Article pubs.acs.org/est
Modeling Environmental Impacts of Urban Expansion: A Systematic Method for Dealing with Uncertainties Yi Liu,*,† Sheng Yang,† and Jining Chen† †
School of Environment, Tsinghua University, Beijing 100084, China S Supporting Information *
ABSTRACT: In a rapidly transitioning China, urban land use has changed dramatically, both spatially and in terms of magnitude; these changes have significantly affected the natural environment. This paper reports the development of an Integrated Environmental Assessment of Urban Land Use Change (IEA-ULUC) model, which combines cellular automata, scenario analysis, and stochastic spatial sampling with the goal of exploring urban land-use change, related environmental impacts, and various uncertainties. By applying the IEA-ULUC model to a new urban development area in Dalian in northeastern China, the evolution of spatial patterns from 1986 to 2005 was examined to identify key driving forces affecting the changing trajectories of local land use. Using these results, future urban land use in the period 2005−2020 was projected for four scenarios of economic development and land-use planning regulation. A stochastic sampling process was implemented to generate industrial land distributions for each land expansion scenario. Finally, domestic and industrial water pollution loads to the ocean were estimated, and the environmental impacts of each scenario are discussed. The results showed that the four urban expansion scenarios could lead to considerable differences in environmental responses. In principle, urban expansion scenarios along the intercity transportation rail/roadways could have higher negative environmental impacts than cluster-developing scenarios, while faster economic growth could more intensely aggravate the environment than in the moderate growth scenarios. flexibility to model and visualize complex, spatially distributed processes at larger spatial scales.8−12 However, due to the weak connections between urban land-use changes and the underlying driving forces, there is a need to integrate multiscale socioeconomic factors into land-use modeling to improve forecasting validity.1,13 Recently, there has been a trend to examine urban environmental effects, including pollution emissions,14−17 soil erosion,18,19 noise,20 and biodiversity and ecological deterioration,21−23 as well as public health 24 in conjunction with landuse modeling. In most of these studies, statistical analysis was applied to acquire the correlation between land use and its environmental consequences. However, due to the absence of integrated methods, the causal relationship between urban sprawl and environmental responses has not been systematically quantified. The complex city-environment system has various inherent and significant uncertainties. In principle, there are at least two major sources of uncertainty that underlie urban land use and potential environmental impacts. First, urban expansion is
1. INTRODUCTION China has been experiencing rapid urbanization that has not been observed elsewhere in the world. According to the sixth national census in China, the population reached 1.34 billion in 2010, and nearly half of the citizens (i.e., 49.68%) resided in 654 cities and 19,322 towns. The urbanization rate has dramatically increased to 1.36% annually since 2000, which implies that over 10 million rural residents become urban residents each year. Considering the impressively rapid growth of the Chinese economy, in which cities and towns have contributed approximately 60% of the total national industrial output, it is not surprising that urban areas have dramatically expanded as the urban population and economy have grown. According to urban construction data, built urban areas have increased at an annual rate of 6.1% from 2000 to 2009, for total growth of 80.8%. Modeling urban land-use change is becoming a priority in the field of urban study. Recently, various agent-based models,1,2 cellular automata models, and application software, such as What-If 3 and UrbanSim,4 have been developed. These programs focus either on individual behaviors between buyers, developers, and governments at the microlevel (i.e., urban block)5,6 or on the macroscale dynamics of urban spatial patterns and underlying societal driving forces.7 The cellular automata (CA) method has become popular for its ability and © 2012 American Chemical Society
Received: Revised: Accepted: Published: 8236
February 27, 2012 June 9, 2012 July 9, 2012 July 9, 2012 dx.doi.org/10.1021/es300766a | Environ. Sci. Technol. 2012, 46, 8236−8243
Environmental Science & Technology
Article
study area. The transportation effect of rail/roadways and the concentrated effects of industrial agglomeration and business centralization were accounted for. Microscale driving forces have localized effects on specific cells and neighbors. The genetic effect and neighborhood effect were applied as two major factors at the microlevel for local infrastructure development and land-use density. The genetic effect describes the cell’s transition probability from the current state into another land-use type, and the neighborhood effect describes how neighbor cells affect the land-use type transitions. The cellular transition probability P is calculated via the “weight of evidence” and can be expressed as
affected by multiple agents including government, developers, investors, individuals, and others. Various driving forces have played different roles over time. Although there is strong governmental intervention in urban planning, future urban sprawl cannot be precisely predicted. In this regard, urban expansion studies are often combined with scenario analysis approaches. Second, there are difficulties in addressing the environmental impacts of urban spatial growth. For example, the environmental impact differs for various industrial sectors. This impact is often represented by the pollution emission load per unit of economic output or product quantity. However, most studies on urban expansion have treated various industrial lands as one single land use type. This simplification leads to a gap between urban studies and impact assessments. Therefore, there is an urgent need to connect various urban expansion scenarios to their potential environmental impacts. In this study, a cellular automata model was developed to examine the causal relationship between various driving forces and urban expansion during different development periods. The findings were applied to different sprawl scenarios. To identify various environmental consequences, a spatial stochastic sampling process was applied to generate plausible industrial development layouts by taking various industrial sectors into consideration. Statistical analyses were conducted to identify various environmental impacts based on given urban expansion scenarios. The article is organized as follows. An integrated methodology that combines CA modeling, scenario development, and uncertainty analysis is discussed in section 2. In section 3, the application of the methodology to a comprehensive case study of a newly developing area of Dalian Municipality in northeastern China is described. The results and conclusions are presented in the final section.
