BIOPLAN POSTING 2002-2-1
BIOPLAN POSTING 2002-2-1
David Duthie
02/25/02 06:40 PM
Dear BIOPLANNERS,
As you should all know, the "devil is in the detail" especially when it comes to big problems. For global biodiversity, tropical deforestation remains "the big problem" - if we do not change land use trends in the tropics, then global biodiversity will continue to decline.
David Kaimowitz has just highlighted a meta-analysis of 152 studies
of tropical deforestation conducted by Helmut Geist and Eric Lambin of
The Land Use and Land Cover Change (LUCC) Project, a Programme Element
of the International
Geosphere-Biosphere Programme (IGBP) and the International Human
Dimensions Programme on Global
Environmental Change (IHDP).
The analysis has recently been summarised in an article in BioScience as "Proximate Causes and Underlying Driving Forces of Tropical Deforestation". For a copy, contact the authors at: lucc.ipo@geog.ucl.ac.be
The entire 136 page report on which the Bioscience article is based "What Drives Tropical Deforestation? A meta-analysis of proximate and underlying causes of deforestation based on subnational case study evidence" LUCC Report Series No. 4, can be downloaded, as a 1.6 MB pdf file, at:
<http://www.geo.ucl.ac.be/LUCC/>
Below, I am pasting Chapter 5: Conclusions, which clearly demonstrate the regional variation in the relative importance of both proximate and ultimate causes of deforestation. What works in one country may not be the best practice in another and, for sure, global solutions are not the answer.
This report (along with David Kaimowitz and Arild Angelsen's 2000 analysis of Tropical Deforestation and Agriculture, available from CIFOR) should be an essential resource for anyone involved in biodiversity planning in a country with significant tropical forest resources.
Best wishes
David Duthie
david.duthie@unep.org
*****************************************************************************
5 Conclusions
Conclusions from the meta-analysis of 152 subnational cases of tropical
deforestation
are drawn in a fourfold manner. First, empirical findings are held
against prevailing
explanations of tropical forest decline. Second, implications are outlined
with view on
the future modelling of the process of deforestation, and, third, the
same is done with
view on policies designed to control deforestation. Fourth, we conclude
with a short
outlook on the future design of case study comparisons exploring the
causes of land use
and land cover change
.
Empirical findings versus prevailing explanations
Looking upon empirical results from the meta-analysis as the first study
which relates
underlying to proximate causes in a systematised manner; see Figure
9 (inlay) which
graphically summarises the results, our main findings are compared
to prevailing explanations
of forest decline in the tropics as follows:
Causes and drivers of tropical deforestation cannot be reduced to a
single variable, or
to a few variables even. Rather, the interplay of several proximate
as well as underlying
factors drive deforestation in a synergetic way. We found that mainly
3- and
4-factor terms of underlying causation are related to 2- and 3-factor
terms of proximate
causation. Among the most frequently found factor combinations are
the
agriculture-wood-road connexus (mainly driven by economic, policy,
institutional
and cultural factors), the agriculture-wood connexus (mainly driven
by technological
factors), and population-driven agricultural expansion. Regional variations
of
synergetic cause connections turned out to be considerable, with the
agriculture-wood
connexus featuring mainly Asian cases, and the road-agriculture connexus
featuring mainly Latin American cases, for example.
While the expansion of cropped land and pasture is clearly the most
important proximate
cause of tropical deforestation, shifting cultivators are not always
the key agents
of deforestation: shifting cultivation is often associated with timber
logging and road
construction as concomitant causes; traditional shifting cultivation
(swidden-fallow
farming) mainly characterises upland and foothill Asian cases, while
colonist shifting
cultivation (slash-and-burn agriculture by in-migrants) is limited
mostly to humid
lowland cases in Latin America, with many of the latter cases driven
by conditions of
poverty. Rather than shifting cultivation, the expansion of permanently
cropped land
for food by smallholders dominates agricultural expansion leading to
deforestation.
Chain-logical causation in the form of simple tandems (or 2-factor
chains) underlies
about two thirds of the proximate as well as underlying causes. It
provides insight in
all underlying/proximate factor interlinkages. On average, 4 to 5 tandems
are associated
with each case of tropical deforestation (since only tandems were considered,
the
actual pattern might be more complex even). At the proximate level,
the infrastructure
(road)-agriculture tandem seems to be the key causative connection.
At the
underlying level, policy and institutional factors – such as formal
state policies, policy
climate (or informal policies) and property right arrangements (or
issues) – exert the
strongest impact upon proximate causes, while economic factors dominate
the overall
frequency pattern of cause occurrence (i.e., including single factor,
concomitant and
chain-logical causation).
