Classifying and mapping the urban transition in Vietnam

Classifying and mapping the urban transition in VietnamThe urban transition almost always involves wrenching social adjustment as small agricultural communities are forced to adjust rapidly to industrial ways of life. Large-scale in-migration of young people, usually from poor regions, creates enormous demand and expectations for community and social services. One im media te problem planners face in approaching this chal lenge is how to define, differentiate, and map what is rural, urban, and transitional (i.e., peri-urban). This project established an urban classification for Vietnam by using national census and remote sensing data to identify and map the smallest administrative units for which data are collected as rural, peri-urban, urban, or urban core. We used both natural and human factors in the quantitative model: income from agriculture, land under agriculture and forests, houses with modern sanitation, and the Normalized Difference Vegetation Index. Model results suggest that in 2006, 71% of Vietnam’s 10,891 communes were rural, 18% peri-urban, 3% urban, and 4% urban core.

Applied Geography 50 (2014) 80e89 Contents lists available at ScienceDirect Applied Geography journal homepage: www.elsevier.com/locate/apgeog Classifying and mapping the urban transition in Vietnam S Saksena a,1, J Fox a, *, J Spencer b, 2, M Castrence a,1, M DiGregorio d, 3, M Epprecht c, N Sultana b, M Finucane a, 4, L Nguyen d, T.D Vien d a East-West Center, 1601 East-West Road, Honolulu, HI 96848, USA Department of Urban and Regional Planning, University of Hawaii, Manoa, Honolulu, HI 96822, USA c Centre for Development and Environment, University of Bern, Switzerland d Hanoi University of Agriculture, Hanoi, Vietnam b a b s t r a c t Keywords: Urban transition Rural transition Peri-urban Vietnam GIS Remote sensing The urban transition almost always involves wrenching social adjustment as small agricultural communities are forced to adjust rapidly to industrial ways of life Large-scale in-migration of young people, usually from poor regions, creates enormous demand and expectations for community and social services One immediate problem planners face in approaching this challenge is how to define, differentiate, and map what is rural, urban, and transitional (i.e., peri-urban) This project established an urban classification for Vietnam by using national census and remote sensing data to identify and map the smallest administrative units for which data are collected as rural, peri-urban, urban, or urban core We used both natural and human factors in the quantitative model: income from agriculture, land under agriculture and forests, houses with modern sanitation, and the Normalized Difference Vegetation Index Model results suggest that in 2006, 71% of Vietnam’s 10,891 communes were rural, 18% peri-urban, 3% urban, and 4% urban core Of the communes our model classified as peri-urban, 61% were classified by the Vietnamese government as rural More than 7% of Vietnam’s land area can be classified as peri-urban and approximately 13% of its population (more than 11 million people) lives in peri-urban areas We identified and mapped three types of peri-urban places: communes in the periphery of large towns and cities; communes along highways; and communes associated with provincial administration or home to industrial, energy, or natural resources projects (e.g., mining) We validated this classification based on ground observations, analyses of multi-temporal night-time lights data, and an examination of road networks The model provides a method for rapidly assessing the ruraleurban nature of places to assist planners in identifying rural areas undergoing rapid change with accompanying needs for investments in building, sanitation, road infrastructure, and government institutions Ó 2014 Elsevier Ltd All rights reserved Introduction important are the transformations of the landscape needed for these structural shifts As economies become industrialized and more people are employed in services, the nature of urban and rural areas changes The transition from predominantly rural to increasingly urban economies is one of the great development challenges of the times (Aoyama & Horner, 2010) Urbanization spurs growth and reduces poverty but can also exacerbate inequalities, increase exposure to certain health risks, degrade environmental quality, lead to food insecurity, and have other deleterious effects Managing the rural-to-urban transition in a way that safeguards equitable and sustainable growth is therefore a major concern of the development community Policy makers around the world are looking for ways to manage the urban transition that ensure beneficial outcomes and minimize risk (Dudwick, Hull, Katayama, Shilpi, & Simler, 2011) Simon Kuznets summarized the structural transition that accompanies economic development, emphasizing “the shift away from agriculture to non-agricultural pursuits and.away from industry to services” (Kuznets, 1992, p 89) Less obvious but no less * Corresponding author Tel.: ỵ1 808 944 7111 E-mail address: foxj@eastwestcenter.org (J Fox) Tel.: ỵ1 808 944 7111 Present address: Planning Department Clemson University Clemson, SC 29634, USA Tel.: ỵ1 864 656 1208 Present address: Asia Foundation Hanoi, Vietnam Present address: Rand Organization Pittsburg, PA, USA Tel.: ỵ1 412 683 2300x4279 http://dx.doi.org/10.1016/j.apgeog.2014.02.010 0143-6228/ể 2014 Elsevier Ltd All rights reserved S Saksena et al / Applied Geography 50 (2014) 80e89 81 Fig Methodological workflow One immediate problem planners face in approaching this challenge is how to define, differentiate, and map what is rural, urban, and transitional (i.e., peri-urban) Statistical definitions of “rural” and “urban” vary from country to country (or even within countries) and can be based on administrative boundaries, size, level of services, or population density (Aoyama & Horner, 2010) In reality there is a rural-to-urban continuum, ranging from sparsely populated, isolated settlements to small towns to secondary cities to megacities; and in any given country there is heterogeneity within areas that are