
sub-models describing the mechanism of conversion in a particular transition type.
Therefore, each sub-model has a link to the main part and could have links with other
3.4.1 Human Population Dynamic

In this land-use change model, the human population plays an important role in
determining the dynamics of the system as a whole. That population is closely related to
the land-use change is evident in many part of the world (Meyer and Turner II 1992).
Population can affect the dynamics of land-use directly through demand of land for
agriculture and settlement as well as indirectly through its role in supplying labor for
many economic activities such as labor for logging companies, rubber tapping and oil
palm plantation.
In this exercise, the population sub-model was constructed in such a way by considering
the dynamics of the system related to inflow and outflow components and their
derivation into components needed by other sub-models. Inflows components include
birth, immigration (spontaneous immigration) and transmigration (government-assisted
immigration), while the outflow components consist of death and emigration. From this
basic structure, information for other sub-models such as labor force and number of
agricultural household can be derived.
In the structure of this sub-model, population is segregated according to sex (male and
female) and age class (5-year interval). Representing population structure in this way will
facilitate the derivation of components related to in-flow and out-flow such as (1)
female’s reproductive age which relates to birth rate (2) labor force supply (3) mortality
rate distribution (4) immigration (transmigration and non-transmigration) and (5)
By segregating female population according to age, we can determine the group of
female that falls into a reproductive age --a period within which a female is capable of
having babies. The data for Muaro Tebo shows that the reproductive age was from age
16 to 45. In this model, we made an assumption that birth rate varies across the age
classes within the reproductive period. Age class 21-25 has the highest birth rate and
then the birth rate decreases progressively toward the oldest age class. In reality, there is
a tendency that birth rate is decreasing overtime, as also happened in Bungo Tebo.
Improvement on female education, family planning program and the change of
perception toward an ideal family size is among the factors that contribute to the
decrease in birth rate. However, we did not incorporate those factors into the model and
assigned constant birth rates instead. The babies born from all females then enter the
lowest age class with the ratio of women and man of 0.5.
The 1990 data for Muaro Tebo also shows that mortality rate varied across age classes
with the highest value for the age class 0-5 amounting to 50 % --one-third of which was
baby death. Based on this information, we assigned a low survival rate for age 0-5 and a
uniform survival rate for age class 6-10 until 71-75. The survival rate for age 76 upward
was assumed to be zero. As in the birth rate, the mortality rate also shows a decreasing
trend, which could associate with the improvement on sanitation, nutrition and medical
services etc.
            START FELLOWSHIP REPORT                                                               11

The migration component of the sub-model can be divided into immigration and
emigration. Including in immigration component is spontaneous migration and
transmigration --a government-assisted migration. Jambi has been a recipient area for
transmigration program since 1950. In this program, each farmers is given a plot of land
(2 ha) and expected to establish a food crop-based farming or a perennial crop-based
farming. The land for this project is usually taken from the state forest. Since it is a
government program, then the number of transmigrant brought into this are will depend
on the government plan. At the present model, we only make an arbitrary estimation on
the number of people brought into the area through this program.
The driving factor for spontaneous migration, either immigration or emigration, would be
likely an economic reason. The availability of opportunity in making a better life in an
area could act as a push or pull factor. We suggest that one way to express this
relationship is by comparing an economy performance variable for local and national
level, which may be represented by Gross Regional Domestic Product (regional) and
Gross Domestic Product (national). In situation where the local economy is better than
other places, it is likely that some people would be pulled to migrate into the area. This
has been the case for Jambi where easy access to available land and the promising
benefit from rubber has attracted people from outside to establish rubber cultivation
(Gouyon, De Foresta et al. 1993). Likewise, when the performance of local economy is
poor, some people would be likely to be pushed to seek a better life elsewhere. In the
present model, the expression of the economy performance relationship is a
hypothetical one.
The spontaneous emigration, immigration and transmigration component is combined
to obtain a migration net flow, referring to the total people migrating into Muaro Tebo.
The migration net flow then becomes an inflow to local population. When distributing
the migrants across the age classes, we made an assumption that the age of migrant is
normally distributed with the peak at age class 35 - 45.
From this structure, some components related to labor force and agriculture activity,
which eventually drives the land-use change, can be derived. Labor force is defined as a
number of people under the productive age, i.e. 15-60. This is the amount of labor that
may engage in agriculture and non-agriculture activities. To derive the agriculture labor
force component, we simply take a proportion from the total labor force.
According to our observation, household can be used to represent the unit of production
in agriculture system. It is at the household level that decision on matters like ‘what to
plant this year’ and ‘where to clear a new field’ are made. Therefore, we may view the
land-use dynamics in the context of relationships between the number of agricultural
household and land-use conversions related to agriculture activities. Taking this view,
the member of an agricultural household could consist of several farmers (e.g. head of
household, wife, son etc) with no or few non-farmers. Then, the number of agricultural
household is derived by dividing the (total) agriculture labor force by the average
number of farmer per household (AFH). This agricultural household variable will then be
exported to other sub-models. Note that in this framework, the agricultural household
derived already takes into account the transmigrants.
3.4.2. Forest Utilization

