Defining urban and rural areas

PROCEEDINGS OF SPIE SPIEDigitalLibrary.org/conference-proceedings-of-spie

Defining urban and rural areas: a new approach

Blanca Arellano, Josep Roca

Blanca Arellano, Josep Roca, "Defining urban and rural areas: a new approach," Proc. SPIE 10431, Remote Sensing Technologies and Applications in Urban Environments II, 104310E (4 October 2017); doi: 10.1117/12.2277902 Event: SPIE Remote Sensing, 2017, Warsaw, Poland Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 8/8/2018 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

Defining urban and rural areas: A new approach Blanca Arellano Josep Roca UPC, Technical University of Catalonia, Av. Diagonal, 08028 Barcelona, Spain - (blanca.arellano, josep.roca)@upc.edu

ABSTRACT The separation between the countryside and the city, from rural and urban areas, has been one of the central themes of the literature on urban and territorial studies. The seminal work of Kingsley Davis [10] in the 1950s introduced a wide and fruitful debate which, however, has not yet concluded in a rigorous definition that allows for comparative studies at the national and subnational levels of a scientific nature. In particular, the United Nations (UN) definition of urban and rural population is overly linked to political and administrative factors that make it difficult to use data adequately to understand the human settlement structure of different countries. The present paper seeks to define a more rigorous methodology for the identification of rural and urban areas. For this purpose it uses the night lights supplied by the SNPP satellite, and more specifically by the VIIRS sensor for the determination of the urbanization gradient, and by means of the same construct a more realistic indicator than the statistics provided by the UN. The arrival of electrification to nearly every corner of the planet is certainly the first and most meaningful indicator of artificialization of land. In this sense, this paper proposes a new methodology designed to identify highly impacted (urbanized) landscapes worldwide based on the analysis of satellite imagery of night-time lights. The application of this methodology on a global scale identifies the land highly impacted by light, the urbanization process, and allows an index to be drawn up of Land Impacted by Light per capita (LILpc) as an indicator of the level of urbanization. The methodology used in this paper can be summarized in the following steps: a) a logistic regression between US Urban Areas (UA), as a dependent variable, and night-time light intensity, as an explanatory variable, allows us to establish a nightlight intensity level for the determination of Areas Highly Impacted by Light (AHIL); b) the delimitation of the centers and peripheries is made by setting a threshold of night-time light intensity that allows the inclusion of most of the centers and sub-centers; c) once identified urbanized areas, or AHIL, it is necessary to delimit the rural areas, or Areas Little Impacted by Light (ALIL), which are characterized by low intensity night light; d) finally, rurban landscapes are those with nightlight intensities between ALIL and AHIL. The developed methodology allows comparing the degree of urbanization of the different countries and regions, surpassing the dual approach that has traditionally been used. This paper enables us to identify the different typologies of urbanized areas (villages, cities and metropolitan areas), as well as “rural”, “rurban”, “periurban” and “central” landscapes. The study identifies 186,134 illuminated contours (urbanized areas). 404 of these contours have more than 1,000,000 inhabitants and can be considered real “metropolitan areas”; on the other hand there are 161,821 contours with less than 5,000 inhabitants, which we identified as “villages”. Finally, the paper shows that 40.26% live in rural areas, 15.53% in rurban spaces, 26.04% in suburban areas and only 18.16% in central areas. Keywords: Gradient of Urbanization; Night-time Lights; Urban Sprawl; Rurbanization Remote Sensing Technologies and Applications in Urban Environments II, edited by Thilo Erbertseder, Nektarios Chrysoulakis, Ying Zhang, Proc. of SPIE Vol. 10431, 104310E · © 2017 SPIE · CCC code: 0277-786X/17/$18 · doi: 10.1117/12.2277902 Proc. of SPIE Vol. 10431 104310E-1 Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 8/8/2018 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

1. INTRODUCTION The second half of the twentieth century has been, without doubt, the time when there has been further development of urbanisation on a global scale. The urban population has grown from 750 million people in 1950 to 2,860 in 2000, more than 50% of the world population [35]. In the Developed World, the urban model has suffered significant changes in recent decades, transforming from a model of urban continuum of medium and high densities to a model of an endless diffuse and sprawled city, driven by technological innovation processes, separation of functions and seeking proximity to nature [32]. This redefinition of the spatial model has based on new communication networks and infrastructures. The pattern has resulted in an expansionist, unsustainable and predatory city, especially in the USA, a paradigmatic example of high land consumption [19]. Therefore, since 1950 there has been a real reversal in the topology of the landscape [24]. Landscapes that are highly artificialised 1 have changed from "islands" within the "rural ocean" to "colonise" almost the entire planet. The process of urban sprawl has relegated to open spaces the role of auxiliary elements within the spatial structure. The diffusion of the Garden City theory, together with the generalisation of the car, which occurred after the Second War, and the dream of having a single family home with grill and pool became an international concept, which spread from the USA to Europe and to the whole world. The sprawl in residential areas is linked to the gradual decentralisation of economic activity, first the industry, and then the services and even the most qualified tertiary activities. Urban sprawl 2, the massive consumption of land, can be found worldwide, although it takes many different forms in different regions and continents. In developing countries, the diffusion of urbanisation is even more pronounced. If in 1950 only 300 million inhabitants lived in cities, it is expected that in 2030 these figures in developing countries will have reached 4000 million, representing 4/5 of the world urban population [33]. The countryside to city emigration has generated an extensive diversity of landscape developments. From semi-rural hamlets close to big cities, or informal peripheries lacking the minimum services, to giant megalopolises. In the early 21st century, 20 of the 26 megacities of more than 20 million inhabitants were in developing countries [1]. Despite the extraordinary growth of cities in recent decades, there is no unanimous academic consensus about what "urbanisation" means. The plural nature of the forms of human settlement make it difficult to identify the urbanisation processes. This difficulty leads to an absence of any global database that would allow comparative studies of the urbanisation processes worldwide. An example of this difficulty lies in the ambiguous definition of urban and rural population of the UN. The diversity of criteria used by nations not allows reliable comparisons about the degree of urbanisation at international scale. The UN have failed to harmonise these differences of opinion between nations: “In preparing estimates and projections of the urban population, the United Nations relies on the data produced by national sources that reflect the definitions and criteria established by national authorities. It has long been recognised that, given the variety of situations in the countries of the world, it is not possible or desirable to adopt uniform criteria to distinguish urban areas from rural areas. An analysis of that set of definitions indicates that 118 of the 231 countries in the world use administrative criteria to make a distinction between urban and rural, 64 of which use it as the sole criterion to make that distinction. In 107 cases, the criteria used to characterise urban areas include population size or population density, solely in the case of 48 countries”. A sample of the huge variety of existing definitions is “the lower limit above which a settlement is considered urban, ranging between 200 and 50,000 inhabitants. Economic characteristics were part of the criteria used to identify urban areas in 33 countries or areas, including all the successor States of the former Union of Soviet Socialist Republics; and criteria related to the functional nature of urban areas, such 1