classified as rural or urban Whether an administrative unit is classified as rural or urban, however, affects how it is governed and the financial resources allotted for governance Of particular concern is the fact that the governance and management of places that are neither rural nor urban are frequently neglected by both rural and urban administrators because such places either lie beyond urban administrative boundaries and thus lack access to urban resources, or they fall under the administration of cities that lack the financial resources to upgrade the planning and infrastructure of transitional areas The urban transition almost always involves wrenching social adjustment as small agricultural communities are forced to adjust rapidly to industrial ways of life Large-scale in-migration of young people, usually from poor regions, creates enormous demand and expectations for community and social services Environmental stresses in peri-urban areas are also significant due to the patchy nature of newer settlements, pollution from a variety of industrial and residential sources, as well as motorization; and inadequate public-sector financial resources to cope with the rapid development (Webster, 2002) Planners and development agencies badly need methods for collecting and analyzing data that enable them to assess variables along the rural-to-urban continuum and to classify and map areas as rural, urban, and peri-urban Without such methods they cannot estimate how much of the landscape is affected by peri-urbanization; they not know how many peoples’ lives are affected; they not know the extent of the environmental and human-health problems; and they cannot address issues of governance and responsibility Numerous studies have used remotely sensed data to map the extent and rate of urban expansion at local (Li & Yeh, 1998; Schneider, Seto, & Webster, 2005), national (APN, 2009), and global scales (Schneider, Friedl, & Potere, 2009, 2010; Seto et al., 2012) These studies, however, tell us little or nothing about the socioeconomic characteristics of places undergoing the rural-tourban transition or the interaction of rural and urban activities Other studies have used household and community data to map local-scale administrative units according to characteristics such as population size and density, communication and transportation networks, educational facilities, and access to health services and markets These studies map the relative urban nature of a place, a broad concept that is often referred to as “urbanicity” (Dahly & Adair, 2007; Jones-Smith & Popkin, 2010; McDade & Adair, 2001; Van de Poel, O’Donnell, & Van Doorslaer, 2009) While these studies show differences in the rural/urban nature of communities across space and time, they have been limited to small- to mediumscale observational studies Novak, Allender, Scarborough, and West (2012) report a multicountry urbanicity scale for Ethiopia, India, and Peru, but their work does not map changes in urbanicity across national space More significantly, the variables used in these studies largely measured “urban” features and failed to distinguish between different levels of “ruralness.” While some studies have used statistical construct validation (Dahly & Adair, 2007; Novak et al., 2012), none have validated the results of their models through ground truthing The objectives of this study are twofold First, we seek to establish an urban classification by using Vietnamese national census and remote sensing data to identify and map communes, the smallest administrative unit for which data are collected, as 82 S Saksena et al / Applied Geography 50 (2014) 80e89 being rural, peri-urban, urban, or urban core.5 Second, we seek to validate this classification based on ground observations, analyses of multi-temporal night-time lights data, and an examination of road networks Ultimately, we seek to demonstrate methods for classifying and mapping the dynamics of the urban transition in Vietnam Theories of peri-urbanization Friedmann (2011) argues that “a general theory of the periurban.escapes us” and “the best we can is provide ‘thick’ descriptive accounts of the events that transform these places” (p 430) Webster (2011) suggests a divide between those who look at urban transitions from the perspective of peri-urban areas as distinctive areas of agricultural and nonagricultural activities emerging adjacent to and between urban cores (Firman, 2000; McGee, 1991; Winarso, 2011) and those who look at the process from the perspective of rural communities and the changes they are experiencing (Harms, 2011; Leaf, 2002; McGregor, Simon, & Thompson, 2006; Rigg, 2006) Peri-urban areas are characterized by patchwork development and mixed land use, with large amounts of land still in agricultural use McGee’s concept of desakota (Indonesian for “village-town”) is perhaps the best known model of the peri-urbanization process (McGee, 1991) McGee identifies six characteristics of a desakota region: (1) a large population of smallholder cultivators; (2) an increase in nonagricultural activities; (3) extreme fluidity and mobility of population; (4) a mixture of land uses including agriculture, cottage industries, and suburban development; (5) increased participation of the female labor force; and (6) a lack of administrative responsibility (i.e., administrative “grey zones”), which encourages informal and illegal activities (McGee, 1991, pp 16e17) In a similar vein, Webster (2002) lists four characteristics of the peri-urbanization process These include: (1) a shift from an agriculturally based to a manufacturing-dominated economy; (2) a shift in employment from agriculture to manufacturing; (3) rapid population growth; and (4) changing spatial development patterns and rising land costs On the other hand, Rigg (2006) approaches the urban transition from the perspective of farming, which is evident in the label he uses for the processd“deagrarianization.” He argues that deagrarianization is characterized by (1) diversification of rural occupations and livelihoods; (2) occupational multiplicity becoming more common and more pronounced; (3) balance of household incomes shifting from farm to