The Indonesian government in 1980’s established a forest policy called ‘Agreed Forest
Use Categories’ (TGHK). This policy classifies all state forestland under following
categories: conservation forest, protection forest, limited production forest, production
forest and conversion forest. This is a policy that regulates state forest allocation for
different forest uses. Forest utilization especially in the form of timber extraction can take
place in the primary forest or regenerated logged-over forest under production forest

            START FELLOWSHIP REPORT                                                               12

In Muara Tebo case study, we suggested that the main driving factor of conversion from
primary forest to secondary forest was logging operation mainly carried out by
concession holders, as presented in Figure 10. The timber market will certainly affect the
rate of the conversion. But, role of government is also significant, since the scale of the
logging operation will be determined by how much primary forest is allocated under
production forest. According to the government regulation, the timber volume that can
be cut in during the concession period (35-year) is 80% of the potential timber volume.
Therefore, annual volume of timber exploited is the above number divided by 35.

            Figure 10. Conversion from primary forest into secondary forest.

However, in many cases, the conversion from primary forest to secondary forest can result
from illegal logging. We suggested that one possible cause of illegal logging was lack of
timber supply, that is, the difference between actual demand for timber and the
allowable cutting--which is determined by the government regulation. Actual demand
for timber will eventually depends on the timber market situation at
local/national/international context, but we did not formulate that factor into the model.
Instead, we assigned a fixed value for the actual timber demand. The total area of
primary forest converted to secondary forest each year is, therefore, the sum of legal
cutting and illegal cutting.

3.4.3. Jungle Rubber Cultivation

Jungle rubber is a non-intensive form of rubber cultivation usually practiced by
smallholder farmers, which fit well with the existing shifting cultivation system. Following
land clearing, farmers plant the young rubber with other annual crops. After two or three
years when the crops yield become very low, farmers abandon the crops and let the
rubber grow with other species with almost no management input. After an average of 10
years, jungle rubber starts producing latex for more than 30 years. When the latex
production become very low, the farmers either abandon the field or replant the field
(Gouyon, De Foresta et al. 1993).

            START FELLOWSHIP REPORT                                                               13
At the present model, we have only constructed the sub-model for conversion from
secondary forest into jungle rubber, as presented in Figure 11. The main driving force
that we suggested for conversion from secondary forest to jungle rubber is market for
rubber. The rubber market itself will be influenced by many factors operating outside the
area, such as saturation market, competition from other rubber producing countries. In
addition to economic motives, establishing a perennial plot like jungle rubber has been
used as a mean to claim land ownership. However, we did not explicitly consider those
factors into account. Instead, we summarize all economic factors regulating the rubber
commodity in the form of Rubber Attractiveness Index (RAI). In this case, a high RAI
means that more agriculture household will consider cultivating rubber.

Figure 11. Conversion from secondary forest into jungle rubber.

As mention before, as a non-intensive form of rubber cultivation, jungle rubber
cultivation would only be possible when the resource (land) is abundant. This
relationship is expressed in the form of Jungle Rubber Attractiveness Index (JRAI). The
value of JRAI is high when both RAI is high and land (secondary forest) is abundant. In
situation where land is scarce, a high RAI would rather attracts farmers to cultivate rubber
intensively (monoculture rubber plantation) rather than jungle rubber. The JRAI will then
determine the proportion of agricultural household engaging jungle rubber cultivation
(jungle rubber household). The number of jungle rubber household is calculated by
multiplying that proportion with the number of agricultural household derived from
population sub-model.