It refers to the urbanised territories. Almost all the land surface of the planet has undergone different processes of human artificialisation. In this sense, it is possible to distinguish between those characterised with high artificialisation (such as urban spaces, but also mining or other) of those (such as farmland and generally rural) whose degree of artificialisation could be considered minor. 2 There is no broad consensus on how to define Urban Sprawl [19, 20, 38]. European Environment Agency has described sprawl as the physical pattern of low-density expansion of large urban areas, under market conditions, mainly into the surrounding agricultural areas. Sprawl is the leading edge of urban growth and implies little planning control of land subdivision. Development is patchy, scattered and strung out, with a tendency for discontinuity. It leap-frogs over areas, leaving agricultural enclaves. Sprawling cities are the opposite of compact cities, full of empty spaces that indicate the inefficiencies in development and highlight the consequences of uncontrolled growth" [14].

Proc. of SPIE Vol. 10431 104310E-2 Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 8/8/2018 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

as the existence of paved streets, water-supply systems, sewerage systems or electric lighting, were part of the definition of urban in 43 cases, but only in ten cases such a criteria was used alone. Lastly, in another 14 cases no definition of “urban” was available and in a further eight the entire population of a country or area was considered to be urban.” [34]. Unlike in the USA, where the Census Bureau precisely defines urban areas 3, it seems that there is no one unified set of criteria to identify urban (and metropolitan) systems on a planetary scale. In Europe alone there seems to be a similar (although not identical) concept to that used in the USA to identify urbanised areas. Eurostat, that initially focused its efforts on delimiting Larger Urban Zones (LUZ), similar to the American metropolitan areas, has recently implemented a method that aims to define the Degree of Urbanisation Classification (DEGURBA) of Europe 4. The very concept of urbanisation is ambiguous. Urbanisation has traditionally been defined as “the process by which towns and cities are formed and become larger as more and more people begin living and working in central areas” (Merriam-Webster Dictionary), “the process of creating towns in country areas” (Collins), “a population shift from rural to urban areas” (Wikipedia) or “the process by which large numbers of people become permanently concentrated in relatively small areas, forming cities” (Britannica). Nevertheless this definition does not completely fit the urbanisation processes initiated in the second half of the 20th Century, when a significant part of the population had left traditional centres, seeking a greater proximity to natural spaces on the edges of the sprawl. This has led academics coming from the field of geography to speak of “counterurbanization” [4], “desurbanization” [3], “città difusa” [22], or other similar terms, suggesting that the urbanisation processes would have come to an end (in the more developed world). Others, however, have spoken of the urbanisation of the countryside, or “rurbanization” [2], disseminating terms such as “rural sprawl” [9] or “ex urban sprawl” [31]. In this context, rather than speaking of the crisis of the urbanisation processes we must refer to the progressive disappearance of the city-country contradiction, in a context of extension of the urbanisation networks throughout the territory [6, 24, 36]. There is a second concept of “urbanisation” that is less widespread than the first. A concept based on the idea of the physical transformation of the landscape, not on the conversion of “countryside” into “city”. According to various dictionaries, urbanisation would be “the action and effect of urbanising”. In a similar way to building process, urbanisation would represent, from this second concept, the transformation of the primordial rural landscape by the incorporation of services networks characterised by a high artificialisation (vehicle access, paving, lighting and other public services, water supply, electricity supply, sewers, etc.). Nevertheless this second sense also incorporates a significant dose of ambiguity. Firstly because urbanisation does not always come before building. In the peripheries in developing countries the land is usually divided into "plots" first and then it is built on (occupying the land) and then the service networks arrive afterwards. But, above all, because urbanisation is not always a single, comprehensive act, but is more commonly a process (especially in the “Third World”) in which the service networks (vehicle access and electricity, first), the water supply, sewerage and later paving are gradually added. In this regard it would be necessary to speak of the gradient or intensity of the urbanisation, more than an integrated or single event. Remote sensing allows detailed analysis of the ground cover and, therefore, the identification of the landscapes highly artificialised by the urbanisation processes. Satellite images, such as Landsat with a spatial resolution of 30 m/pixels and a high spectral resolution, make it possible to identify the urbanised areas. Nevertheless, despite remote sensing 3

The US Census Bureau [37] defines an urban area as “(it) will comprise a densely settled core of census tracts and/or census blocks that meet minimum population density requirements, along with adjacent territory containing non-residential urban land uses as well as territory with low population density included to link outlying densely settled territory with the densely settled core. To qualify as an urban area, the territory identified according to criteria must encompass at least 2,500 people, at least 1,500 of which reside outside institutional group quarters. The Census Bureau identifies two types of urban areas: a) Urbanized Areas (UAs) of 50,000 or more people; b) Urban Clusters (UCs) of at least 2,500 and less than 50,000 people” [7]. For 2010 Census there were 3,573 UA in USA, with a population of 308,745,538 inhabitants (80.7% of total population). 4 The Degree of Urbanisation (DEGURBA) is a classification that indicates the character of an area [15]. The latest update of the classification is based on 2011 population grid and the 2014 Local Administrative Units (LAU) boundaries. This classification was established by Eurostat in 2010 [16] and is based on the demographic density of the grid of 1 km2 of the European region. DEGURBA differentiates three degrees of urbanisation according to the density: a) High-density cluster or urban centre (contiguous grid cells of 1 km2 with a density of at least 1,500 inhabitants per km2 and a minimum population of 50,000; b) Urban clusters (clusters of contiguous grid cells of 1 km2 with a density of at least 300 inhabitants per km2 and a minimum population of 5,000; and c) Rural grid cells (Grid cells outside urban clusters).