P = CLandsuit × CTime × (α1WInherit + α2WNeighbor + α3WTraffic + α4WCBD + α5WIDC) × IPlan × IPolicy × RI 5
∑ αi = 1 αi ≥ 0 i=1
(1)
where CLandsuit, CTime, IPlan, and IPolicy denote the land use suitability, current land-use duration restriction, planning intervention, and policy governance, respectively, which are the four macroscale driving forces; WTraffic, WCBD, and WIDC are the roadway transportation effect, the business centralization effect and the industrial agglomeration effect, respectively, which are the three mesoscale driving forces; WInherit and WNeighbor are the genetic effect and neighborhood effect, respectively, which are the microscale driving forces; RI is the random perturbation, which must be calibrated; and αi (i = 1, 2, ..., 5) is the weight coefficient, where the values sum to 1. The Matching Rate (MR) and Kappa coefficient are often applied as criteria to examine the fitness of CA model output to the real world.29,30 The MR is defined as the percentage of the total number of simulated cells that have the same land-use types as those found in the actual study area. The Kappa coefficient, widely used in remote sensing studies, was developed based on a confusion matrix to assess the accuracy level of land-use classification (cf. the SI). To calibrate all of the parameters, data for at least two different times are needed. Because the driving forces change over time, the more historical data that are available on land use and socioeconomic development the better one can understand the uncertainty of the parameters. The computing efficiency of a CA model, which is determined by the number of cells and numerical algorithms, is another crucial issue. The size of each cell normally ranges from 50 × 50 m to 200 × 200 m, based on the territorial magnitude of the study area. For these sizes, Monte Carlo methods31,32 and more efficient sampling methods (e.g., Latin Super Cubic Sampling33,34) are highly desired. 2.2. Uncertainty Analysis of Newly Developed Industrial Land. Because the environmental impact of different industrial sectors varies significantly, it is necessary to depict various industrial layouts. This approach is especially important for newly developed areas where the goal is to quantify the environmental impacts. However, due to the prevalence of inconsistent data on the classification of urban and industrial lands, most ordinary surveys of land use do not provide a sophisticated data set of various industrial land uses in urban areas.35 Therefore, the ILL-UA module (a specific version of the SIMULAND model described by Liu et al.36)
2. METHODS An Integrated Environmental Assessment of Urban Land Use Change Model (IEA-ULUC) was developed to quantitatively assess the environmental impacts of urbanization via a comprehensive, analytical process that includes forecasting urban land-use changes, allocating newly developed industrial lands, and identifying environmental response. The model consists of three subsystems: an urban expansion module based on CA (ULE-CA), an industrial land-use module based on a spatially stochastic sampling and uncertainty analysis (ILL-UA), and an environmental assessment module that is based on emissions per unit of land and a statistical analysis (EIA-LU). These subsystems are discussed in the following sections. 2.1. An Urban Expansion Module Based on Cellular Automata. A CA model is a discrete, dynamic system in which the study area is divided into regular spatial units called cells. The state or value of the cells may change over time (a year is often the base unit). Each cell has one of a finite number of states, which are regularly updated in simulation loops that rely on the current state and the states of neighbor cells at the previous time step. Determination of the transition rules, which can be quantified by analysis of the underlying driving forces,25−28 is crucial for CA modeling. Macroscale driving forces function at large spatial scales and are highly dependent on socioeconomic factors. This study considers land-use planning, governmental spatial regulations, the suitability of land for construction, and property tenure. Mesoscale driving forces that play an active role in certain spatial scope have moderate impacts over the 8237
dx.doi.org/10.1021/es300766a | Environ. Sci. Technol. 2012, 46, 8236−8243
Environmental Science & Technology
Article
mental overload. The HSY algorithm was originally developed for application to poorly defined systems with insufficient information.38−40 Different from seeking for the unique best fittingness between the model simulation outputs and the reality which becomes less promising as the complexity of environmental system models dramatically increases, the method recognizes the variation of solutions of a certain model to approximate the observations.36 Therefore, identification of parameter sensitivity, i.e., differences of statistical configurations of model parameters in acceptable and unacceptable simulations given certain criterion of system behavior, becomes predominant in environmental system modeling.41
was developed to generate possible industrial land-use scenarios based on a spatial stochastic process. The stochastic process is implemented in the cells where the land-use type may be changed into industrial land based on the ULE-CA outputs. The generation of various layouts on the new industrial lands is a pseudorandom process. The governmental preference concerning future industrial land development, especially on some large industrial projects and priority areas, provided in urban master plans and industrial development plans, are considered in the process. The probability of each cell developing a specific industrial sector advocated by spatial regulatory instruments can be calculated as Wpy =