non-farm; (4) livelihoods and poverty becoming delinked from land (and from farming); (5) lives becoming more mobile and livelihoods correspondingly delocalized; (6) remittances playing a growing role in rural household incomes; (7) average age of farmers rising; and (8) cultural and social changes being implicated in livelihood modifications While differing in whether they view peri-urbanization as emerging from urban cores or farm fields, the studies by McGee (1991), Rigg (2006), and Webster (2002) provide operational definitions of peri-urbanization with which the relative “rural” or “urban” nature of the environments can be characterized The variables they identify capture both the loss and the The 59 provinces and five cities under the central government are divided into districts, provincial towns, and provincial cities Districts are further divided into communes (rural areas) and towns, and provincial towns and cities are divided into wards (urban subdistricts) and communes (see Table 3) For the purpose of simplicity we will henceforth use the term “commune” to refer to the smallest administrative unit whether it is a commune, town, or ward Table Pearson correlation (r) between the variables used in cluster analysis.a Income Land Sanitation NDVI Fraction of houses with income mainly from agriculture Fraction of land under agriculture, forests, and aquaculture Fraction of houses with modern sanitation NDVI À0.168 À0.720** 0.083 0.425** 0.034** À0.337** Note: **means p < 0.001 a Correlations were computed separately based on whether communes were surveyed fully or partially diversification of agriculture systems (percent of households whose major source of income is from agriculture, agricultural population density, and livestock density) and the changing nature of the built environment (types of building materials and water and sanitation systems) These variables can be used to define and map differences in the rural/urban nature of communities across space and time Methods Data Vietnam makes digital national population and agricultural census data available at local levels of administration (Epprecht & Heinimann, 2004; Epprecht & Robinson, 2007) We used the 2006 Vietnam Agricultural Census, which surveyed 100% of rural households, asking a wide range of questions regarding household characteristics, farming practices, and household infrastructure, among other matters While unique in its breadth of information at the commune level, the census is incomplete in urban areas, where only those households engaged in agriculture were surveyed Hence, in some urban communes we have only partial survey coverage if less than 100% of households were engaged in agriculture Additionally, in communes within big cities, agricultural census interviews were not conducted at all In 2006, there were 269 such communes, which we labeled as urban core In addition, there were another 649 communes with missing or erroneous data Together, these 918 communes represent 8% of the communes in Vietnam Analysis Our methodological workflow is illustrated in Fig The study builds on the variables identified by McGee (1991), Rigg (2006), Spencer (2013), and Webster (2002) to define and map the relative “urban” nature of the environment McGee (1991, p 20) argues that to map the urban transition two pieces of information are sufficient: “(1) what is the contribution of agricultural and nonagricultural activities to the gross domestic product of a given spatial unit (nation, province, and so on)? And (2) what is the proportion of the working labor force employed in agricultural and nonagricultural work in a given spatial unit?” Spencer (2013) used the 1999 Vietnam Census of Population and Housing to develop a “settlement coherence” index as a quantitative measure of the urban transition in Hanoi and Ho Chi Minh City The variables in his index included house building materials, water supplies, sanitation, and a composite index of all three variables We included a variation of McGee’s two variables, Spencer’s sanitation variable, and a S Saksena et al / Applied Geography 50 (2014) 80e89 83 Table Commune classification based on agricultural income; toilets; land under agriculture, forests, and aquaculture (homes and enterprises); and NDVI Class Number of communes Percentage of communes (%) Fraction of houses with income mainly from agriculture Fraction of land under agriculture, forests, and aquaculture Fraction of houses with modern sanitation NDVI Rural Peri-urban Urban Urban core Communes with missing/ erroneous data 7686 1909 378 454 464 71 18 4 0.82 0.43 n/a n/a n/a 0.43 0.42 0.10 n/a n/a 0.11 0.57 0.89 n/a n/a 0.63 0.55 0.41 0.36 n/a remotely sensed measure of vegetation density (an independent indicator of agricultural land).6 In our work, we characterized urbanicity based on:  Fraction of households whose main income is from agriculture, forestry, and aquaculture;  Fraction of land under agriculture, forestry, and aquaculture, across household and enterprise ownership;  Fraction of houses using modern forms of toilet (pour flush or septic); and  Vegetation density, measured by the Normalized Difference Vegetation Index (NDVI) derived from NASA’s Moderate Resolution Imaging Spectroradiometer (MODIS).7 Table shows the correlations among the variables we selected to include in the model The generally low (except between sanitation and income) bi-variate correlations provide confidence that no single factor is dominating the results of the model We used a spatial data mining approach (Linoff & Berry, 2011) to integrate, select, clean, and transform data; and to recognize and evaluate patterns We then used cluster analysis to classify the communes into three categories Cluster analysis involves splitting a data set into a number of groups of observations that are distinct in terms of typical group values of the variables (Everitt, Landau, Leese, & Stahl, 2011) The aim is to maximize between-group variance and to minimize within-group variance Cluster analysis is a classification technique where any number of variables may be used to classify members of the sample We used a hierarchical agglomeration algorithm for clustering We used the squared