ORI GIN AL PA PER

Implications of population growth and urbanizationon agricultural risks in China

G. Fischer • W. Winiwarter • G. Y. Cao •

T. Ermolieva • E. Hizsnyik • Z. Klimont • D. Wiberg •

X. Y. Zheng

Published online: 10 May 2011

� Springer Science+Business Media, LLC 2011

Abstract Growing population, rapid urbanization, rising incomes, and changing

consumption preferences stimulate intensification of livestock production and

excessive fertilization of crops in China. We present an innovative approach that

sheds light on options to prevent negative environmental consequences of food

production. Trends indicate that agricultural production expansion will take place in

‘‘profitable’’ locations around densely populated areas, where there are generally

insufficient natural resources to recycle production wastes. This will likely lead to

increased environmental impacts and risks to human health, with the largest impacts

in close proximity to population hotspots. We identify trends in Chinese agricultural

production and devise and compare feasible mitigation scenarios. We present a

spatial allocation procedure that facilitates management of agricultural production

expansion, accounting for environmental and health constraints. This procedure,

based on behavioral principles, uses a spatial risk preference structure induced by

local conditions, including environment, production, and demand, with important

research and policy implications.

Keywords Population growth � Urbanization � Agriculture � Intensification �Environment � Pollution � Health � Spatial planning of production expansion �China � Demography

G. Fischer � W. Winiwarter � G. Y. Cao � T. Ermolieva (&) � E. Hizsnyik �Z. Klimont � D. Wiberg

International Institute for Applied Systems Analysis (IIASA),

Schlossplatz 1, 2361 Laxenburg, Austria

e-mail: ermol@iiasa.ac.at

W. Winiwarter

Austrian Institute of Technology (AIT), Donau-City Str. 1, 1220 Vienna, Austria

X. Y. Zheng

Population Institute, Beijing University, Beijing, China

123

Popul Environ (2012) 33:243–258

DOI 10.1007/s11111-011-0134-4

Introduction

Economic–demographic growth, urbanization and, consequently, changing con-

sumption preferences have altered the structure of agricultural production systems

in China (Ma et al. 2004). These factors have furthermore promoted industrial

agriculture geared toward making use of economies of scale to produce the highest

output at the lowest cost. Although such intensification has brought about many

positive effects, it also carries significant disadvantages, risks, and costs. Environ-

mental impacts and health hazards have increased awareness and underlined the

need to identify steps to be taken to achieve sustainable agriculture.

One of the main contributors to agricultural pollution in China is intensive

livestock production. Traditional livestock farming was based on a natural farming

cycle. Lack of nitrogen for agriculture required minimization of wastage. Thus,

manure was used as the main fertilizer, preventing leakage of nitrogen compounds

and avoiding environmental degradation. However, the natural farming cycle has

been replaced by large-scale livestock production, in particular pig and poultry

production. Intensive livestock production enterprises are located in close proximity

to meat and feed markets near urban areas, and such livestock concentration often

exceeds the limits for adequate disposal of manure and waste on land to be recycled.

Combined with intensive crop cultivation, the excess nutrients problem is further

exacerbated by imbalanced fertilizer application.

Intensive agricultural land use and the management of crop and livestock

production cause significant emissions of methane (CH4) and nitrous oxide (N2O)

into the atmosphere, both of which are radiatively active and contribute to global

warming. Excess leaching of nitrogen into groundwater and rivers, as well as the

volatilization of ammonia (NH3) and nitrogen oxides (NOx), contributes to regional

and local air quality problems, causes the acidification and eutrophication of

ecosystems, and compromises human health. Pollution of the atmosphere, water,

and soil resources by residues of intensive crop and livestock production has already

become a serious environmental issue in China (Edwards and Daniel 1994;

Steinfeld et al. 1997; Wang 2005). Without appropriate measures, the trend will

become severe and may even become irreversible in some areas. As we show below,

even with efforts to control environmental impacts, China’s national-level

environmental deterioration may increase by roughly one-third to nearly half (see

also Ermolieva et al. 2009; Fischer et al. 2010).

In this paper, we compile the results of recent studies (Ermolieva et al. 2005,

2009; Fischer et al. 2007, 2008, 2009, 2010) to analyze how the dynamic and rapid

pace of China’s economic growth and demographic transition have affected the

agricultural sector, the extent and geographical heterogeneities of production, and

the spatial distribution of induced environmental impacts. The following section

highlights the main factors and trends underlying agricultural intensification in

China. We use these factors as indicators to guide production allocation within

acceptable health, air, and water pollution thresholds. The subsequent section

summarizes our numerical results, which are described in detail in the studies of

Fischer et al. (2007, 2008, 2009) based on different model settings, with the

objective of illustrating how the inclusion of risk indicators in the agricultural

244 Popul Environ (2012) 33:243–258

123

planning model can decrease environmental pollution and exposure to health risks.

We present our conclusions in the final section.