∑ θPy(i) × PPy(i)
(2)
3. CASE STUDY 3.1. Study area. Dalian, located at the southern tip of the Liaodong Peninsula, is an economic hub and the biggest port city in northeastern China. From 2000 to 2010, urban areas and the industrial added value increased annually by 5.6% and 21.7%, respectively. In 2010, the Gross Domestic Product (GDP) per capita reached 11,700 United States dollars, almost 5 times the value in 2000. The petrochemical industry, the electronics industry, machine manufacturing, transport equipment manufacturing, the refined chemical industry, steel refinement, the textile industry, food processing, the cement industry, light industry, and the electric power industry are the 11 pillar industries which were examined in this study. To remove the geographical barriers to urban growth, the local government has advocated an ambitious spatial development plan since 2003. The Jinzhou region, to the north of the city center, has been promoted as the largest new urban development area. The territory of the Jinzhou region spans 1,360 km2 and is divided into six watersheds that flow into Dalian Bay, Dayaowan Bay, Xiaoyaowan Bay, Pulandian Bay, Jinzhou Bay, and Eastern Offshore Areas (Figure S1 in the SI). According to local monitoring data, inorganic nitrogen has been the primary pollutant affecting offshore water quality. Therefore, this study adopted inorganic nitrogen as the critical environmental indicator to illustrate the application of the IEAULUC model. 3.2. Urban Land Expansion. Based on a balance between computing efficiency and accuracy, the study area was divided into 33,981 grids with 200 × 200 m cells. For each cellular neighborhood expansion area, a Moore neighborhood (a square space composed of 5 × 5 cells in which each central cell is surrounded by 24 neighbor cells) was adopted. Six land-use types were taken into consideration in this study. These types were based on the Standard for Land Use Classification (GB/T 21010-2007) and included built urban land, industrial land, rural residential land, green space (including forestry, grass, and garden lands), cultivated land, and water space. GIS-based data on topography, regional land use, and intercity transportation as well as hydro-meteorological data and data concerning natural and water conservation areas in 1986, 1995, and 2005 were collected. This study considered the regulatory priorities and official government preferences on spatial expansion that were found in the regional land-use and urban development plans. The topographical limitation, as illustrated by the land construction feasibility,42 was regarded as a constant constraint during the study period (Figure S2 in the SI). On the mesoscale, the effective distances for the railway (the Shenyang-Dalian railway), highways (the Shenyang-Dalian and Dalian-Dandong highways), industrial agglomerations (the
where WPy is the policy intervention weight; PPy(i) denotes the governmental preference provided in the plan i concerning the future spatial development of each industrial sector; and θPy(i) represents the weight for the regulatory power of each plan. According to Weber’s industrial location theory, transportation, labor resources, and industrial clusters have significant effects on the spatial agglomeration of industrial development. Based on an input-output table, the interactive relationship between an industrial sector and the location factors can be quantified.37 The new industrial cells are defined by the ULE-CA module; whereas, the probability of transition of a certain industrial sector located in an industrial cell WLt can be analyzed as WLt = UTrans × PTrans + ULabor × PLabor + UIndcl × PIndcl (3)
where PTrans, PLabor, and PIndcl represent the effects of rail/ roadways transportation, labor provision, and existing industrial agglomerations, respectively, and UTrans, ULabor, and UIndcl represent their corresponding weights. The probability of transition of a new industrial land to a certain industrial sector can be calculated as WInd . Layout = θPy ln WPy′ + θLt ln WLt′
(4)
where θPy denotes the weight of policy/planning intervention, θLt denotes endogenous dynamics of industrial development, WPy′ is the standardized value of WPy, and WLt′ is the standardized value of WLt. The probability of transition for each industrial sector in a cell is a number on the interval [0,1]. Given appropriate convergence criteria (e.g., the average industrial output in all of the cells achieves an acceptable, stable level), the stochastic process will terminate. Thereafter, the artificially generated layouts can be regarded as approximate representations of all of the possibilities of the spatial distribution of newly developed industries.35,36 2.3. Identification of the Environmental Impacts of Urban Land Change. The EIA-LU module provides a statistical analysis of potential environmental overload and identification of key industrial sectors that are likely to have a significant environmental impact. Using emission coefficients for each industrial sector, a grid-based pollution emission value can be calculated based on the various industrial land layout scenarios generated by the ILL-UA module. Because the newly developed industrial land could vary significantly spatially, the pollution emission value is expressed in a statistical manner. Furthermore, a HSY algorithm is applied to identify key industrial sectors that are responsible for potential environ8238
dx.doi.org/10.1021/es300766a | Environ. Sci. Technol. 2012, 46, 8236−8243
Environmental Science & Technology
Article
Figure 1. Urban expansion scenarios in 2020.
