Socio-economic and climate change impacts

Downloaded from http://rstb.royalsocietypublishing.
org/ on January 14, 2015
Phil. Trans. R. Soc. B (2005) 360, 2067–2083

doi:10.1098/rstb.2005.1744
Published online 24 October 2005
Socio-economic and climate change impacts on

agriculture: an integrated assessment, 1990–2080
Günther Fischer1,*, Mahendra Shah1, Francesco N. Tubiello1,2
and Harrij van Velhuizen1
1
International Institute for Applied Systems Analysis (IIASA), Laxenburg, Austria
2
Goddard Institute for Space Studies, Columbia University, NY, USA
A comprehensive assessment of the impacts of climate change on agro-ecosystems over this century is
developed, up to 2080 and at a global level, albeit with significant regional detail. To this end an
integrated ecological–economic modelling framework is employed, encompassing climate scenarios,
agro-ecological zoning information, socio-economic drivers, as well as world food trade dynamics.
Specifically, global simulations are performed using the FAO/IIASA agro-ecological zone model, in
conjunction with IIASAs global food system model, using climate variables from five different general
circulation models, under four different socio-economic scenarios from the intergovernmental panel
on climate change. First, impacts of different scenarios of climate change on bio-physical soil and
crop growth determinants of yield are evaluated on a 5 0 !5 0 latitude/longitude global grid; second,
the extent of potential agricultural land and related potential crop production is computed. The
detailed bio-physical results are then fed into an economic analysis, to assess how climate impacts
may interact with alternative development pathways, and key trends expected over this century for
food demand and production, and trade, as well as key composite indices such as risk of hunger and
malnutrition, are computed. This modelling approach connects the relevant bio-physical and socio-
economic variables within a unified and coherent framework to produce a global assessment of food
production and security under climate change. The results from the study suggest that critical impact
asymmetries due to both climate and socio-economic structures may deepen current production and
consumption gaps between developed and developing world; it is suggested that adaptation of
agricultural techniques will be central to limit potential damages under climate change.
Keywords: agriculture; crop production; climate change; food security; sub-Saharan Africa;
risk of hunger
1. INTRODUCTION and possibly increased frequency of extreme events

There is significant concern about the impacts of such as drought and floods, will probably combine to
climate change and its variability on agricultural depress yields and increase production risks in many
production worldwide. First, issues of food security world regions, widening the gap between rich and poor
figure prominently in the list of human activities and countries (e.g. IPCC 2001a,b). A consensus has
ecosystem services under threat of dangerous anthro- emerged that developing countries are more vulnerable
pogenic interference on Earth’s climate (Watson et al. to climate change than developed countries, because of
2000; IPCC 2001a,b; Ecosystem Millennium Assess- the predominance of agriculture in their economies,
ment 2005; see also article II, UNFCCC). Second, the scarcity of capital for adaptation measures, their
each country is naturally concerned with potential warmer baseline climates and their heightened
damages and benefits that may arise over the coming exposure to extreme events (Parry et al. 2001). Thus,
decades from climate change impacts on its territory as climate change may have particularly serious con-
well as globally, since these will affect domestic and sequences in the developing world, where some 800
international policies, trading patterns, resource use, million people are undernourished. Of great concern is
regional planning and ultimately the welfare of its a group of more than 40 ‘least-developed’ countries,
people. mostly in sub-Saharan Africa, where domestic per
Current research confirms that while crops would capita food production declined by 10% in the last 20
respond positively to elevated CO2 in the absence of years (e.g. UN Millennium Project 2005).
climate change (e.g. Kimball et al. 2002; Jablonski et al. Many interactive processes determine the dynamics
2002; Ainsworth & Long 2005), the associated impacts of world food demand and supply: agro-climatic
of high temperatures, altered patterns of precipitation conditions, land resources and their management are
clearly a key component, but they are critically affected
by distinct socio-economic pressures, including current
* Author for correspondence ([email protected]). and projected trends in population growth, availability
One contribution of 17 to a Discussion Meeting Issue ‘Food crops in and access to technology and development. In the last
a changing climate’. three decades, for instance, average daily per capita
2067 q 2005 The Royal Society

Downloaded from http://rstb.royalsocietypublishing.org/ on January 14, 2015
2068 G. Fischer and others An integrated assessment of agriculture, 1990–2080
intake has risen globally from 2400 to 2800 cal (from ecological–economic analysis
10 to 12 MJ ), spurred by economic growth, improved
production systems, international trade and globaliza-
tion of food markets. Feedbacks of such growth
patterns on cultures and personal taste, lifestyle and AEZ
demographic changes have in turn led to major dietary
changes—mainly in developing countries, where shares
of meat, fat and sugar to total food intake have
increased by about 40% (e.g. Fischer et al. 2002b).
climate climate impact
Given the virtual impossibility to test experimen-
model response relations
tally, or to simply try to sum up in a linear fashion all
relevant agro-climatic and socio-economic factors
involved in determining long-term future trends, it is
no surprise that the scientific literature is replete with production demand
modelling studies attempting to assess at least some of
the aspects likely to characterize the impacts of climate SRES
change on future agricultural production. There is a BLS
scenario
wealth of site-specific, regional and/or national short- global food system
and long-term assessments of climate change impacts trade
(basic linked system)
performed to date (e.g. Rosenzweig et al. 2002;
Tubiello et al. 2002; Olesen & Bindi 2002; Reilly
et al. 2003). Global assessments of climate change on
world market
food production have been less frequent (i.e. Rosenz-
weig & Parry 1994; Fischer et al. 2002b; Parry et al. Figure 1. Graphic description of the AEZ–BLS modelling
2004), due in part to the difficulties of gathering framework. Socio-economic SRES scenarios determine both
comprehensive global agro-climatic datasets, in part to climatic and market conditions under which AEZ and BLS
the need to employ global trade economic models (see are run. Climatic impacts on agricultural production—
below), as well as due to the significant computer computed with AEZ, are passed on to the agricultural
resources required. economics and trade model, BLS, to determine overall
Such assessment studies have focused mainly on impacts on world food systems.
agro-climatic components, including simplified
approaches to simulating adaptation responses, i.e.
changes in agro-technology that enable farmers to growth and climate change. In short, the questions
minimize risks and/or maximize profits under changed we address herein are as follows: what are the likely
climates. Focus of assessment studies with adaptation impacts of climate change on the world’s agricultural
has either been on explicit, simple farm-level adap- resources? How do climate impacts compare to socio-
tation strategies; or on implicit, market-driven response economic pressures over this century? Where and how
functions (see, for discussion, Rosenzweig & Tubiello do significant interactions arise? A few previous studies,
in press). The first approach is local, allowing for agro- notably Rosenzweig & Parry (1994), FAO (2003) and
technological detail, but lacking key regional market Parry et al. (2004), have employed various components
feedbacks. It includes on-site evaluation of strategies of the FAO–IIASA methodology to address these same
such as early planting, use of cultivars better adapted to questions; we believe, however, that this is the first time
altered climates or modifications to water and/or that a fully coherent, unified data and modelling system
fertilizer levels. The second approach better includes has been used.
agro-economic dynamics over a region, but cannot Specifically, we employ the FAO–IIASA agro-
provide specific technical solutions. It includes gener- climatic database and modelling framework known as
alized strategies such as regional shifting of cropping the agro-ecological zones or AEZ, model (e.g. Fischer
systems and management; responses based on pesticide et al. 2002a,b), in conjunction with four socio-
and/or fertilizer use, etc. Furthermore, local studies economic scenarios defined by IPCC and the IIASA
with crop models allow for better calibration and world food system or basic linked system (BLS)
validation compared to regional approaches. (Fischer et al. 1988, 2002b; Parry et al. 1999) (figure 1).
Clearly, both methodologies are necessary in order The main simulation results of the study herein
to conduct more realistic regional studies. In addition, presented include climate change impacts on agro-
and importantly, local to regional effects of global food climatic resources, potential arable land and related
trade would need to be included within either changes in crop production patterns. Our economic
approach. This is because international trade can analyses assess over this century changes to food
greatly modify the regional dynamics of food demand, demand, production, trade and prices and the scale
production and supply under present climate, and thus and location of risk of hunger.
significantly modulate impacts under climate change
(e.g. Reilly et al. 2001; Tubiello 2005). 2. MATERIAL AND METHODS
This paper presents an integrated ecological–econ- The combination of a spatially detailed bio-physical/
omic modelling framework for the assessment of the agronomic assessment tool and a global food system model
world food system in the twenty-first century, under provides an integrated ecological–economic framework for
various future scenarios of population, economic the assessment of the impacts of climate change and
Phil. Trans. R. Soc. B (2005)

