Download Integrating Biophysical and Socioeconomic Factors in Günther F

Present and Future of Modeling Global Environmental Change: Toward Integrated Modeling,
Eds., T. Matsuno and H. Kida, pp. 271–292.
© by TERRAPUB, 2001.
Integrating Biophysical and Socioeconomic Factors in
Modeling Impacts of Global Environmental Change
Günther F ISCHER
International Institute for Applied Systems Analysis, 2361 Laxenburg, Austria
Abstract—This paper presents examples of modeling studies carried out at the
International Institute for Applied Systems Analysis (IIASA) where an
integration of biophysical and socioeconomic factors in modeling impacts of
global environmental was successfully achieved. The first example is taken
from an integrated analysis of the impacts of alternative future energy paths on
the regional supply and trade of agricultural products. Second, three
complementary approaches for analyzing and projecting regional land use are
discussed. The methods have been developed by the Land Use Change (LUC)
Project at IIASA and have been applied to assess land-use trajectories and
prospects of China’s food and agriculture sectors. It is concluded that land-use
policy alternatives aimed at minimizing adverse impacts of global environmental
change can adequately be analyzed in an economic framework using optimization
when this is well informed by incorporating a representation of climate and
biophysical conditions in the model specification.
INTRODUCTION
Various approaches for analyzing and projecting regional impacts of global
environmental change have been developed at the International Institute for
Applied Systems Analysis (IIASA). Initially these studies were geared towards
assessing the potential impacts of global climate change. For example, the
analysis of the impacts of alternative future energy paths on the regional supply
and trade of agricultural commodities has been part of an integrated assessment
study undertaken jointly by the ECS (Environmentally Compatible Energy
Strategies), TAP (Transboundary Air Pollution) and LUC (Land Use Change)
projects at IIASA. The complexity and multiple interrelationships of the problem
at hand require an analysis that takes into account the relevant physical and
economic relationships governing the world food system. To achieve consistency
among the various research groups, the assessment models have been harmonized
through a process of soft-linking.
More recently, changes in land use and land cover were recognized as being
central to the study of global environmental change (Turner et al., 1995). In
addition to their cumulative long-term global dimensions and their potential
responsiveness to global environmental change, such changes can have profound
271
272
G. F ISCHER
Fig. 1. Demographic changes 1995–2030 and 2030–2050.
regional environmental implications during the life span of current generations,
including various themes central to the debate of sustainability such as reduced
biodiversity, changed land productivity, climate feedbacks, robustness of the
land use systems with respect to economic or environmental shocks, or enhanced
water scarcity problems, e.g., the lowering of groundwater tables.
The importance of these topics, and the need for innovative and
interdisciplinary approaches to study the nature of land-use and land-cover
changes, have prompted IIASA to establish its project on Modeling Land-Use and
Land-Cover Changes in Europe and Northern Asia (LUC). The project has two
main goals: first and foremost, to develop new concepts that address the
methodological challenges of projecting complex human-environment systems;
and second, to apply these concepts within regional assessments to identify
economically viable options for land use and food policy. The current application
Integrating Biophysical and Socioeconomic Factors in Modeling Impacts
273
focus has been on China.
Global environmental change issues are long-term and are related to questions
of resource development and investment planning. Current demographic and
socioeconomic trends suggest that the next 30–50 years will be decisive for
managing viable transitions towards sustainable land-use systems (see Fig. 1).
The LUC project therefore is concentrating its analysis on the period up to 2050.
For instance, China’s population growth will most likely come to a halt around
2030, and pressures on the food system will ease subsequently.
In the first part of this paper as summary is provided of the linkage
mechanisms used in IIASA’s integrated analysis of the impacts of alternative
future energy paths on regional supply and trade of agricultural products. While
the process of soft-linking may be criticized for being achieved by sacrificing
certain rigorous formalism, the benefits clearly accrue in being able to exploit
detailed and specialized assessment models.
In the second part of the paper we illustrate the integration of biophysical and
socioeconomic factors with three complementary approaches in land-use analysis,
which were applied within LUC’s study of China’s food and land-use prospects.
First, an enhanced Agro-ecological zones (AEZ) assessment model is used to
provide a spatially explicit measure of crop suitability and land productivity
potentials for rain-fed and irrigated conditions. This provides inputs to scenario
analysis, based on an extended Input-Output (I-O) model, to quantify future land
requirements associated with the projected demographic trends and economic
activities. Finally, changes in land and water use are viewed as dependent on how
these resources are transformed and managed by human activity. The underlying
decision problem is cast in the form of a welfare optimum model to elaborate
socially desirable and economically efficient trajectories of resource uses and
transformations. For this, special emphasis was placed on specifying a set of
production relations of the land-based economic sectors such that these would
permit the integration of spatially varied biophysical factors into the economic
model.
ANALYZING IMPACTS OF ALTERNATIVE FUTURE ENERGY PATHS
ON REGIONAL FOOD SYSTEMS
The analysis of the impacts of alternative future energy paths on the regional
supply and trade of agricultural commodities has been part of an integrated
assessment study undertaken at IIASA. For the agricultural study, results from
the MESSAGE-MACRO energy models of IIASA’s Environmentally Compatible
Energy Strategies (ECS) project and from the regional air pollution model
RAINS developed by IIASA’s Transboundary Air Pollution (TAP) project were
compiled to define the economic and environmental conditions for a number of
simulation experiments with the BLS, IIASA’s world agriculture model. IIASA’s
research has provided a framework for analyzing the world food system, viewing
national agricultural systems as embedded in national economies, which in turn
interact with each other at the international level.
274
G. F ISCHER
Fig. 2. Scheme of integrated study on energy-environment-economy impacts of alternative energy
pathways on regional food provision.
To achieve consistency among the various research groups, the assessment
models have been harmonized through an approach that we term soft-linking. A
general scheme of the assessment is provided in Fig. 2.