Urban transformations and agricultural developments in China

Agricultural intensification processes in China are largely attributable to the

country’s economic growth. Among the key factors for rapid economic develop-

ment are population growth and structural labor transformations, in particular the

migration of workers from agricultural inland regions to manufacturing jobs in

coastal China (Huang et al. 2003). Despite the large-scale reallocation of labor from

agricultural to non-agricultural sectors and the migration from rural to urban areas,

the productivity disparity between urban industry and the agricultural sector is still

substantial, continues to increase (Keyzer and van Veen 2005), and will likely result

in further urbanization. Alternative scenario settings project that urbanization levels

could reach 48–51% by 2015 and 56–62% by 2030 (Toth et al. 2008). A rapid

increase in urbanization and incomes is expected, especially in the wealthier coastal

areas where approximately 70% of China’s population reside (coastal provinces of

China in this context include Beijing, Tianjin, Hebei, Liaoning, Shanghai, Jiangsu,

Zhejiang, Fujian, Shandong, Guangdong, and Guangxi, which are primarily located

in the North, East, South, and Southwest geographical regions, see Table 1).

Although the overall trend in China is toward a gradual increase in incomes,

regional and interregional income disparity is expected to rise owing to the rapid

expansion of non-agricultural activities and decreasing/diminishing role of agricul-

ture (Keyzer and van Veen 2005).

The geographical and social heterogeneities of the population, urbanization, and

income patterns have implications on the patterns of demand, allocation, and the

structure of livestock production. Demand in urban higher-income areas is rapidly

increasing, with preference for reduced fat and higher quality poultry and pork

Table 1 Population development and urbanization by region

Region 2000 2015 2030 Change 2000–2030

Total

(million)

Urban

(%)

Total

(million)

Urban

(%)

Total

(million)

Urban

(%)

Total

(%)

Urban (Percentage

points)

North 309 33 332 46 333 57 8 24

Northeast 106 52 109 61 102 69 -3 17

East 197 42 209 55 207 66 5 23

Central 166 32 172 43 164 54 -1 22

South 129 50 219 58 212 72 65 22

Southwest 20 26 202 37 234 48 -2 22

Plateau 8 27 9 37 10 47 -30 20

Northwest 110 32 124 43 129 53 17 21

China 1,263 36 1,376 49 1,392 59 10 23

Source Toth et al. (2008)

Popul Environ (2012) 33:243–258 245

123

products produced by specialized agricultural enterprises. Between 1990 and 1999,

per capita consumption of pork increased by roughly 30%, while that of poultry

meat more than doubled (see Table 2). Poultry now accounts for about 21% of total

meat production in China. Pork, on the other hand, still makes up around two-thirds

of total meat consumption.

China is among those countries with the highest densities of pigs and poultry in

the world.1 Livestock production is typically structured in three production system

categories: Traditional backyard production (1–5 pigs), specialized farms/house-

holds (5–1,000 pigs), and industrial farms ([1,000 pigs per enterprise). Estimates of

future livestock production rest on income-consumption elasticities for livestock

products. According to studies by the State Statistical Bureau (SSB) and

International Food Policy Research Institute, the income elasticity for all livestock

products ranged between 0.75 and 0.85 in the early 1990s and is expected to

increase only slightly by the 2030s. Milk has the highest income elasticity of over

1.5, followed by fish and poultry. Beef and mutton have the lowest income elasticity

of 0.34 for rural and 0.69 for urban consumers.

Projections show that future demand patterns for cereals and other staple grains

are expected to only modestly fluctuate between 2000 and 2030 due to the fact that

the per capita consumption of cereals is lower among urban residents than rural

residents. While urbanization has slowed cereal consumption, it has accelerated the

increase in meat consumption. Figures 1 and 2 show that the urban population’s

projected meat consumption will just about double between 2000 and 2030 (Huang

et al. 2003; Ermolieva et al. 2005; Fischer et al. 2009). The projected demand for

meat in general will also rise considerably. Table 3 presents the meat demand

projections in China’s economic regions.

Agricultural production planning model

Geographical distribution and the management characteristics of agricultural

activities to a large extent determine the levels of air, water, and soil pollution

arising from agriculture. While intensification has a number of advantages,

Table 2 Per capita consumption of pork, poultry, and eggs

Rural (kg/person/year) Urban (kg/person/year)

1980 1990 1999 1980 1990 1999

Pork 8.5 10.9 13.8 17.4 21.6 29.3

Poultry 0.8 1.4 3.3 2.0 3.8 7.8

Eggs 1.5 3.0 6.3 5.5 8.9 14.8

Source CCAP (2002)

Consumption estimated in retail weight

1 Only five other countries have higher densities of both pigs and poultry: Vietnam, Korea, Belgium-

Luxemburg, the Netherlands, and Denmark (FAOSTAT 2002).

246 Popul Environ (2012) 33:243–258

123

especially in terms of productivity increase, associated environmental loads and

health risks raise concerns about how to reconcile environmental standards with the

demands of future agricultural development in China.