An integrated assessment of agriculture, 1990–2080 G. Fischer and others 2069
agricultural vulnerability (figure 1). The focus has been on (mean monthly minimum temperature, mean monthly
analysing the current and future availability and use of maximum temperature, precipitation, solar radiation, vapour
regional and global land resources, in the context of pressure deficit) for the period 1901–1996, and including a
technological and environmental changes, including climate monthly mean climatology based on the decades 1960–1990.
change and climate variability. In AEZ, the CRU data are transformed into daily data and
analysed vis-à-vis crop requirements (e.g. Fischer et al.
2002a). Finally, AEZ employs a land-cover/land-use layer
(a) AEZ modelling methodology specifying distributions of aggregate land-cover classes, as
Because of the complex interactions between climate, agro- derived from global 1 km land-cover datasets from NOAA
ecosystem dynamics and human management at farm to AVHRR and GLC2000, respectively. The AEZ global land
regional level, assessment studies of the impacts of climate resources database incorporates spatial delineation and
change on agriculture have involved the use of computer accounting of forest and protected areas. In terms of key
simulations that link climate predictions of general circulation socio-economic datasets, AEZ employs a global population
models (GCMs) together with crop models and land data set calibrated for the year 2000, including estimates of
management decision tools of various kinds (e.g. Tubiello & spatially explicit population distribution and densities for
Ewert 2002). Specific dynamic crop models such as DSSAT each country.
(e.g. Tsuji et al. 1994; Rosenzweig et al. 1995), EPIC (e.g. In summary, the AEZ framework contains the following
Williams et al. 1984), etc. or simpler, modified ecosystem basic elements:
models, such as TEM (e.g. Feltzer et al. 2004), have been
employed for computing crop growth, water dynamics and (i) land resources database, containing geo-referenced
harvest yield, as a function of soil, climate and management climate, soil and terrain data;
data. The first set of models is more agronomic in nature, (ii) land utilization types (LUT) database of agricultural
allowing for computation of many details of a crop life cycle production systems, describing crop-specific environ-
that might be important for final yield—such as differential mental requirements and adaptability characteristics,
climatic and management impacts during vegetative versus including input level and management;
flowering stages, etc.—and that are often missed by the (iii) mathematical procedures for matching crop LUT
simpler ecosystem models. Likewise, the first set of crop requirements with AEZs data, including potentially
models allows for simulation of more realistic field manage- attainable crop yields estimates, by land unit and grid-
ment activities such as type of water and fertilizer manage- cell (AEZ global assessment assesses 2.2 million grid
ment, sowing and harvesting operations, etc. For these cells, covering a 5 0 !5 0 latitude/longitude grid, based on
reasons, dynamic crop models require many input data for a 1 : 5 000 000 scale global soil map);
calibration and validation, and perform best at local to (iv) assessments of crop suitability and land productivity;
national scales, provided sufficiently detailed ‘representative and
sites’ can be found to cover the area of study (e.g. Tubiello (v) applications for agricultural development planning.
et al. 2002). On the other hand, simplified ecosystem models
employ generalized crop algorithms and are better suited to As a part of its assessment, the AEZ model computes
large-scale simulations, requiring less detailed input data to amounts of non-arable and arable land, as a function of
run; yet they may be prone to larger errors and problematic environmental constraints. Land is classified as having severe
validation, due to lack of both crop and management detail. constraints (either too cold, too wet, too steep or having
The AEZ modelling framework synthesizes essential serious soil quality constraints); moderate, slight or no
components of both the crop and ecosystem models constraints to cultivation. Classification is also made between
described above. It uses detailed agronomic-based knowledge rainfed and irrigated land, depending on water deficits
to simulate land resources availability and use, farm-level computed internally as precipitation minus evapo-
management options and crop production potentials; at the transpiration.
same time, it employs detailed spatial bio-physical and socio- The AEZ model had been validated for use in agricultural
economic datasets to distribute its computations at fine resource assessment and employed in many studies, both
gridded intervals over the entire globe (e.g. Fischer et al. regionally and globally (e.g. Fischer & Sun 2001; Fischer et al.
2002a). This land-resources inventory is used to assess, for 2002a; ). Importantly, AEZ is one of the main tools used by
specified management conditions and levels of inputs, the FAO for analyses of present and future land resources, both
suitability of crops in relation to both rainfed and irrigated regionally and globally (e.g. FAO 2003).
conditions, and to quantify expected attainable production of In this study, following the approach of several previous
cropping activities relevant to specific agro-ecological con- assessments of world food production, we have chosen to
texts characterizing the study area. The characterization of focus on cereal-production (including wheat, maize and other
land resources includes components of climate, soils, land- coarse grains) as a proxy for global world agricultural
form and present land-cover. Crop modelling and environ- production. Cereals, in fact account for over two-thirds of
mental matching procedures are used to identify crop-specific average caloric intake, and provide most human protein
environmental limitations, under various levels of inputs and supply either via direct consumption or indirectly via cereal
management conditions. livestock feed (e.g. Fischer et al. 2002b; FAO 2003).
Specifically, AEZ employs the FAO/UNESCO digital soil
map of the world (DSMW) as the underlying reference for its (b) World agricultural trade and economic modelling
own land surface database, consisting of more than 2.2 In addition to land resource assessment and computation of
million grid cells at 5 0 !5 0 latitude/longitude, i.e. with a size of potentially attainable yield, the analysis includes an agro-
about 10!10 km at the equator. In addition, a global digital economic model for the estimation of actual regional
elevation map (DEM) and derived slope distribution production and consumption. The BLS developed at IIASA
database is linked to DSMW. AEZs current climate database comprises a series of national and regional agricultural
is based on the climate research unit (CRU) of the University economic models. It provides a framework for analysing the
of East Anglia, consisting of historical monthly mean data world food system, viewing national agricultural components

as embedded in national economies, which in turn interact drivers of food systems might evolve over the same period. To
with each other at the international trade level (e.g. Fischer this end, we used plausible socio-economic development
et al. 2002a,b). The BLS model consists of 34 national and paths, as specified in the IPCC special report on emissions
regional geographical components covering the globe, scenarios (SRES) (IPCC 2000, 2001a,b). The SRES
calibrated and validated over past time windows (e.g. Fischer scenarios have been constructed to explore socio-economic
et al. 1988, 1994) and successfully reproduces regional development and related pressures on the global environment
consumption, production and trade of major agricultural in this century, with special reference to emissions of
commodities in 2000. Several applications of the BLS to greenhouse gases into the atmosphere (see notes in table 1).
climate-change impact analysis have been published, e.g. in Within this context, climate change is clearly seen as the
Rosenzweig & Parry (1994), Fischer et al. (1996) and Parry consequence of complex social, economic and environmental
et al. (1999, 2004). interactions, possibly modulated by the capacity to mitigate
The individual national/regional models are linked and adapt regionally and globally. Emissions of greenhouse
together by means of a world market, where international gases connected to specific SRES scenarios are translated into
clearing prices are computed to equalize global demand with projections of climate change over this century by using
supply. The BLS is formulated as a recursively dynamic GCMs. We used projections available from IPCC data
system, working in successive annual steps. Each individual distribution centre (DDC) and the CRU, corresponding
model component focuses primarily on the agricultural to five emission scenarios: A1FI, A1B, A2, B1 and B2 (see
sector, but attempts to represent the whole economy as table 1).
necessary to capture essential dynamics among capital, The following methodology was developed for application
labour and land. For the purpose of subsequent international of GCM and SRES scenarios to the AEZ–BLS framework.
linkage, production, consumption and trade of goods and For use in AEZ, projected GCM changes were interpolated to
services are aggregated into nine main agricultural sectors, a grid of 0.58!0.58 latitude/longitude and applied to the
though individual regional models have more detail. The nine baseline climate period of 1961–1990 (see also following
agricultural sectors include: wheat, rice, coarse grains, bovine section). Application of SRES scenarios to BLS were realized
and ovine meat, dairy products, other meat and fish, protein via the following steps. First, UN-based SRES population
feeds, other food, non-food agriculture. The rest of the growth rates were either incorporated for individual countries
economy is coarsely aggregated into one simplified non- or aggregated to BLS regions. In order to maintain
agricultural sector. Agricultural commodities may be used consistency with SRES structure, the BLS 34 regions were
within BLS for human consumption, feed, intermediate further aggregated to 11, following SRES. We then let BLS
consumption and stock accumulation. The non-agricultural dynamically compute allocation of labour and capital
commodity may contribute as investment, and for processing between agriculture and non-agriculture as a function of
and transporting agricultural goods. All physical and financial specified economic conditions. Second, BLS runs were
accounts are balanced and mutually consistent: the pro- harmonized with SRES specifications. The approach chosen
duction, consumption and financial ones at the national level was to harmonize rates of economic growth generated in the
and the trade and financial flows at the global level. BLS with those projected in the IPCC-SRES scenarios,
Within each regional unit, the supply modules allocate through adjustment of capital investment (saving rates) and
land, labour and capital as a function of the relative of rates of technical progress in non-agricultural sectors. The
profitability of its different economic sectors. In particular, harmonization of production factors and GDP, individually
actual cultivated acreage is computed from agro-climatic land for each decade during the period 1990–2080, was carried
parameters (derived from AEZ) and profitability estimates. out on a region-by-region basis.
Once acreage, labour and capital are assigned to cropping and In order to address the research questions formulated in
livestock activities, actual yields and livestock production is the introduction of this paper, namely the impacts of climate
computed as a function of fertilizer applications, feed rates change on agriculture and the interactions with concomitant
and available technology. socio-economic pressures over this century, we devised a two-
Population growth and technology are key external inputs step simulation strategy involving AEZ–BLS simulations
to BLS. Population numbers and projected incomes are used under different SRES scenarios, with and without climate
to determine demand for food for the period of study. change. The latter set includes the combined impacts of socio-
Technology affects BLS yield estimates, by modifying the economic variables and climate, in particular higher tem-
efficiency of production per given units of input (e.g. Fischer perature, modified rainfall patterns and elevated carbon
et al. 2002a,b). For simulations of historical periods up to the dioxide (CO2) concentrations, with respect to current
present, population data are taken from official UN data at conditions. In either case, modifications to potentially
country-level, while the rate of technical progress can be attainable yields computed in AEZ were input into the BLS,
estimated from past agricultural performance. For simu- and resulting changes were analysed globally and regionally,
lations into the future, scenarios of socio-economic develop- with special attention to food balances in sub-Saharan Africa.
ment and population growth must be chosen in order to
inform BLS computations. Another key external input to
AEZ/BLS is climate and environment, which determine crop (d) Climate change scenario generation
suitability and potential yields used by the economic model as GCMs represent one powerful tool to generate character-
an input in resource allocation. Thus, projected climate istics of future climates under anthropogenic forcing, i.e.
change affects BLS results indirectly yet significantly, via its under present and projected future emissions of greenhouse
impacts on agro-climatic land resources computed by AEZ. gases (e.g. IPCC 2001a). Their use in climate-change impact
assessment studies is widespread (e.g. Reilly et al. 2001;
IPCC 2001b). Importantly, GCMs provide internally
(c) Socio-economic scenario generation coherent climate dynamics, by solving globally all climate-
In order to assess agricultural development over this century, relevant physical equations. Yet, it is well known that GCM
with or without climate change, it is necessary to first make projections present significant uncertainties, due in part to
some coherent assumptions about how key socio-economic issues of scale resolution, leading to incomplete model

