Page 1
1
Preprint of: Kaimowitz, D., P. Mendez, A. Puntodewo and J. Vanclay, 2002. Spatial regression analysis of
deforestation in Santa Cruz, Bolivia. In: C.H. Wood and R. Porro (eds) Land Use and Deforestation in the Amazon.
University Press of Florida, p. 41-65. ISBN 0-8130-2464-1.
Spatial Regresion Analysis of Deforestation in Santa Cruz, Bolivia
David Kaimowitz1, Patricia Mendez2, Atie Puntodewo1, Jerry Vanclay1
1. Center for International Forestry Research, Bogor, Indonesia, Fax +62 251 622100,
cifor@cgiar.org, http://www.cgiar.org/cifor
2. Land Use Planning Technical Office (OT-PLUS), Departmental Government (Prefectura), Santa
Cruz, Bolivia
Abstract
This paper applies a spatial economic regression model to analyze the relation between
deforestation in the period from 1989 to 1994 and access to roads and markets, ecological
conditions, land tenure, and zoning policies in Santa Cruz, Bolivia. The data comes from a
Geographic Information System (GIS) compiled by the Natural Resources Department of the Santa
Cruz Government. Locations closer to roads and the City of Santa Cruz and that have more fertile
soils and higher rainfall have a greater probability of being deforested. The same also applies to
colonization areas. National parks and areas occupied by indigenous people do not have
significantly less deforestation than sites with similar acess and ecological conditions. Forest
concessions, on the other hand seem to protect forests.
1. Introduction
International interest in the issue of tropical deforestation has grown rapidly during the last
twenty years, but major uncertainties persist regarding when, where, and why it occurs. In a recent
survey by Kaimowitz and Angelsen (1998), the authors identified more than 150 quantitative
models that researchers have developed to answer these questions, most of them since 1990.
These models assess the impact of over 115 variables that potentially influence deforestation, but
in many instances the direction and magnitude of their effect on deforestation remains uncertain.
This reflects the inherent complexity of the issue and limited data availability, as well as various
methodological weaknesses.
-----------------------------------------------------------------------------------------------------------------------------------
Acknowledgements
This paper represents a collaborative effort between the Center for International Forestry Research
(CIFOR) in Bogor, Indonesia, the Natural Resources Department of the Government (Prefectura)
of Santa Cruz, Bolivia, and the Bolivian Sustainable Forest Management Project (BOLFOR). The
authors wish to express their gratitude to individuals in each of these institutions that have
contributed to our efforts, including Sergio Antelo, Andreas Carstens, Osvaldo Escalante Saldaña,
Francisco Kempff, John Nittler, Christian Vallejos, and Roderich von Offen. Other useful comments
and suggestions have come from Ken Chomitz, Ivan Morales, Gery Nelson, and Tom Tomich.
Page 2
2
The Kaimowitz and Angelsen study evaluated the potential of different modeling
approaches for improving our understanding of tropical deforestation and concluded household
and regional-level studies showed more promise than national and global models. It expressed
particular enthusiasm for the opportunity the growing availability of spatial databases presents for
examining the relation between deforestation and spatial variables such as access to markets,
land tenure, land use zoning policies, and ecological factors. Most spatial models use relatively
reliable data and large sample sizes and are particularly well suited for predicting where
deforestation will occur. In addition, researchers can often test the models’ robustness by
measuring what percentage of the time they accurately predict which areas will be deforested.
This paper presents a spatial econometric model of deforestation in the Department of
Santa Cruz, Bolivia between 1989 and 1994. We chose Santa Cruz for several reasons:
• The Bolivian tropics have historically had low deforestation, but this has changed rapidly in
recent years, particularly in Santa Cruz. Spatial models may be able to help us understand why;
• Unlike many Latin America regions, large-scale mechanized agriculture plays a major role
in forest loss in Santa Cruz. Thus, the causal pathways that influence deforestation there may
differ greatly from other locations;
• CIFOR has an on-going project comparing the effects of different policies and social trends
on tropical forests in Bolivia, Cameroon, and Indonesia; and
• The Government of Santa Cruz (prefectura) had an existing Geographic Information
System (GIS) with much of the data we needed to model the influence of spaital variables on
deforestation trends.
This paper first briefly describes spatial econometric deforestation models and reviews the
conclusions of previous models made for other tropical areas. Then, we give background
information on deforestation in Santa Cruz. Next, we present the model and data we used in this
study and finally we discuss our results.
Page 3
3
2. Spatial Econometric Deforestation Models
Spatial econometric regression models measure the correlation between land use and
other geo-referenced variables such as:
• Transportion costs (distance from markets and road, railways, and rivers),
• Ecological conditions (topography, soil quality, precipitation, and forest fragmentation), and
• Land tenure and land use zoning categories (national parks, forest concessions,
colonization areas, and indigenous territories).
The models can either focus on land use in a single time period or the change in land use
over two or more periods. The majority of models relate the state of the explanatory (independent)
variables in the first period to the probability the forest in that location is removed between the first
and second periods.
Some model makers include only locations covered with forest during the initial period in
their samples. Others include both locations that had forest cover in the initial period and those that
did not.