A critical step in this process is linking the results of the integrated energy
model MESSAGE-MACRO (Nakicenovic et al., 1997) and IIASA’s model of the
world food and agriculture system, the Basic Linked System of National
Agricultural Policy Models (BLS) (Fischer et al., 1988). First, the demographic
and economic projections of the BLS are calibrated to match the output from the
macroeconomic energy model component MACRO. Second, the yield impact
module of the BLS is parameterized according to emission trajectories calculated
by MESSAGE (Model for Energy Supply Strategy Alternatives and their General
Environmental Impact), the systems-engineering component of the IIASA energy
model, and according to global temperature changes derived from GCM
experiments, if available, or obtained with a robust simple climate model
MAGICC (a Model for the Assessment of Greenhouse-gas Impacts and Climate
Change; Wigley and Raper, 1992; Hulme et al., 1995). Third, results from RAINS
(Regional Acidification INformation and Simulation model; Amann, 1993;
Amann et al., 1995; Cofala and Dörfner, 1995) were used to inform regional yield
damage functions in the BLS accounting for the effects of increasing SO2
emissions and deposition. This proved to be especially relevant in high emission
fossil energy scenarios.
The world agriculture model system BLS
The BLS is a world level general equilibrium model system. It consists of
some thirty-five national and/or regional models. The individual models are
Integrating Biophysical and Socioeconomic Factors in Modeling Impacts
275
linked together by means of a world market module. A detailed description of the
entire system is provided in Fischer et al. (1988). Various results obtained with
the system are discussed in Parikh et al. (1988), Rosenzweig and Parry (1994),
Fischer et al. (1990, 1994, 1996), and Fischer and Rosenzweig (1996).
The general equilibrium approach upon which the BLS is constructed
necessitates that all economic activities are represented in the model. Financial
flows as well as commodity flows within a country and at the international level
are consistent in the sense that they balance. Whatever is produced will be
demanded, either for human consumption, feed or intermediate input; it might be
traded or put into storage. Consistency of financial flows is imposed at the level
of the economic agents in the model (individual income groups, governments,
etc.), at the national as well as the international level. This implies that total
expenditures cannot exceed total income from economic activities and from
abroad, in the form of financial transfers, minus savings. On a global scale, not
more can be spent than what is earned.
The country models are linked through trade, world market prices and
financial flows. The system is solved in annual increments, simultaneously for all
countries. Since these steps are taken on a year-by-year basis, a recursive
dynamic simulation results.
Although the BLS contains different types of models, all adhere to some
common specifications. The models contain two main sectors: agriculture and
non-agriculture. Agriculture produces nine aggregated commodities. All nonagricultural activities are combined into one single aggregate sector. Production
is critically dependent on the availability of the modeled primary production
factors, i.e., of land, labor and capital. The former is used only in the agricultural
sector, while the latter two are determinants of output in both the agricultural and
the non-agricultural sectors.
For agricultural commodities, acreage or animal numbers and yield are
determined separately. Yield is represented as a function of fertilizer application
(crops) or feeding intensity (livestock). Technical progress is included in the
models as biological technical progress in the yield functions of both crops and
livestock. Rates of technical progress were estimated from historical data and, in
general, show a decline over time. Mechanical technical progress is part of the
function determining the level of harvested crop area and livestock husbandry.
Linking BLS with MESSAGE-MACRO
MESSAGE-MACRO is a recently developed integrated systems-engineering
and macroeconomic energy model, based on two components, MACRO and
MESSAGE, which can also be run independently. MACRO is a macroeconomic
model derived from 11R, an eleven-region adaptation of the Global 2100 model
(Manne and Richels, 1992). This model, in several variants, has been widely used
for economic studies of the global implications of CO2 reductions. MACRO is a
dynamic nonlinear optimization model used for the analysis of long-term CO2energy-economy interactions. Its objective function is the total discounted utility
of a single representative producer-consumer.
276
G. F ISCHER
MACRO generates internally consistent projections of global and regional
gross domestic product (GDP), as well as trajectories of regional investment,
labor, and primary energy consumption. A high degree of correspondence with
the BLS in key variables for modeling the economy makes it feasible to harmonize
the scenario analysis undertaken with the MESSAGE-MACRO and BLS models.
The approach chosen for linking was to synchronize rates of economic
growth generated in the BLS with those projected by MACRO through adjustment
of production factors and of autonomous technical progress. The thirty-four
model components of the BLS were aggregated into ten world regions as closely
as possible matching the regionalization of MACRO. Then, the harmonization of
production factors and GDP for the period 1990 to 2050 was carried out on a
region-by-region basis, by calibrating the economic growth generated in the BLS
for the sub-periods 1990 to 2010, 2010 to 2030, and 2030 to 2050. To keep the
analysis transparent and focused on energy-environment-economy interactions,
it was decided to use only one common population projection (Lutz et al., 1996).
Another cornerstone of the integrated assessment was MESSAGE, a dynamic
systems engineering optimization model used for medium to long-term energy
system planning and energy policy analysis. MESSAGE uses a bottom-up
approach to describe the full range of technological aspects of energy use, from
resource extraction, conversion, transport and distribution, to the provision of
energy end-use services. The model keeps a detailed account of pollutant
emissions such as of CO 2 and SO2.
The emissions projected in MESSAGE-MACRO scenario runs were input to
MAGICC (Hulme et al., 1995), a simple climate change model that has been
widely used for assessments reported by the IPCC. MAGICC accounts for the
climate feedback due to CO2 fertilization, and for negative radiative forcing due
to sulphate aerosols and stratospheric ozone depletion. Emissions are converted
to atmospheric concentrations by gas models, and the concentrations are converted
to radiative forcing potentials for each gas. The net radiative forcing is then
computed and input into a simple upwelling-diffusion energy-balance climate
model. This produces global* estimates of mean annual temperature.
Temperature and CO 2 yield impacts
To estimate crop yield changes for a variety of conditions, we employed
geographically detailed information generated within earlier climate impact
studies (see Rosenzweig and Parry, 1994; Rosenzweig and Iglesias, 1994;
Fischer et al., 1994, 1996; ASA, 1995; Strzepek and Smith, 1995). It is assumed
that yield impacts estimated in detail for different GCM climate scenarios
describe response relationships that can be scaled to the levels of temperature
change and atmospheric CO2 projected in the alternative energy emission scenarios.
Let ∆y 0t,j denote the estimated aggregate yield changes per degree of warming in
region j (of the BLS) due to climate change only, and ∆y0c,j is a vector of
*MAGICC estimates temperature change separately for the northern and southern hemispheres.