Since the agricultural household variable produced by population sub-model is in a
cumulative form, we then subtract the number of current household with those from the
previous year to obtain a number of new jungle rubber household. It is this new
household that determines the area to be cleared for that year. The total area cleared
for each year is calculated by multiplying the number of new jungle rubber household
and the average area cleared per household. The later variable may also be influenced
by factors like land accessibility or land clearing technique, but we did not include them
into the present model.

3.4.4. Monoculture Rubber Cultivation

As a non-intensive form of rubber cultivation, the productivity of jungle rubber is very low
amounting more or less 500 kg/year compared with 1500-2000 in a more intensive
monoculture plantation. Lack of capital and access to better planting material is one
reason why farmers engage this form of rubber cultivation. Actually, there has been some

            START FELLOWSHIP REPORT                                                               14

efforts initiated by government to assist smallholder farmers to increase the productivity
through credit schemes, such as Smallholder Rubber Development Project (SRDP). This
financial assistance allows farmers to purchase high-yielding rubber clones. Since
rubber clones are adapted to the weed-free environment, the consequence of this effort
is to convert the multi-species jungle rubber to mono-species rubber plantation.
However, such a financial assistance has only reached a small proportion of the
smallholders (Gouyon, De Foresta et al. 1993).

In this modeling exercise we made an assumption that the conversion from jungle
rubber to rubber plantation is mainly driven by availability of capital, as shown in Figure
12. The capital availability was expressed as a credit extension rate. Credit extension rate
refers to the percentage of total jungle rubber household that receives credit each year.
The value of the rate was an arbitrary one since we did not have the information. The
credit extension rate is multiplied by number of jungle rubber household to obtain the
number of household intending to move to rubber plantation (rubber plantation
household). Then, the result is multiplied by the land cleared per household to calculate
the actual area to be converted each year. However, this present framework has not yet
adjusted the effect the reduction of jungle rubber household back to the jungle rubber

                                          Figure 12. Conversion from jungle rubber into
                                          rubber plantation.

3.4.5. Food Crop Cultivation and Homegarden

The mosaic land-cover recognized by satellite in ASB benchmark area mostly refers to
land cultivated for food crops cultivation, homegarden and settlement. Wetland rice,
upland rice, corn, soybean, chili, and cassava are the most widely planted crop in Jambi
peneplain. Meanwhile, in homegarden farmers plant a variety of fruit trees, timber trees,
vegetables, spices and sometimes cash crop such as coffee or cinnamon. Part of the
production from these fields are sold, but the revenue from food crops is generally
smaller compared with the revenue from selling cash crop like rubber (Ginting,
Sumarhani et al. 1996), (Zaini, Basa et al. 1996). Most of the productions from these
systems are used for household consumption. Therefore, it can be said that these land-
uses are closely associated with the existence of agricultural household itself as
indicated by the role of these systems to support the basic consumption need. It seems
that having cropland and homegarden is a kind of necessity in an agricultural

Taking this view in the model construction, we made an assumption that all newly
established agricultural households will posses a piece of land-used for crop cultivation.
Therefore, a household can have one or more cropland and no perennial plot, but not
the other way around. In addition, economic value of the food crop commodity could
become a reason to open cropland. But, if we look at the types of commodity which are
            START FELLOWSHIP REPORT                                                               15

mostly planted (i.e. rice, soybean, chili, corn) and the revenue gained out of them which
is very small, it is less likely that the attractiveness of the crops commodities would be a
strong reason why people establish croplands. Therefore, we would suggest that the
conversion from secondary forest to mosaic land be closely related to the addition of
new agricultural household. Figure 13 shows the implementation of this concept in the

                                                                        Figure              13.
                                                                        Conversion        from
                                                                        secondary       forest
                                                                        into mosaic of crop.