The values of the best-fit key parameters, including the five weighting coefficients, the two intervention factors and the random disturbance as described in eq 1, varied significantly between the two modeling periods (Tables S1 and S2 in the SI). In principal, the mesoscale effects of transportation rail/ roadways, industrial and business centralization have become more influential on new urban development, while microscale effects of genetic and neighborhood transitions have weakened as the average weight coefficients decreased from 0.403 to 0.302. This implies investment on infrastructures and industrial development has played an increasingly active role in stimulating urban expansion during the second period. Given the fact that China’s urbanization rate was considerably higher during the second period, the local governments, by means of policy instruments and spatial planning, showed an impaired ability to direct urban growth (additional descriptions of local urbanization trends and implications from the simulation results can be found in Tables S3 and S4 in the SI). However, the variation among the model parameters suggests inherent uncertainties of the complex urban system and that significant, if not insurmountable, problems exist in model calibration and validation. It is acknowledged that uncertainties are universal and sometimes dominate a system-
Gangjingzi district and the Dagushan industrial development zone), and business centers (the Gangjingzi district and the town of Jinzhou) were 5 km, 10 km, 30 km, and 30 km, respectively. The threshold of the centralized effect was calculated based on a GIS platform (Figure S3 in the SI). The genetic matrix and the neighborhood matrix were calculated according to historical data (Figure S4 in the SI). The ULE-CA module was operationalized independently for two periods (1986−1995 and 1996−2005) with the goal of better understanding the interrelations between urban expansion and various underlying driving forces. A Monte Carlo sampling process accelerated by a Latin Super Cubic Sampling technique was applied to search for the ‘best-fit’ outputs. The Monte Carlo process was implemented 20,000 times, and 140,000 samples were generated, which guaranteed an acceptable convergence (the change in sample values was lower than 1% whenever an increase in sample numbers). The simulation results suggested moderately good approximations to reality (Figure S5 in the SI). The MR was 90.6% and 93.0% for the period 1986−1995 and 1996−2005, respectively. The Kappa coefficient also reached an acceptable level for both periods: 0.853 for 1986−1995 and 0.896 for 1996−2005. 8239
dx.doi.org/10.1021/es300766a | Environ. Sci. Technol. 2012, 46, 8236−8243
Environmental Science & Technology
Article
modeling process.43−46 Because it is difficult to verify whether the best-fit parameters are reliable for forecasting land use, this study adopted a scenario analysis approach to develop typical urban expansion layouts based on historical data of the dynamics of urban land-use changes. 3.3. Scenarios of Urban Expansion. Two probable urban growth patterns were identified as a basis for constructing future urban land-use scenarios. The most likely models are ring and cluster-based expansion and intercity transportoriented expansion. A second consideration was the influence of local policy and spatial planning. The urban growth rate was also taken into consideration as the third factor that affects the demand for urban residential and industrial land. The effects at micro- and mesoscales and governmental intervention were used to account for the differences in the spatial development patterns. The determination of the key parameter values is discussed in the Supporting Information. Four urban expansion scenarios were developed for the period from 2005 to 2020: • Scenario I: ring and cluster-based expansion at a moderate growth rate under high governmental intervention; • Scenario II: ring and cluster-based expansion at a high growth rate under moderate governmental intervention; • Scenario III: intercity transport-oriented expansion at a moderate growth rate under high governmental intervention; • Scenario IV: intercity transport-oriented expansion at a high growth rate under moderate governmental intervention. The results suggested a significant variety of spatial expansion of the Jinzhou region by 2020, as shown in Figure 1. All of the scenarios showed rapid sprawl in the western areas but not in the eastern areas. The development of the two cluster-based scenarios would mainly be concentrated in the western coastal areas including the towns of Daweijia and Ershilipu, which are adjacent to the current business center in the south. The development of the intercity transport-oriented scenarios would claim more land in the town of Shanshilipu in the northwestern part of the region along the Shenyang-Dalian railway and highway, in the town of Dengshahe and near the eastern coastal areas along the Dalian-Dandong highway. Intercity transport-oriented development has facilitated an increase of the urban lands. As shown in Table 1, given the