Integrating Biophysical and Socioeconomic Factors in Modeling Impacts
277
respective yield changes due to CO 2 fertilization at 555 ppmv CO2 as derived
from crop modeling studies (Rosenzweig and Iglesias, 1994). Then, for global
climate conditions resulting from any particular energy scenario s, i.e., a
combination of projected temperature change and increase of CO2 concentration
(∆ts, ∆cs) defined in the climate model, the effective yield impact ∆yj is calculated
by interpolation according to:
∆yj(∆ts, ∆cs) = ∆y0t,j·∆ts + ∆y0c,j·g(∆cs).
Multiplier g(∆c) is specified as a quadratic function parameterized such that
it yields zero at base year concentrations, unity at a CO2 concentration level of
555 ppmv, and reaches saturation at 800 ppmv. In the BLS simulations, projected
temperature changes were read in 5-year intervals starting in 1990. Annual values
within each 5-year interval were obtained by linear interpolation. This approach,
which combines geographically explicit information on yield impacts derived
from GCM climate experiments with transient temperature change and CO2
projections, is the best that can be done given usually the lack of GCM transient
climate change results simulated with the range of emission scenarios of interest.
SO2 yield impacts
Key input parameters for defining regional sulfur damage functions in the
BLS were obtained from RAINS (Amann, 1993; Amann et al., 1995). RAINS is
a modular simulation system originally designed for integrated assessment of
alternative strategies to reduce acid deposition in Europe (Alcamo et al., 1990).
The model quantifies sulfur emissions from given activity levels in the energy
sector, both production and end-uses, traces the fate of these emissions using
atmospheric transport and chemical transformation models, calculates the amount
of sulfur deposition and estimates their impacts on soils and ecosystems. RAINS
generates results in a geographically explicit manner on a grid of 1 × 1 degree
along latitude and longitude. To parameterize the yield damage caused by dry
deposition of SO 2 , the gridded estimates of sulfur deposition and SO 2
concentrations for South and East Asia projected by RAINS-Asia were evaluated
for areas with crop cultivation, using a linear damage function:
e( x ) − 30


0.01
∆ysS, j (e( x )) = − max 0,


2.67
where
x geographic location (i.e., pixel of 1 × 1 degree along latitude and longitude);
e(x) mean annual SO2 concentration in µgm–3 at location x;
∆y sS,j yield change caused by SO2 at mean annual concentration of e(x).
The quantification of SO2 impacts on crops is difficult and controversial.
278
G. F ISCHER
Nevertheless, it was decided to attempt quantification of possible damages from
sulfur deposition in the BLS runs, because omitting these effects would have
created an unacceptable bias in the assessment. However, we must stress, there
is great uncertainty as to the magnitude of the possible SO2 damage. We use a SO2
concentration threshold of 30 µgm–3 as established for Europe (see Ashmore and
Wilson, 1993). In accordance with experiments cited in Fitter and Hay (1987) and
Conway and Pretty (1991), we have adopted the assumption that crop yield
damage increases linearly with SO2 concentration levels exceeding the threshold
such that yield is reduced by 10 percent for each 10 ppbv (i.e., each 10 ppbv ≅ 26.7
µgm–3) increase of mean annual sulfur dioxide concentrations beyond the critical
level. The estimates of crop damage by grid-box were then aggregated for the
main agricultural areas of major countries in the study region of RAINS-Asia
(e.g., China, India, Pakistan, Thailand, Indonesia). Estimates of crop damage
from SO2 deposition were also included for the former Soviet Union (FSU) and
North America (NAM) using the regional trajectories of sulfur emissions calculated
by MESSAGE in each of the energy scenarios. Consequently, the climate change
yield impact equation was amended to also include a term accounting for SO2
damage:
∆yj(∆t s, ∆cs, es) = ∆y0t,j·∆ts + ∆y0c,j·g(∆cs) + ∆ysS,j(es).
What was learned by integration?
Crop experiments and studies of the impact of climate change on crop
productivity have resulted in the understanding that global warming (i.e., the
climate effect only), on a broad regional level, will have negative impacts on
agriculture especially in tropical and sub-tropical regions. This effect is mitigated
and will often be more than compensated by beneficial effects on plants of
increasing CO2 levels, through enhancing photosynthesis and water use efficiency.
For a number of reasons, agriculture in temperate zones is expected to fare better
under climate change than tropical agriculture.
The release of pollutants in high fossil fuel emission scenarios, notably of
SO2, poses a number of environmental risks including human health effects,
acidification of soils and water bodies, fumigation of crops and forests, and
damage to buildings and engineering materials. Unlike in the debate on climate
change impacts, where the regions mainly responsible for the increase in
atmospheric CO2 concentrations may be different from those most affected by it,
the damage caused by air pollution stays more closely with the region of origin,
at least when analyzing the effects in terms of broader world regions.
While numerous studies of climate change impacts on agriculture have
arrived at similar conclusions as mentioned above, the insights offered below
could hardly be obtained without integrating biophysical, agronomic, technological
and economic factors across a wide range of models:
• In all scenario variants projected for 2050 and 2100, developing countries
experience a worsening of agricultural production trends relative to simulations
Integrating Biophysical and Socioeconomic Factors in Modeling Impacts
279
with current climate and atmospheric conditions. Outcomes for developing
regions are most negative in scenarios with high and unabated fossil fuel
combustion where both warming and SO2 pollution cause sizeable damages.
• Agricultural production in developed regions increases relative to projections
assuming current climate and atmospheric conditions. However, scenario outcomes
for developed regions are mainly determined by economic conditions resulting
from production changes and import needs in developing countries rather than
direct impacts due to changing environmental conditions.
• Sulfur emission abatement, in terms of agricultural and environmental
impacts, is a regional issue much more than a global one. While there is relatively
little difference between outcomes at the global level, regional results vary
greatly between scenarios. Hence, from a regional perspective, sulfur abatement
appears to be foremost in the interest of the polluters themselves.
MODEL-BASED ANALYSIS OF FUTURE LAND-USE/LAND-COVER CHANGE
Land use and food systems represent a critical intersection of the economy
and the environment. While studies of the Earth system are concerned with landcover changes and alterations of biochemical cycles of carbon, nitrogen, etc.,
social science disciplines and political attention relate to food security, rural
development, and sustainability of land-use systems. Land-use changes are most
often directly linked with economic decisions. This recognition has led LUC to
choose an economic framework as the organizing principle, resulting in a broad
set of project activities geared towards providing a biophysical and geographical
underpinning to the representation of actors and land-based economic sectors in
modeling land and water use decisions.