technology enabling the uniform analysis of land cover on a worldwide scale, the complexity of the urbanising phenomenon and the high cost of computing have not enabled the generation of an integrated database on a world scale. The effort made in this regard does not have sufficient resolution to analyse the plural nature of landscape developments. The Global Land Cover (GLC) project, for example, with a resolution of 1 km2, cannot differentiate the gradient of urbanisation [18, 21, 23] with sufficient detail to identify and classify different landscapes artificialised by urbanisation. There is no single methodology to measure the phenomenon of urbanisation [28, 30]. The plural nature of forms of human settlement greatly limits the identification of the urbanised area. The forms of human settlement differ markedly all over the planet. From the predominantly rural structures of Africa and part of Asia and South America to the immense world megalopolises. From the big cities of Mesoamerica and Mexico to the infinite peripheries of the sprawl in the United States and central Europe. From the città difusa of the Po Valley, to the compact Mediterranean cities. Linear cities, following axes of communication; grid-based cities, by way of new contemporary centurions; monocentric structures vs. polycentric urban systems. From the immense “pseudo-urbanised” suburbs of India, Africa and Latin America to the fragmented regions developed in Europe based on administrative divisions (parishes, communes, municipalities, etc.). But it is not only the plural nature of the developed landscapes on the planet that limits identification and classification using remote sensing techniques. The different interaction between the diverse types of land covers also hinders the task of delimiting urbanised spaces. Urban areas, for example, include large spaces that cannot be built upon, green areas and even open spaces. The pattern of urban development is usually characterised by fragmentation; “leap frog” that hinders the identification of urban areas by merely identifying buildings. The scattered nature of urbanisation represents, therefore, an intrinsic difficulty in the identification and delimitation of urban areas by means of remote sensing. This difficulty is increased if we consider that the pixels resulting from the satellite images usually present “confusion” as they are a combination of various real types of land covers [27]), which has led to a great variety of techniques aimed at optimising the classification of the ground cover. In the absence of an integrated approach of how identify urban areas, this paper uses the information derived from nighttime lights to delineate the areas impacted by urbanisation on a world scale. The night-time lights have been used in widespread scientific contributions, from building human development indices [13] to identify megalopolis [1, 17]. Night-time images obtained by remote sensing can represent a useful way to define urbanised landscapes [11]. A new methodology is proposed, using night-time lights, to study the impact of urbanisation on the world, especially in urban and metropolitan areas. The application of this methodology on a global scale delineates the land highly impacted by light and allows an index to be drawn up of Land Impacted by Light per capita (LILpc) as an indicator of the level of urbanisation. With this index it is possible to identify the different typologies of the urbanised areas (villages, cities and metropolitan areas), as well as “rural”, “rurban”, “periurban” and “central” landscapes.

2. METHODOLOGY The research is based upon the following key questions: How can we identify urbanisation? How can we measure it? Is it useful to use night-time lights to measure the degree of urbanisation? Is it possible to define a unified method to delimit urbanised areas at worldwide scale using night-time light images? In addition, could other types of human landscapes be understood using nightlights images? How could those landscapes be classified? The study assumes that night-time lights satellite imagery provides valuable information for the identification of human landscapes, such as rural and urbanised areas. The "dark" landscapes are certainly related to rural settlements. The landscapes of light and darkness detect more clearly than traditional statistics based on the percentage of urban/rural forms of human settlement on the world population, with the advantage, in turn, of allow it to be studied on a subnational level, which is not possible when simply using official statistics. Concerning the "lit" landscapes, it clearly identifies areas of the world characterised by high human artificialisation. The electricity supply, along with the division of land


into plots and the “lines” of the streets 5, represent the first steps in the process of urbanisation. The almost universal access to electrical energy as well as the diverse intensity of its use makes the analysis of night-time images an exceptional tool for studying the urbanisation gradient on a world scale. The research develops a methodology to detect urbanised areas, Areas Highly Impacted by Light (AHIL) as they are called in this paper, also identifying central (compacted) and periurban (sprawled) settlements. After that, rural landscapes (identified as Areas Little Impacted by Light or ALIL) have been detected. Finally, intermediate patterns of human occupation of territory, such as ex-urban sprawl or rurban landscapes, have been analysed. For more than forty years the U.S. Air Force Defense Meteorological Satellite Program (DMSP) Operational Linescan System (OLS) has been collecting global low light imaging data 6. In 2011 NASA [25] and NOAA launched the Suomi National Polar Partnership (SNPP) satellite carrying the first Visible Infrared Imaging Radiometer Suite (VIIRS) instrument 7. The VIIRS collects low light imaging data and has several improvements over the capabilities of the OLS [12]. There is a major difference in the pixel footprint between both sensors. The VIIRS day-night band is ten to fifteen times better than the OLS system at resolving the relatively dim lights of human settlements and reflected moonlight. Each pixel shows 742 metres (0.46 miles) across, compared to the 3-kilometre footprint (1.86 miles) of DMSP. Beyond the resolution, the sensor can detect dimmer light sources. And since the VIIRS measurements are fully calibrated, scientists now have the precision required to make quantitative measurements of clouds and other features. VIIRS DNB uses sixty-four detector aggregation zones (32 on each side) to maintain at a constant 742 metres from nadir out to edge of scan (http://earthobservatory.nasa.gov/Features/IntotheBlack/). In this paper, a preliminary version of the data obtained by the VIIRS sensor, supplied by NASA under the name Black Marble in 2013 8, has been used. The reason for using Black Marble imaging instead of NOAA/NGDC is because at the time this study was carried out, the SNPP-VIIRS data was not yet available on the NOAA website. Given the absence of an international standard adopted to identify urban areas, the definition given by the US Census Bureau has been used. The urban areas obtained in this study therefore include noncontiguous territory separated by exempted territory (like bodies of water), via hops and jumps (maximum hop distance 0.5 miles, maximum jump distance 2.5 miles, and no hops after jumps) as well as other “natural” spaces highly impacted by urbanisation [7]. The methodology used in this paper can be summarised in the following steps: 1.

Conversion of the file supplied by NASA, which offers in the visible spectrum three images (RGB), in a conventional greyscale palette (0-255). The image conversion from greyscale to elevations allows contours to be developed at different intensity levels, capable of identifying different hypotheses of urbanised areas.

2.