Table 1. Summary of SRES development scenarios and GCM outputs used in this study.
(A total of 14 combinations of socio-economic and climate scenarios were used. Notes on climate models: HadCM3 is a coupled
atmosphere–ocean GCM developed at the UK Hadley Centre for Climate Prediction and Research and is described by Gordon
et al. (2000) and Pope et al. (2000). The coupled global model ECHAM4/OPYC3 was developed in co-operation between the
Max-Planck-Institut Für Meteorologie (MPI) and Deutsches Klimarechenzentrum (DKRZ) in Hamburg, Germany. (Roeckner
et al. 1992; Oberhuber 1993; Roeckner et al. 1996). The Commonwealth Scientific and Industrial Research Organization
(CSIRO) coupled model involves global atmospheric, oceanic, sea-ice, and biospheric sub-models (Gordon & O’Farrell 1997;
Hirst et al. 1997). The Canadian Center for Climate Modelling and Analysis obtained results for SRES emission scenarios with
the second version of the Canadian Global Coupled Model (CGCM2) (Flato et al. 2000). The Parallel Climate Model (NCAR-
PCM) operated at the National Center for Atmospheric Research (NCAR) has been sponsored by the US Department of
Energy (DOE) and is a joint effort involving several research laboratories in the United States. Details of the PCM control run
were described in Washington et al. (2000). Notes on SRES scenarios. Scenario A1 represents a future world of very rapid
economic growth, low population growth, and rapid introduction of new and more efficient technologies. Major underlying
themes are economic and cultural convergence and capacity building, with a substantial reduction in regional differences in per
capita income. The A1 scenario family develops into three groups that describe alternative directions of technological change in
the energy system: fossil-intensive (A1FI), non-fossil energy sources (A1T), or a balance across all sources (A1B). Scenario A2
portrays a very heterogeneous world. The underlying theme is that of strengthening regional cultural identities; with high
population growth rates, and less concern for rapid economic development. Scenario B1 represents a convergent world with
rapid change in economic structures, ‘dematerialization’, and introduction of clean technologies. The emphasis is on global
solutions to environmental and social sustainability, including concerted efforts for rapid technology development,
dematerialization of the economy, and improving equity. Scenario B2 depicts a world in which the emphasis is on local
solutions to economic, social and environmental sustainability. It is again a heterogeneous world with less rapid, and more
diverse technological change.)
scenario/model HADCM3 ECHAM CSIRO CGCM2 NCAR-PCM
A1FI !
A1B !
A2 ! ! ! ! !
B2 ! ! ! ! !
B1 ! !
representation of regional climate systems; and in part to following standard methodological solution. For three
imperfect understanding of key climate dynamics, such as 30-year periods up to 2100 (the 2020s: years 2010–2040,
water vapour–cloud feedbacks (e.g. IPCC 2001a). For the 2050s: years 2040–2070 and the 2080s: years
instance, the earth climate sensitivity, defined as mean global 2070–2100), climate change parameters were computed at
planetary temperature response to a doubling of CO2 levels each grid point by comparing GCM monthly mean
(ca 560 ppm) in the atmosphere, is thought to be in the prediction for that decade, to those corresponding to the
1.5–4.5 8C range (IPCC 2001a). Though GCMs simulations GCM ‘baseline’ climate 1960–1990. Such changes (i.e. delta
fall squarely within this range, future climate projections with differences for temperature; ratios for precipitation, etc.) were
GCMs corresponding to lower and upper values may be then applied to the observed climate of 1960–1990, used in
different in terms of projected global warming. More AEZ, to generate future climate data. AEZ was then run for
importantly, even among GCMs with similar temperature each future period, and its results compared to its climatic
change simulations, predictions of regional precipitation baseline.
responses may vary significantly, due in part to the intrinsic
chaotic nature of climate, and in part to differences in model (e) Limitations of modelling framework
approach to resolving local to regional atmospheric Simulation models provide a valid, and often the only
dynamics. available, tool used in investigations of complex interactions
As a final methodological issue, we developed climate and feedbacks of many variables. As discussed, simulation
change scenarios based on five GCMs: HadCM3, ECHAM4, impact assessment studies are widely used for projecting
CSIRO, CGCM2 and NCAR, for application to AEZ climate change and socio-economic effects on human
(table 1). Climate sensitivity of the first four sets of models activities. It is thus important to discuss limits and
is in the upper range previously discussed, while the NCAR uncertainties linked to such exercises, so that results can be
model has lower climate sensitivity (R. Dickinson 2003, better interpreted and used by the wider public.
personal communication). This climate sensitivity factor and First, as discussed in §2d, there is uncertainty in the
related temperature–CO2 correlation were shown visually by magnitude of climate change and its spatial and temporal
pooling mean annual temperature predictions of all models distribution. For these reasons, GCMs results must be
against CO2 concentrations (a proxy for time, as CO2 may considered as representative of physically plausible future
increase at a rate of 0.5–1% per annum over this century) over climates, rather than exact predictions. As a consequence,
land. At the same time, there was some correlation between when developing assessment studies for climate change for
climate projections of mean annual temperature and agriculture, and given the prominent role played by rainfed
precipitation over land at mid-to-high latitudes (figure 2a), production worldwide, it is very important not to depend
where most developed countries are located, but none at on one GCM alone, but to use several climate predictions
tropical latitudes (figure 2b), where most developing (Tubiello & Ewert 2002). Particularly in agriculture, the
countries are located. direction of predicted precipitation can largely shape
In order to run simulations with the AEZ model under regional results. At the same time, the GCM simulations
climate change, for each GCM considered we applied the used herein do not contain information on potential

(a) 12
6 HAD3
GCM-projected mean annual

CSIRO 10
precipitation change (%)

temperature change (˚C)
5
CGCM2
8 HAD3
4 NCAR
CSIRO
ECHAM
3 6 CGCM2
NCAR
NCAR
CSIRO
2 4 ECHAM
HAD3
1 CGCM2 2
ECHAM
0 0
300 400 500 600 700 800 900 0 1 2 3 4 5 6
(b) 6 6

HAD3
precipitation change (%)

5 CSIRO 4
CGCM2
4 2 HAD3
NCAR
CSIRO
ECHAM
3 0 CGCM2
NCAR
NCAR
2 CSIRO
–2 ECHAM
HAD3
1 CGCM2 –4
ECHAM
0 –6
300 400 500 600 700 800 900 0 1 2 3 4 5 6
CO2 concentration ( ppm)
Figure 2. GCM-predicted mean annual temperature change over land, as a function of corresponding CO2 concentrations
(a proxy for time from 1990 to 2080) and against GCM-predicted mean annual precipitation change. Graphs show results for all
14 scenarios considered in this study, computed for all land in (a) developed world and (b) developing world.
changes in the frequency of extreme events, so that our 3. ANALYSIS OF RESULTS

projected results might be greatly modified, probably in (a) Impacts of climate change on agricultural
negative terms, had climate variability changes been resources
included. We first employed climate change projections from five
Second, the AEZ modelling simulations have been GCMs to investigate how higher temperatures, modi-
validated in many places, but the global nature of simulations fied precipitation patterns and elevated CO2 level
signifies that validation has of course not been possible at predicted for the next decades, up to 2100, might
every grid point considered. In addition, the gap between combine to affect key agronomic parameters such as
current actual yields and potentially attainable yields in many land suitability of cereal-production and attainable
developing countries is much larger than the impacts of yields, in combination with management responses.
climate changes on potential yields, which increases uncer- Specifically, AEZ automatically simulated new crop
tainty as to what actual effects of climate change on yields will calendars and altered planting and harvesting oper-
be. Likewise, AEZ-simulated crop response to elevated CO2, ations based on the new climate scenarios. In addition,
derived from experimental data, may be larger than actually AEZ allowed for automatic shifting of crops and
possible at farm levels.
cultivars, based on changes in temperature, precipi-
Third, our simulation results depend strongly on BLS
tation and soil moisture conditions. The latter is
dynamics. Although this model has been validated for past
computed as the difference between precipitation and
periods, various additional and uncertain assumptions are
evapo-transpiration, including infiltration and runoff as
needed to obtain food system projections (e.g. technical
a function of soil type.
progress in crop yields; regional irrigation development;
changes in food preferences, etc.). Nonetheless, as discussed
for GCM models, BLS provides an internally coherent (i) Land resources
picture of socio-economic dynamics, so that its predictions The agro-climatic assessment with climate projections
represent plausible futures for scenario analysis. of five GCMs shows a northward shift of thermal
Finally, the SRES scenarios considered herein have regimes, reducing significantly boreal and arctic
recently been criticized for their regional economic growth ecosystems (60% reduction of current total 2.1 billion
patterns, regarded as too strong when compared to historical hectares). In contrast, tropical zones expand, which
data (e.g. Nakicenovic et al. 2003). It may be noted that we means, for example, that apart from a very small stretch
have also simulated lower economic growth scenarios. In in South Africa and a narrow fringe along the
these, the ability to adjust to climate change computed by Mediterranean coast, the tropics will cover almost all
BLS has been smaller, leading to more negative impacts than of Africa.
those presented herein. In addition, SRES A2 is regarded as Four of five major GCMs project consistent
having population growth rates beyond the current UN high expansion of arid areas in developing countries.
projections. The results obtained under this socio-economic Currently, almost 1 billion people live worldwide in
path may thus be regarded as providing a worst case scenario for arid lands; more than 180 million people in Africa
our food system analysis. alone. In Africa, under current climate, AEZ estimates