IV. New development of industrial land could dominate future urban expansion. 3.4. Quantifying the Environmental Impact of the Urban Expansion. Mapping industrial development to the newly urbanized land is essential to quantifying the environmental impact of urban expansion. Given the land use intensity and the pollution emission intensity data for different industrial land types,36 the ILL-UA module was implemented to generate various industrial layouts over the newly developed urban land. Two assumptions were made to operationalize the ILL-UA module. First, the amount of industrial lands in all scenarios was made to match the industrially added value proposed by the local planning organization. Second, the land-use intensity of different industrial sectors was equally improved during future development. Consequently, there were 5,000 land-use layouts by industrial sector for each scenario generated (refer to Figure S6 in the SI for the spatial distribution probability of the petrochemical industry as an example). The convergence of the sampling process was guaranteed via a K−S two-tail test at a confidence threshold of 0.1. In agreement with the Standards for Urban Land Use Classification and Planning Construction (GBJ137-90), this study applied 100 m2 as a requirement for per capita urban residential land uses. Applying domestic water consumption of 255 m3 per capita and an emission coefficient of 0.8, the amount of domestic wastewater was calculated to be 930,800 m3 km−2 each year. The total inorganic nitrogen emissions (domestic and industrial) were estimated by using industrial land-use intensity, industrial wastewater discharge (Tables S5 and S6 in the SI), the capacity of urban sewage infrastructure and discharge permits for wastewater treatment plants (GB8978-1996). The uncertainty of the impacts on offshore water quality, due to the variety of urban expansion and industrial spatial development, were quantified by the overload probability of emission quantities to the environmental permits (as indicated by carrying capacity) as illustrated in Figure 2 (cf. Table S7 in the SI for tabulated statistical results). The results showed that the Xiaoyaowan Bay would become the most polluted offshore area because in all four of the scenarios the emission quantities of inorganic nitrogen will overload the local environmental permits. The maximum overload ratio could rise to approximately 6 times greater than the environmental permits allow. This fact strongly suggested that this area is clearly not a priority for development. The Pulandian Bay and the Eastern Offshore Areas are likely to have the second worst water quality. The emission quantities will exceed the local permits in Scenarios II−IV. Even in Scenario I, the overload probabilities (51.8% and 41.8%), indicated by the percentage of the number of environmental unaccepted samples to the total samples, will be significantly high (Table S7 in the SI). While the water quality of the Dayaowan and Jinzhou Bays will be maintained, the quality of the Dalian Bay could become worse, especially for Scenarios II and IV. In contrast, Scenario I is the least environmentally intense development model and is recommended as the highest priority for future urban growth. Furthermore, the key industries that most severely threaten the offshore water quality can be identified. This identification can aid in implementing precautionary pollution controls in future urban expansion and industrial development regions. It is apparent that strict spatial regulation of rapid urban expansion, especially industrial development in Xiaoyaowan Bay area and
Table 1. Urban and Industrial Land Areas in 2020 for Four Scenarios (km2) Scenario Scenario Scenario Scenario
I II III IV
built urban land
industrial land
total urban areas
97.5 109.6 96.9 113.8
122.8 143.6 134.6 196.7
220.3 253.2 231.5 310.5
same intervention efforts and economic growth rate, the amount of industrial land could increase by 5% compared with Scenarios I and III and by 22.6% compared with Scenarios II and IV. Furthermore, the results also supported the idea that faster economic growth and less governmental intervention efforts lead to more rapid urban expansion. Given the same expansion model, the amount of urban lands increased by 14.9% from Scenarios I to II and 34.1% from Scenarios III to 8240
dx.doi.org/10.1021/es300766a | Environ. Sci. Technol. 2012, 46, 8236−8243
Environmental Science & Technology
Article
Figure 2. Probability distribution of emission quantity of inorganic nitrogen versus environmental capacity for the urban expansion scenarios in 2020.