Recent studies investigated this question using a stochastic and dynamic

simulation model to analyze agricultural production in China at both county

(approximately 3,000 administrative units) and provincial level (31 provinces),

running for a time span of 30 years with a 5-year time step (Ermolieva et al. 2005;

Fischer et al. 2007, 2008, 2009).2

Using alternative economic and demographic projections (Huang et al. 2003;

Toth et al. 2003, 2008), the model estimates demand increases and consumption of

major agricultural products, e.g., cereals, meat, milk, etc. Demand patterns differ

between urban and rural areas as well as between geographical regions, and vary by

income. That is, a rise in household income affects consumer preference for higher

Rural

Urban

0

20000

40000

60000

80000

100000

2000 2005 2010 2015 2020 2025 2030

1000

mt

Fig. 1 Projected urban and rural meat demand

Other meat

Pork

Poultry

0

20000

40000

60000

80000

100000

2000 2005 2010 2015 2020 2025 2030

1000

mt

Fig. 2 Projected meat demand by type

2 The model was developed within the scope of the projects ‘‘Policy Decision Support for Sustainable

Adaptation of China’s Agriculture to Globalization’’ (CHINAGRO) (Keyzer and van Veen 2005),

‘‘Chinese Agricultural Transition: Trade, Social and Environmental Impacts’’ (CATSEI) (Fischer et al.

2006), and ‘‘Integrated Nitrogen Management in China’’ (INMIC) (Ermolieva et al. 2009).

Popul Environ (2012) 33:243–258 247

123

quality and low-fat meats, e.g., poultry. A schematic of the model is presented in

Fig. 3.

The evolution of demand is modeled as a function of group-specific per capita

income increases based on income elasticities and the distinction between urban and

rural consumers. Agricultural supply is represented by areas with major cereals and

livestock commodities. The projection of crop production in given locations starts

from the base year data and takes account of productivity and resource constraints

and relies on information about attainable yields (Fischer et al. 2002) under different

practices and fertilization rates. Livestock production distribution is projected from

the base year data for both the provincial and county levels and by management

system according to the following principles:

• Projections of livestock distribution of confined3 traditional systems have been

linked to expected changes in the rural population.

• The confined specialized and industrial livestock systems have been modeled to

meet the provinces’ projected demand for livestock products. Hence, such

systems compensate for the decrease in traditional systems and continually

evolve with the demand growth projected at provincial level.

• The geographical distribution of pastoral livestock has been projected in

accordance with the availability and productivity of grasslands.

Table 3 Projected meat demand (106 tons) by region

Region 2000 2015 2030 Change 2000–2030 (%)

North 8.7 13.5 17.3 78

Northeast 3.7 5.0 5.8 52

East 7.8 11.4 13.7 59

Central 7.0 10.0 11.9 67

South 9.7 16.0 22.3 111

Southwest 9.2 12.8 15.6 67

Plateau 0.3 0.4 0.6 97

Northwest 2.8 4.4 5.8 90

China 49.2 73.4 93.1 77

Productionallocation procedure

Population,Economic growth,

Incomes,Demand increase

Crops andlivestock bylocation andmanagement

system

Environmental and

human health riskindicators

Iterative adjustmenttoward risks minimization

Fig. 3 Schematic of the model

3 A confined system may include post-harvest stubble grazing as opposed to pure grazing systems based

on pastures.

248 Popul Environ (2012) 33:243–258

123

Environmental pollution indicators associated with agricultural activities are

calculated at the county level for two major types/forms of agricultural nutrient

losses, i.e., (1) point-source losses resulting from location-specific sources such as

livestock housing/stables, manure storage facilities, and liquid manure disposal, and

(2) non-point losses attributable to fertilizer and manure applications to cultivated

land or to grazing livestock in pasture areas. Technical coefficients on manure

production, storage, and recycling are derived from regional studies summarized in

Menzi (2002) and NuFlux (2001). The pollution level is estimated with the help of

agricultural, environmental, and biophysical indicators characterizing production

intensity, water, soil, and air quality. Health risks are measured in terms of

population exposure to different levels of environmental pollution. The feasible

domains of the indicators’ variables are subdivided into subdomains of different

degrees of impact, severity, and suitability. The variables may be combined in risk

functions to reflect the levels of different risks in areas associated with agricultural

production (Ermolieva et al. 2009; Fischer et al. 2008).

The main components of production increases to meet demand include the

establishment of new and/or the expansion of existing facilities. In some areas,

especially in the vicinity of cities, the indicators may already divulge that a further

allocation of production is impossible. The question is how to adjust the production

facilities in response to increasing demand without exacerbating environmental and

health problems. The summary of the production allocation algorithm can be found

in the Annex, and additional details are available in Fischer et al.’s studies (2007,

2008, 2009).