Table 2. Transition matrix of changes in agro-ecological constraints to agriculture in sub-Saharan Africa, predicted by AEZ for
the 2080s under scenario HadCM3-A1FI.
(Current land conditions are summarized in the first column. The values in each row indicate, for each different land classes, the
projected change to different conditions.)
HadCM3-A1FI, 2080s
reference climate 1000 km2 no constraint slight moderate severe
no constraint 535 457 66 6 6

slight 2704 11 2395 262 36
moderate 6061 3 67 5379 612
severe 15 128 0 0 80 15 048
471 2528 5727 15 702
developed, all land developed, current cultivated land

150 130
HAD3 HAD3
index (1961–90 = 100)
140
CSIRO 120 CSIRO
CGCM2 CGCM2
130
NCAR NCAR
110
HAD3 HAD3
120
CGCM2 NCAR
NCAR 100 CGCM2
110
CSIRO CSIRO
100 90
developing, all land developing, current cultivated land
115 115
HAD3 HAD3
index (1961– 90 = 100)
110 CSIRO 110 CSIRO

CGCM2 CGCM2
NCAR NCAR
105 105
HAD3 HAD3
CGCM2 NCAR
100 NCAR 100 CGCM2
CSIRO CSIRO
95 95
300 400 500 600 700 800 900 300 400 500 600 700 800 900
CO2 ppm CO2 ppm
Figure 3. Changes in potentially attainable cereal-production predicted by AEZ under different GCM climate change scenarios,
versus CO2 concentration. Projections are for either current cultivated land (right), or all available under future climates (left),
and pooled into developed (top) and developing (bottom) countries. Results are expressed against an index of climate change
(Z100 in 1990), a proxy for time from 1990 to 2080.
1.1 billion hectares of arid and dry semi-arid land, i.e. scenario, indicate that with rapid climate change
with less than 120 days length of growing period, these constraints may change respectively to 5.2,
defined as the number of days in a year with 29.0, 1.1, 5.7 and 24.5%. The agro-ecological
temperature and soil moisture conditions favourable changes due to climate change will result in water
to crop cultivation (Fischer et al. 2002a). Under the deficits in some areas and surplus in others as well as
climate change scenarios considered, and by 2080s, increased or reduced infestation of disease pathogens
AEZ estimates of arid and dry semi-arid areas in Africa and parasites.
increase by about 5–8%, or 60–90 million hectares. Under climate change and by the 2080s, regional
analyses of AEZ results indicate expansions of land area
Environmental constraints with severe constraints as follows: Central America and
Under current climate, AEZ computes that two-thirds Caribbean (1–3% increase; AEZ simulations for
of the global land surface—some 8.9 billion hectares— current climate: 270 million hectares); Oceania and
suffer severe constraints for crop cultivation: 13.2% is Polynesia (0.5–4.5%), northern Africa (2–3.5%; AEZ
too cold, 26.5% is too dry, 4.6% is too steep, 2.0% is simulations for current climate: 550 million hectares)
too wet and 19.8% has poor soils. Climate change will and Western Asia (up to 1%; AEZ simulations for
have positive and negative impacts, as some constraints current climate: 435 million hectares). In southern
will be alleviated while others may increase. The Africa, AEZ projects up to an additional 11% of a total
results for the Hadley HadCM3 climate model and the land area of 265 million hectares to be at risk of being
IPCC A1F1 scenario, representing a high-emission severely constrained for crop agriculture.

The case of sub-Saharan Africa (ii) Rainfed cereal-production

Simulations with AEZ suggest the following overall For developing countries, under NCAR climate
picture for sub-Saharan Africa: decrease of constraint- projections, AEZ computes substantial gains in poten-
free prime land with highest suitability for crop tial cereal-production in both developed and develop-
cultivation; increase in land with moisture stress; and ing countries (figure 3). Under climate projections with
expansion of land with severe climate, soil or terrain the other GCMs, AEZ simulations produce mixed
constraints, by 30–60 million hectares, in addition to results for developing countries. For instance, on
the 1.5 billion hectares already unfit for rainfed average among GCMs and specifically for wheat
agriculture under current climate. production in the 2080s, AEZ projects substantial
Under the rather dramatic climate changes of decreases in attainable wheat production, in the order
scenario SRES A2, AEZ computes a decrease of good of 15–45% (figure 4). Specifically, AEZ computed
land (the sum of very suitable and suitable land) under decreases in south Asia (20–75%); southeast Asia
all five GCM climate projections, by an average of (10–95%); and in South America (12–27%); the
K6.3%, ranging from K8.2 million hectares (NCAR) same simulation results suggest that land suitable for
to K27.3 million hectares (CSIRO). Land with severe wheat production might virtually disappear in Africa.
climate, soil or terrain constraints, increases in 12 of the Under all climate change scenarios, AEZ results
14 climate projection considered, by 26–61 million indicated that declines in cereal-productivity potentials
hectares. Under the NCAR model, the extent of sub- of more than 5% will be realized by 2080 in a group of
Saharan land with severe environmental constraints to more than 40 countries worldwide, with mean losses of
about K15%. Cumulative population projected for
crop agriculture declines by about 15 million hectares
these countries ranges between 1.6 and 3.8 billion
for both the simulated A2 and B2 scenarios, due to
people by 2080 (depending on simulated impacts for
significant increases in precipitation and milder tem-
India and thus whether India is included in this group).
perature increases compared to the other GCM results
Fourteen countries account for 80% of current cereal-
considered in this study.
production of the developing world. China, the world’s
A snapshot of AEZ-derived environmental con- largest cereal producer, is projected by AEZ to increase
straints in sub-Saharan Africa is presented in table 2, its cereal-production potential by 5–23% across climate
showing a transition matrix from current baseline projections, except for the CSIRO climate, under
climate to the 2080s, as projected by HadCM3-A1FI. which AEZ predicts a loss of 5–7%. For India, the
Current conditions are summarized in the transition second-largest cereal producer, Brazil and Thailand,
matrix by the row totals shown in the first numeric results vary across climate projection. Argentina gains
column. Column totals, in the last row of the matrix, production potential by 7–24% for the HadCM3 and
denote class extents calculated for future climate. The NCAR models, while production potential decreases
values in each row of the transition matrix indicate for by 10–30% under both CSIRO and CGCM2 projec-
different land classes the fate of extents classified under tions. AEZ computes cereal-production potential
current conditions. For instance, AEZ results indicate decreases in South Africa under all climate projections,
that only 8 million hectares of land currently under except under NCAR-A2, where a small gain occurs.
severe environmental constraints are expected to Results obtained in this study thus indicate, in
improve under climate change, whereas more than 60 agreement with previous studies, that a wide range of
million hectares of land currently under moderate outcomes is to be expected for many countries, due to
constraints would further degrade to severe environ- heterogeneity of their agro-climatic resources as well as
mental constraints. of projected climate change. In addition, grid-based
simulations in AEZ make it possible to aggregate
results spatially to any unit of interest. In particular, we
Cultivated land
derived within-country impact statistics. Results
Under the climate change scenarios considered, the
pooled in figure 5 show for select countries, under
AEZ model predicts gains in potential agricultural land HadCM3 projections an for the 2080s, the distribution
globally. In particular, large gains are predicted for of impacts with respect to changes in cereal suitability
North America and the Russian Federation, due to under to climate change, considering currently culti-
longer planting windows and generally more favourable vated land, as well as all land potentially available under
growing conditions under warming. Significant losses future climates (where available, as in Russia). Such
are by contrast predicted in northern and southern distributions clearly indicate the inherent heterogeneity
Africa, due to a worsening of growing conditions from of impacts, even within a country. National-level
increased heat and water stresses. averaged results may thus correspond to narrow normal
Specifically, for the developed nations AEZ predicts distributions around an impact-neutral mean, such as
under all climate projections an expansion of potential the case of AEZ projections for India; present
land suitable for crop cultivation, mostly at higher asymmetries towards negative impacts, such as com-
latitudes. Increases are mainly located in North puted for South Africa; or present positive means, but
America (40% increase over the 360 million hectares with substantial winners and losers (Russia and China).
under current baseline climate); northern Europe In the case of the latter countries, strong positive ‘tails’
(16% increase over current 45 million hectares); were also projected, owing to either additional land
Russian Federation (64% increase over 245 million brought into cultivation in future, warmer climates
hectares) and in East Asia (10% increase over 150 (Russia) or to strong shifts within currently used land,
million hectares). from previous marginal to more suitable conditions