Transformation of night-time lights imagery into a point file allows a logistic regression to be performed with the 2010 US Urban Areas (UA) as a dependent variable, and night-time light intensity as an explanatory variable. The logistic regression, adjusted for 18 million points (3.6% of which correspond to UA), allows us to establish a night light intensity level of 164 (in a greyscale from zero to 255 levels of intensity) for the determination of those Areas Highly Impacted by Light (AHIL). The results of this model can be seen in Table 1. The model has an effectiveness level of 86.4% to explain the UA pattern in USA. Figure 1 shows the result of

5

We understand here as "lined streets" the simple layout of the roads, as well as a slight compaction that allows the access to the parcels resulting from the division of rural property. Not the urbanization in an integral sense (pavement of the vials, construction of sidewalks, street lighting, ...). 6 NOAA provides the DMSP-OLS databases (version 4.1992-2013) on the website http://ngdc.noaa.gov/eog/dmsp/downloadV4composites.html (accessed on 1 November 2015). The files are cloud-free composites made using all the available archived DMSP-OLS smooth resolution data for calendar years. The products are 30 arc second grids, spanning -180 to 180 degrees longitude and -65 to 75 degrees latitude. 7 NOAA currently provides the SNPP-VIIRS data (version 1; January 2014 - March 2016) at the website http://ngdc.noaa.gov/eog/viirs/download_monthly.html (accessed on 8 August 2016). The Earth Observations Group (EOG) at NOAA/NGDC is producing a version 1 suite of monthly average radiance composite images using nighttime data from the Visible Infrared Imaging Radiometer Suite (VIIRS) Day/Night Band (DNB). This version 1 series of composites has not been filtered to screen out lights from aurora, fires, boats, and other temporary lights. This separation is under development and will be included in a later version of this time series. Version 1 spans the globe from 75N latitude to 65S. The products are produced in 15 arc-second geographic grids. 8 Black Marble imagery, http://earthobservatory.nasa.gov/NaturalHazards/view.php?id=79803 (accessed on 1 October 2013).


the model for the UA in Southern California (red = UA; blue = 164 intensity level contour). The Kappa coefficient reaches a value of 0.925, which proves the efficiency of the proposed model 9. Table 1: Logistic Regression Model Model Summary Step 1

-2 log likehood

Classification Table

Cox & Snare Nagelkerke R R square square

840036,373

,231

Predicted Observado

0.869

Artificial 0

Step 1

Artificial

Percentage Correct

1

0

17362503

0

1

87931

559718

Overall Percentage

100,0 86.4 99.5

a. The cut value is .500

Variables in the equation B Step 1

grid_code Constant

E.T. 0.067 0.000

-10.842

.013

Wald

gl

Sig.

Exp(B)

588521.721

1

.000

1.069

652029.337

1

.000

.000

a. Variable(s) entered in step 1: grid_code.

Source: Self prepared from NASA 2013, Census Bureau (USA)

Figure 1: Highly light Impacted (intensity => 164) and Urban Areas in Southern California

si

o

-44- 4

o

Source: Self prepared from NASA 2013, Bureau of Census (USA)

3.

The extension of this methodology at global scale allows identifying AHIL around the world. Figure 2 shows worldwide AHIL. The study identifies 186,134 illuminated contours (urbanised areas) at 164 or higher intensity of night light.

9 The method used to build the logistic model was progressive. Firstly a model was adapted for all of the points analysed, which represented obtaining a “cut-off” light intensity of 198, with an accuracy ratio (with respect to the total of urban areas) of 71.1%. Then the points outside of the UA with a light level over 198 were disregarded, adapting a new logistic regression that allowed us to reduce the level of night-time light intensity to 184, with an accuracy ratio of 79.9%. The successive repetition of this procedure enabled us, after 7 steps, to adjust the model to the aforementioned intensity of 164, with an accuracy ratio of 86.4%. The discarded points, outside of the UAs and characterised by a high night-time light intensity (>164), represented 2.5% of the total of the “rural” points, according to the definition of the USA Census.


Figure 2: Highly Light Impacted Areas

Source: Self prepared from NASA 2013

4.

The delimitation of the centres and peripheries of urban agglomerations is made by setting a threshold of nighttime light intensity that allows the inclusion of most of the centres and sub-centres, defined as the localities with a population over 50,000 inhabitants (according to the criteria of the US Census Bureau to identify the metropolitan centres). This threshold corresponds approximately to a light intensity of 230 10. Moreover, the identification of core areas was established through the division into natural breaks (Jenks) in ArcGis, with five classifications, where “central landscapes” is the highest class, the 230 intensity level, and other intensities (between 164 and 230) represent the Areas Highly Impacted by Light the “periurban landscapes”, as shown in Figure 3. Figure 4 represents the results of such methodology used in the case of Southern California. Figure 3: Natural breaks of intensity night lights (>= 164) Classification

X

Classification

Classification Statistics

Method:

Natural Breaks (Jenks)

Count:

4826476 165

Minimum:

Classes:

Maximum:

Data Exdusion

255

980,250,405

Sum:

Exdusion

203.0985765

Mean:

Sampling ...

21.8235362

Median: Columns:

100

d

['Show Std. Dev.

Show Mean

Break Values

10000

o

181

I_

198

80000

214 231 255

60000

4000

2000

165

187.5

210

OK

232.5

I

255 Cancel

Snap breaks to data v_atues

10

In this study one single threshold (>230) at a world level was used. A subsequent improvement could consider a set of thresholds, depending on the variety of human settlements. For example, if the adjustment is made by standard deviations and continents, the “cut-off” of the central landscapes (mean + 1 standard deviation) at a world level would be 225 of night-time light intensity, with North and South America reaching 228, Europe 222, Asia 223, Africa 224 and Oceania 227. These thresholds would represent a slight increase in the dimensions of the centres, with the consequent reduction of the peripheries. This increase of the centres would mainly affect Europe, and to a lesser extent Asia and Africa. In an equivalent way they would have less of an effect on North and South America, as well as Oceania.


Source: Self prepared

Figure 4: Centres and Peripheries in Southern California

Source: Self prepared from NASA 2013, LandScan 2008

5.

Calculating the population of the AHIL by overlapping information on population of the LandScan database [5] developed by the Oak Ridge National Laboratory [26]. This database allows us to analyse the population structure of different environments on the urbanised planet, with a close approximation to reality. Figure 5 shows the overlap between contours of intensity 164 or greater with LandScan population in Southern California (green =< 50 inhabitants per pixel; orange and red >= 250 inhabitants per pixel).

Figure 5: Overlap between highly light Impacted Areas (164 contours) with LandScan in Southern California

Source: Self prepared from NASA 2013, LandScan 2008

6.