Euclidean distance measure and the average linkage between groups cluster method There was no need to transform the data because each classification criterion (parameter/variable) in this case is measured in the same units (a fraction between and in this case) We used IBM SPSS (version 19.0.0.2) statistical software to perform the calculations We labeled the three categories produced from the cluster analysis as rural, peri-urban, and urban In addition, we had two categories for which no data existeddurban core and communes with missing data Post-analysis, we characterized as rural those communes primarily devoted to agriculture; as peri-urban those communes with a mixture of agricultural and nonagricultural activities that often stretch along corridors between urban cores; as urban, communes near urban cores with limited agriculture (many While population density has also often been used to classify the ruraleurban nature of places (Gibbs, 1966; McDade & Adair, 2001; Sudhira, Ramachandra, & Jagadish, 2004; etc.) we could not assess population density from the 2006 agricultural census because some urban communes were partially sampled and urban core communes were not surveyed Population density data from other surveys were not available for the year 2006 Gap-filled mean monthly NDVI from MODIS (MCD43GF) was provided as 30 arc-second grids subset to Vietnam (upper left: 102E, 21N; lower right: 110E, 8N) by Crystal Schaaf and Qingsong Sun of the Center for Remote Sensing at Boston University urban communes were only partially surveyed in the 2006 census because many households did not practice agriculture); and as urban core, communes where the 2006 agricultural census was not conducted because no households were known to practice agriculture Results The urban transition Model results showed that in 2006, 71% of the communes in Vietnam were rural, 18% peri-urban, 3% urban, and 4% urban core (Table 2) Maps based on the model for the entire country and the regions of Hanoi and Ho Chi Minh City are shown in Fig More than 7% of Vietnam’s land area can be classified as peri-urban, and approximately 13% of its population (more than 11 million people) lives in peri-urban areas (Table 3) In 2006, the Vietnam government classified 61% of the peri-urban communes as rural and 39% as urban As expected, as the fraction of houses whose main income is agricultural decreases along the rural-to-urban continuum, so does the vegetation density as measured by the NDVI Also as expected, the fraction of houses using modern sanitation increases along this continuum Rural and peri-urban communes have similar levels of land under agriculture and forestry (average about 43%) Vietnam forbids the conversion of land classified as paddy into non-paddy uses; indeed, our analysis showed that the percentage of agricultural land under paddy was significantly higher (F ¼ 70, p < 0.001 based on one-way ANOVA) in peri-urban areas (73%) as compared to rural areas (67%) This suggests that farmers in peri-urban communes are selling or converting non-paddy agricultural land to other land uses Peri-urban communes Further analysis of the communes the model identified as periurban suggests three types: (1) communes on the periphery of large towns and cities; (2) communes with settlements showing the linear features associated with being built along the sides of roads; and (3) communes in rural areas associated with activities such as provincial administration, mining, and dams Figs and show examples of these three types of peri-urban communes overlaid on current Google Earth images A third (33%) of the 1909 communes the model classified as peri-urban are within 50 km of the two largest cities in Vietnam, Hanoi and Ho Chi Minh City An analysis of spatial autocorrelation in the results showed that peri-urban communes are highly spatially correlated in the Red River and Mekong deltas; this is consistent with national polices that support manufacturing in the deltas and their proximity to Hanoi and Ho Chi Minh City During our ground truthing field trips, we noticed that a number of communes that the model classified as peri-urban were built along highways In the focus group interviews, we learned that 84 S Saksena et al / Applied Geography 50 (2014) 80e89 Fig Urban classification map of Vietnam some of these settlements paid to build a road to their settlement and in other communes people had relocated their houses to be near roads built by the government Interview data suggest that rural communes in which a highway had been constructed often witnessed the establishment of small commercial enterprises and services along the highway Owners of these establishments live above or alongside their shops and modernize their homes more rapidly than households not located along the highway The diversity of occupations among people who live along highways and the modernization of their homes cause these communes to be modeled as transitional Table Area mapped as peri-urban based on 2006 Agricultural Census data Finally, an examination of Fig shows a number of peri-urban communes in predominantly rural areas Table shows that 92% of towns under rural districts are modeled as being peri-urban Many of these towns are home to industrial, energy, or natural resource projects (e.g., mining) These peri-urban communes in predominantly rural areas reinforce the argument that proximity to towns in itself does not define peri-urban; rather, it is the coexistence of both rural and urban characteristics, ruraleurban linkages, and the flows of goods and services between rural and urban areas As Bowyer-Bower (2006) maintains, what constitutes the peri-urban is co-existing rural and urban land uses, which may be in continuous or fragmented units in any one area; further, this juxtaposition of rural and urban land uses can geographically occur anywheredin the core of the city, at its periphery, or in a village Item National total Mapped as peri-urban Percent of national total Model validation Land area (sq km.) Population (2006a) Number of communes Wards under urban districts Wards under provincial towns Communes under provincial towns Towns under rural districts Communes under rural districts 