(China), often brought about by more favourable simulations without climate change, representing a
precipitation regimes under the projected climates. reference case against which simulations including
climate change are analysed. In order to then run
Current food-insecure countries BLS with climate change, AEZ-derived projections of
The total population of the over 80 poor food-insecure changes in land production potentials and attainable
countries currently amounts to some 4.2 billion, more yields for future decades are used to modify in a simple
than 70% of current world population (FAO 2001); multiplier fashion the BLS yield production function.
about 20% of this population is considered under- In either set of simulations, BLS simulations start in
nourished. By the 2080s, the total population of these 1990 and were carried out in 1-year increments from
countries is projected to increase to 6.8 billion, or about 1990–2080. Analyses of projected changes are made
80% of the world population at that time. Individual relative to the 1990s.
country outcomes in this group are reason for concern.
For example, the AEZ results based on a suite of Global cereal-production
HadCM3 projections (scenario A1FI to B1) indicate For the baseline decade 1990, BLS computes world
that 20–40 poor and food-insecure countries, totalling cereal-production (rice is included as milled equival-
1–3 billion people, may lose on average 10–20% of their ent; conversion factor is 0.67 from rice paddy) at 1.8
cereal-production potential under climate change. billion metric tons (G ton), about equally divided
Results for developing countries in terms of cereal- between developed and developing countries, in good
production potential for gaining and losing countries agreement with current statistics. Effects of socio-
are highlighted in figure 6, showing broad groups of economic scenarios are substantial, and results vary in a
countries where production capacity increases or range with lower and upper values corresponding to
decreases by at least 5%, compared to current climate. SRES B1 and A2, respectively. By 2080, BLS projects
With the exception of the results for the NCAR global cereal-production in the range 3.7–4.8 G ton,
model, simulations under all remaining climate scen- depending on SRES scenario. Production in the
arios indicated that Sudan, Nigeria, Senegal, Mali, developed countries ranges 1.4–1.6 G ton; thus BLS
Burkina Faso, Somalia, Ethiopia, Zimbabwe, Chad, computes for the developing countries up to threefold
Sierra Leone, Angola, Mozambique and Niger, would increases in production from the 1990 baseline levels,
lose cereal-production potential by the 2080s, across all with fivefold and higher increases projected for Africa
the emission scenarios. These countries currently have in all the scenarios, as a consequence of the substantial
87 million undernourished, equivalent to 45% of the economic development assumed in SRES.
total undernourished in sub-Saharan Africa. In con-
trast, Zaire, Tanzania, Kenya, Uganda, Madagascar, Cultivated land
Côte d’Ivoire, Benin, Togo, Ghana and Guinea were all Total cultivated land in developed countries in
projected to gain cereal-production potential by the 1994–1996 amounted to about 600 million hectares.
2080s. These eight gaining countries currently have 73 During the same period, land in developing countries
million undernourished, equivalent to 38% of the amounted to 870 million hectares, 30% of which
undernourished population in sub-Saharan Africa. located in southeast Asia.
Figure 7 indicates the spatial heterogeneity of impacts Under the four socio-economic scenarios con-
on cereal-production capacity, allowing for crop- sidered, BLS results show that land under cultivation
switching and changes in crop calendars, in the 2080s in developed countries likely remains close to the 1990
relative to current climate, for climate projections by levels, while additional production comes from
different GCMs, under SRES A2. Two important increased productivity. In developing countries, by
factors arise from further aggregating such results. contrast, cultivated land is projected by BLS to increase
First, the net balance of changes in cereal-production by 2080. Projected increases are about 10% globally in
potential for sub-Saharan Africa was projected to be SRES A1 and B1; higher increases of cultivated land
negative, with net losses of up to 12%. Second, there are projected under SRES A2 and B2, about 30 and
will be large variations in outcomes, with up to 40% of 20%, respectively. BLS results locate most of the
sub-Saharan countries losing significant shares of their additional cropland either in Africa or Latin America.
agricultural resources (tables 3 and 4). In southeast Asia, due to lack of suitable land and land-
use competition from other sectors, only some 30–40
(b) BLS projections of the world food system: million hectares may additionally be brought under
1990–2080 cultivation across socio-economic scenarios.
While climate and farm management are key determi- BLS calculations also show interesting time
nants of food production locally, and can be computed dynamics associated with changes in cultivated land,
via AEZ, agro-economics and world trade combine to with land amounts and trends, generally, following
shape regional productivity significantly. The following population. For instance, the lowest and highest
sections describe results obtained with BLS simu- amounts of global cultivated land in 2080 are
lations of world food systems. computed under SRES B1 (1.6 million hectares) and
SRES A2 (1.8 million hectares) respectively, in
(i) Baseline simulations under SRES development paths, no correspondence to the lowest and highest population
climate change projections, respectively. In addition, in SRES A1 and
We next assess the impacts of four SRES socio- B1, land under cultivation computed by BLS for the
economic scenarios on world agricultural economy 2050s is higher than for the 2080s. This is in line with
using the BLS framework. First, we present results for peaks in projected population numbers, which in A1

developed, all land developed, current cultivated land

150 130
HAD3 HAD3
index (1961–90 = 100)
140 120 CSIRO

CSIRO
CGCM2 CGCM2
130 NCAR
NCAR 110
HAD3 HAD3
120 NCAR
CGCM2
NCAR
100 CGCM2
110
CSIRO CSIRO
100 90
developing, all land developing, current cultivated land

110 110
HAD3 HAD3
index (1961–90 = 100)
100 100
CSIRO CSIRO
90 CGCM2 90 CGCM2
NCAR NCAR
80 80
HAD3 HAD3
70 CGCM2 70 NCAR
NCAR CGCM2
60 CSIRO 60 CSIRO
50 50
300 400 500 600 700 800 900 300 400 500 600 700 800 900
CO2 ppm CO2 ppm
Figure 4. Changes in potentially attainable wheat production predicted by AEZ under different GCM climate change scenarios,
versus CO2 concentration. Projections are for either current cultivated land (right), or all available under future climates (left),
and pooled into developed (top) and developing (bottom) countries. Results are expressed against an index of climate change
(Z100 in 1990), a proxy for time from 1990 to 2080.
southern Africa,
Russia, all suitable land current cultivated land
% of land in impact class
18 16
16 14
14 12
12 10
10
8 8
6 6
4 4
2 2
0 0
China, current cultivated land India, current cultivated land

14 35
12 30
10 25
8 20
6 15
4 10
2 5
0 0
– 100 – 70 – 40 – 10 20 50 80 –100 –70 –40 –10 20 50 80
SI changes by impact class SI changes by impact class
Figure 5. AEZ-simulated distribution of climate impacts on cereal productivity in the 2080s, under HadCM3-A1FI climate
projections. The diagrams show the distribution of land with respect to cereal suitability changes under climate change. Bars
shown to the right of the red line indicate land pixels where suitability increased; bars to the left denote negative changes.
SIZsuitability index for potential cereal-production computed by AEZ.
and B1 occur by mid-century, after which total scenarios, in particular depending on projected income
population, and food demand, start declining. levels and their regional distribution, as well as on
population projections.
Food security The BLS estimate is based on FAO data (FAO
The BLS estimates of the number of people at risk of 2001) and relies on a strong empirical correlation
hunger vary greatly with socio-economic SRES between the shares of undernourished in the total

Hadley, 2080 CSIR0, 2080

% change of regional total 8 6
6 4
59 50 42 49
4 2 43 48
56 36
2 B1 0 B1
B2 B2
0 –2 A2
A2
27 37 30 30
–2 A1F –4 A1B
29 35
–4 40 52 –6
–6 –8
gainers losers gainers losers
Figure 6. Impacts of climate change on rainfed cereal-production potential of developing countries, for currently cultivated land
in the 2080s, according to HadCM3 and CSIRO climate projections.
CGCM2 CSIRO
undefined
> 50 %
25 to 50 %
5 to 25 %
–5 to 5 %
–25 to –5 %
–50 to –25 %
< –50 %
not suitable
water
HADCM3 ECHAM4
undefined
> 50 %
25 to 50 %
5 to 25 %
–5 to 5 %
–25 to –5 %
–50 to –25 %
< –50 %
not suitable
water
Figure 7. Geographic distribution of percentage changes of AEZ-simulated potentially attainable cereal output per grid cell in
the 2080s, SRES A2, and different GCM climate projections, relative to current climate.
population and the ratio of average national food aggregate food supply exceeds aggregate national
supply, including imports, relative to aggregate food requirements by 30%. Hunger is nearly elimi-
national food requirements, as plotted in figure 8. For nated for index values of food supply over requirements
instance, the diagram suggests that the share of higher than 160.
undernourished in the total population will fall below The BLS simulation results indicate that in spite of
20% for an index value of about 130, i.e. when relatively high levels of economic growth in SRES

Table 3. AEZ-estimated impacts on potentially attainable rainfed cereal-production in the 2080s HadCM3 climate projections,
for currently cultivated land of sub-Saharan African countries.
(G, countries gaining 5% or more; N, small change of K5 to C5%; L, countries losing 5% or more.)
number of countries projected population (billions) change in cereal-production potential (%)
scenario G N L G N L G N L total
A1FI 17 8 16 0.62 0.43 0.55 5.1 K0.6 K8.6 K4.1

A2 14 11 16 0.58 0.23 0.78 5.3 0.7 K7.9 K1.9
B2 13 15 13 0.56 0.57 0.46 4.4 0.1 K7.1 K2.6
B1 17 10 14 0.65 0.22 0.72 5.0 0.1 K6.4 K1.3
Table 4. AEZ-estimated impacts on potentially attainable rainfed cereal-production in the 2080s, SRES A2, for currently
cultivated land of sub-Saharan African countries.
(G, countries gaining 5% or more; N, small change of K5 to C5%; L, countries losing 5% or more.)
number of countries projected population (billions) change in cereal-production-potential (%)

climate
model G N L G N L G N L total
HadCM3 14 11 16 0.58 0.23 0.78 5.3 0.7 K7.9 K1.9

CSIRO 16 8 17 0.63 0.24 0.73 3.3 0.1 K15.1 K11.7
CGCM2 13 14 14 0.79 0.32 0.49 6.0 0.0 K8.7 K2.7
NCAR 25 13 3 1.07 0.39 0.13 8.2 0.4 K0.8 7.8
scenarios, there is only modest progress in reducing