The overlapping allows us to typify world urban areas (or AHIL) based on their population size. This paper differentiates metropolitan areas (with a population > 1,000,000 inhabitants), medium cities (from 100,000 to 1,000,000 inhabitants), small cities (10,000-100,000 inhabitants), villages (between 1,000 and 10,000 inhabitants), and other types of urban settlements (< 1,000 inhabitants).

7.

Once identified urbanised areas, or AHIL, it is necessary to delimit the rural areas, or Areas Little Impacted by Light (ALIL), which are characterised by low intensity night light. For this purpose, it sets a cut-off below which there is evidence of being in rural areas. This cut-off is the threshold of intensity = 164), which indicates the distribution shown in Table 2. Table 2: Population distribution of the AHIL (intensity 164) Population Int. 164 Num. Areas Population % Pop < 1,000 inhab. 130,572 24,077,383 0.82% 1,000-10,000 inhab. 39,479 130,600,243 4.43% 10,000-100,000 inhab. 12,944 403,349,958 13.69% 100,000-1,000,000 inhab. 2,735 764,918,473 25.96% 1,000,000-10,000,000 inhab. 370 952,285,041 32.32% > 10,000,000 inhab. 34 670,749,830 22.77% TOTAL 186,134 2,945,980,928 100.00 Source: NASA, LandScan, self-prepared

11

The definition of ex-urban landscape is particularly complex [29]. The studies carried out to date have been principally based on the density as well as the commuting with urban and metropolitan centres [8]. The absence of information on a world scale on daily home/work trips has determined the use of the criterion used in other studies [1] to delimit the megalopolis, a concept which includes both a considerable artificialisation of the region as well as the relative proximity to urban and metropolitan areas. 12 In the X-axis, the different light intensities (from 0 to 255). In the Y-axis, the accumulated world population.


Four hundred and four of these illuminated areas, 2.2 per thousand of the total areas, can be defined as metropolises that reach a population of over one million inhabitants, which concentrates 1.623 million people, 24.45% of the population worldwide. Thirty-four of those areas are the “proto-megalopolis” with a population exceeding tens of millions. They represent the "seeds" of the megalopolitan structures [1, 17]. On the opposite end of the scale to those giants, 70.15% of the AHIL (with an intensity equal to or greater than 164) do not exceed 1,000 inhabitants and represents 0.36% of the world population. Figure 7 shows, at the same scale, the results of the identification of AHIL and its rurban hinterlands of the five major world agglomerations. Figure 7: Top-five Megalopolises

Sources: NASA, LandScan, self-prepared


The AHIL or urbanised area of Nile is in first place with a population of 78,363,600 inhabitants and a density of 1,975.98 inhabitants per square kilometre. This area extends from the delta, with Alexandria as the main urban centre, passing through Cairo and Aswan and reaches the first cataract. Secondly, the agglomeration of Pearl River in China reached 45,296,306 inhabitants, with a density of 2,878.85 inhabitants per square kilometre. The Pearl River Delta with the Shanghai (36,853,534 inhabitants) and Beijing agglomerations (11,642,291 inhabitants) are the most important metropolitan areas in China. The former, including cities like Hong Kong, Shenzhen, Guangzhou and Macao, is the second urban agglomeration in the world. Shanghai, with Nantong, Changzhou, Wuxi and Suzhou as major cities, is the third metropolitan system in our ranking based on night-time lights. It is clearly differentiated from the Nanjing (6,606,566 inhabitants) and Hangzhou (12,356,984 inhabitants) agglomerations. In the case of Beijing, it does not include de metropolitan areas of Tianjin (7,406,135 inhabitants) or Tangshan (1,827,716 inhabitants). All these metropolises have high-medium and medium densities: 3,349.71 inhabitants per square kilometre in the case of Beijing, 2,878.85 Pearl River Delta, and 2,127.04 Shanghai. Then comes the agglomeration of Tokyo-Yokohama reaching 35,514,940 inhabitants. The metropolitan area of Tokyo appears clearly differentiated from the other urban agglomerations of Hokkaido Island, like Nagoya (10,528,167 inhabitants) or Kyoto-Osaka-Kobe (17,234,951 inhabitants). Tokyo, with 4,274.20 inhabitants/km2, is a high density metropolis, ranking 12 of the 404 world metropolises delimited in this study. Finally this top-five group comes to a close with the metropolitan area of New York - Philadelphia. The metropolitan system of New York (26,581,672 inhabitants) includes Philadelphia, Trenton and Hartford, but not Washington-Baltimore (6,986,575 inhabit.) or Boston (5,243,601 inhabit.), representing the fourth densest metropolis in USA with 1,535.62 inhabitants per km2. However, it is ranked 323 of the 404 world metropolises, showing increased consumption of land per inhabitant associated with processes of urban sprawl 13. Table 3 summarises the data of the thirty-four agglomerations of more than 10 million inhabitants. Table 3: Proto-Megalopolis Rank 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

Metropolis Nile Pearl River Shanghai Tokyo New York Sao Paulo Jakarta Seoul Delhi Taipei Mumbai Mexico DF Bruss-Amst Manila Osaka Calcutta Los Angeles

Population 78,363,600 45,296,306 36,853,534 35,514,940 26,581,672 24,529,896 24,020,441 23,522,132 22,851,644 21,060,702 20,953,305 20,378,028 19,995,071 19,326,521 17,234,951 16,959,857 15,777,380

Density 1,975.98 2,878.85 2,127.04 4,274.20 1,535.62 3,387.03 5,374.38 3,521.00 4,348.95 2,076.97 10,618.07 4,909.35 977.37 1,2319.92 4,435.52 8,572.58 1,987.94

Rank 18 19 20 21 22 23 24 25 26 27 28 28 30 31 32 33 34

Metropolis Buenos Aires Moscow Tehran Karachi Rio de Jan. Bangkok Hangzhou Istanbul Lahore Nord Italy Beijing London Paris Chicago Nagoya Dacha Near Orient

Population 14,272,542 13,423,467 12,975,631 12,534,887 12,529,447 12,428,160 12,356,984 11,988,123 11,881,112 11,858,956 11,642,291 11,420,350 10,672,304 10,567,277 10,528,167 10,303,030 10,147,122

Density 2,994.15 2,774.32 2,355.80 10,438.34 3,706.67 1,538.46 1,889.72 4,498.48 6,346.08 813.34 3,349.71 2,798.50 2,876.65 1,061.83 2,147.45 13,580.16 1,252.08

Source: NASA, LandScan, self-prepared

Once the metropolitan areas had been delimited on a world scale, an attempt was made to characterise these areas (identifying agglomerations with more than 1,000,000 inhabitants) according to their degree of urban sprawl. To do so, the Light Impacted Land per capita (LILpc) indicator, obtained by dividing the area of the night-time light intensity >= 164 by population and expressed in square metre per inhabitant, can serve to identify the Metropolitan Areas subject to high Urban Sprawl. 13 In contrast to New York, the metropolitan area of Los Angeles (15,777,380 inhabitants), despite its reputation as sprawled agglomeration, is the densest metropolis of the USA (and number 298 in the world ranking) with 1,987.94 inhabitants/km2. Metro Los Angeles includes Orange and San Bernardino Counties, but not the metropolis of San Diego, which absorbs Tijuana, in Mexico (4,192,676 inhabit.).