LUC has been aiming to fill this niche with theoretically sound yet practical
new approaches, including integration of diverse statistical and geographical data
sets within a Geographical Information System (GIS), agro-ecological assessments
of environmental constraints and land productivity potentials, and development
of decision tools for evaluating policy options concerning land use and agriculture
(Fischer and Makowski, 2000).
Crop suitability and land productivity assessment based on AEZ modeling
LUC has recently completed a new implementation of a series of land
evaluation steps, originating from the Food and Agriculture Organization of the
United Nations (FAO) and widely know as the AEZ methodology (for details of
the AEZ approach see “http://www.iiasa.ac.at/Research/LUC/GAEZ”). In
applying this methodology to the territory of the FSU, Mongolia, and China, LUC
has extended the AEZ from its original focus on tropical and sub-tropical
conditions to the seasonal temperate and boreal climate zones. The system works
with recent digital databases to quantify agricultural production risks as expressed
by historical variability and considers possible impacts of climate change on the
prevalence of constraints to crop production and of agricultural potentials.
Another objective was to generate geographically explicit information that could
280
G. F ISCHER
be embedded in LUC’s economic analysis and that would allow consistent
linkage to water availability assessments. The choice of applying the principles
of the AEZ methodology (FAO, 1984, 1985; UNDP/SSTC/FAO/SLA, 1994)
within the land productivity component of LUC is based on the fact that AEZ
follows an environmental approach. It provides a geographic framework for
establishing a spatial inventory and database on land resources and crop/grassland
production potential. The data requirements are sufficiently limited to enable full
coverage of a country or larger region, and it uses readily available data to the
maximum. Moreover, it is comprehensive in terms of coverage of factors
affecting production. In its simplest form, the AEZ framework contains three
elements:
(i) Selected agricultural production systems with defined input/output
relationships, and crop-specific environmental requirements and adaptability
characteristics (Land Utilization Types (LUT));
(ii) Geo-referenced land resources data (climate, soil and terrain data); and
(iii) Models for the calculation of potential yields and procedures for
matching crop/LUT environmental requirements with the respective environmental
characteristics contained in the land resources database, by land unit and gridcell.
Agro-ecological zoning involves the inventory, characterization and classification
of the land resources, to enable assessments of the potential of agricultural
production systems. This characterization includes all components of climate,
soils and landform, which are basic for the supply of water, energy, nutrients and
physical support to plants.
A water-balance model is used to quantify the beginning and duration of the
period when sufficient water is available to sustain crop growth. Soil moisture
conditions together with other climate characteristics (radiation and temperature)
are used in a simple crop growth model to calculate potential biomass production
and yield. This potential yield is then combined in a semi-quantitative manner
with a number of reduction factors directly or indirectly related to climate (e.g.,
pest and diseases) and soil and terrain conditions. The reduction factors, which
are successively applied to the potential yields, vary with crop type, the environment
(in terms of climate, soil and terrain conditions) and assumptions on level of
input/management. The final results consist of attainable crop yields under sets
of pre-defined standardized production circumstances, referred to as LUTs.
Results have been classified in five basic suitability classes according to
attainable yield: VS—very suitable (80–100 percent of maximum attainable
yield); S—suitable (60–80 percent); MS—moderately suitable (40–60 percent);
mS—marginally suitable (20–40 percent); NS—not suitable (<20 percent of
maximum attainable yield).
For each crop type and grid-cell the starting and ending dates of the crop
growth cycle are determined individually to obtain best possible crop yields,
separately for rain-fed and irrigated conditions. This procedure also guarantees
maximum adaptation in simulations with year-by-year historical weather
conditions, or under climate distortions applied in accordance with various
Integrating Biophysical and Socioeconomic Factors in Modeling Impacts
Fig. 3. AEZ index SI of land suitability for cereal production.
281
282
G. F ISCHER
climate change scenarios.
Adequate agricultural exploitation of the climatic potentials or maintenance
of productivity largely depends on soil fertility and the use and management of
the soil on an ecological sustained basis. The AEZ agro-edaphic suitability rating
scheme has been intensively used by the FAO and other organizations, at various
scales, and in numerous countries and regions; it passed through several
international expert consultations, and hence it constitutes the most recent
consolidation of expert knowledge. In this system suitability classifications are
proposed for each soil unit, by individual crops at defined levels of inputs and
management circumstances.
The model systematically tests the growth requirements of about 150 crop
types against a detailed set of agro-climatic and soil conditions. For China the
model operates on a 5 by 5 kilometer grid; so the total grid matrix has 810 by 970
cells, of which about 375 thousand grid-cells cover the mainland of China.
Results of crop suitability analysis have been summarized in tabular and map
form. On these maps, suitability results of each 5 kilometer grid-cell of the China
resource inventory are represented by a combination of the suitability index
(SI)*, reflecting the level of suitability of the part of each grid-cell considered
suitable, and by the percentage suitable in that particular grid-cell. The results for
China, assuming an intermediate level of management and input conditions, are
shown in Fig. 3.
AEZ modeling establishes a platform for developing various applications. In
LUC’s study of China, besides providing land availability scenarios and the
representation of agro-climatic and biophysical conditions for production function
estimation, other recent examples deal with China’s food prospects with special
reference to water resources (Fischer and Wiberg, 2000; Heilig et al., 2000). The
AEZ modeling clearly brings out and quantifies the vast regional differences of
China’s rain-fed and irrigated grain production potential and irrigation water
requirements.
Extended Input-Output modeling and land-use scenario analysis
Extended I-O models have been widely used for natural resource accounting,
material balances, and scenario analyses in the area of ecological and environment
assessments (e.g., Duchin, 1998). The fundamental purpose of an I-O model is to
analyze the interdependence of economic sectors. Extensions of the basic I-O
model include representations of social institutions (cf. Stone, 1970) and of the
environment (e.g., Daly, 1968; Ayres and Kneese, 1969).
The basic I-O model presents the state of an economy during a single
accounting period (generally a year) and analyzes the changes from one state to
another that have taken place in the past. Dealing with discrete and explicit
changes in economic structure through rigorous accounting constitutes the most
distinguished feature of I-O modeling. Through the evaluation of scenarios that
*The suitability index is defined as SI = 0.9 VS + 0.7 S + 0.5 MS + 0.3 mS.