Numerical results

Several scenarios designed to estimate the environmental impacts of current

agricultural intensification trends in China have been formulated and compared, and

the benefits of feasible mitigation measures evaluated. Details on the development

of the scenarios and more extended discussion of results focusing on environmental

pollution and health exposure indicators have been reported by Fischer et al. (2010):

• a ‘‘business-as-usual’’ scenario for cases in which the rise in production is

proportionally allocated to demand increase, which is concentrated in the

vicinity of densely populated urban areas;

• a reallocation scenario which combines the demand-driven preference structure

of the business-as-usual scenario with information on population density and

urban agglomerations to reduce risks caused by livestock production;

• a fertilizer-saving scenario (nutrients are supplied to crops through mineral

fertilizer only after manure has been applied), and

• a scenario combining optimized fertilizer use with specific ammonia abatement

technologies (‘‘minimized ammonia’’ scenario).

The advantages of the four scenarios are compared in terms of indicators related

to nitrogen release: leaching of nitrate into groundwater, atmospheric emissions of

N2O, and ammonia NH3. With reference to regional human and ecosystem health,

Popul Environ (2012) 33:243–258 249

123

excess leaching of nitrogen into groundwater and rivers, as well as the volatilization

of ammonia (NH3) and nitrogen oxides (NOx), contributes to regional and local air

quality problems, acidification, and eutrophication of ecosystems and damages

human health. With regard to the global environment and climate change, nitrogen

attributable to agriculture produces significant emissions of nitrous oxide (N2O) gas

into the atmosphere. Our aim is not to identify a scenario allowing for a complete

nitrogen balance, but rather to derive a set of robust nitrogen-related indicators that

enable measurement of environmental impacts from agricultural activities and to

estimate pollution mitigation options without compromising food security targets.

The indicators are computed for approximately 3,000 administrative units at

county level (Table 4 presents regional aggregates). The simulation of alternative

scenarios corresponds with different production allocation priors of the algorithm

summarized in the Annex. The following formula is used to calculate nitrogen

surplus from agriculture:

N Surpl ¼ Nmnr þ Nfert þ Nfix þ Ndep � Nupt;

where N_Surpl denotes the surplus of nitrogen in the given field, Nmnr represents

nitrogen in available manure to fertilize the field (net losses in stables); Nfert

signifies nitrogen in chemical fertilizers; Nfix is nitrogen fixed by N-fixing crops;

Ndep is nitrogen deposition; and Nupt stands for the nitrogen uptake by all crops (net

nitrogen in recycled crop residues). To estimate nitrogen leaching into groundwater,

we adopted the MITERRA model (Velthof et al. 2008), which is based on a simple

methodology of a combined water and nitrogen balance to derive leaching

indicators for a broad range of soil types aggregated into seven classes (sandy, clay,

gleyic, stagno-gleyic, peat, loam, and paddy soils) with different leaching

characteristics (Shi et al. 2004). Soils are also distinguished by crop water

management, that is, separately for irrigated and for rain-fed land. This method

estimates the share of nitrogen surplus that enters the groundwater, i.e., the leaching

fraction for each soil class, climate condition (e.g., precipitation, temperature), and

land use type.

Table 4 Projected environmental pressure indices for China by scenario

In kt 2000 2015 2030 Change

2005–2030 (%)

Region Lch N2O–

N

NH3–

N

Lch N2O–

N

NH3–

N

Lch N2O–

N

NH3–

N

Lch N2O NH3

Business-as-usual 701 855 7,469 953 1,163 9,822 1,101 1,282 10,878 57 50 46

Sustainable

reallocation

701 855 7,469 950 1,162 9,825 1,077 1,278 10,848 54 49 45

Fertilizer saving 231 652 5,426 293 887 7,003 321 956 7,487 39 47 38

Minimized

ammonia

231 654 5,168 298 897 6,461 328 963 6,884 43 46 33

Data in kt N

250 Popul Environ (2012) 33:243–258

123

The emission of N2O into the atmosphere is estimated by applying a

methodology developed for the International Institute for Applied Systems Analysis

(IIASA) Greenhouse gas–Air pollution Interactions and Synergies (GAINS) model

(Winiwarter 2005). N2O emissions are computed as the product of an emission

factor times the respective activity data (livestock and crop production operations).

This methodology is based on Intergovernmental Panel on Climate Change (IPCC)

guidelines, where IPCC default emission factors are applied (IPCC 2000).

Ammonia emissions (NH3) from livestock production in a given area primarily

occur at four stages: Animal housing, manure storage, land application of manure,

and livestock grazing. These stages are explicitly distinguished in the China model

(Fischer et al. 2007–2010) and in the GAINS methodology (Klimont 2001; Klimont

and Brink 2004).