320,390 83,311,200 10,242b 182 504 368 439 8480 23,329 11,175,467 1909 38 303 160 406 1002 7.3 13.4 19 21 40 43 92 12 We used three different methods to validate the results of our model These included ground truthing visits to a sample of communes, an analysis of satellite images, and an examination of road networks Ground truthing a Demographics of Vietnam, 2014 We have subtracted the communes with missing or erroneous data from the total number of communes b We randomly selected 49 communes to test the accuracy of the model These communes were mainly located in the Red River S Saksena et al / Applied Geography 50 (2014) 80e89 85 Fig Example of peri-urban communes near Hanoi Delta and Mekong River Delta We purposely oversampled periurban communes because we believed the model accurately identified extremely rural and urban communes In July and August of 2012 and August 2013, we conducted walk-through surveys in the sampled communes observing land-use types, types of agriculture, extent of industrialization, types of buildings and houses, transportation infrastructure, markets, and so forth We chose random spots within each place and took 360 panoramic photographs We examined both current and historical Google Earth images of these communes In addition, we interviewed key informants (local authorities and statistical officers) about their views on the extent of urbanization that had occurred in their jurisdiction Fig Top: Peri-urban communes in towns under rural districts in the North East region Bottom: Example of peri-urban communes along a highway in the Mekong Delta region 86 S Saksena et al / Applied Geography 50 (2014) 80e89 Table Model validation results based on observation Model class Observed class Total Rural Peri-urban Urban Total Rural Peri-urban Urban 16 0 16 24 26 17 27 49 Finally, we examined data in the commune’s archives The timeand labor-intensive nature of this exercise limited the sample to these communes Based on the data and insights gained from the above methods, the team assessed whether each of these 49 communes had been correctly classified by the model One way to measure the accuracy of the national model’s results is to determine if the ground truth exercise’s results matched the model’s classification Of the 49 communes, the ground truth was discordant with the model classification in only five cases, resulting in a model accuracy rate of 90% Table shows the cross-classification Overall, the error analysis shows that the model tends to classify communes as slightly more urban compared to how we classified them based on interviews and observations Had we also included more extreme rural and urban communes in our sample, the accuracy would have been higher than the current estimate of 90% Remote sensing In addition to ground truthing, we used satellite image analysis to validate our classification model Visible and infrared data collected by the Defense Meteorological Satellite ProgramOperational Linescan System (DMSP-OLS) can be used to measure the intensity of night time lights (NTL), including illumination from human settlements, gas flares, fires, and fishing vessels The DMSPOLS is a weather sensor that was designed to detect daily and nightly cloud cover and cloud top temperatures, and its coarse spatial resolution cannot measure light variability at sub-kilometer scales However, because of their association with human settlements, NTL data have been used to map and monitor urban growth (Baugh, Elvidge, Ghosh, & Ziskin, 2010) Light saturation and overglow in urban centers can cause NTL data to overestimate urban extent (Small, Elvidge, Balk, & Montgomery, 2011; Zhang & Seto, 2011), and variations in the intensity of lights may indicate differences in socioeconomic development or energy consumption rather than differences in population densities (Elvidge et al., 2010; Henderson, Yeh, Gong, Elvidge, & Baugh, 2003) Despite these limitations, the DMSP-OLS is currently the only sensor that provides global coverage of NTL data for the past two decades (Elvidge et al., 2007), which makes it useful for studying changes over time.8 To overcome the aforementioned problem of saturation in urbanized areas, some researchers have combined NTL with NDVI data (Lu, Tian, Zhou, & Ge, 2008; Zhang, Schaaf, & Seto, 2013; Zhang & Seto, 2011); however, since our communebased cluster analysis relies on MODIS NDVI as a key variable, we chose to conduct a multi-temporal analysis of NTL data alone to provide an independent comparison Global, cloud-free annual composites of NTL from 1992 to 2009 are available from the National Geophysical Data Center’s Earth Observation Group (NGDC/EOG) as 30 arc-second grids (Baugh et al., 2010) After subsetting these data to the national Improvements in detecting and resolving small areas of low lights have been included in the new VIIRS sensor aboard the NOAA-NASA Suomi NPP satellite, but it was only launched in late 2011 boundaries of Vietnam, we calculated per-pixel mean and standard deviation (SD) over the 18-year time period A simple side-by-side comparison of the resulting maps from our multi-temporal remote sensing analysis and census-based cluster analysis showed qualitative similarities in spatial patterns We then took a 10% sample of all communes to generate time-series profiles, summarizing the NTL data set by standard deviations within each category.9 The differences in SD values between categories illustrate different levels of economic development More importantly, the trend lines for both rural and urban categories have similar, relatively low slopes, while the peri-urban trend line is twice as steep, indicating a greater rate of change (Fig 5) Roads and peri-urban communes Finally, roads and road density are frequently used measures of urbanization (Zhu, Xu, Jiang, Li, & Fan, 2006) To test for the impact of roads on peri-urban settlements, we used data on highways from Vietnam’s General Statistics Office to analyze various metrics such as presence/absence of a highway, minimal distance to a highway, length of a highway passing through a commune, and the highway density in a commune (length/area of commune) We found a statistically significant association with