Integrating Biophysical and Socioeconomic Factors in Modeling Impacts
283
Table 1. Schematic representation of an extended Input-Output table.
reflect current thinking and by pinpointing the inadequacies and inconsistencies
in these scenarios, as a basis for improving them, scenario analysis based on I-O
modeling can stimulate new insights in the search for promising future
developments.
Table 1 provides the scheme of an extended I-O table. The matrixes of interindustrial flows (zij), of final deliveries (uis), and of factor inputs (vkj), correspond
to a standard I-O table used in economic analysis. For application in environmental
analysis, the standard I-O table is augmented by a representation of inputs and
outputs of land, water and other environmental resources, as shown in Table 1.
The rows in the bottom panel show the quantities of different resources used
by the various economic sectors (Lrj), and by households and other institutions
(Lrs). The columns in matrix (dir) account for the depreciation and degradation of
these resources caused by industrial, household and other social activities. For the
land-use scenario analysis, we included various land types as inputs (denoted as
Lj and Ls hereafter) in the enhanced I-O table. These include the uses by economic
sectors of the major land categories such as rain-fed and irrigated cropland,
grassland, forestland, and land used for industry and mining, transportation,
residence and services.
For a single accounting period, the input-output relations are represented by
fixed coefficients. When dealing with another state of the economy, corresponding
to a different accounting period, usually also a different set of coefficients is
established to represent the altered structure of the economy.
284
G. F ISCHER
These changes are derived from scenarios developed around each question
to be explored. Structural changes include the technologies in use in different
sectors, changes in the relative size of different sectors, changes in the composition
and magnitude of various final demand sectors, and changes in the availability
and quality of different environmental resources. A central piece of information
in scenario development is technical literature, expert knowledge, and, where
possible, modeling of sub-systems to provide insights into current and potential
future production processes, population and other social trends, and the
environment. Given a specific research topic, modelers need to first identify what
the important contributing factors and issues are, usually through a qualitative
analysis based on literature and expert opinions. Then the task is to identify a
comprehensive yet minimal set of variables for modeling, and to quantify
parameters. This step provides great flexibility for embedding biophysical
aspects into the economic analysis although it is usually not possible to maintain
full geographical detail.
Steps in implementing an I-O based scenario analysis
The procedure of the land-use scenario analysis undertaken for China
consists of seven logical steps (for details see Hubacek and Sun, 1999):
Step 1: Stylize various scenarios of population growth, urbanization, changes
of lifestyles, and per capita income growth.
Step 2: Define quantitative scenarios according to bundles of changes, which
are considered to be the most interesting or most representative ones. This is done
mainly based on literature surveys across different research fields as well as
international comparisons.
Step 3: Translate the selected social and economic scenarios into
corresponding future states of final demand by different economic sectors.
Step 4: Referring to Table 1, calculate the I-O matrix of technical coefficients,


A = aij = zij  ∑ zij + ∑ vkj  ,
 i

k
( )
for the selected future years in the scenario schedule. For example, let us consider
year 2025. The future sectoral structure requires a new inter-sectoral technical
structure. Because of path dependence and economic inertia, it is plausible to
assume that the future technical structure will have a close linkage with the
current technical structure. To establish a consistent I-O technical coefficient
matrix for year 2025, we minimize the difference between the current coefficient
matrix (Acurrent) and the matrix representing the economy in the target year, matrix
Integrating Biophysical and Socioeconomic Factors in Modeling Impacts
285
(A2025). The distance to be minimized is defined as*:

  aij (2025)   
D[ Acurrent : A2025 ] = ∑ ∑ aij (2025)ln
 − 1 .
  aij (current )   
i j 

This generates the “least surprising” representation of matrix (A2025) as it fully
incorporates both the historical information (Acurrent) and the projected structural
information of sectoral output (X2025), intermediate deliveries (U2025), and
intermediate purchases (V2025).
Step 5: Using the Leontief inverse matrix, (I – A)–1, we obtain estimates of
sectoral total outputs, x = (I – A)–1y. Here, x and y denote, respectively, the vectors
of output (xi) = ( ∑ zij + yi ) and final demand (yi) = ( ∑ uis ). Letter I indicates the
j
x
identity matrix, and A is the matrix of I-O coefficients as before.
Step 6: Establish scenarios of land supply based on the results of the AEZ
assessment as well as other technical changes in non-agricultural sectors, and
calculate the vector of sectoral land requirement coefficients, (cj) = (Lj/xj). The
inverse of a coefficient cj measures the output of a given sector per unit of land
employed, i.e., sectoral land productivity. In order to link land requirements
associated with changes in economic sectors to changes in land categories (∆L),
the vector representing changes in sector output (∆x) is pre-multiplied by a
)
diagonal land requirement coefficient matrix ( C ) and a land distribution matrix
(R). Future land use (LF) is calculated by adding the present land uses (LP) and the
changes in land use (∆L) caused by the changes in output:
)
∆L = RC∆x and LF = LP + ∆L.
The land distribution matrix R constitutes the mapping relationship between land
uses in economic sectors and the natural categories of land, and the coefficients
in R are the shares of the former in the latter.
Step 7: Compare the results from different land-use scenarios against
various land availability limits, and calculate the necessary land productivity
increases that would keep the future sectoral land requirements consistent with
the availability of land. Alternatively, productivity increases can be limited, and
*This distance function is named the “information gain entropy function” and is employed in
the RAS technique (Theil, 1967; Miller and Blair, 1985). The term RAS method refers to a
mathematical procedure for adjusting sequentially the rows and columns of a given I-O coefficient
matrix, A(0), in order to generate an estimate of a revised matrix, A(1), when only the new structural
information of sectoral output, intermediate deliveries, and intermediate purchases, are known.
Once the procedure converges, the final outcome is usually written as A(1) = RA(0)S (hence the name
RAS method), where R and S denote diagonal matrices constructed by the algorithm.
286
G. F ISCHER
either imports can supplement domestic production, or final demand must be
curtailed.
The seven-step procedure, as outlined above, was applied in an aggregate
national I-O model of China as well as in a multi-region model covering seven
(out of eight; the Plateau region had to be dropped for lack of data) major
economic regions for which a regional I-O table has been established by the LUC
project.