80
hunger to 2020. This implies, regardless of climate
change, that specific targeted programmes for hunger
reduction will be essential to meet the millennium goals 60
% under nourished
(UN Millennium Project 2005) of reducing hunger by

half in 2015 (figure 9).
In the period 2020–2050, BLS results suggest that 40
hunger is substantially reduced in scenarios A1, B1;
somewhat less in B2. In the case of scenario A2, with 20
(perhaps unrealistically) high-assumed population
growth and thus slower per capita income growth
compared to the other SRES scenarios, hunger 0
persists. 80 100 120 140 160 180 200
index of food supply/food requirement
(ii) Simulations under SRES development paths and Figure 8. BLS-based estimation of the percent share of
undernourished people in the total population of any given
climate change
country, based on food balances, based on domestic
The following section describes results of coupling BLS production as well as on international trade.
simulations of global and regional production with
AEZ-derived climate-change impacts on land and crop In general, in these simulations, modifications to
potentials. To this end, percentage changes indicated in food production caused by climate change create
the text refer to comparisons of BLS results without market imbalances with respect to the baseline case,
and with climate change. modifying international prices leading to reallocation of
Some 60 simulations were performed with the AEZ– capital and human labour. Thus, BLS computes
BLS framework (Fischer et al. 2002b). The simulations certain adaptation strategies automatically, by search-
were related to four main research questions: magni- ing for solutions that optimize new opportunities under
tude of climate change for different future socio- climate-change via model-calculated economic
economic paths, uncertainty of results in view of adjustments.
differences in climate projections of different GCM
groups, robustness of results with regard to altered
economic growth assumptions and sensitivity of results Impact of climate change on prices and agricultural GDP
to different assumptions with regard to physiological Generally, BLS simulated crop price changes under
effects of atmospheric CO2 enrichment on yields. As climate change are moderate, due to relatively small net
for the ‘no climate change’ case described in §3b(i), global impacts on crop-production potentials. For the
BLS simulations with climate change started in the range of scenarios, in the case of HadCM3 climate
base year 1990 and proceeded up to 2080. Three projections, cereal prices increase 2–20% (scenario B1
separate snapshots are provided for concise discussion to A1FI); for CSIRO the increase is 4–10%, relative to
of results: 2020s, 2050s and 2080s. baselines BLS projections. For the remaining GCM

projections aggregate price changes due to climate- consumption of cereals in developing countries. For
change impacts were even less. HadCM3, human cereal consumption in developing
BLS simulation results suggest that impacts of countries declines by 2–4%, i.e. in the order of 40–80
climate change on agricultural GDP are also small at million metric tons, compared to simulated total direct
global level, i.e. between K1.5% (in HadCM3-A1FI consumption between 1.6 G ton (scenario B1) and
scenario) and C2.6% (in NCAR-A2 scenario); this 2.1 G ton (scenario A2) in the BLS reference projec-
comparison refers to total global GDP of agriculture in tions. Consumption changes in Asian developing
BLS baseline projections, ranging from US$ 2.9 to 3.6 countries accounted for two-thirds of this amount.
trillion (1990 US $). Consumers in Latin America were least affected in
At the same time, BLS results indicated that model simulations.
agriculture in developed countries as a group would Simulations of the SRES reference scenarios
benefit under climate change. Among developed (without climate change) revealed a growing depen-
regions, simulations indicate that North America dence of developing countries on net cereal imports,
gains in all GCM scenarios (in particular, 3–13% totalling in 2080 between 170 million ton (B1) and 430
under SRES A2, for different GCM projections); million ton (A2). According to the AEZ–BLS runs,
agricultural GDP mostly increases in the Former climate change will add to this dependence, increasing
Soviet Union (up to 23% in scenario A2); while only net cereal imports of developing regions by 10–40%
Western Europe loses agricultural GDP, across all across GCM climate projections. The largest increases
GCM scenarios (K6 to K18% under SRES A2). occurred under CSIRO climate projections. Even
By contrast, BLS results indicated decreases in under NCAR projections, resulting in overall positive
agricultural GDP in most developing regions, with the impacts on agricultural productivity, the comparative
exception of Latin America. For Asia, by 2080, advantage for producing cereals is predicted to shift
agricultural GDP losses amount to about K4%, towards developed countries, and net imports of
under SRES A1 and A2, and HadCM3 and CSIRO developing countries increase by about 25%, i.e.
climate (BLS baseline projections: US$ 1.1–1.3 around 90 million tons of additional cereal imports in
trillion). Aggregate projections for Africa are also scenario B1, and an additional 110 million tons in
negative, K2 to K8% for HadCM3 and CGCM2 scenario A2.
climate, and K7 to K9% for CSIRO (BLS baseline
projections: US$ 0.6–0.7 trillion). Climate change and food security: number of people
at risk of hunger
Impact of climate change on cereal-production Some fairly robust conclusions emerge from the
AEZ–BLS model results present a fairly consistent analysis of climate-change impacts on the number of
response pattern of regional cereal-production to people at risk of hunger. First, climate change will most
climate change. At the global level, taking into account likely increase the number of people at risk of hunger.
economic adjustments of actors and markets, simu- Second, the importance and significance of the climate-
lated cereal-production changes fall within 2% of the change impact on the level of undernourishment
results for the reference simulations without climate depends entirely on the level of economic development
change. However, aggregation produces deceivingly assumed in the SRES scenarios.
small numbers. The developing countries consistently For the wealthy societies of SRES scenario A1,
experience reductions in cereal-production in all where even currently poor regions are assumed to reach
climate scenarios. Decreases of 5–6% are most economic levels exceeding in per capita terms current
pronounced in simulations based on CSIRO climate OECD incomes, hunger is a marginal issue and
projections. In this case, production moves to remains so even with climate change. In contrast, the
developed regions, notably North America and the outcome of simulations is quite different for the high-
Former Soviet Union, where increases of 6–9% are population SRES scenario A2. Under this set of
observed. The most significant negative changes occur demographic and economic assumptions, the level of
in Asian developing countries, where production undernourished, even without considering climate
declines in all scenarios, ranging from about K4% for change, remains at a high level throughout the entire
CGCM2 and NCAR to K6 to K10% for HadCM3 simulation period to 2080. In the reference projection
and CSIRO. (without climate change), the number of undernour-
ished was estimated at 768 million for 2080. With
Impact of climate change on cereal consumption and trade considering climate change, this number increases by
In the SRES worlds of the 2080s, consumers are nearly 120 million, equivalent to some 15%, under
assumed to be much richer than they are today; they are both HadCM3 and CSIRO climate projections.
largely separated from agricultural production pro- Climate change projections with the Canadian
cesses, dwelling in cities and earning incomes in the CGCM2 model result in an additional 50 million
non-agricultural sectors. As in today’s developed undernourished.
countries, consumption levels depend largely on food In the simulation experiments, it is possible to
prices and incomes rather than on changes in domestic impose different levels of climate change (i.e. for the
agricultural production. full range of emission scenarios and climate model
BLS simulations for 2080 predicted, in response to outputs) on any socio-economic development path.
climate change across all climate models and emission This allowed us to establish a relationship between level
scenarios—with the exception of the NCAR climate of climate change and estimated number of people at
projections, a fairly uniform decline in direct human risk of hunger with regard to different development

(a) developing countries (b) African countries

900 350
people at risk of hunger (millions)
800 300
700
250
600 A2 A2
500 B2 200 B2
400 B1 150 B1
300 A1 A1
100
200
100 50
0 0
2000 2010 2020 2030 2040 2050 2060 2070 2080 2000 2010 2020 2030 2040 2050 2060 2070 2080
Figure 9. BLS-projected number of people undernourished, under all four SRES scenarios and no climate change, computed for
(a) all developing countries and (b) African developing countries.
path assumptions. The full range of climate-change
additional under-nourished due

175
to climate change (millions)

impacts on crop suitability and productivity for all four
150
IPCC emission scenarios was assessed in AEZ. For all
four levels of climate change, production impacts were 125
imposed on the socio-economic development path of 100
scenario A2. For instance, for the large climate changes 75
projected by the HadCM3-A1FI scenario for the 50
2080s, additional undernourished people amount to
25
175 million. This indicates that under socio-economic
conditions of development path A2, the number of 0
300 400 500 600 700 800 900
undernourished may increase more than linear with the atmospheric CO2 concentration (ppm)
level of climate change. Figure 10 summarizes the
results, showing the additional number of people at risk Figure 10. Additional number of people at risk of hunger in
of hunger in 2080 plotted against different levels of 2080, plotted against different levels of atmospheric CO2
atmospheric CO2 concentrations and associated cli- concentrations and associated climate change, for SRES A2
mate changes. development path and simulation results with HadCM3 and
CSIRO climate projections.
We nonetheless chose to include it alongside the other