In all scenarios, the significant increase in food demand will further intensify the

problem of nitrogen emissions and the impacts on China’s environment (population

growth will continue and is coupled with increased meat consumption which, in

0

100

200

300

400

500

600

700

kg/ha cult land

Nu

mb

er o

f co

un

ties

0%

20%

40%

60%

80%

100%

0

100

200

300

400

500

600

0 25 50 75 100 125 150 175 200 225 250 300 400 More

0 25 50 75 100 125 150 175 200 225 250 300 400 More

Nu

mb

er o

f co

un

ties

0%

20%

40%

60%

80%

100%

kg/ha cult land

(a)

(b)

Fig. 4 Severity of ammonia emissions by number of counties affected: kg per ha of cultivated land in2000 (a) and 2030 (b), and cumulative share of counties, business-as-usual scenario

Popul Environ (2012) 33:243–258 251

123

turn, increases the amount of nitrogen fertilizer required). A qualitative comparison

across the scenarios reveals that the absolute values of the environmental indicators

do not change much in the sustainable reallocation scenario in comparison with the

business-as-usual scenario. Exposure to environmental and health impacts,

measured in terms of population and severity (i.e., of ammonia emissions),

decreases considerably in the second scenario due to a production shift away from

population centers to areas with a better production resource balance. Fischer et al.

(2007) found that the reallocation scenario offers a solution that corresponds to the

0

50

100

150

200

250

300

350

> 300

250 - 300

200 - 250

150 - 200

100 - 150

50 - 100

< 50

N NE E C S SW NW 0.0 0.2 0.4 0.6 0.8 1.0

N

NE

E

C

S

SW

NW

Fig. 5 Absolute (million inhabitants) and relative (share of total population) distribution of populationaccording to severity of environmental impact in 2000. Measured in terms of kg nitrogen in ammoniaemitted per ha of cultivated land in 2000. The indications on the horizontal axis refer to China’s regions:N, NE, E, C, S, SW, NW, namely North, Northeast, East, Center, South, Southwest, Northwest,respectively, business-as-usual scenario

North North-East East Central

0

50

100

150

200

250

300

350

0

50

100

0

50

100

150

0

50

100

150

> 300

250 - 300

200 - 250

150 - 200

100 - 150

50 - 100

< 50

South South-West North-West China

0

50

100

150

200

250

0

50

100

150

200

0

50

100

BU RA OM MA BU RA OM MA BU RA OM MA BU RA OM MA

BU RA OM MA BU RA OM MA BU RA OM MA0

200

400

600

800

1000

1200

1400

BU RA OM MA

> 300

250 - 300

200 - 250

150 - 200

100 - 150

50 - 100

< 50

Fig. 6 Number of inhabitants by severity of ammonia losses (kg nitrogen in ammonia emitted per hatotal area), by economic regions and scenarios BU business-as-usual, RA reallocation, OM optimizedmanure; MA minimized ammonia in 2030

252 Popul Environ (2012) 33:243–258

123

cross-entropy maximization principle. The fertilizer-saving approach is a no-regret

solution. A substantial amount of chemical fertilizers can be saved by increasing the

utilization of manure nutrients. The result reveals the extent of achievable nitrogen

reduction if the practice is implemented at the national level.

It must be noted that the large-scale application of manure poses numerous

management challenges (timing of application and quality of fertilization), i.e., the

advantages of this scenario may perhaps not be fully attainable. Moreover,

compounds in manure (heavy metals, pathogens, as well as drugs added to fight

such pathogens) may compromise the quality of manure as a fertilizer. The

reduction in ammonia emissions in combination with the optimized use of fertilizers

is significant, but may cause an increase in N2O emissions. Measures to reduce the

emission of one of these gases may lead to an increase in the emission of others

because of the interaction between them. Substituting urea with, for instance,

ammonium nitrate has the potential to reduce ammonia emissions from synthetic

fertilizer use without increasing emissions from nitrous oxide or methane.

Environmental risk indicators, as illustrated in Table 4, measure the level of

pollution by administrative region or area. It is not so much about the adverse

environmental situation in a given region or area, but rather about the impact on the

health of the population. To estimate this risk, we need to assess the population’s

exposure. A differentiation between area-based and population-based risk can be

applied to any risk indicator. As the population is not, in general, evenly distributed

over a given area, the population-weighted risk may differ considerably from area-

based risk. Figures 4 and 5 illustrate population exposure to ammonia emissions in

2000. Figures 5 and 6 compare the same indicator for the four alternative scenarios

by region and for China as a whole.

The severity of health impacts from ammonia emissions depends, of course, on

the route of exposure, the dose, and the duration of exposure. This gas can be

harmful in several ways. In humans, it primarily affects the respiratory tract. Long-

term exposure may cause asthma, chronic bronchitis, and allergies. Ammonia

emissions also affect animal performance and can reduce livestock productivity.

The deposition of ammonia compounds has other consequences as well: (1) In

addition to being toxic, ammonia can reduce the availability of nutrients to plants;

(2) the accumulation of ammonia in soil and water leads to eutrophication and

acidification, thus damaging fauna and flora.

A gradual improvement of the situation across the different scenarios (Fig. 6) is

visible when mitigation measures are implemented and additional measures derived

from the business-as-usual scenario are incrementally applied to the minimizedammonia scenario, thus reducing the population’s exposure to health risks.