3.5. Discussion. The IEA-ULUC model was developed as a bottom-up system based on land use to explore the potential environmental impacts of urban expansion in a systematic manner. The ULE-CA module considers three driving forces: spatial regulation on the macroscale, location and distance effects on the mesoscale, and neighborhood and genetic effects on the microscale. An industrial land allocation module (ILLUA) was introduced to generate new industrial land distributions based on the CA-generated land-use layout. This approach is of particular importance for China and many newly developing economies in which industrialization, as it is strongly coupled with and sometimes even dominates the rapid urbanization process, has heavily influenced natural environments. The EIA-LU module was used to assess environmental impacts or potential risks via statistical methods
rapid sprawl of urban residential land and industrial land in Pulandian Bay area and the Eastern Offshore areas (cf. Table S8 in the SI), is required for those scenarios in which the emissions exceed the environmental permits at a 100% probability. Adjustments of industrial structure are likely to remain promising for the lower pollution scenarios, including Pulandian Bay and the Eastern Offshore Areas in Scenario I and Dalian Bay in Scenarios II and IV in which industries will significantly contribute to the total pollution emission (Table S8 in the SI). Applying the HSY algorithm and K−S test approach, it is found that the petrochemical, refined chemical, and electric power industries will be the key sectors (Table S9 in the SI); therefore, limiting the scale of these three key industries will largely determine local offshore water quality in the future. 8241
dx.doi.org/10.1021/es300766a | Environ. Sci. Technol. 2012, 46, 8236−8243
Environmental Science & Technology
Article
simulation of multi-actor spatial planning. J. Environ. Manage. 2004, 72 (1−2), 43−55. (3) Klosterman, R. E. The what if? collaborative planning support system. Environ. Plann. B 1999, 26, 393−408. (4) Kakaraparthi, S. K.; Kockelman, K. M. Application of UrbanSim to the Austin, Texas, Region: Integrated-Model Forecasts for the Year 2030. J. Urban Plann. Dev. 2011, 137 (3), 238−247. (5) Esbah, H. Land use trends during rapid urbanization of the City of Aydin, Turkey. Environ. Manage. 2007, 39 (4), 443−459. (6) Fox, J.; Rindfuss, R. R.; Walsh, S. J.; Mishra, V. People and the environment: Approaches for Linking Household and Community Surveys to Remote Sensing and GIS; Kluwer: Boston, 2002. (7) Verburg, P. H.; Veldkamp, A.; Willemen, L.; Overmars, K. P.; Castella, J. C. Landscape level analysis of the spatial and temporal complexity of land-use change. In Ecosystems and Land Use Change; De Vries, R.,Houghton, R., Eds.; American Geophysical Union: Washington, DC, 2004. (8) Batty, M.; Xie, Y.; Sun, Z. Modelling urban dynamics through GIS-based cellular automata. Comput. Environ. Urban Syst. 1999, 23 (3), 205−233. (9) Clarke, K. C.; Hoppen, S.; Gaydos, L. J. A self-modifying cellular automata model of historical urbanization in the San Franciso Bay area. Environ. Plann. B 1997, 24, 247−261. (10) Braimoh, A. K.; Onishi, T. Spatial determinants of urban land use change in Lagos, Nigeria. Land Use Policy 2007, 24 (2), 502−515. (11) Liu, Y.; Phinn, S. R. Modelling urban development with cellular automata incorporating fuzzy-set approaches. Comput. Environ. Urban Syst. 2003, 27 (6), 637−658. (12) Veldkamp, A.; Verburg, P. H. Modelling land use change and environmental impact. J. Environ. Manage. 2004, 72 (1−2), 1−3. (13) Kok, K. The role of population in understanding Honduran land use patterns. J. Environ. Manage. 2004, 72 (1−2), 73−89. (14) Chen, J.; Avise, J.; Guenther, A.; Wiedinmyer, C.; Salathe, E.; Jackson, R. B.; Lamb, B. Future land use and land cover influences on regional biogenic emissions and air quality in the United States. Atmos. Environ. 2009, 43 (36), 5771−5780. (15) Mukerjee, S.; Smith, L. A.; Johnson, M. M.; Neas, L. M.; Stallings, C. A. Spatial analysis and land use regression of VOCs and NO2 from school-based urban air monitoring in Detroit/Dearborn, USA. Sci. Total Environ. 