The standardised LILpc value is a good indicator of level of urbanisation and urban sprawl. Values above zero mean they are above the LILpc average. Values above 1 mean LILpc > average + 1 Std. deviation. Negative values mean areas with below-average sprawl. Forty-seven of the 404 metropolitan areas (MA) can be considered sprawled agglomerations (with a LILpc > average +1 standard deviation, that is more than 918 m2/inhabitant). Most of these affected areas are located in USA (27). Oil producing countries also bring a number of significant MA with a high LILpc, perhaps the effect of oil exploitation. Figure 8 shows agglomerations with standardised LILpc greater than 2. Figure 8: Standardised LILpc vs. Rank Population 8,00000

Ahvaz Kuwait

6,00000-

Charlotte Dubai

4,00000

Abha

o

Tripoli

Atlanta

Pittsburgh Génova

Cincinnati

Cleveland

o

®Reasa:WCßUen

CI : 4.4:tie M :: T L O

,00000

O

O

8

O

®

e

%

O

00

O

O O

0

^O

O

00

O

O

°oouoó 0 00 O

9g

Op

O

Oo 'coo o

'07'e0

O

O

OO

O

6,v o ct 00On C

Jacksonville

Oslo

Alicante- Murcia

2,00000 O

Memphis

Indianapolis O

0

St. Louis

Tampa

Nashville O

Raleigh

fr_ etk, O

OO

O

O

00 00

oo

0

OOÓ)

8 co3' ,O 00

OO 0 0 0 00

0O 0. OO.r

- 2,00000-

0

100

200

300

400

Rank_pop

Source: NASA, LandScan, Self-prepared

Leaders in urban sprawl (standardised LILpc > 2) are the 24 following metropolises: Ahvaz (Iran), Kuwait (Kuwait), Charlotte (United States), Dubai (United Arab Emirates), Nashville (United States), Abha (Saudi Arabia), Raleigh (United States), Tripoli (Libya), Memphis (United States), Jacksonville (United States), Pittsburgh (United States), Indianapolis (United States), Atlanta (United States), Cincinnati (United States), Bologna-Modena-Parma (Italy), Tampa (United States), Alicante-Murcia (Spain), Kansas City (United States), Reynosa-McAllen (United States-Mexico), Oslo (Norway), Cleveland (United States), St. Louis (United States), Austin (United States), and Columbus (United States). Figure 9 shows two metropolitan areas with similar population (20 million inhabitants); Brussels-Amsterdam and Mexico DF. The first one impacts 20,458.10 square kilometres, the second just 4,150.86. The LILpc indicator goes from 203.69 m2 per inhabitant, in Mexico City, to 1,023.16, in central the European agglomeration. Figure 9: Brussels-Amsterdam vs. Mexico DF

Source: NASA, LandScan, Self-prepared


The methodology allows, beyond analysing the size of urban and metropolitan systems and their degree of sprawl, a rigorous comparison of the degree of urbanisation of different territories. Figure 10 compares the night-light footprint of India and United States. In the X-axis night-light intensity from zero to 255 and in the Y-axis the population. As can be seen, while in US (left image) the population remains at very low levels until the light intensity of 164, growing exponentially to the intensity of 230, and then decreasing, in India (right image) the peak population occurs in the intensity 44 before falling sharply until it is overcome at the intensity of 164. The degree of urbanisation or ruralisation, can be measured quite accurately by the surface (or population cumulative) below night-light curves between the standards used in this study (64, 164 and 230). This measure is certainly much closer to reality than the statistics of rural/urban population provided by United Nations, which suffer from a high variation of criteria based on national criteria. Figure 10: Footprint of night-time light of US and India

I


In this regard, Figure 11 shows the diversity of patterns of night-time light footprint between continents. The image presents the light intensity on the X-axis, and the cumulative population for continents on the Y-axis. Africa and Asia concentrate most of their population in the darkest areas, denoting their still strongly rural character. North America, South America, Oceania and Europe instead concentrate most of their population in brightly lit landscapes, denoting high levels of urbanisation. The analysis of the different morphology of the night-time lighting footprint between different territories provides a potential for research that will certainly enrich the knowledge of the degree of urbanisation of human landscapes. Figure 11: Night-light footprint (cumulative) by continent Continent 1,0000-