4. DISCUSSION AND CONCLUSIONS SRES scenarios, as the scope of this work was to test the
The extensive AEZ–BLS simulation results presented sensitivity of the world food system under ‘what-if’ type
herein provide a full albeit lengthy analysis of the questions, including the full range of scenarios
combined effects of alternative socio-economic considered by IPCC. The reader is advised to interpret
development paths and associated climate change on results from A2 discussed herein as being representa-
agricultural production worldwide. A number of tive of a worst case scenario.
conclusions can be gathered and are offered in this BLS baseline results indicated that differences in
section. We then conclude by making a number of more assumed socio-economic development—in this study
generic considerations on the nature of future agricul- represented by the four IPCC-SRES scenarios—can
tural needs in order to improve long-term sustainability
significantly impact global agriculture. Against the
of food systems under multiple challenges.
current backdrop of about 1.8 G metric tons of
In general, our simulations were carried out under
cereal-production worldwide, BLS computed by 2080
the assumption that assessment studies simply based
a range between 3.7 and 4.8 G metric tons, with
on current agricultural economies, over which climate
signals are superimposed without economic re-adjust- scenarios B1 and A2 representing the lower and
ment, cannot produce sensible results vis à vis the state upper prediction limits. These projections represent a
of global agricultural economics over coming decades, near doubling of current global production, in response
as climate change and socio-economic pressures to the projected rise in population and incomes. The
progressively become interlinked. Within a coherent context behind these figures is that globally, land and
agro-ecological and economic framework, the question crop resources, together with technological progress,
of relative importance among these different drivers can appear to be sufficient to feed a world population of
be properly analysed. Specifically, the baseline BLS about 9 billion people (13 billion in A2) in 2080
runs without climate change provided a way to set a (nevertheless with great uncertainty in some develop-
reference, against which climate-change impacts could ing regions).
be assessed in their absolute and relative importance. It The impact of different SRES scenarios is significant
should be noted here that current opinion considers the at both global and regional levels. Globally, BLS
population component of SRES A2 ‘out of the range’, projects cereal-production of 4.3G0.5 G metric tons.
as its associated population growth rates seem now The variation across scenarios is thus about 10–15%.
too high (e.g. W. Lutz 2004, personal communication). Regionally, BLS results show a more heterogeneous

picture, as discussed. Of particular relevance to projected demographic and socio-economic changes

regional food security is the case of sub-Saharan Africa, that substantially reduced the risk of under-nutrition.
where a growing share of people considered under- In conclusion, the AEZ–BLS results provided for in-
nourished is located. BLS baseline results indicate a depth analysis of the potential interactions between
significant reduction in both the absolute number and socio-economic pressures and climate-change impacts
percentage (compared to world population) of under- over the coming decades. Results suggest that socio-
nourished people, i.e. at risk of hunger, for all SRES economic development over this century will greatly
scenarios by 2080, except for A2, due to significantly alter production, trade, distribution and consumption
larger populations and slower per capita income growth of food products worldwide, as a consequence of
in that scenario. Under A2, BLS predicts the number of population growth, economic growth, and diet changes
undernourished in 2080 to be 768 million, virtually in developing countries. Climate change will addition-
equivalent to today’s figures. Even taking A2 out of the ally modify agricultural activities, probably increasing
picture, BLS average results, across scenarios, indicate any gaps between developing and developed countries.
150G50 million undernourished people by 2080. Adaptation strategies, both on-farm and via market
Variability across scenarios, of about 30%, is thus mechanisms, will be important contributors to limiting
much larger than computed for global food production. the severity of impacts. Additional climate change
Importantly, all scenarios indicated that reduction of pressures, not investigated in this study, may arise by
world hunger will be small, until 2030 globally, and changes in the frequency of extreme precipitation
perhaps not until 2040 for sub-Saharan Africa. These events such as floods and droughts, which may
results are important in terms of UN millennium diminish the capacity of countries to adapt, especially
development goals, i.e. the current commitment of in poor tropical regions.
developed nations to reduce hunger by half by 2015. The climate change issue is global, long-term and
BLS results suggest that global economic development involves complex interaction between demographic,
can ‘trickle down’ to poor regions to nearly eradicate climatic, environmental, economic, health, political,
world hunger by the 2050s and 2080s, provided high institutional, social and technological processes. It has
levels of economic growth are accompanied by significant international and intergenerational impli-
moderate population increase. Yet, under all scenarios cations in the context of equity and sustainable
considered, BLS results indicate that millennium development. Climate change will impact on social,
development goals cannot be met simply through
economic and environmental systems and shape
market mechanisms, and that additional targeted
prospects for food, water and health security (Steffen
actions would be necessary.
et al. 2004; WHO 2003). The capacity to mitigate and
The impacts of climate change were found to be
to adapt to climate-change impacts is strongly related
critically dependent upon SRES development scenario
to the future development paths. The socio-economic
and on GCM regionally, but not so much globally. At
and, even more so, the technological characteristics of
the global level simulation results indicate only small
different futures strongly affect emissions, hence the
percentage changes from the baseline reference case
extent and pace of the impacts of climate change, as
with respect to cereal-production. Analyses of the
AEZ–BLS framework thus indicate that the two levels well as the capability of societies to adapt to and
of adaptation considered in our simulations, i.e. mitigate climate change. Recent developments in
autonomous adaptation at the field level, such as geographical information systems, data acquisition by
changing of crop calendars and cropping systems as a remote sensing, and the increasing quality and spatial
function of climate; and market adjustments at both coverage of global resource databases, has enabled
regional (re-distribution of capital, labour and land) modelling to identify potentials and environmental
and global (trade) levels, can successfully combine to constraints to crop production at regional and national
reduce otherwise larger negative impacts, as computed levels. The integration of these assessments in the
by AEZ alone. Regionally, climate change effects were global food economy together with projections of future
to further widen the gap between developed and climate change and variability enables evaluation of
developing countries, in addition to results analysed impacts of climate change on agriculture and provides a
under the BLS reference cases. The conclusion is that basis of prioritizing regional and commodity specific
climate change impacts will be more severe in tropical agricultural research for adaptation and mitigation.
semi-arid developing countries. The AEZ–BLS results Climate change and variability may result in
suggest reductions in production in developing irreparable damage to arable land and water resources
countries, in the 5–10% range, accompanied by similar in some regions, and with serious local consequences
increases of production in northern developed regions, for food production. These losses will be felt most
especially North America and Russia. Finally, in terms profoundly in developing countries with low capacity to
of food security, AEZ–BLS results indicate significant cope and adapt. While the international community
climate change effects would occur only if levels of has focused on climate change mitigation, the issue of
undernourishment were already high under ‘no climate adaptation to climate change is an equally pressing
change’ conditions, e.g. as in SRES A2. In particular issue and must be put on the international negotiation
for SRES A2, projected climate change resulted in an agenda. This is of critical importance to many
additional 15% increase in the number of under- developing countries that have contributed little to
nourished people. Under the other three SRES greenhouse gas emissions thus far and yet, it is these
scenarios, simulated climate-change impacts on food countries that will bear the brunt of the negative
security were rather small (in absolute terms) due to impacts of climate change and variability.