2009, 407 (16), 4642−4651. (16) Zhao, N.; Liu, Y.; Chen, J. Regional industrial production’s spatial distribution and water pollution control:A plant-level aggregation method for the case of a small region in China. Sci. Total Environ. 2009, 407 (17), 4946−4953. (17) Polebitski, A.; Palmer, R. Impacts of climate change and growth on water demands in the Puget Sound Region. In World Environmental and Water Resources Congress 2010: Challenges of Change, Providence, RI, United States, 2010. (18) Zorn, M.; Petan, S. Interrill soil erosion on flysch soil under different types of land use in Slovenian Istria. In 24th Conference of the Danubian Countries on the Hydrological Forecasting and Hydrological Bases of Water Management, Bled, Slovenia, 2008. (19) Gu, S. The study on the relationship between land use and soil erosion in the Hengduan mountainous region of the Eastern Tibet. In 2011 International Conference on Remote Sensing, Environment and Transportation Engineering, Nanjing, China, 2011. (20) Xie, D.; Liu, Y.; Chen, J. Mapping Urban environmental noise: A land use regression method. Environ. Sci. Technol. 2011, 45 (17), 7358−7364. (21) de Koning, G. H. J.; Benítez, P. C.; Muñoz, F.; Olschewski, R. Modelling the impacts of payments for biodiversity conservation on regional land-use patterns. Landscape Urban Plann. 2007, 83 (4), 255− 267. (22) Trisurat, Y.; Alkemade, R.; Verburg, P. H. Projecting land-use change and its consequences for biodiversity in northern thailand. Environ. Manage. 2010, 45 (3), 626−639. (23) Polasky, S.; Nelson, E.; Pennington, D.; Johnson, K. A. The impact of land-use change on ecosystem services, biodiversity and
while considering the various uncertainties in the previous two modules. This integrated methodology overcomes, to a certain degree, the inherent weakness of most current land-use studies, which is the failure to quantitatively link urban expansion to environmental response. The IEA-ULUC model enables a systematic identification of environmentally sensitive areas and the key industries most likely to affect those areas before significant expansion occurs. This early identification is one of the central concerns in China when developing urban spatial development strategies, such as urban planning, local land-use planning, and regional industrial development planning. Various uncertainties in urban land-use change were systematically analyzed in this paper, including industrial distribution changes that can dominate urban growth in China. Although the potential environmental impacts caused by each one of the uncertainties were not discussed in depth, it is clear that smart urban governance should move toward regulating future environmental impacts by avoiding the unexpected urban sprawl scenarios at a planning level. It was found that the ring and cluster-based urban expansion model in conjunction with strict permits on the key industries most prone to damaging the environment appear to be the most promising in dealing with environmental pollution. However, not all uncertainties could be accounted for in this study. Some factors that may change over time and may considerably affect the conclusions, including variations of the generalized land use buffer area based on Weber’s theory, landuse efficiency, resource consumption, and pollution emission per residential unit or industrial land, were not taken into consideration as model input uncertainties. It is notable that the changes in cell size and neighborhood style and size have considerable impact on a CA model’s outputs.47,48 Furthermore, if an environmental quality model is introduced, the uncertainties can be expected to increase significantly. There is also a need to explore and integrate other environmental restricting factors, i.e. offshore phosphates emission permit and urban air pollution, which might have different responses to and form different constraints of the urban sprawl scenarios.
■
ASSOCIATED CONTENT
S Supporting Information *
A detailed description of the IEA-ULUC model, extensive figures and tables that are related to the background information of this case study, and extensive discussions of interest mainly to specialists. This material is available free of charge via the Internet at http://pubs.acs.org.
■
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Notes
The authors declare no competing financial interest.