Asia Africa Oceania North America South America Europe

,8000-

,4000-

,2000-

,0000r

r

r

0

50

700

r

750

200

Hum



250

The analysis of the night-time lights enabled us to make an initial approximation of the spatial distribution to the different landscapes of human settlements around the world (Table 4). The study classifies four different landscapes, rural areas ( = 164, that rose to 86% for urbanised landscapes, and 97.5% for rural. It can therefore be concluded, with regard to the USA, that the analysis of the results obtained in the delimitation of the AHIL has allowed a positive evaluation of the methodology used in the study. The extent of this method on a world scale has made it possible to identify the urbanised areas of the whole planet, and characterise them in virtue of their population and degree of impact of their urbanisation (the latter measured by means of the indicator Light Impacted Land per capita). In this regard different human landscapes have been identified, characterised by a different impact of the night-time lights. Compact landscapes have been differentiated, typical of the central nuclei of the urban and metropolitan systems; the suburbs characteristic of urban sprawl as well as of suburban developments; the rurban landscapes, where intense processes of rural/urban transformation occur and where processes coexist that may be characterised as ex urban and rural sprawl; as well as the dark, intrinsically rural landscapes. The study has therefore enabled the construction of a method of urban/rural classification that we believe to be more efficient than the current distinction supplied by the United Nations. In addition, this method has the advantage of enabling the analysis of the urban and rural population at a subnational level. The rural/urban segmentation can no longer be understood as dual. The diffusion of the urbanisation networks in practically all of the developed world leads to the need to go beyond this dual vision, contrasting between rural and urban. Thus the notion of urbanisation gradient emerges as an essential concept in analysing urbanised landscapes. This relative success of using night-time lights as a proxy for urbanisation does not allow us to give any assurances, nevertheless, that the method is likewise valid for evaluating the great wealth of human settlement existing on the planet. The great diversity of land occupation processes, together with the different degree of electrification throughout the geography of the planet, suggests the need for further analysis to test the thresholds used in this research. It is difficult to verify on a world scale the validity of the models constructed as there is no reference database on a world scale on urbanised areas. The verification must therefore be carried out on a local scale and, even so, with enormous difficulty. For the case of Europe, for example, it is possible to verify the results, comparing the AHIL as well as the rest of the lit landscapes with the resulting degree of urbanisation from the Eurostat DEGURBA database. For the rest of the world, the contrast is even more difficult, as there are no global studies of ground uses and cover; we must resort to strictly local analyses. Therefore further studies are needed to confirm the performance of the model on other continents, such as Asia or Africa.

CONCLUSIONS The dispersion of the urbanisation networks therefore seems to observe different patterns throughout the geography of the planet. While in Europe this dispersion occurs in rurban landscapes, in Africa a high percentage of the population still resides in rural landscapes, which represent the main potential of urbanisation and will generate massive migration processes from rural to urbanised landscapes. Meanwhile, Asia would be in an intermediate position, receiving still strong contingent of rural-urban migration, but also observing an accelerated process of converting the rurban interstices into fully urbanised landscapes. North America seems to already have a weak capacity for increasing urbanisation. North America is not only the most urbanised area in the world, is also the region showing the lowest proportion of the population living in rural areas. Urban development potential lies mainly in the rurban landscapes. Finally, South America seems to have the potential to increase its urbanising process. We can summarise the main conclusions of the research as follows: 1.

The study shows that satellite night-time lights have a high potential for analysing urbanisation. The arrival of electrification to nearly every corner of the planet is certainly the first and most meaningful indicator of


artificialisation of land. In this sense, the paper develops a new methodology designed to identify the highly impacted landscapes worldwide based on the analysis of satellite imagery of night-time lights. 2.

The methodology used in the paper allows us to detect urban areas worldwide, or Areas Highly Impacted by Light (AHIL). The study identifies 186,134 illuminated contours at 164 or higher intensity of night-time light. 404 of these contours, 2.2 per thousand of the total, can be considered real “metropolitan areas”, concentrating 1.623 million people, 24.45% of the population worldwide. 34 of those areas, or proto-megas, exceed 10 million, representing the "seeds" of the megalopolitan structures. In contrast to those giants, 70.15% of the AHIL do not exceed 1,000 inhabitants, accounting for 0.36% of the world population.

3.

We have used the Light Impacted Land per capita (LILpc) as an indicator of Urban Sprawl and considered sprawled agglomerations those with a LILpc > average + 1 standard deviation, in other words more than 918 m2/inhabit. 47 of the 404 Metro Areas are in that situation. Most of these affected areas are located in the USA (27). Oil producing countries also bring a number of significant MA with a higher LILpc, perhaps the effect of oil exploitation. Metropolises like Kuwait (Kuwait), Dubai (United Arab Emirates), Nashville, Memphis, Jacksonville, Pittsburgh, Indianapolis, Atlanta, Cincinnati, Austin, Kansas City, St. Louis, Cleveland or Columbus (United States), BolognaModena-Parma (Italy), and Alicante-Murcia (Spain) are the best examples of sprawled agglomerations.

4.

The methodology developed allows us to compare the degree of urbanisation of the different territories and countries. The degree of urbanisation (or ruralisation) can be measured quite accurately by the population accumulated under night-light curves between the standards used in this study (64, 164 and 230). This measure is certainly much closer to reality than the statistics of rural/urban population provided by UN-HABITAT, which, as we have seen, suffer from a high variation of criteria to be based on national criteria.

5. This paper allows the identification of the different typologies of human settlement: “rural”, “rurban”, “periurban” and “urban” landscapes. Of the worldwide population 40.26% lives in rural areas, 15.53% in rurban spaces, 26.04% in suburban areas and only 18.16% in central areas. In Africa and Asia rural landscapes are notable, with 63.29% and 43.59% of the population living in dark areas, compared to Europe and North America, where the percentages of the population living in areas of low intensity night light are 17.65% and 13.02% respectively. Urbanised landscapes stand out in North America (76.99%), Oceania (70.11%), South America (68.25%) and Europe (67.76%), while in Asia (37.01%) and especially Africa (30%) they are still in the minority. Finally, rurban landscapes have a special role in Asia (19.4%) and, to a lesser extent in Europe (14.58%).

REFERENCES [1] Arellano, B. and Roca, J., “Las megaestructuras urbanas del siglo XXI: un ensayo para la identificación y caracterización de las megalópolis”, ACE, Architecture, City and Environment, 9 (25), 127-158 (2014). [2] Bauer, G. and Roux, J. M., [La rurbanisation ou la ville éparpillée]. Paris, 1976. Éditions du Seuil (1976). [3] Berg L. van den, Drewett, R, Klaasen, L. H., Rossi, A. and Vijverberg, C. H. T., [Urban Europe: a study of growth and decline], Elmsford, N.Y./Oxford, England, Pergamon Press, xxii, (1982). [4] Berry, B. J. L., [Urbanization and counterurbanization], Beverly Hills, California. Sage Publications (1976). [5] Bhaduri, B., Bright, E., Coleman, P., and Urban, M., “LandScan USA: A High Resolution Geospatial and Temporal Modeling Approach for Population Distribution and Dynamics”, GeoJournal, 69, 103-117 (2007). [6] Brenner, N. and Keil, R., “From Global Cities to Globalized Urbanization”, Glocalism: Journal of Culture, Politics and Innovation. 2014, 3 (2014).