The structure of this sub-model is more or less similar to the conversion to jungle rubber
(3.4.3). In this case, it was assumed that all-new agricultural household need to establish
(mosaic) cropland. In situation where the access to establish an agricultural field in
secondary forest is open (due to communal land ownership), it is only the technology
used and availability of labor that will determine the achievement of a household in
terms of area that they are able to clear. The low technology such as 'traditional' slash-
and-burn will only be capable of opening relatively small areas. The availability of
better technology, such as chainsaw, will soon lead to increasing size of field.

Note that although transmigration program is certainly part of the land-use conversion in
this area, it is not explicitly represented in the structure of the sub-model. The reason for
this is because the derivation of the number of agricultural household has already taken
into account the number of transmigrant. However, a modification to include
transmigration can be made when necessary.

3.4.6. Oil Palm Cultivation

Indonesia is the second largest palm oil exporter in 1996 providing almost 20 % of the
world’s production. At the moment, the only rival is Malaysia, which provides 60 % of the
world’s production. This achievement of Indonesia is as a result of the large expansion of
oil palm plantation since early 1990s, which mostly took place in Sumatra. It is very
likely that the trend of oil palm expansion in Sumatra will continue considering that,
firstly, the Indonesia’s palm oil processing industry has not reached their full-production
capacity. At the moment it is operating at 30 % of their capacity. In addition, with the
present level of production increase, it has been projected that Indonesia will become
the largest exporter by the year 2003 (Kompas, 13/10/97), replacing Malaysia. In
Malaysia, the expansion of oil palm plantations now become limited, whereas in
Indonesia, there is still a vast area available that ‘can’ be converted into oil palm

            START FELLOWSHIP REPORT                                                               16

Secondly, many part of Sumatra provides a good environmental condition (i.e. soil and
climate) to grow oil palm and also have a good road infrastructure compared with the
other outer islands. Good road infrastructure is an absolute requirement in this business.
This is because the bunches of fruits harvested have to be delivered to the processing
factories within 24 hours, otherwise the oil quality will drop. Jambi in particular is an
attractive region in this regard because there has been an increase in road expansion in
this area especially following the completion of Trans Sumatra Highway. Therefore it is
likely that the increasing demand of oil palm-related product would continue the
pressure to convert other land-uses into oil palm.

The parties involved in the oil palm plantation business include smallholder farmers,
private companies and state-owned companies. The later two usually operate a large-
scale plantation. Figure 6 indicates that most of the land-use conversion to oil palm took
place in secondary forest. The dominant actor of this conversion may have been large
company. As in the case of logging operation, government play an important role in
controlling the conversion from secondary forest into oil palm plantation. This is
because the government through TGHK regulates the conversion of secondary forest
that is classified as state forest, into non-forest use such as plantation. Under the TGHK,
the conversion of forest to non-forest use can be done only in a forest declared as
‘conversion forest’.

For smallholder farmers, the promising benefit of oil palm provides farmers with a new
option. Oil palm plantation has become a competitor for jungle rubber and mosaic
cropland cultivation. In this situation, it is very likely that some farmers will give up their
current land-use and convert it into oil palm plantation. This has been the case in North
Sumatra where 15 % of the 500 farmers of Salapian sub-district converted their jungle
rubber into oil palm plantation (Kompas, 2/5/1997). They now prefer to grow oil palm
since it has shorter time to reach harvestable stage i.e. three or four years and also it is
easier to get financial assistance. However, since the establishment of oil palm
plantation requires a good road network, farmers living in remote areas may not consider
this option.

One simple way that may help to assess how farmers consider these cultivation options is
by comparing the expected income from oil palm with the expected income from
jungle rubber or mosaic cropland. If the expected income from oil palm is much higher
than those from mosaic cropland or jungle rubber, some farmers may consider the oil
palm option seriously. However, since oil palm plantation is a capital-intensive
cultivation, not all interested farmers could convert their current land-use to oil palm.
This will depend on capital availability such as access to credit or other form of financial
assistance. Figure 14 present the implementation of this concept for conversion from
jungle rubber to oil palm plantation. The conversion from mosaic to oil palm plantation
could also be expressed in the same way.