In summary, the increases in final demands and sectoral outputs corresponding
to a range of scenarios would drive the associated land requirements to exceed the
availability of suitable land areas for certain categories unless an optimistic
technology scenario is assumed. In other words, China may not be able to support
the increasing demand for land-intensive products with its land base without
significant improvements in land productivity and/or increasing imports. When
imposing a continuing self-sufficiency in grain and food, a high annual growth in
land-productivity of about 2 percent is required to match the required farmland
with available resources. Such growth is higher than what is usually expected for
the next 30 years. Hence, a moderate grain import in general, and of feed grains
in particular—some 10 to 15 percent of China’s current total grain output—would
significantly reduce the pressure on China’s cropland and water resources*.
From the extended I-O model to a dynamic welfare optimum model
The I-O scenario analysis discussed in the previous section provides an
interesting initial assessment of the feasibility of land-use trends with respect to
selected scenarios of possible future directions of the Chinese economy and
society. The analysis showed clearly the stringent scarcity of land resources and
made a strong appeal for rapid technology improvements. The results of the I-O
assessment call for extending the analysis to include the adaptation of economic
actors and the resulting consequences for land and water use.
The LUC economic model of China is designed to establish a tightly
integrated assessment of the spatial and intertemporal interactions among various
socioeconomic and biophysical factors that drive land-use and land-cover change.
The applied general equilibrium (AGE) framework (Ginsburgh and Keyzer,
1997) makes use of the typical I-O accounting tables as the initial representation
of the economy. Moreover, in dynamic AGE modeling, the results from extended
I-O analysis can serve as a sound initialization of the dynamic optimization.
The basic justification for adopting the dynamic AGE framework in land-use
analysis is as follows. From an economic perspective, interactions between
climate, land resources, and vegetation largely depend on how these resources are
transformed and managed by human activity. The objective of a dynamic welfare
*As determined by the AEZ analysis, an import of 50 million tons (i.e., 10 percent of current
grain output) is equal to about 40 percent of the production increases estimated to result from full
application of irrigation in China’s northern regions where water scarcity has become a major
constraint to expansion of agriculture.
Integrating Biophysical and Socioeconomic Factors in Modeling Impacts
287
optimum model is to elaborate socially desirable and economically efficient
trajectories of resource uses and transformations. In mathematical terms, the
welfare optimum levels of resource uses and transformations are a function of the
initial state of the economy and of resources, of the parameterization of consumer
preferences and production relations, and of (exogenously) specified dynamics
and constraints. Though optimization may be regarded as too idealistic, it helps
to distinguish problems attributable to incompatibilities of fundamental
relationships, for example population dynamics and resource availability, from
those problems induced by specific modes of social organization and institutional
settings. The results of AGE modeling are thus particularly helpful and relevant
for policy analysis.
The integrated agricultural production function
To transform the theoretical framework, a key requirement of socioeconomic
and environmental land-use analysis in China is to establish an integrated
agricultural production function for use in the AGE framework (Keyzer, 1998).
For use in optimization, it should permit consistent aggregation over farm units
within each economic region. Furthermore, once the optimal solution has been
determined in the regional AGE model, the production relations should then
allow for a disaggregation of the results to the basic sub-regional observation
units (i.e., the county level). The specification of the LUC agricultural production
relations meets all of these requirements.
The basic structure of the agricultural production relations consists of an
output function and an input response function, linked by means of an agricultural
output index. This output index is a quantity aggregation of crops produced. It is
based on a standard aggregation rule used in economics, namely a constantelasticity-of-transformation (CET) function. As a starting point for estimation,
we derived a database of gross value of crop production per hectare of cultivated
land in 1990. The compilation was based on province-level prices and countylevel statistical information on output quantities, land survey data, and mapped
distribution of cultivated land, as shown in Fig. 4.
Crop production is co-determined by the biophysical potential of land, and
by the level of factor inputs per unit of cultivated land. The potential output is
derived from an AEZ assessment of agro-climatic and biophysical conditions.
The input response function is specified as a generalization of a popular yield
function in the agronomic literature, called the Mitscherlich-Baule yield function
(Franke et al., 1990; Llewelyn and Featherstone, 1997). The rationale behind this
specification is that the observed actual crop output level represents a certain
fraction of the biophysical potential and is determined by the factor input levels
per unit of land, and by the technology employed. In mathematical terms,
aggregate output Q is obtained as,
(
Q = C0 ∏ f j V , H ( A; δ ); β j
j
θ
) N ( A, y ( x );ν )
j
288
G. F ISCHER
Fig. 4. Average value of crop output per hectare of cultivated land in 1990 (Yuan/ha).
where N(·) specifies an aggregate potential output index incorporating the
maximum attainable yield estimates y , for given soil and climatic conditions x
and derived through AEZ. Inputs of power and nutrients are denoted by V, an
aggregate index of land types per observation unit is given as H(A; δ ). All other
symbols denote parameters and should be determined by empirical estimation. In
the Mitcherlich-Baule yield curve, function f is specified as saturating exponential
function. The theoretical background and specification details of the agricultural
production relations are discussed in Albersen et al. (2000).
Output and input response functions of China’s cropping agriculture in 1990
were estimated for seven of the eight economic regions distinguished in the LUC
model. The basic observation units are counties (2358 in total). The parameter
estimations were deliberately conducted in the form of primal regressions, even
though estimation of such non-linear problems can be difficult from a numerical
perspective. Estimation of the production relations in primal form provides a
methodological advantage in terms of modeling farmers’ behavior in a transition
economy like China, where both state-controlled and market prices coexist for
major agricultural products, and furthermore in which the statistical data on
prices, if available, are averaged ones without details as to how the averaging is
conducted. The primal estimation uses only the physical quantities of outputs and
inputs. The dual variables associated with the primal optimization problem
represent economically meaningful signals, usually termed as “shadow prices”.
Integrating Biophysical and Socioeconomic Factors in Modeling Impacts
289
The estimation procedures and corresponding significance tests were implemented
in GAMS (Brooke et al., 1992); their convergence was facilitated through
elaborating step-wise programming and proper loop structures (see Albersen et
al., 2000).
The specification of the agricultural production function assumes that the
representative farmer in each region determines optimal output and input mixes
across all suitable crops and key input factors. Decision-making is subject to
varying conditions of climate, soil, landform and other natural factors across the
counties in the region. This means that the specification captures farmers’
adaptation to variations in natural conditions. It incorporates detailed agronomic
knowledge, and thus also provides an ideal means for assessing the impact of
climate change. This is important, as detailed biophysical assessments of cropping
impacts and water availability indicate great differences in climate impact
consequences across different regions in China (Fischer and Wiberg, 2000).