[7] Bureau of the Census, “Proposed Urban Area Criteria for the 2010 Census”, Federal Register, Typology 75, No. 163 / Tuesday, August 24, 2010 / Notices, 2010. [8] Clark, J., Munroe, D. and Irving, E., 2006, “Exurban Settlement Pattern and the Exurban Condition: A of US Metropolitan Areas”, 53rd Annual North American Meetings of the Regional Science Association, Toronto, November 16, 2006. [9] Daniels, T., “What to do about rural sprawl?”, American Planning Association Conference. Seattle, WA, 1999. [10] Davis, K. and Hertz Golden, H, “Urbanization and the Development of Pre-Industrial Areas”, Economic Development and Cultural Change, 3 (1), 6–26,1954. [11] Elvidge, C.D., Imhoff, M.L., Baugh, K.E., Hobson, V.R., Nelson, I., Safran, J., Dietz, J.B. and Tuttle, B.T., “Nighttime lights of the world: 1994-1995”, ISPRS Journal of Photogrammetry & Remote Sensing, 56, 81-99, 2001. [12] Elvidge, C.D., Baugh, K., Zhizhin, M. and Hsu, F. C., 2013, “Why VIIRS data are superior to DMSP for mapping night-time lights”, Proceedings of the Asia-Pacific Advanced Network, 35, 62-69, 2013. [13] Elvidge, C.D., Baugh, K.E., Anderson, S.J., Sutton, P.C. and Ghost, T., “The Night Light Development Index (NLDI): a spatially explicit measure of human development from satellite data”, Social Geography, 7, 23-35, 2012. [14] European Environment Agency, [Urban sprawl in Europe: the ignored challenge], Report Nº 10/2006, Copenhagen, EEA, 2006. [15] European Commission, [A harmonised definition of Cities and Rural Areas: The new Degree of Urbanisation], Working Paper 1/2014, 2014. [16] Eurostat, European Commission, [Regional Year Book 2010], Luxembourg, Publications Office of the European Union, 2010. [17] Florida, R., Gulden, T. and Mellander, C., “The rise of the mega-region”, Cambridge Journal of Regions, Economy and Society, 1 (3), 459-476, 2008. [18] Friedl, M. A., McIver, D. K., Hodges, J. C. F., Zhang, X. Y., Muchoney, D., Strahler, A.H., Woodcock, C. E., Gopal, S., Schneider, A., Cooper, A., Baccini, A., Gao, F. and Schaaf, C., “Global land cover mapping from MODIS: algorithms and early results”, Remote Sensing of Environment, 83, 287–302, 2002. [19] Fulton, W., Pendall, R., Nguyen, M. and Harrison, A., [Who Sprawls Most? How Growth Patterns Differ Across the U.S. Center on Urban & Metropolitan Policy], The Brookings Institution, Survey Series. Washington, D.C., 2001. [20] Galster, G., Hanson, R., Ratcliffe, M., Wolman, H., Coleman, S. and Freihage, J., “Wretsling sprawl to the ground: Defining and measuring an elusive concept”, Housing Policy Debate 12 (4), 681-718, 2001. [21] Hansen, M. C., Defries, R. S., Townshend, J. R. G., and Sohlberg, R., “Global land cover classification at 1 km spatial resolution using a classification tree approach”, International Journal of Remote Sensing, 21(6– 7), 1331– 1364, 2000. [22] Indovina, F., [La città diffusa], Quaderno Daest n. 1, IUAV, Venezia, 1990. [23] Loveland, T. R., Reed, B. C., Brown, J. F., Ohlen, D. O., Zhu, Z., Yang, L. and Merchant, J. W., “Development of a global land cover characteristics database and IGBP DISCover from 1 km AVHRR data”, International Journal of Remote Sensing, 21:6-7, 1303-1330, 2010. [24] Margalef, R., “Acelerada inversión de la topología de los sistemas epicontinentales humanizados”, in Naredo, J.M., Gutiérrez, L., [La incidencia de la especie humana sobre la faz de la tierra (1955-2005)], Universidad de Granada/Fundación César Manrique, Granada, 217-219, 2005.


[25] NASA Earth Observatory, “Night Lights 2012 - The Black Marble”, http://earthobservatory.nasa.gov/NaturalHazards/view.php?id=79803 (30 October, 2013). [26] Oak Ridge National Laboratory, “LandScan™”, http://web.ornl.gov/sci/landscan/, 2008(3 March, 2014). [27] Ridd, M. K., “Exploring a V-I-S (vegetation - impervious surface - soil) model for urban ecosystem analysis trough remote sensing: comparative anatomy for cities”, International Journal of Remote Sensing, 16- 12, 2165-2185,1995. [28] Short Gianotti, A.G., Getson, J.M., Hutyra1, L.R. and Kittredge, D.B., “Defining urban, suburban, and rural: a method to link perceptual definitions with geospatial measures of urbanization in central and eastern Massachusetts”, Urban Ecosyst 2016-19, 823–833, 2016. [29] Taylor, L., “No boundaries: exurbia and the study of contemporary urban dispersion”, GeoJournal 7(2011) 76: 323339, 2011. [30] Theobald D.M., “Placing exurban land-use change in a human modification framework”, Ecol Environ 2(3), 139– 144, 2004. [31] Theobald, D. M., “Landscape Patterns of Exurban Growth in the USA from 1980 to 2020”, Ecology and Society, 10(1), 32. URL: http://www.ecologyandsociety.org/vol10/iss1/art32/, 2005. [32] Tsai, Y. H., “Quantifying Urban Form: Compactness versus ‘Sprawl’”, Urban Studies, 42(1), 141-161, 2005. [33] United Nations, [World Urbanization Prospects: The 2001 Revision], United Nations publication, ESA/P/WP 173, New York, 2002. [34] United Nations, [World Urbanization Prospects: The 2011 Revision], United Nations publication, ST/ESA/SER.A/322, New York, 2012 [35] United Nations, [World’s population increasingly urban with more than half living in urban areas], New York, United Nations publication, UN http://www.un.org/en/development/desa/news/population/world-urbanization-prospects2014.html, 2014. [36] United Nations, [Urbanization and Development: Emerging Futures], World Cities Report, United Nations Human Settlements Programme (UN-Habitat), 2016. [37] United States Census Bureau, [2010 Census Urban and Rural Classification and Urban Area Criteria], https://www.census.gov/geo/reference/ua/urban-rural-2010.html (22 September, 2014). [38] Wolman, H., Galster, G., Hanson, R., Ratcliffe, M., Furdell, K. and Sarzynski, A., “The Fundamental Challenge in Measuring Sprawl: Which Land Should Be Considered?”, The Professional Geographer, 57 (1), 94-105, 2005.