Rather than single or direct fundamental causes, underlying driver
tandems were
identified to be most important. These are mainly economy-, policy
and institution-,
and culture-driven tandems impacting upon the proximate level (especially
the latter
tandem is not treated well in the deforestation literature). However,
next to each case
has its own and very specific type of interlinkage, and hardly any
generalisations are
possible on the underlying tandems. There are only few exceptions:
the land-migration
tandem dominates Latin American cases, while the population
growth-intensification tandem was found to prevail in African and Asian
cases.
Among broad, aggregate economic and policy/institutional causes, multiple
and interactive
factors were found to drive cases of tropical deforestation.
Population pressure in the form of natural increases in number of population
due to
high fertility is clearly not the major underlying driving force at
the scale of a few
decades, when taken as a direct cause in isolation from other factors.
Rather, inmigration
into forested (not natural increase in densely populated) areas plays
an
important role in cases of frontier colonization. In all cases, however,
population does
not operate as a single force but is interlinked with other underlying
factors. In
underlying driver combinations, policy/institutional, economic, socio-political
(or
cultural) and technological factors play the major part.
With view on variations of cause frequencies and causality patterns,
there is empirical
indication that in cases with high rates of annual deforestation pre-disposing
biophysical factors are at work or shape the pattern of deforestation.
Namely, these
are low relief and f lat topography in combination with good soil quality
and high
water availability. In contrast, proximate causes that cannot be assimilated
to biophysical
conditions are more associated with cases featuring considerably lower
rates
of annual deforestation (no such equivalent was found at the underlying
level).
The explanatory power of PAT variables (population, aff luence, technology)
is astonishingly
poor. This model obviously fails to consider policy and institutional
factors
shaping market opportunities as powerful explanatory factors of tropical
deforestation.
Though difficult to code, we consider the meta-analysis to be the first
study – not
only to relate underlying driving forces to proximate causes in a systematised
manner,
but also – to quantify the impact of cultural or socio-political factors
upon the process
of tropical deforestation. It has contributed to sharpen and partly
revise the six broad
patterns commonly associated with certain deforestation processes,
which are agricultural
expansion, wood extraction and infrastructure extension at the proximate
level, and economic, policy and institutional, cultural, technological
and demographic
factors at the underlying level. Feedbacks from the proximate upon
the underlying level turned out to
be fairly weak (and only mattered in the case of infrastructure
extension inducing in-migration and fostering the economic valorisation
of
areas affected by road construction).
The multiple factors intervening in tropical deforestation also make
it particularly difficult
to develop generic and widely applicable policies that best attempt
to control
the process. Many land-use policies are underlain by simplifications
on the drivers of
change. Such simplification suggest simple technical solutions and
sometimes may
serve the interests of critical groups (Lambin et al., 2001). From
the results of the
meta-analysis it is clear that any universal policy or global attempt
to control deforestation
(e.g., through poverty alleviation) is doomed to failure.
Implications for modelling and controlling deforestation
Implications from our our empirical findings for both the future modelling
of the
process of deforestation and for policies to control deforestation
are as follows:
Deforestation is a complex, multiform process which cannot be represented
by a
mechanistic approach. This has implications for modelling as many of
the simulation
models of land-use change tend to be mechanistic. Mechanistic models
are built on
the belief that we know the processes by which a system operates (Elston
and
Buckland, 1993) and that individual processes can be modelled using
scientific laws,
or rules, described by simple equations. Given the large number of
interacting factors
driving deforestation, and given interactions at different levels of
causality – underlying
forces, trigger events, mediating factors, proximate causes – only
a system
approach seems to be appropriate. System models are mathematical descriptions
of
several complex, interacting processes. They emphasise the interactions
among all
components within an ecosystem by implementing the “whole system“ concept.
System models sometime sacrifice parsimony and abstraction in favour
of an apparent
attempt to simulate the socio-natural system in its entirety (ibd.,
1993).
Do we understand with a sufficient level of generality and clarity
the causes of deforestation
to be able to represent these in a system model? The case study evidence
examined in this meta-analysis clearly suggests that we still lack
an overarching theory
to encompass the different factors which intervene in the processes
of deforestation.
The role of a theory is to explain experimental findings and to predict
new results.
Specific relationships underlying deforestation and the processes involved
have been
effectively addressed through theories of particular components of
a land use system.
These include, but are not limited to, household economics, smallholder
and peasant
behaviour, land allocation, technological innovation, fertility change,
policy/institutional
regimes associated with land resource management, national markets,
and
international accords (Brown and Pearce 1994; Kasperson et al. 1995;
Palo and Mery
1990). The complex and multidimensional character of deforestation
processes
requires understanding and modelling that incorporates the principles
from such theories.