We thank two anonymous reviewers for their valuable IPCC 2000 Summary for policymakers, emissions scenarios.
comments, which greatly improved clarity and substance of A Special Report of IPCC Working Group III, Inter-
this paper. F. N. T. was supported in part by NOAA Research governmental Panel on Climate Change. Cambridge, UK:
grant GC02-333, and by a US–Italy Climate Change Grant Cambridge University Press.
from INGV-Bologna. IPCC, 2001a Climate change 2001: the scientific basis.
Contribution of working group I to the third assessment
report of the Intergovernmental panel on climate change,
ISBN 0521-014-95-6.
REFERENCES
IPCC 2001b IPCC climate change 2001: impacts, adaptation,
Ainsworth, E. A. & Long, S. P. 2005 What have we learned
and vulnerability. Contribution of working group II to the
from 15 years of free-air CO2 enrichment (FACE)?
third assessment report of the intergovernmental panel on
A meta-analysis of the responses of photosynthesis,
climate change. Cambridge, UK: Cambridge University
canopy properties and plant production to rising CO2.
Press.
New Phytol. 165, 351–372. (doi:10.1111/j.1469-8137.
Jablonski, L. M., Wang, X. & Curtis, P. S. 2002 Plant
2004.01224.x.)
reproduction under elevated CO2 conditions: a meta-
FAO 2001 The State of Food Insecurity in the World, 2001. Food
analysis of reports on 79 crop and wild species. New Phytol.
and Agriculture Organization of the United Nations,
156, 9–26. (doi:10.1046/j.1469-8137.2002.00494.x.)
Rome, Italy, ISBN 92-5-104628-X.
Kimball, B. A., Kobayashi, K. & Bindi, M. 2002 Responses of
FAO 2003 World Agriculture: Towards 2015/1030. In A FAO
agricultural crops to free-air CO2 enrichment. Adv. Agron.
perspective (ed. Jelle Bruinsma), p. 432. Rome: Food and
Agricultural Organization of the United Nations. 77, 293–368.
Felzer, B., Kicklighter, D. W., Melillo, J. M., Wang, C., Millennium Ecosystem Assessment 2005 Ecosystems and
Zhuang, Q. & Prinn, R. 2004 Effects of ozone on net human well-being: synthesis. Washington, DC: Island Press.
primary production and carbon sequestration in the Nakicenovic, N. et al. 2003 IPCC SRES revisited: a response.
conterminous United States using a biogeochemistry Clim. Change 14, 187–214.
model. Tellus 56B, 230–248. Oberhuber, J. M. 1993 Simulation of the Atlantic circulation
Fischer, G. & Sun, L. 2001 Model-based analysis of future with a coupled sea-ice mixed layer-isopycnal general
land-use development in China. Agr. Ecosyst. Environ. 85, circulation model. Part I: model description. J. Phys.
163–176. (doi:10.1016/S0167-8809(01)00182-7.) Oceanogr. 13, 808–829. (doi:10.1175/1520-0485(1993)
Fischer, G., Frohberg, K., Parry, M. L. & Rosenzweig, C. 023!0808:SOTACWO2.0.CO;2.)
1994 Climate change and world food supply, demand and Olesen, J. E. & Bindi, M. 2002 Consequences of climate
trade: who benefits, who loses? Global Environ. Change 4, change for European agricultural productivity, land use
7–23. (doi:10.1016/0959-3780(94)90018-3.) and policy. Eur. J. Agron. 16, 239–262. (doi:10.1016/
Fischer, G., Frohberg, K., Parry, M. L. & Rosenzweig, C. S1161-0301(02)00004-7.)
1996 Impacts of potential climate change on global and Parry, M. L., Rosenzweig, C., Iglesias, A., Fischer, G. &
regional food production and vulnerability. In Climate Livermore, M. T. J. 1999 Climate change and world food
change and world food security, vol. 137 (ed. E. T. Downing), security: a new assessment. Global Environ. Change 9,
pp. 115–159. Berlin: Springer-Verlag. 51–67. (doi:10.1016/S0959-3780(99)00018-7.)
Fischer, G., Frohberg, K., Keyzer, M. A. & Parikh, K. S. Parry, M. L. et al. 2001 Millions at risk: defining critical
1988 Linked national models: a tool for international policy climate change threats and targets. Global Environ. Change
analysis. Dordrecht, The Netherlands: Kluwer Academic. 11, 181–183. (doi:10.1016/S0959-3780(01)00011-5.)
Fischer, G., van Velthuizen, H., Shah, M. & Nachtergaele, Parry, M. L., Rosenzweig, C., Iglesias, A., Livermore, M. &
F. O. 2002a Global agro-ecological assessment for agriculture in Fischer, G. 2004 Effects of climate change on global food
the 21st century: methodology and results. IIASA RR-02-02. production under SRES emissions and socio-economic
Laxenburg, Austria: IIASA. scenarios. Global Environ. Change 14, 53–67. (doi:10.
Fischer, G., Shah, M., & van Velthuizen, H. 2002b Climate 1016/j.gloenvcha.2003.10.008.)
Change and Agricultural Vulnerability, Special Report to Pope, V. D., Gallani, M. L., Rowntree, P. R. & Stratton, R. A.
the UN World Summit on Sustainable Development, 2000 The impact of new physical parametrizations in the
Johannesburg 2002. Laxenburg, Austria: IIASA. Hadley Centre climate model—HadAM3. Climate Dyn.
Flato, G. M., Boer, G. J., Lee, W. G., McFarlane, N. A., 16, 123–146. (doi:10.1007/s003820050009.)
Ramsden, D., Reader, M. C. & Weaver, A. J. 2000 The Reilly, J., Tubiello, F. N., McCarl, B. & Melillo, J. 2001
Canadian centre for climate modelling and analysis global Climate change and agriculture in the United States. In
coupled model and its climate. Climate Dyn. 16, 451–467. Climate change impacts on the United States: foundation,
(doi:10.1007/s003820050339.) USGCRP (ed. J. Melillo, G. Janetos & T. Karl),
Gordon, H. B. & O’Farrell, S. P. 1997 Transient climate pp. 379–403. Cambridge, UK: Cambridge University
change in the CSIRO coupled model with dynamic sea ice. Press.
Mon. Weather Rev. 125, 875–907. (doi:10.1175/1520- Reilly, J. et al. 2003 U.S. Agriculture and climate change: new
0493(1997)125!0875:TCCITCO2.0.CO;2.) results. Climatic Change 57, 43–69.
Gordon, C., Cooper Sr., C. A., Banks, H., Gregory, J. M., Roeckner et al. 1992 Simulation of the present-day climate with
Johns, T. C., Mitchell, J. F. B. & Wood, R. A. 2000 The the ECHAM 4 model: impact of model physics and resolution.
simulation of SST, sea ice extents and ocean heat Max-Planck Institute for Meteorology, Report No. 93,
transports in a version of the Hadley centre coupled p. 171, Hamburg, Germany.
model without flux adjustments. Climate Dyn. 16, Roeckner, E. et al. 1996 The atmospheric general circulation
147–168. (doi:10.1007/s003820050010.) model ECHAM-4: model description and simulation of
Hirst, A. C., Gordon, H. B. & O’Farrell, S. P. 1997 Response present-day climate. Max-Planck Institute for Meteorology,
of a coupled ocean-atmosphere model including oceanic Report No. 218, p. 90, Hamburg, Germany.
eddy-induced advection to anthropogenic CO2 increase. Rosenzweig, C. & Parry, M. L. 1994 Impacts of climate
Geophys. Res. Lett. 23, 3361–3364. (doi:10.1029/ change on world food supply. Nature 367, 133–138.
96GL03234.) (doi:10.1038/367133a0.)

Rosenzweig, C. & Tubiello, F. N. In press. Interactions of Tubiello, F. N., Jagtap, S., Rosenzweig, C., Goldberg, R. &
adaptation and mitigation strategies in agriculture. Jones, J. W. 2002 Effects of climate change on U.S. crop
Mitigation and adaptation strategies for climate change. production from the National assessment. Simulation
Rosenzweig, C., Tubiello, F. N., Goldberg, R. A., Mills, E. & results using two different GCM scenarios. Part I: wheat,
Bloomfield, J. 2002 Increased crop damage in the U.S. potato, corn, and citrus. Climate Res. 20, 259–270.
from excess precipitation under climate change. Global Tubiello, F. N. & Ewert, F. 2002 Modeling the effects
Environ. Change 12, 197–202. (doi:10.1016/S0959- of elevated CO2 on crop growth and yield: a review.
3780(02)00008-0.) Eur. J. Agr. 18, 57–74. (doi:10.1016/S1161-0301(02)
Rosenzweig, C., Allen , L. H., Harper, L. A., Hollinger, S. e. 00097-7.)
& Jones, J. W. (eds) 1995 Climate change and agriculture: Investing in development. A practical plan to achieve the
analysis of potential international impacts. ASA Special Millenium development goals. Report to the UN Secretary
Publication No. 59. Madison, WI: ASA, CSSA and SSSA.
General. 2005. ISBN 1-884707-217-1, New York.
Steffen, W., Sanderson, A., Tyson, P. D., Jäger, J., Matson,
Washington, W. M. et al. 2000 Parallel climate model (PCM)
P. A., Moore III, B., Oldfield, F., Richardson, K.,
control and transient simulations. Climate Dyn. 16,
Schellnhuber, H. J., Turner, B. L. & Wasson, R. J. (eds)
2004 Global change and the Earth system: a planet under 755–774. (doi:10.1007/s003820000079.)
pressure. Berlin, Heidelberg, New York: Springer-Verlag. Watson, R. T., Noble, I. R., Bolin, B., Ravindranath, N. H. &
Tsuji, G. Y., Uehara, G. & Balas, S. (eds) 1994 DSSAT v3, Verardo, D. J. (eds) 2000 Land use, land-use change, and
volumes 1, 2 and 3. Honolulu, Hawaii: International forestry. Special report of the intergovernmental panel on
Benchmark Sites Network for Agrotechnology Transfer, climate change. Cambridge, UK: Cambridge University
University of Hawaii. Press.
Tubiello, F. N. 2005 Climate variability and agriculture: WHO 2003 Climate change and human health—risks and
perspectives on current and future challenges. In Impact of responses. Geneva: WHO.
climate change, variability and weather fluctuations on crops Williams, J. R., Renard, K. G. & Dyke, P. T. 1984 EPIC—a
and their produce markets (ed. B. Knight), Impact Reports, new model for assessing erosion’s effect on soil pro-
Cambridge, UK. ductivity. J. Soil Water Conserv. 8, 381–383.

Socio-economic and climate change impacts

Uploaded by

Copyright:

Available Formats

Socio-economic and climate change impacts

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Socio-economic and climate change impacts

Uploaded by

Copyright:

Available Formats

Downloaded from http://rstb.royalsocietypublishing.

org/ on January 14, 2015

Phil. Trans. R. Soc. B (2005) 360, 2067–2083

Socio-economic and climate change impacts on

1. INTRODUCTION and possibly increased frequency of extreme events

2067 q 2005 The Royal Society

Phil. Trans. R. Soc. B (2005)

Phil. Trans. R. Soc. B (2005)

Phil. Trans. R. Soc. B (2005)

scenario/model HADCM3 ECHAM CSIRO CGCM2 NCAR-PCM

Phil. Trans. R. Soc. B (2005)

GCM-projected mean annual

precipitation change (%)

GCM-projected mean annual

precipitation change (%)

changes in the frequency of extreme events, so that our 3. ANALYSIS OF RESULTS

Phil. Trans. R. Soc. B (2005)

reference climate 1000 km2 no constraint slight moderate severe

no constraint 535 457 66 6 6

developed, all land developed, current cultivated land

110 CSIRO 110 CSIRO

Phil. Trans. R. Soc. B (2005)

The case of sub-Saharan Africa (ii) Rainfed cereal-production

Phil. Trans. R. Soc. B (2005)

Phil. Trans. R. Soc. B (2005)

developed, all land developed, current cultivated land

140 120 CSIRO

developing, all land developing, current cultivated land

China, current cultivated land India, current cultivated land

Phil. Trans. R. Soc. B (2005)

Hadley, 2080 CSIR0, 2080

Phil. Trans. R. Soc. B (2005)

number of countries projected population (billions) change in cereal-production potential (%)

A1FI 17 8 16 0.62 0.43 0.55 5.1 K0.6 K8.6 K4.1

number of countries projected population (billions) change in cereal-production-potential (%)

HadCM3 14 11 16 0.58 0.23 0.78 5.3 0.7 K7.9 K1.9

scenarios, there is only modest progress in reducing

(UN Millennium Project 2005) of reducing hunger by

Phil. Trans. R. Soc. B (2005)

Phil. Trans. R. Soc. B (2005)

(a) developing countries (b) African countries

path assumptions. The full range of climate-change

additional under-nourished due

to climate change (millions)

We nonetheless chose to include it alongside the other

Phil. Trans. R. Soc. B (2005)

picture, as discussed. Of particular relevance to projected demographic and socio-economic changes

Phil. Trans. R. Soc. B (2005)

Phil. Trans. R. Soc. B (2005)

Phil. Trans. R. Soc. B (2005)

You might also like