highway density (results of one-way ANOVA: F ¼ 244, p < 0.001 for national highways; F ¼ 190, p < 0.001 for provincial highways) The national and provincial highway densities in rural communes were 57 km/km2 and 11 km/ km2, respectively, while in peri-urban communes they were 123 km/km2 and 40 km/km2, respectively The highway density metric best approximates the number of establishments and houses along the highway, in a standardized form Discussion This study demonstrates a robust method for using Vietnamese national census and remote sensing data to identify and map communes as being rural, peri-urban, urban, or urban core; and validates the classification based on ground observations, analyses of multi-temporal night-time lights data, and examination of road networks Peri-urbanization is one of the major issues facing planners in Vietnam today These areas face intense pressures on resources, slum formation, lack of adequate services such as water and sanitation, degradation of farmland, and a host of other issues These areas also face demands from users with contrasting lifestyles and conflicting interests that range from agriculture to residential, industrial, and commercial, to the development of green belts and nature reserves The data platform and model built in this project provide a method for rapidly assessing the ruraleurban nature of communes, and, as new census data become available (we are currently acquiring the 2011 National Agricultural Census), of updating the assessment The model can also be used to test theories about the nature of development, urbanization, periurbanization, deagrarianization, and other issues Vietnam is unique in mainland Southeast Asia in making digital national population and agricultural census data available in a format that can be linked to administrative boundaries; other nations, however, are beginning to develop similar capacities National population census data are available for Laos at the village level, although village boundaries have not been mapped Demographic and socioeconomic data are also available at the local Our 10% validation sample did not include communes that were smaller than the km pixel size of our remote sensing imagery; however, these small communes, most of which were located around urban core areas, represented only 2% of the entire nation S Saksena et al / Applied Geography 50 (2014) 80e89 87 Fig Time-series profiles of the NTL SD by commune classification level of administration (tambon) in Thailand, but there does not appear to be a single source for these data, making it difficult to compile a complete data set Hence, in the short term, the application of this method to other countries in Southeast Asia may be limited although workarounds may be available (e.g., using Thiessen polygons to approximate village boundaries in Laos as demonstrated by Heinimann et al., 2013) Theoretically, government agencies in most countries should have access to national census data Their ability to access these data and link it to administrative boundaries within a digital database, however, remains problematic Problems include obstacles to sharing data among agencies, lack of capacity for analyzing and mapping the data, and not being identified as a high priority task by any government agency These problems, rather than the availability of data, may prove the most difficult to overcome We have demonstrated a method for classifying and mapping the urban transition; whether or not that has meaningful consequences remains to be seen Kontgis et al (submitted for publication) used Landsat remote sensing imagery to map land-cover and land-use change in a 50km radius surrounding Ho Chi Minh City between 1990 and 2012 Of the total urban expansion mapped, over 62% occurred in communes we classified as peri-urbanddouble the amount of urban expansion that occurred in urban communes and over 16 times the amount of urban expansion that occurred in rural and urban core communes Using population census data from 1999 to 2009 and the classification system presented in this paper, Kontgis et al found that 79% of the communes with the fastest population and urban expansion rates were defined as peri-urban communes in our classification The two projects (the current project and that conducted by Kontgis et al.) provide alternative ways to define and map the urban transition Analysis of census data provides insights into the socioeconomic characteristics of places undergoing the rural-to-urban transition or the interaction of rural and urban activities, while analysis of remotely sensed data provides insights into the biophysical manifestations (i.e., land use and land cover changes) associated with such transitions and interactions Both census data and remotely sensed data can be used independently to map the urban transition but perhaps provide the greatest insights into change when used together The work by Kontgis et al (submitted for publication) suggests that if we can use remote sensing data to establish a good approximation of the urban classification we demonstrated here for all of Vietnam (not just Ho Chi Minh City), we may be able to apply the same remote sensing methods to other countries and parts of the region Limitations Secondary data not exist on some of the variables hypothesized as important in the urban transition theories proposed by McGee (1991), Rigg (2006), and others Examples include mobility, remittances, and female participation in the labor force; hence, we could not include these in our model However, considering that we have created a parsimonious model that has an accuracy of at least 90%, we not anticipate that including more explanatory variables will improve the accuracy significantly But it is possible that in the future, as development trajectories change, some of these neglected variables may gain importance in model development as urbanization and peri-urbanization become more nuanced The major uncertainty in our data relates to measurements of total land within the smallest administrative unit’s boundaries While the agricultural census keeps track of land-use under different major agricultural and forest categories, it does not include a