However, the variability in how these principles come together in a
place or
region at particular times is not conducive to research strategies
aimed at a test of
simple hypotheses that might equate deforestation to population, economic
struc-
tures, technology, political structures, or biophysical attributes
(Lambin et al. 2001).
These and other drivers of deforestation are always present but interact
differently
according to the temporal and spatial dynamics of particular regional
or local situations.
A thorough understanding of these complex interactions is a prerequisite
to
generate realistic projections of land-cover changes based on simulation
models.
Empirical evidence shows that the complexity and diversity of driving
factors of
deforestation is reduced when looking at specific processes – e.g.,
subsistence agriculture,
commercial agriculture, colonisation activities or logging activities
– and
specific geographic situations – i.e., frontier areas, roadside areas,
peri-urban areas.
Thus, while the development of a “universal“ model of deforestation
is probably out
of reach, a collection of specific models which represent the particular
interactions
between a reduced set of dominant driving forces for a given process
of deforestation,
in a given geographic situation, is a feasible approach. While the
essence of modelling
land-use change processes is “[...] to transcend the complexity of
context, seeking to
identify broad and universally applicable forces of change that crosscut
the circumstances
of place and period“ (Turner and Meyer 1991, p. 672), one should also
recognise the importance of that very complexity and the uniqueness
of particular
cause-impact relationships in specific situations (Turner and Meyer,
1991). Thus,
models must recognize the variability by region or place of the human
activities
driving land-cover changes (Turner et al. 1990). This can best be achieved
by applying
different models to regions which are relatively homogeneous with respect
to the
phenomenon being modelled.
The overall dominance of the broad cluster of agricultural expansion
is well perceived
in the modelling of tropical deforestation (e.g., Kendall and Pimentel
1994; UNEP
1997). However, this is not always the case when it comes more specific
agricultural
uses and other than agricultural land uses such as logging (e.g., Hamilton
1997). This
meta-analysis provides indicative values for these proportions causing
deforestation.
For example, the emphasis given to traditional shifting cultivators
in agent-based
modelling (e.g., Walker 1999) or heuristic methods (e.g., Amelung and
Diehl 1992;
Döös 2000) certainly has to be revised in favour of modelling
the gradual expansion
of permanently cultivated land for both commercial and subsistence
needs, or in
favour of infrastructure extension and wood extraction. In addition,
chain-logical
modelling along the lines of the logging-shifting agriculture tandem
is less fruitful
than considering the more dominant infrastructure (road)-agriculture
tandem.
Further, any effort to apply, for example, cattle development as a
major variable in
global models attaining scenario development is likely to be misleading,
since forest
conversion for pasture creation almost exclusively concerns lowland
cases under
humid climates in mainland South America.
Implications for future case study comparisons
We conclude that any future case study, aimed at understanding the causes
of tropical
deforestation in any particular place, should follow a standard protocol
to allow for
improved comparisons in the future.
The LUCC research framework (Turner et al. 1995, pp. 20-22; Lambin
et al. 1999)
proved to be a fruitful platform from which to proceed to develop a
general understanding
of the drivers of land use and land cover change, and from which to
conduct
a systematic comparison of a large number of subnational case studies.
Theorizing, for
example, of proximate as well as underlying biophysical causes has
started, and the
typification of certain proximate and underlying variables as filter
variables, modifiers,
contextual factors or triggers (catalytic factors) will certainly help
to improve
the conceptual understanding. Probably, the proximate/underlying divide
will have
to be overcome, so as to allow for more complex interplays of human
agency and
structure in processes of land change.
Concerning regional representation of case studies in future comprehensive
comparisons,
we found that weighting bias in our meta-analysis was low, but we also
found
indication that future work will have to include considerably more
African cases and
reduce, by far, the proportion of Asian cases (given most recent forest
cover dynamics
in these regions). Latin American cases, holding about half of all
cases, probably
continue to feature quite well.
Concerning weighting bias in terms of agents involved in the process
of deforestation,
one needs to have a better understanding of logging company behaviour
and/or
industrial forestry plantation activities (as compared to farming communities)
– given
the considerable role of wood extraction involved in many synergetic
driver combinations,
and given that especially state-run and illegal (illicit or undeclared)
commerical logging activities were found to play a major role.
Finally, we believe that a systematic comparison of local-scale case
studies is an
extremely productive methodology to extract generalities on the causes
and processes
of land-use change at multiple scales. It is labour-intensive and requires
a high level
of rigour in its implementation, but the insights it reveals are much
more realistic and
richer than cross-national statistical analyses, and more representative
and general
than research on single cases. Global change research, thus, needs
more of such synthesis
activities based on case study comparisons.