Conclusions

This paper summarizes and extends the results of prior work by our group, focusing

on how the demographic and economic forces driving urbanization and production

industrialization in China may lead to severe environmental pollution. We estimate

that economic and demographic drivers will affect demand for agricultural products

Popul Environ (2012) 33:243–258 253

123

in China. While rapid economic growth and urbanization are likely to slow cereal

consumption, the increase in meat consumption will accelerate. Urban diets include

higher consumption of meat, and per capita meat consumption is strongly linked to

income growth. We project that the consumption of livestock will have roughly

doubled by 2030. We contend that both income growth and the rising number of

consumers who adopt an urban lifestyle will reinforce an increase in meat

consumption. At the projected levels of per capita meat consumption, China would

approach the current per capita meat consumption in industrialized countries.

According to current trends, production expansion to meet this increased demand

will take place in ‘‘profitable’’ locations around densely populated areas, which are

generally characterized by an insufficiency of natural resources to recycle

production waste. Consequently, an increase in environmental impacts is expected.

Four scenarios were used to estimate future environmental impacts, on the one hand,

and to determine the level of remediation that is feasible, on the other. The scenarios

describe an economic and ecological condition without providing details on the

process of implementation. Hence, drivers and consequences of agricultural

intensification were used to compile indicators of environmental pollution and

health risks. Health risks were measured as a coincidence of several factors

including population density, urbanization, production intensification level, and

available livestock facilities.

To measure environmental pollution, nitrate leaching into groundwater and

nitrous oxide and ammonia emissions into the atmosphere were used. Although the

results of the four scenarios clearly differ for the individual economic regions, the

level of pollution is likely to increase in all cases. Even in scenarios (3) and (4)

which carry considerable improvements, China’s environmental deterioration

increases by roughly one-third to nearly half. Nitrogen losses in the form of

ammonia pose the biggest problem by far in all four scenarios. Losses in the form of

N2O follow, while nitrate leaching is of less importance. This is attributable to the

huge losses in the gas phase, while the leaching fractions (which are relatively

small) only apply to excess soil nitrogen not already lost or used elsewhere. Also,

leaching does not include nitrate runoff to surface water. Under such inherent

uncertainties, the estimates of nitrogen fluxes serve as indicators of environmental

risks in China’s environment. Therefore, changes which are attributable to the

implemented measures are considered robust. Combining the indicators allows for

an assessment of the population’s exposure to environmental loads from agriculture

under alternative demographic or socioeconomic scenarios and to make policy

recommendations. Relying on the estimates of these indicators, it is also possible to

supervise future production expansion based on the mitigation and minimization of

environmental and health risks.

The procedure to geographically allocate livestock production according to

different criteria is based on behavioral principles and a spatial risk preference

structure. The proposed algorithm converges to the solutions maximizing a cross-

entropy function, which provides results that reflect the maximum likelihood

solutions (Fischer et al. 2007, 2008). When applying a top-down model to China,

numerous details are obviously overlooked. The resulting indicators provide for a

better understanding of potential loads and assign the preference structure to areas

254 Popul Environ (2012) 33:243–258

123

where further studies are necessary or where mitigation measures have to urgently

be geared toward specific goals. The general preference structure for the efficiency

of specific measures (combined in different scenarios) can also be deduced based on

our approach.

Annex: Production allocation algorithm

Let us briefly summarize the production allocation algorithm (for more details, see

Fischer et al. 2007–2009). The objective is to allocate new supply facilities in the

best way possible to meet the projected increases in national demand di for livestock

products i among the production activities/locations k, k ¼ 1;K while considering

various risk indicators. In the following model, the risks are treated as constraints on

production expansion (similar to ambient targets in pollution control models).

Therefore, the problem is formulated to determine suitable activity levels yik given

the constraints:X

k

yik ¼ di; ð1Þ

yik� 0; ð2ÞX

i

yik � bk; i ¼ 1;m; k ¼ 1;K; ð3Þ

where bk denotes thresholds for environmental and health risks and imposes

limitations to an increase in production of system or location k, k ¼ 1;K. Apart

from bk, there may be additional limits on yik, yik B rik which may be associated

with legislation, for example, to restrict production i to a production ‘‘belt’’ or to

exclude production i from urban or protected areas, etc. Thresholds bk and rik may

indicate that production exceeding these values is strictly prohibited. The procedure

may also allow for the thresholds to be exceeded while imposing taxes or requiring a

premium to be paid for the mitigation of certain risks. Equations 1–3 are well

established in the literature on type of transportation problems (Kantorovich 1942;

Koopman 1947). There may be more general constraints of typeP

i aikxik� di,

which require the proposed rebalancing approach to be expanded (see discussion in

Fischer et al. 2007).

In general, there may be infinite solutions to the Eqs. 1–3. The aim is to derive a

solution that provides for an appropriate balance between efficiency and risks. Here,

we can distinguish between two types of uncertainties that generate potential risks:

Behavioral or endogenous uncertainties which are addressed in the procedure and

associated with the allocation of new production capacities, and exogenous

uncertainties which are related to data and the model’s parameters are not addressed

in the current framework (see Fischer et al. 2007).