question on total land area in each commune We used data on total land from other government sources In some cases, the two datasets were inconsistent, leading to illogical values in some of the derived variables This was noticed more frequently in 88 S Saksena et al / Applied Geography 50 (2014) 80e89 the smaller communes and wards We applied data quality checks to filter out such cases Smaller communes also posed problems when matching remote sensing data (see footnote 8), and thus had to be excluded from analysis A number of researchers have produced continuous-scale metrics of urbanicity by conducting primary surveys of small populations on variables hypothesized as important (Dahly & Adair, 2007; Jones-Smith & Popkin, 2010; McDade & Adair, 2001; Van de Poel et al., 2009) Our approach uses secondary data to produce a nominal-scale metric that is relatively quicker to produce, even for an entire province or nation Future studies should aim to simultaneously adopt both approaches for the purpose of cross comparison Conclusions This project applied theories of the ruraleurban transition to spatially-explicit census and remote sensing data to build, test, and validate a model that classifies communes according to their socioeconomic and biophysical characteristics Our results suggest that more than 7% of the country’s land area and roughly 13% of its population resides in peri-urban places Hence, a significant portion of Vietnam’s landscape and population are affected by the social, economic, and institutional dynamics of both urban and rural ways of living and infrastructures We found that 61% of the communes we classify as peri-urban, that is, places with both rural and urban features, are classified and managed by the Vietnamese government as rural areas We also found that peri-urban communes could be further classified into three types: (1) communes on the periphery of large towns and cities; (2) communes with settlements showing the linear features associated with being built along the sides of roads; and (3) communes in rural areas associated with activities such as provincial administration, mining, and dams This finding is important for reiterating the fact that proximity to towns in itself does not define the peri-urban; rather, it is the co-existence of both rural and urban characteristics Acknowledgments This project was funded by the United States National Science Foundation Grant No DEB-0909410 We thankfully acknowledge our graduate students Charles Phan Nguyen, Duong Huu Nong, and Chinh Cong Tran for their assistance in the lab and in the field We are also thankful for the assistance of Dr Nghiem Tuyen for the excellent work she did in translating our questions Finally, we could not have completed the field work without the professional assistance of Dr Trinh Dinh Thau, Ms Dung Pham, Ms Huong Tran, Ms Trang Tran, Ms Phuong Le, and other employees of the Center for Agricultural Research and Ecological Studies (CARES), Hanoi University of Agriculture References Aoyama, Y., & Horner, R (2010) World development report 2009: Reshaping economic geography, by the World Bank Wiley Online Library Retrieved from http:// onlinelibrary.wiley.com/doi/10.1111/j.1467-9787.2010.00709_1.x/full APN (2009) Peri-urban development and environmental sustainability: Examples from China and India (No ARCP2009-05CMY-Sellers) Kobe, Japan: Asia Pacific Network Baugh, K., Elvidge, C., Ghosh, T., & Ziskin, D (2010) Development of a 2009 stable lights product using DMSP-OLS data Proceedings of the 30th Meeting of the Asia Pacific Advanced Network, 30, 114e130 Bowyer-Bower, T A S (2006) The inevitable illusiveness of “sustainability” in the peri-urban interface: the case of Harare In D McGregor, D Simon, & D Thompson (Eds.), The peri-urban interface: Approaches to sustainable natural and human resource use (pp 151e164) Retrieved from http://books.google.com/ books?hlẳen&lrẳ&idẳyZUJ9K_nKbQC&oiẳfnd&pgẳPA313&dqẳMcGregorỵ periurban&otsẳrVttlyIgbM&sigẳvfVc18krIqpXjH_v6gX6HGMCtGE Dahly, D L., & Adair, L S (2007) Quantifying the urban environment: a scale measure of urbanicity outperforms the urbanerural dichotomy Social Science & Medicine, 64(7), 1407e1419 Demographics of Vietnam (2014, January 15) In Wikipedia, the free encyclopedia Retrieved from http://en.wikipedia.org/w/index.php?title¼Demographics_of_ Vietnam&oldid¼590039374 Dudwick, N., Hull, K., Katayama, R., Shilpi, F., & Simler, K (2011) From farm to firm: Rural-urban transition in developing countries Washington, DC: World Bank Elvidge, C D., Baugh, K E., Sutton, P C., Bhaduri, B., Tuttle, B T., Ghosh, T., et al (2010) Who’s in the darkdsatellite based estimates of electrification rates In X Yang (Ed.), Urban remote sensing: Monitoring, synthesis and modeling in the urban environment (pp 211e224) Chichester, UK: John Wiley & Sons Elvidge, C D., Safran, J., Tuttle, B., Sutton, P., Cinzano, P., Pettit, D., et al (2007) Potential for global mapping of development via a nightsat mission GeoJournal, 69(1e2), 45e53 http://dx.doi.org/10.1007/s10708-007-9104-x Epprecht, M., & Heinimann, A (2004) Socioeconomic atlas of Vietnam: A depiction of the 1999 population and housing census Hanoi: Cartographic Publishing House/ Berne: Swiss National Centre of Competence in Research (NCCR) North-South, University of Berne Epprecht, M., & Robinson, T (2007) Agricultural atlas of Vietnam: A depiction of the 2001 rural agriculture and fisheries census Retrieved from http://agris.fao.org/ agris-search/search/display.do?f¼2009/XF/XF0809.xml;XF2008436453 Everitt, B S., Landau, S., Leese, M., & Stahl, D (2011) Cluster analysis (5th ed.) Chichester, UK: John Wiley &Sons Firman, T (2000) Rural to urban land conversion in Indonesia during boom and bust periods Land Use Policy, 17(1), 13e20 Friedmann, J (2011) Becoming urban: periurban dynamics in Vietnam and Chinadintroduction Pacific Affairs, 84(3), 425e434 Gibbs, J P (1966) Measures of urbanization Social Forces, 45(2), 170e177 Harms, E (2011) Material symbolism on Saigon’s edge: the political-economic and symbolic transformation of Ho Chi Minh City’s periurban zones Pacific Affairs, 84(3), 455e473 Heinimann, A., Hett, C., Hurni, K., Messerli, P., Epprecht, M., Jørgensen, L., et