Much of the structural information on China’s agriculture was revealed by
the cross-section estimation carried out within each economic region. The results
strongly support the view that it is both possible and worthwhile to integrate
information from biophysical process understanding within an economic model.
CONCLUDING REMARKS
The paper discusses examples of methodologies from IIASA’s global
environmental change research where integration of biophysical and
socioeconomic factors is of crucial importance to the analysis. A first application
provided an integrated assessment of the impacts of alternative energy paths on
global environmental change and regional food provision, the second application
illustrates how spatially explicit biophysical factors can be embedded in an
economic model based on optimization.
In the first case, an integrated systems-engineering and macroeconomic
energy model, a regional air pollution model, and a global food system model
were combined in a process termed soft-linking, i.e., a harmonization of basic
assumption and transfer of relevant outputs among models. Furthermore, the
procedure ensured that results from climate and crop models was consistently
embedded in the economic analysis.
In the second example, three complementary methodologies were applied
within a land-use change and food systems study of China. While each method by
itself can provide useful information and is well regarded in its discipline, only
their combination provides a powerful and integrative tool for policy analysis. An
AEZ assessment was carried out, (i) to provide land availability estimates,
employed in land-use scenario analysis based on an extended I-O model, and
(ii) to generate an agro-environmental characterization for use in the specification
and estimation of a spatially explicit agricultural production function. This
production function in turn plays a key role in LUC’s regionalized applied
general equilibrium model of China. An extended I-O model is used to obtain an
initial assessment of the feasibility of land-use trends with respect to a range of
scenarios of possible future directions of China’s economy and society. The
290
G. F ISCHER
results from the I-O modeling also serve as a sound initialization of the dynamic
general equilibrium model.
The issues of global environmental change, in particular also of land use and
cover change have been studied by many scientific disciplines. A better
communication between these disciplines has often been desired and postulated
as a prerequisite for further development of improved methodologies. On the
basis of rich modeling experience accumulated at IIASA, we argue that a solution
is not likely to be found in a theory of everything but rather in improved ways of
channeling and translating disciplinary information among natural and social
science disciplines. We conclude that models based on economic theory and
rational behavior are perhaps the most appropriate tools to modeling adaptation
to and consequences of global environmental change. However, when dealing
with global environmental change, economists and their models need to pay more
attention to three aspects: first, a stronger integration of biophysical conditions
and process characteristics in specifying economic activities; second, a more
adequate embedding of geographical features and spatial relations; and third, a
more careful consideration of spatial scale, spatial heterogeneity and aggregation.
Acknowledgments—The integrated energy-environment-economy study was a major
collaborative effort of three research projects at IIASA: the Environmentally Compatible
Energy Strategies (ECS) project led by Neboja Nakicenovic, the Transboundary Air
Pollution (TAP) project under the leadership of Markus Amann, and the Land Use Change
(LUC) project, with major contributions from Cynthia Rosenzweig and Ana Iglesias
(Columbia University, NY, USA). The research of the LUC project is a multidisciplinary
and collaborative effort. It has involved researchers at IIASA and in various collaborating
institutions in China, Europe, Japan, Russia, and United States. For the work presented in
this paper, the author is grateful and owes due recognition to the researchers who have
developed and significantly contributed to the various themes: Harrij T. van Velthuizen
(IIASA, LUC) contributed to the AEZ modeling. Klaus Hubacek (IIASA, LUC and
Rensselaer Polytechnic Institute, Troy, NY, USA) and Laixiang Sun (IIASA, LUC)
contributed to developing and implementing the extended I-O analysis; Peter Albersen
and Michiel A. Keyzer (Free University, SOW-VU, Amsterdam, The Netherlands) and
Laixiang Sun (IIASA, LUC) designed and greatly contributed to implementing and
estimating the agricultural production functions used in the welfare optimum model. Li
Xiubin, Liu Yanhua, Zhao Mingcha (Institute of Geography, Chinese Academy of
Sciences, Beijing, China) and Zheng Zenyuan (State Land Administration, Beijing,
China) greatly supported the provision, interpretation and compilation of data.
REFERENCES
Albersen, P., Fischer, G., Keyzer, M. and Sun, L. (2000). Estimation of agricultural production
relation in the LUC Model for China. Interim Report IR-2000-027, IIASA, Laxenburg, Austria.
Alcamo, J., Shaw, R. and Hordijk, L. (eds.) (1990). The RAINS Model of Acidification, Science and
Strategies in Europe. Kluwer Academic Publishers, Dordrecht, Netherlands.
Amann, M. (1993). Transboundary Air Pollution: The RAINS model, Options, Winter ’93,
International Institute for Applied Systems Analysis, Laxenburg, Austria.
Amann, M., Cofala, J. and Hordijk, L. (1995). Integrated Assessment. In: RAINS ASIA: An
Assessment Model for Air Pollution in Asia, Chapter 2 of the Report on the World Bank
Sponsored Project “Acid Rain and Emission Reductions in Asia”, International Institute for
Integrating Biophysical and Socioeconomic Factors in Modeling Impacts
291
Applied Systems Analysis, Laxenburg, Austria.
ASA (1995). Climate change and agriculture: Analysis of potential international impacts. ASA
Special Publication Number 59. American Society of Agronomy, Madison, Wisconsin, USA.
Ashmore, M. R. and Wilson, R. B. (eds.) (1993). Critical Levels of Air Pollutants for Europe.
Creative Press, Reading, UK.
Ayres, R. and Kneese, A. (1969). Production, consumption, and externalities. American Economic
Review 59(3), 282–297.
Brooke, A., Kendrick, D., and Meeraus, A. (1992). GAMS: A User’s Guide, Release 2.25. Boyd &
Fraser Publishing Company, Danvers, MA.
Cofala, J. and Dörfner, P. (1995). Constructing regional energy scenarios from global energy
models. WP-95-112, International Institute for Applied Systems Analysis, Laxenburg, Austria.
Conway, G. R. and Pretty, J. N. (1991). Unwelcome Harvest: Agriculture and Pollution. Earthscan,
London, UK.
Daly, H. (1968). On economics as a life science. Journal of Political Economy 76(3), 392–406.
Duchin, F. (1998). Structural Economics: Measuring Change in Technology, Lifestyles, and the
Environment. Island Press, Washington, D.C.