■ ■
ACKNOWLEDGMENTS The authors thank the anonymous reviewers for their instructive comments. REFERENCES
(1) Aspinall, R. Modelling land use change with generalized linear modelsa multi-model analysis of change between 1860 and 2000 in Gallatin Valley, Montana. J. Environ. Manage. 2004, 72 (1−2), 91−103. (2) Ligtenberg, A.; Wachowicz, M.; Bregt, A. K.; Beulens, A.; Kettenis, D. L. A design and application of a multi-agent system for 8242
dx.doi.org/10.1021/es300766a | Environ. Sci. Technol. 2012, 46, 8236−8243
Environmental Science & Technology
Article
returns to landowners: A case study in the state of Minnesota. Environ. Resour. Econ. 2011, 48 (2), 219−242. (24) Corburn, J. Urban land use, air toxics and public health: Assessing hazardous exposures at the neighborhood scale. Environ. Impact Assess. Rev. 2007, 27 (2), 145−160. (25) Alonso, D.; Solé, R. V. The DivGame simulator: A stochastic cellular automata model of rainforest dynamics. Ecol. Modell. 2000, 133 (1−2), 131−141. (26) Pinto, N. N.; Antnes, A. P. Cellular automata and urban studies: a literature survey. Archit. City Environ. 2007, 1 (3), 368−399. (27) Li, X.; Yeh, A. G. Neural-network-based cellular automata for simulating multiple land use changes using GIS. Int. J. Geog. Inf. Sci. 2002, 16 (4), 323−343. (28) M, T. P.; Sullivan, D. O. Cellular automata and urban simulation: Where do we go from here? Environ. Plann. B 2001, 28, 163−168. (29) van Vliet, J.; Bregt, A. K.; Hagen-Zanker, A. Revisiting Kappa to account for change in the accuracy assessment of land-use change models. Ecol. Modell. 2011, 222 (8), 1367−1375. (30) Kim, J. O.; Yu, K.; Heo, J.; Lee, W. H. A new method for matching objects in two different geospatial datasets based on the geographic context. Comput. Geosci. 2010, 36 (9), 1115−1122. (31) Melchers, R. E.; Ahammed, M. A fast approximate method for parameter sensitivity estimation in Monte Carlo structural reliability. Comput. Struct. 2004, 82 (1), 55−61. (32) Kuczeraa, G.; Parent, E. Monte Carlo assessment of parameter uncertainty in conceptual catchment models: the Metropolis algorithm. J. Hydrol. 1998, 211 (1−4), 69−85. (33) Wyss, G. D.; Jorgensen, K. H. In A User’s Guide to LHS: Sandia’s Latin Hypercube Sampling Software, Sandia; Sandia National Laboratories: 1998. (34) Owen, A. B. Latin supercube sampling for very high-dimensional simulation. ACM Trans. Model Comput. Simul. 1998, 8 (1), 71−102. (35) Wu, H. Integrated Environmental Assessment Methodology and Case Study based on Urban Land Use Simulation. Master Thesis, Tsinghua University, Beijing, China, 2008. (36) Liu, Y.; Chen, J.; He, W.; Tong, Q.; Li, W. Application of an Uncertainty Analysis Approach to Strategic Environmental Assessment for Urban Planning. Environ. Sci. Technol. 2010, 44 (8), 3136−3141. (37) Yan, X.; Liu, Y.; Chen, J.; Wu, H. The Impact of Road Transportation system on regional industrial distribution. Econ. Geogr. 2009, 29 (5), 746−751. (38) Young, P. C.; Hornberger, G. M.; Spear, R. C. Modeling badly defined systems-some further thoughts. In Proceedings SIMSIG Simulation Conference; Australian National University: Canberra, Australia, 1978; pp 24−32. (39) Hornberger, G. M.; Spear, R. C. Eutrophication in Peel Inlets I. The problem-defining behavior and a mathematical model for the phosphorus scenario. Water Res. 1980, 14, 29−42. (40) Spear, R. C.; Hornberger, G. M. Eutrophication in Peel Inlet II. Identification of critical uncertainties via generalized sensitivity analysis. Water Res. 1980, 14, 43−49. (41) Beck, M. B. Water quality modeling: a review of the analysis of uncertainty. Water Resour. Res. 1987, 23 (8), 1393−1442. (42) Liu, Y.; Li, T.; Chen, J.; Zhang, L.; Wang, W.; Tong, Q.; Lv, C. Ecologically feasible space for urban development: A methodology and case study on Dalian Municipality, Northeastern China. China Environ. Sci. 2007, 27 (1), 34−38. (43) Oreskes, N.; Shrader-Frechette, K.; Belitz, K. Verification, Validation, and Confirmation of Numerical Models in the Earth Sciences. Science 1994, 263 (5147), 641−646. (44) Rockafellar, R. T.; Wets, R. J. B. Scenario and policy aggregation in optimization under uncertainty. Math. Oper. Res. 1991, 16 (1), 119−147. (45) Christensen, K. S. Coping with Uncertainty in Planning. J. Am. Plann. Assoc. 1985, 51 (1), 63−73. (46) Beck, M. B. Environmental Foresight and Models: A Manifesto; Elsevier: Amsterdam, 2002.
(47) Ménard, A.; Marceau, D. J. Exploration of spatial scale sensitivity in geographic cellular automata. Environ. Plann. B 2005, 32 (5), 693− 714. (48) Kocabas, V.; Dragicevic, S. Assessing cellular automata model behavior using a sensitivity analysis approach. Comput. Environ. Urban Syst. 2006, 30 (6), 921−953.
8243
dx.doi.org/10.1021/es300766a | Environ. Sci. Technol. 2012, 46, 8236−8243