Therefore, farmers who most likely establish oil palm plantation are those who have
access to capital (credit) and live in accessible area. But, those areas could also attract
outside investors to locate their oil palm plantations. In this case, the farmers may have
further option that is establishing their own plantations or selling their land to companies.

The trade-off between these options may be assessed by comparing the land value with
the expected income from oil plantation. In situation where the farmers can establish
their own plantation, they may only give up their land for selling if the land value is far
higher than the expected income of operating their own plantation. The purchasing
land by companies may not affect the total area cleared but could affect the distribution
of money. Because if the plantations are operated by local farmers the money would
pretty much stay in the region. But we do not know how significant the purchasing land
by companies from smallholders is. It could be that the a company would be more
            START FELLOWSHIP REPORT                                                               17

comfortable in dealing with government --in terms of acquiring state land-- rather than
with individual farmer. Therefore, we did not implement this concept into the model.

Figure 14. Conversion from mosaic of crop into oil palm plantation.

3.4.7. Fire Disturbance

Fire disturbance plays an important role in determining the dynamic of a landscape.
The establishment of large scale Imperata cylindrica (alang-alang) grassland in many
parts of Sumatra and Kalimantan has been attributed to the presence of repeated fires.
As far as the cause of fire is concerned, we may distinguish between naturally ignited fire
--wild fire-- and human-ignited fire. The later type can refer to fire, which is used as a
land clearing method. Fire as a land clearing method has been practiced for years by
farmers and recently by companies to clear the land for establishing plantation.

As far as land-use change issue is concerned, it is likely that fire as a method of land
clearing, given it is properly controlled, will not directly link to land-use change.
Because, soon after the area has been burnt out, the land will be used as whatever the
purpose of the clearing. However, in many cases, the burning could become out of
control. Uncontrolled fire could easily escape and spread outside the intended area,
thus becoming a wildfire. In such situation, fire could lead to permanent land-cover
conversion, such as set a primary forest back to early succession stage (bush) or Imperata
grassland formation.

            START FELLOWSHIP REPORT                                                               18
In general, many factors determine the magnitude and the intensity of fire, among them
are climatic condition, soil, topography and vegetation. Some vegetation types are
easier to burn than the others. Likewise, the extreme climatic event like El Nino will
make the change of fire event higher. In this model, we do not explicitly include those
factors into the model. Instead, the fire disturbance was simply modeled as a random

            Figure 15. Big forest fire in Sumatra.

The fire disturbance sub-model is presented in Figure 16. It represents the fire
disturbance that converts the primary forest into secondary forest. In this sub-model, fire
event is simulated as random event with a probability of 0.5. Then, the size of fire (small,
medium or big) is also randomly selected. The areas cleared by small, medium and big
fire are 0-10, 10.1-50, 50.1-1000 hectare respectively.

                                                                        Figure             16.
                                                                        Conversion       from
                                                                        primary              to
                                                                        secondary       forest
                                                                        due      to        fire

            START FELLOWSHIP REPORT                                                               19
3.4.8. Concluding Remarks

During the model construction, there are some problems associated with the way in
which the mechanisms of conversion are expressed that need to be resolved. One of
them is the presence of multiple ways of conversion leading to formation of the same
land-use. Figure 6, indicates that one type of activity may involve several land-use
conversion types. For example, land converted into oil palm plantation was originally
from several land-use types (secondary, mosaic, and jungle rubber). At the present
model, we still handled it as separate issues.

Other problem is whether the model needs to consider a successional sequence and
cultivation sequence issue and how to handle it. Because in the case of jungle rubber,
the early stage of field establishment will likely be recognized as mosaic croplands.
Then, the conversion to jungle rubber (as recognized by satellite) is just a part of the
cultivation sequence. There is no new decision associated with that conversion.
Likewise, at the later stage, some of over mature jungle rubber field would be ‘converted’
into secondary forest simply due to succession.

In summary, the present model constructed during the workshop is far from complete.
Not all-possible mechanisms of conversions have been identified. Moreover, there are
many links between sub-models that are not yet carefully examined--in terms of logical
relationship. Therefore, the overall structure of the model is still open to modification.
However, many things can be learnt by doing this exercise. One of them is the
importance of interdisciplinary approach in modeling land-use change. It is hoped that
there will be an opportunity to complete the model construction in the future with a
deeper and comprehensive study by involving experts from many fields.