FAO (1984). Guidelines: Land evaluation for rain-fed agriculture. FAO Soils Bulletin 52. Food and
Agricultural Organization of the United Nations, Rome.
FAO (1985). Guidelines: Land evaluation for irrigated agriculture. FAO Soils Bulletin 55. Food and
Agricultural Organization of the United Nations, Rome.
Fischer, G. and Makowski, M. (2000). Land-use planning. In: Model-Based Decision Support
Methodology with Environmental Applications, Wierzbicki, A. P., Makowski, M., and Wessels,
J. (eds.), Kluwer Academic Publishers, Dordrecht/Boston/London, pp. 333–365.
Fischer, G. and Rosenzweig, C. (1996). The impacts of climate change, CO 2 and SO2 on agricultural
supply and trade: An integrated assessment. International Institute for Applied Systems Analysis,
Laxenburg, Austria.
Fischer, G. and Wiberg, D. (2000). Climate Change Impacts on Water-stressed Agriculture in
Northeast China. Paper presented at Workshop on Natural Environment Management and
Applied Systems Analysis, September 2000. International Institute for Applied Systems Analysis,
Laxenburg, Austria.
Fischer, G., Frohberg, K., Keyzer, M. A., and Parikh, K. S. (1988). Linked National Models: A Tool
for International Policy Analysis. Kluwer Academic Publishers, Netherlands.
Fischer, G., Frohberg, K., Keyzer, M. A., Parikh, K. S., and Tims, W. (1990). Hunger—Beyond the
reach of the Invisible Hand, IIASA, Laxenburg.
Fischer, G., Frohberg, K., Parry, M. L., and Rosenzweig, C. (1994). Climate change and world food
supply, demand and trade: Who benefits, who loses? Global Environmental Change 1994 4(1),
7–23.
Fischer, G., Frohberg, K., Parry, M. L., and Rosenzweig, C. (1996). Impacts of potential climate
change on global and regional food production and vulnerability. In: Climate Change and World
Food Security, Downing, E. T. (ed.), Springer-Verlag, Berlin, Germany.
Fitter, A. H. and Hay, R. K. M. (1987). Environmental Physiology of Plants. Academic Press,
London, UK.
Franke, M. D., Beattie, B. R., and Embleton, M. F. (1990). A comparison of alternative crop response
models. American Journal of Agricultural Economics 72, 597–602.
Ginsburgh, V. and Keyzer, M. (1997). The Structure of Applied General Equilibrium Models. The
MIT Press, Cambridge, MA.
Heilig, G. K., Fischer, G., and van Velthuizen, H. (2000). Can China feed itself? An analysis of
China’s food prospects with special reference to water resources. International Journal of
Sustainable Development and World Ecology, 7, 153–172.
Hubacek, K. and Sun, L. (1999). Land-use change in China: A scenario analysis based on inputoutput modeling, Interim Report IR-99-073, IIASA, Laxenburg, Austria.
Hulme, M., Raper, S., and Wigley, T. M. L. (1995). An integrated framework to address climate
change (ESCAPE) and further development of the global and regional climate modules
(MAGICC). Energy Policy, volume 23(4/5), 347–356.
292
G. F ISCHER
Keyzer, M. (1998). Formulation and spatial aggregation of agricultural production relationships
within the Land Use Change (LUC) model. Interim Report IR-98-092, IIASA, Laxenburg,
Austria.
Llewelyn, R. V. and Featherstone, A. M. (1997). A comparison of crop production functions using
simulated data for irrigated corn in western Kansas. Agricultural System 54(4), 521–538.
Lutz, W., Sanderson, W., Scherbov, S., and Goujon, A. (1996). World population scenarios for the
21st century. In: The Future of the World, IIASA and Earthscan, Lutz, W. (ed.), Laxenburg,
Austria and London, UK.
Manne, A. and Richels, R. (1992). Buying Greenhouse Insurance: The Economic Costs of CO 2
Emission Limits. MIT Press, Cambridge, USA.
Miller, R. E. and Blair, P. D. (1985). Input-Output Analysis: Foundations and Extensions. PrenticeHall, Inc., New Jersey.
Nakicenovic, N., Amann, M., and Fischer, G. (1997). Global Energy Supply and Demand and Their
Environmental Effects, Report to the Central Research Institute of the Electric Power Industry
(CRIEPI). Internal document. International Institute for Applied Systems Analysis, Laxenburg,
Austria.
Parikh, K. S., Fischer, G., Frohberg, K., and Gulbrandsen, O. (1988). Toward Free Trade in
Agriculture. Martinus Nijhoff, The Hague, Netherlands.
Rosenzweig, C. and Parry, M. L. (1994). Potential impacts of climate change on world food supply.
Nature, 367, 133–138.
Rosenzweig, C. and Iglesias, A. (eds.) (1994). Implications of Climate Change for International
Agriculture: Crop Modeling Study. United States Environmental Protection Agency, Office of
Policy, Planning and Evaluation, Washington, D.C., USA.
Stone, R. (1970). Demographic input-output: an extension of social accounting. In: Contributions
to Input-Output Analysis, Vol. 1, Carter, A. P. and Brody, A. (eds.), North-Holland, Amsterdam,
pp. 293–319.
Strzepek, K. M. and Smith, J. B. (eds.) (1995). As Climate Changes: International Impacts and
Implications. Cambridge University Press.
Theil, H. (1967). Economics and Information Theory. North Holland, Amsterdam.
Turner, B. L., II, Skole, D., Sanderson, S., Fischer, G., Fresco, L., and Leemand, R. (1995). LandUse and Land-Cover Change: Science/Research Plan. IGBP Report No. 35 and HDP Report No.
7. IGBP, ICSU, HDP and ISSC, Stockholm and Geneva.
UNDP/SSTC/FAO/SLA (1994). Land Resources, Use and Productivity Assessment in China. Main
Report and Technical annex (Vol. 1–9) of Project CPR/87/029/B/01/R. United Nations
Development Program, State Science and Technology Commission of the People’s Republic of
China, Food and Agriculture Organization of the United Nations, State Land Administration of
the People’s Republic of China, Beijing.
Wigley, T. M. L. and Raper, S. C. B. (1992). Implications for climate and sea level of revised IPCC
emissions scenarios. Nature, 357, 293–300.
G. Fischer (e-mail: [email protected])