Based on economic theory, model makers hypothesize that farmers decide whether to
deforest an area based on if the net present value of the returns they receive from putting the land
into agriculture outweighs the cost of forest clearing and any benefits they might obtain from forest
products and services. Areas with soils, rainfall, and topography more suited to agriculture should
provide higher returns to farming and thus farmers would be more likely to deforest them. Likewise,
government subsidies for deforestation in colonization areas increase the likelihood farmers will
clear forests there. On the other hand, higher transportation costs reduce the net returns from
converting an area to agriculture. Similarly, government land use restrictions, such as the
designation of a location as a protected area or forest concession, should raise the expected cost
to farmers of deforesting the area and make forest clearing there less likely. Some researchers
also argue that indigenous people are less likely to deforest their land because of their distinct
belief systems and traditions or their more limited access to capital.
The data used in spatial econometric models comes from maps. Most models have
obtained that data by selecting a random sample of locations (points on the map) and determining
Page 4
4
the state of each variable in that location. They then treat the characteristics of that point as one
observation, as if it were one individual in a household survey.
Typically, modelers use samples of several thousand points or more. Chomitz and Gray
(1995), for example, used a random sample of 10,000 data points for their Belize study. Tom
Tomich of the International Center for Agroforestry Research (personal communication 1998)
worked with 49,000 points in his model of Jambi, Sumatra. Gerald Nelson from the University of
Illinois (personal communication 1998) made a similar study with raster data that yielded 25,000
sample points.
Most land use information comes from national forest inventories, remote sensing, aerial
photographs, and ground truthing. GIS programs generate the information on distance to roads
and markets using the maps in their databases. The remaining information comes largely from
local government departments.
3. Previous Model Results
Previous models have generally confirmed the hypotheses that land holders convert more
forest to agricultural use in locations that have better access to markets, ecological characteristics
more favorable for farming, and no government restrictions on forest clearing. (See Table 1.)
Table 1. Conclusions from Previous Spatial Regression Models
About the Effects of Different Variables on Deforestation
Study Country More
roads
Closer to
markets
Better soils
&/or drier
Nearer
forest edge
Brown et al. (1993) Malaysia NA NA NA Increase
Chomitz & Gray (1995) Belize Increase Increase Increase NA
Deininger and Minten (1996) Mexico Increase NA Increase NA
Gastellu-Etchegorry &
Sinulingga (1988)
Indonesia NA NA Increase NA
Liu et al. (1993) Philippines Increase NA NA Increase
Ludeke et al. (1990) Honduras Increase NA Increase Increase
Mamingi et. al (1996) Cameroon
and Zaire
Increase Increase* Increase NA
Mertens and Lambin (1997) Cameroon Increase Increase NA Increase
Nelson and Hellerstein (1995) Mexico Increase Increase NA NA
Rosero-Bixby and Palloni
(1996)
Costa Rica Increase NA Increase Increase
Sader and Joyce (1988) Costa Rica Increase NA Increase NA
* Only in Cameroon. No effect in Zaire
Page 5
5
Forests are more likely to be cleared when they are closer to roads in physical distance and
traveling time (Chomitz and Gray, 1995; Deininger and Minten 1996; Liu et al.,1993; Ludeke et al.,
1990; Mamingi et. al. 1996; Mertens and Lambin, 1997; Nelson and Hellerstein, 1995; Sader and
Joyce, 1988; Rosero-Bixby and Palloni, 1996). Most studies show forest clearing declines rapidly
beyond distances of two or three kilometers from a road. Liu et al. (1993) and Mamingi et. al.
(1996), however, report that forest clearing and distance to roads remain strongly correlated at
much greater distances in Cameroon, the Philippines and Zaire. Proximity to railroads is also
positively associated with deforestation in Cameroon and Zaire (Mamingi et. al. 1996).
Chomitz and Gray (1995) found locations near urban markets have less remaining forest in
Belize and Mertens and Lambin (1997) say most deforestation occurs less than ten kilometers
from the nearest town in Eastern Cameroon. Nelson and Hellerstein (19997) found that distance to
villages had a much more significant effect on land use than distance to urban areas.
Farmers are also more likely to clear areas with higher quality soils (flat, adequate
drainage, and high soil fertility) and drier climates (Chomitz and Gray, 1995; Gastellu-Etchegorry
and Sinulingga, 1988; Sader and Joyce, 1988; Rosero-Bixby and Palloni; 1996).
Forest fragments have a higher risk of being lost than forests in large compact areas, with
those close to the forest edge especially likely to be cleared (Brown et al., 1993; Liu et al., 1993;
Ludeke et al., 1990; Mertens and Lambin, 1997; Rosero-Bixby and Palloni, 1996).
The effect of roads and environmental conditions interact, so roads induce greater forest
clearing in areas with good soils and favorable climatic conditions. In Belize, for example, Chomitz
and Gray (1995) show the probability of an area being used for agriculture (rather than natural
vegetation) on high quality land next to a road was 50%, while lands next to roads with marginal
soils had only a 15% probability of being deforested. Mamingi et. al. (1996) obtained similar results
in Cameroon and Zaire.
Mertens and Lambin (1997) note that variables affect forest clearing differently depending
on the type of deforestation process. In peri-urban deforestation, forest clearing exhibits a circular
pattern around the towns, and distance to towns and roads strongly affects forest clearing but
proximity to forest edge does not. Along roads you get a “corridor" pattern of deforestation where
End of preview.