            START FELLOWSHIP REPORT                                                               20

4. Achievements

   In terms of the overall trip to USA, it was a fruitful trip. We had an opportunity
   to establish contacts with many people in the field of ecological modeling,
   some of them have expressed an interest to look for the possibility to do a
   modeling work in Indonesia. In fact, Prof. H.H. Shugart will present this
   model along with the model done by UVA students in LUCC Conference in
   Barcelona, March 1998. In addition, the visit to UVA gave us an experience
   of the academic environment in a US university. In AGU Conference, we
   also had an opportunity to observe the state-of-the-art of ecological
   modeling, especially the fields related to global change issues.

   START FELLOWSHIP REPORT                                                               21

5. Future Plan

    We are very eager to continue this modeling exercise in the future. One step
    towards that direction will be communicating the present results hoping that
    we will get a lot of critics and suggestions, particularly on the design of the

    The next step would be looking for the possibility to do a research project on
    this topic. Considering the complexity of the problem, the research team will
    involve people from many fields related with land-use change issues. In that
    project, in addition to collaboration with scientists, there will also be a
    communication with 'practitioners' especially ones directly involve in land-
    use change policy in order to get insights of the practical problems as an
    input in model design, so that the constructed model would have a practical

    START FELLOWSHIP REPORT                                                               22

6. Acknowledgements

   This work was supported by the International START Secretariat through US-
   AID via grant #9529649. We would like to express our sincere thanks to Prof.
   Roland Fuchs, Dr. Hassan Virji, Ms. Anne Phelan, Prof. H.H. Shugart, and Dr
   Louis Lebel, who gave us the opportunity to take part in this workshop. Our
   special thanks are also extended to Mrs. Ramona Shugart and Ms. Regina
   Carlson and Ms. Amber Hill for their hospitality and help that made our visit
   to Charlottesville very enjoyable. Finally, we would like to address a special
   thanks to all members of UVA students' team for their corporation in this work
   and their wonderful companions. Last but not least, we would also like to
   thank Ms. Titiek Setyawati and Ms. Endah Sulistyawati (BIOTROP-GCTE/IC-
   SEA fellowship grantees) for their cooperation and their kindness in spending
   the time in Charlottesville.

   START FELLOWSHIP REPORT                                                               23

7. References

   FAO (1990). Situation and outlook of forestry sector in Indonesia. Volume 1:
         issues, finding and opportunities. Jakarta, Ministry of Forestry,
         Government of Indonesia; Food and Agriculture Organization of the
         United nation.

   Ginting, A. N., Sumarhani, et al. (1996). Characterization of Production and
          Land-use System at the Rantau Pandan Benchmark Area, Indonesia.
          Workshop on Alternative to Slash-and-burn. Bogor.

   Gouyon, A., H. De Foresta, et al. (1993). “Does 'jungle rubber' deserve its
        name? An analysis of rubber agroforestry systems in southeast
        Sumatra.” Agroforestry System 22: 181-206.

   Lusiana, B., Suyanto, et al. (1997). Documentation of ASB-Household
         Survey Data In Indonesia. Bogor, ICRAF.

   Meyer, W. B. and L. Turner II (1992). “Human population growth and global
         land-use/cover change.” Annu. Rev. Ecol. Syst 23: 39-61.

   Murdiyarso, D. and U. R. Wasrin (1995). “Estimating land-use change and
         carbon release from tropical forest conversion using remote sensing
         technique.” Journal of Biogeography 22: 715-721.

   Turner, B. L., D. Skole, et al. (1995). Land-Use and Land-Cover Change :
          Science/Research Plan. Stockholm, IGBP and HDP.

   van Noordwijk, M., T. P. Tomich, et al. (1995). Alternative to Slash-and-Burn
         in Indonesia : Summary Report of Phase I. Bogor, Indonesia, ASB-

   Zaini, Z., I. Basa, et al. (1996). Characterization of Production and Land-use
           System at the Sitiung Benchmark Area, Indonesia. Workshop on
           Alternatives to Slash-and Burn. Bogor.

   START FELLOWSHIP REPORT                                                               24

To top