The available information on the current production facilities, the projected

demand increases, environmental thresholds, and costs are used to derive a prior

probability qik reflecting the assumption that a unit of demand di for product i should

be supplied by activity/location k. For instance, it is reasonable to allocate more

Popul Environ (2012) 33:243–258 255

123

livestock to areas with a higher demand increase, higher productivity, or better access

to animal feeds. This preference structure is expressed in prior qik,P

k qik ¼ 1 for all

i. The use of priors is consistent with spatial economic theory (see discussion, e.g., in

Fujita et al. 1999; Karlqvist et al. 1978). The likelihood qik can be modeled as

inversely proportional to production costs and inherent risks bk and rik. An initial

estimate of production i allocated to k can be derived as qikdi. This could, however,

result in a violation of applicable restrictions (3). Sequential rebalancing (Fischer

et al. 2006) proceeds as follows: We assume that—based on prior probability qik—

the expected initial allocation of di to k is y0ik ¼ qikdi i ¼ 1;m. As this allocation

may not comply with constraintP

i y0ik� bk j ¼ 1; n; the relative imbalances b0

k ¼bk=P

i y0ik are derived and updated z0

ik ¼ y0ikb

0k ; i ¼ 1;m. Now constraintP

i yik � bk is met, k = 1, 2,…, but the estimate z0ik may cause an imbalance for

relation (1), i.e.,P

k z0ik 6¼ di. Continue calculating a0

i ¼ di=P

k z0ik; i ¼ 1;m and

updating the imbalances y1ik ¼ z0

ika0i , etc. The estimate ys

ik can be represented as

ysik ¼ qk

ikdi, qsik ¼ qikb

s�1k

� �=P

j qikbs�1k

� �; i ¼ 1;m; k ¼ 1; 2; . . .. Assume ys ¼

ysik

� �has been calculated. Find bs

k ¼ bk=P

i ysik and qsþ1

ik ¼ qikbsj=P

i qikbsj

� �; i ¼

1;m; k ¼ 1; 2; . . .; etc.

In this form the procedure can be considered a redistribution of required supply di

among producers k = 1, 2,… by applying the sequential adjustment qsþ1ik , e.g., by

using a Bayesian-type rule to update the prior distribution: qsþ1ik ¼ qikb

sk=P

i qikbsk,

q0ik ¼ qik.

The iterative update of qik is based on an ‘observation’ of imbalances of the basic

constraints rather than calculated for observations of random variables. A

rebalancing procedure, similar to the one mentioned above for Hitchcock-

Koopmans’ transportation constraints (1)–(3), was proposed by G. V. Sheleikovskii

(for verification and references, see Bregman 1967) for the estimation of passenger

flows between regions. Verification of its convergence to the optimal solution

maximizing the cross-entropy functionX

i;kyik ln

yik

qikð4Þ

is provided in the study by Fischer et al. (2006) for general forms of constraints.

The alternative scenarios introduced in Sect. 4 correspond to different production

allocation priors qik, i ¼ 1;m; k ¼ 1;K, that guide the allocation procedure. The

priors’ aggregate variables include demand, environmental constraints, critical

health thresholds, and norms.

References

Bregman, L. M. (1967). Proof of the convergence of Sheleikhovskii’s method for a problem with

transportation constraints. Journal of Computational Mathematics and Mathematical Physics, 7(1),

191–204. Zhournal Vychislitel’noi Matematiki, Leningrad.

256 Popul Environ (2012) 33:243–258

123

Edwards, D. R., & Daniel, T. C. (1994). Quality of runoff from fescue grass plots treated with poultry

litter and inorganic fertilizer. Journal of Environmental Quality, 23, 579–584.

Ermolieva, T., Fischer, G., & van Velthuizen, H. (2005). Livestock production and environmental risks in

China: Scenarios to 2030. FAO/IIASA Research Report, International Institute for Applied Systems

Analysis.

Ermolieva, T., Winiwarter, W., Fischer, G., Cao, G.-Y., Klimont, Z., Schopp, W., Li, Y., & Asman, W.

A. H. (2009). Integrated nitrogen management in China. IIASA Interim Report IR-09-005,

International Institute for Applied Systems Analysis.

FAOSTAT. (2002). FAOSTAT data. Online access at http://faostat.fao.org, FAO, Rome.

Fischer, G., Cao, G. Y., Zheng, X. Y., Ermolieva, T., & Sun, L. (2008). Urbanization and livestock

production: Risk-adjusted approaches. Journal of Population and Developments, 14(6), 1668–1674.

Fischer, G., Ermolieva, T., Ermoliev, Y., & Sun, L. (2007). Integrated risk management approaches for

planning sustainable agriculture. In C. Huang, C. Frey, & J. Feng (Eds.), Advances in studies on riskanalysis and crisis response (pp. 24–30). Paris: Atlantis Press.

Fischer, G., Ermolieva, T., Ermoliev, Y., & Sun, L. (2009). Risk-adjusted approaches for planning

sustainable agricultural development. Stochastic Environmental Research and Risk Assessment,23(4), 441–450.