al (2013) Socio-economic perspectives on shifting cultivation landscapes in Northern Laos Human Ecology, 41(1), 51e62 Henderson, M., Yeh, E T., Gong, P., Elvidge, C., & Baugh, K (2003) Validation of urban boundaries derived from global night-time satellite imagery International Journal of Remote Sensing, 24(3), 595e609 Jones-Smith, J C., & Popkin, B M (2010) Understanding community context and adult health changes in China: development of an urbanicity scale Social Science & Medicine, 71(8), 1436e1446 Kontgis, C P., Schneider, A., Fox, J., Saksena, S., Spencer, J., Castrence, M Monitoring peri-urbanization in the greater Ho Chi Minh City metropolitan area Journal of Applied Geography, under review Kuznets, S S (1992) Modern economic growth: findings and reflections In A Lindbeck (Ed.), Economic sciences, 1969e1980: The Sveriges Riksbank (Bank of Sweden) prize in economic sciences in memory of Alfred Nobel (pp 87e102) Singapore/River Edge, NJ: World Scientific Leaf, M (2002) A tale of two villages: globalization and peri-urban change in China and Vietnam Cities, 19(1), 23e31 Li, X., & Yeh, A G O (1998) Principal component analysis of stacked multi-temporal images for the monitoring of rapid urban expansion in the Pearl River Delta International Journal of Remote Sensing, 19(8), 1501e1518 Linoff, G S., & Berry, M J (2011) Data mining techniques: For marketing, sales, and customer relationship management Retrieved from http://books.google.com/ books?hlẳen&lrẳ&idẳAyQfVTDJypUC&oiẳfnd&pgẳPR37&dqẳDataỵMiningỵ Techniques&otsẳKVNyxqNTDF&sigẳWCZcHBQ4eUnnwNf0H4UPm_NexV0 Lu, D., Tian, H., Zhou, G., & Ge, H (2008) Regional mapping of human settlements in southeastern China with multisensor remotely sensed data Remote Sensing of Environment, 112(9), 3668e3679 McDade, T W., & Adair, L S (2001) Defining the “urban” in urbanization and health: a factor analysis approach Social Science & Medicine, 53(1), 55e70 McGee, T G (1991) The emergence of desakota in Asia: expanding a hypothesis In N Ginsburg, B Koppel, & T G McGee (Eds.), The extended metropolis: Settlement transition in Asia (pp 15e55) Honolulu, HI: University of Hawai’i Press McGregor, D., Simon, D., & Thompson, D (2006) The peri-urban interface in developing areas: the research agenda In D McGregor, D Simon, & D Thompson (Eds.), The peri-urban interface: Approaches to sustainable natural and human resource use (pp 313e319) Retrieved from http://books.google.com/ books?hlẳen&lrẳ&idẳyZUJ9K_nKbQC&oiẳfnd&pgẳPA313&dqẳMcGregorỵ periurban&otsẳrVttlyIgbM&sigẳvfVc18krIqpXjH_v6gX6HGMCtGE Novak, N L., Allender, S., Scarborough, P., & West, D (2012) The development and validation of an urbanicity scale in a multi-country study BMC Public Health, 12(1), 530 Retrieved from http://www.biomedcentral.com/1471-2458/12/530 Rigg, J (2006) Land, farming, livelihoods, and poverty: rethinking the links in the rural South World Development, 34(1), 180e202 Schneider, A., Friedl, M A., & Potere, D (2009) A new map of global urban extent from MODIS satellite data Environmental Research Letters, 4(4) http:// dx.doi.org/10.1088/1748-9326/4/4/044003 Schneider, A., Friedl, M A., & Potere, D (2010) Mapping global urban areas using MODIS 500-m data: new methods and datasets based on “urban ecoregions” Remote Sensing of Environment, 114(8), 1733e1746 Schneider, A., Seto, K C., & Webster, D R (2005) Urban growth in Chengdu, Western China: application of remote sensing to assess planning and policy outcomes Environment and Planning B: Planning and Design, 32(3), 323e345 S Saksena et al / Applied Geography 50 (2014) 80e89 Seto, K C., Reenberg, A., Boone, C G., Fragkias, M., Haase, D., Langanke, T., et al (2012) Urban land teleconnections and sustainability Proceedings of the National Academy of Sciences, 109(20), 7687e7692 Small, C., Elvidge, C D., Balk, D., & Montgomery, M (2011) Spatial scaling of stable night lights Remote Sensing of Environment, 115(2), 269e280 Spencer, J H (2013) The urban health transition hypothesis: empirical evidence of an avian influenza Kuznets curve in Vietnam? Journal of Urban Health, 90(2), 343e357 Sudhira, H S., Ramachandra, T V., & Jagadish, K S (2004) Urban sprawl: metrics, dynamics and modelling using GIS International Journal of Applied Earth Observation and Geoinformation, 5(1), 29e39 Van de Poel, E., O’Donnell, O., & Van Doorslaer, E (2009) Urbanization and the spread of diseases of affluence in China Economics & Human Biology, 7(2), 200e 216 Webster, D (2002) On the edge: Shaping the future of peri-urban East Asia Asia/ Pacific Research Center Retrieved from http://iis-db.stanford.edu/pubs/20031/ Webster2002.pdf 89 Webster, D (2011) An overdue agenda: systematizing East Asian peri-urban research Pacific Affairs, 84(4), 631e642 Winarso, H (2011) Urban dualism in the Jakarta metropolitan area In A Sorensen, & J Okata (Eds.), Megacities: Urban form, governance, and sustainability (pp 163e 191) Retrieved from http://link.springer.com/chapter/10.1007/978-4-43199267-7_8 Zhang, Q., Schaaf, C., & Seto, K C (2013) The Vegetation Adjusted NTL Urban Index: a new approach to reduce saturation and increase variation in nighttime luminosity Remote Sensing of Environment, 129, 32e41 Zhang, Q., & Seto, K C (2011) Mapping urbanization dynamics at regional and global scales using multi-temporal DMSP/OLS nighttime light data Remote Sensing of Environment, 115(9), 2320e2329 Zhu, M., Xu, J., Jiang, N., Li, J., & Fan, Y (2006) Impacts of road corridors on urban landscape pattern: a gradient analysis with changing grain size in Shanghai, China Landscape Ecology, 21(5), 723e734  ... for classifying and mapping the dynamics of the urban transition in Vietnam Theories of peri-urbanization Friedmann (2011) argues that “a general theory of the periurban.escapes us” and the. .. define and map the urban transition Analysis of census data provides insights into the socioeconomic characteristics of places undergoing the rural-to -urban transition or the interaction of rural and. .. coherence” index as a quantitative measure of the urban transition in Hanoi and Ho Chi Minh City The variables in his index included house building materials, water supplies, sanitation, and a composite