Fischer, G., Ermolieva, T., Ermoliev, Y., & van Velthuizen, H. T. (2006). Sequential downscaling

methods for estimation from aggregate data. In: K. Marti, Y. Ermoliev, M. Makowski, & G. Pflug

(Eds.), Coping with uncertainty, modeling and policy issues (pp. 155–169). Berlin, Germany:

Springer-Verlag.

Fischer, G., van Velthuizen, H. T., Shah, M. M., & Nachtergaele, F. O. (2002). Global agro-ecological

assessment for agriculture in the 21st century: Methodology and results. Research Report RR-02-02,

International Institute for Applied Systems Analysis.

Fischer, G., Winiwarter, W., Ermolieva, T., Cao, G.-Y., Qui, H., Klimont, Z., et al. (2010). Integrated

modeling framework for assessment and mitigation of nitrogen pollution from agriculture: Concept

and case study for China. Agriculture, Ecosystems & Environment, 136, 116–124.

Fujita, M., Krugman, P., & Venables, A. J. (1999). The spatial economy: Cities, regions, andinternational trade. Cambridge: The MIT Press.

Huang, J., Zhang, L., Li, Q., & Qiu, H. (2003). CHINAGRO project: National and regional economicdevelopment scenarios for China’s food economy projections in the early 21st century. Center for

Chinese Agricultural Policy, Beijing: Chinese Academy of Sciences.

IPCC. (2000). Good practice guidance and uncertainty management in national greenhouse gas

inventories. In J. Penman, D. Kruger, I. Galbally, T. Hiraishi, B. Nyenzi, S. Emmanuel, L. Buendia,

R. Hoppaus, T. Martinsen, J. Meijer, K. Miwa, & K. Tanabe (eds), IPCC National Greenhouse Gas

Inventories Programme, Institute for Global Environmental Strategies.

Kantorovich, L. V. (1942). On the translocation of masses. Compt. Rend. Acad. Sci. URSS, 37, 199–201.

Karlqvist, A., Lundqvist, L., Snickars, F., & Weibull, J. W. (1978). Studies in regional science and urbaneconomics: Spatial interaction theory and planning models (3rd ed.). Amsterdam: North-Holland

Publishing Company.

Keyzer, M. A., & van Veen, W. (2005). A summary description of the CHINAGRO-welfare model.CHINAGRO report, SOW-VU. Amsterdam: Free University.

Klimont, Z. (2001). Current and future emissions of ammonia in China. Paper presented at the 10th

annual emission inventory conference: one atmosphere, One inventory, many challenges. May 1–3,

Denver, USA.

Klimont, Z., & Brink, C. (2004). Modeling of emissions of air pollutants and greenhouse gases from

agricultural sources in Europe. IIASA interim REPORT IR-04-048, International Institute for

Applied Systems Analysis.

Koopman, T. C. (1947). Optimum utilization of the transportation system. Proceedings of theinternational statistical conferences Vol V. Washington, D.C.

Ma, H., Huang, J., Rozelle, S., & Rae, A. (2004). Chinese animal products consumption in the 1990s.

Australian Journal of Agricultural and Resource Economics, 48(4), 569–590.

Menzi, H. (2002). Estimation of excretions and average weight of different livestock categories in Asiaand Latin America: Methodology and results. Internal Working Paper prepared for FAO/LEAD.

Version February 13, 2000.

NuFlux. (2001). NuFlux-AWI, user manual for the NuFlux-AWI nutrient balance calculation program.

Main Manual, Zollikofen: Swiss College of Agriculture (SCA).

Popul Environ (2012) 33:243–258 257

123

Shi, X. Z., Yu, D. S., Warner, E. D., Pan, X. Z., Petersen, G. W., Gong, Z. G., et al. (2004). Soil database

of 1:1,000,000 digital soil survey and reference system of the Chinese genetic soil classification

system. Soil Survey Horizons, 45, 129–136.

Steinfeld, H., De Haan, C., & Blackburn, H. (1997). Livestock-environment interactions: Issues andoptions. General for Development: Report to European Commission Directorate.

Toth, F., Cao, G.-Y., & Hizsnyik, E. (2003). Regional population projections for China. IIASA Interim

Report IR-03-042, International Institute for Applied Systems Analysis.

Toth, F., Cao, G.-Y., & Hizsnyik, E. (2008). Regional population projections for China—Revision 2008.

International Institute for Applied Systems Analysis.

Velthof, G. L., Oudendag, D. A., Witzke, H. P., Asman, W. A. H., Klimont, Z., & Oenema, O. (2008).

Integrated assessment of nitrogen emissions from agriculture in EU-27 using MITERRA-EUROPE.

Journal of Environmental Quality, 38, 1–16.

Wang, X. (2005). Diffuse pollution from livestock production in China. Chinese Journal ofGeochemistry, 24/2, 189–193.

Winiwarter, W. (2005). The GAINS model for greenhouse gases—Version 1.0: Nitrous oxide. IIASA

Interim Report IR-05-55, International Institute for Applied Systems Analysis.

258 Popul Environ (2012) 33:243–258

123