1 Introduction

Since its reforms, following economic liberalization and surge of foreign direct investment, Indian cities have observed rapid growth of population and spatial expansion beyond their municipal boundaries [1]. In the post-economic reform period, most of the large cities in India (urban settlements with a population size of 1 lakhs and more) have observed rapid outward expansion which leads to the actual spatial extension of these cities far beyond its territorial limits [2, 3]. The lateral spread of larger cities in India resulted in bio-physical, morphological, socio-demographic, cultural, economic and functional transformations [4]. Therefore, monitoring, measurement and obtaining accurate and timely information about the dynamics of spatial expansion of larger cities is crucial to promote an effective urbanization, better governance and sustainable urban development in India.

Census of India is the only official source to obtain demographic, socio-economic, amenities and areal expansion data. Census of India gives information on demographic, the socio-economic and areal expansion of the urban settlements in India with ten yeas interval. Indian census has a unique concept of urban agglomerations (UAs) to describe the outward spatial expansion of existing cities. Urban agglomeration (UA) denotes a continuous urban spread and generally comprises of a town (statutory or census town) and its adjoining outgrowths (part of the settlement that acquired urban characteristics in terms of infrastructure and amenities). According to 2011 census, the number of Urban Agglomerations (UAs) in India is 475 (which was 384 UAs in census 2001), of which 53 UAs/Cities are with a population of one million or more (Million Plus cities). These million plus cities are distributed across states of Uttar Pradesh (7), Kerala (7), Maharashtra (6), Madhya Pradesh (4), Gujarat (4), Tamil Nadu (4), Rajasthan (3), Jharkhand (3), Andhra Pradesh(3), Punjab (2), West Bengal (2), Chhattisgarh (2), J&K (Jammu and Kashmir) (1), Chandigarh (1), Haryana (1), NCT (National Capital Region) Delhi (1), Bihar (1) and Karnataka (1). The mega cities in India are Greater Mumbai (18.4 million), Delhi UA (16.3 million) and Kolkata UA (14.1 million). The other important cities are Chennai UA (8.7 million) and Bangalore UA (8.5 million) and Hyderabad (7.6 million). Besides these, other urban agglomerations like Ahmedabad, Jaipur, Poona, Kanpur, Lucknow and Surat currently have more than 2.5 million habitants. It is projected that the number of Indian megacities will increase from three (Mumbai, Delhi and Kolkata) to six (including Bangalore, Chennai and Hyderabad) by 2021.

The increasing number of urban agglomeration and growth of population in the existing UAs is indicative of rapid spatial expansion. But the problem with census approach is that for describing the lateral spread of the cities in India, it only considers the adjoining areas of a city/town only if it is a statutory town or settlements which satisfy the threefold criteria of 5000 population, density of 400 persons per sq. km and 75% of the male population engaged in the nonfarm sector or a part of the adjoining settlements that have acquired urban morphological characteristics and similar infrastructural facilities that of a urban settlement. In doing so it completely bypasses the adjoining transitional settlements (popularly termed as peri-urban, urban fringe etc.) which are important and most dynamic parts of the lateral spread of the existing cities. In that sense, the census data lack in spatial information and therefore it is difficult to monitor the spatial expansion of urban settlements. Of recent medium to high-resolution satellite images become very popular to study dynamics of urban expansion of individual cities or metropolitan areas in India [5,6,7]. Satellite images (Landsat TM, Modis data, SPOT HRV) contains a large amount of spatial information enables continuous monitoring of spatial expansion of individual cities or metropolitan areas in India and also helps to overcome the limitations of census data. But widely used satellite imageries (Landsat TM, SPOT HRV) are limited in its geographical coverage and for that reason, it is difficult to obtain information of spatial expansion of towns/cities on a larger spatial scale.

The defence meteorological satellite program (DMSP) has an operational line-scan system (OLS), which is a new and reliable data source for exploring the dynamics of urban expansion at a city level/regional level/country level/global level [8, 9].

The DMSP–OLS night time light data are widely used and considered as most preferred and reliable source data to study the urban expansion process in China but in India, some experimentation has been done with night time data in the study of urban expansion but limited work is done using any threshold methods. This study has used night time DMSP/OLS satellite data for the years 1992, 2001, 2011 over three Megacities (Greater Mumbai, Delhi and Kolkata) and One Incipient mega-city (i.e. Chennai being the largest among incipient mega-city and the others are Bangalore and Hyderabad) to detect urban footprints and their changes over time. The time period is chosen consciously to keep comparability with the census of India. These four cities are situated in four parts of India, from Delhi (North), Chennai (South), Kolkata (East) and Mumbai (West) and are known urban identities since the colonial period.

2 Data

DMSP operates in sun-synchronous orbits and passes through the Indian region in between 7 and 10 pm on local time. The OLS sensor is different from those used to detect ground objects based on the reflection characteristics of solar radiation such as Landsat TM, SPOT HRV, and NOAA AVHRR.

DMSP–OLS dataset contain three types of data: (1) cloud-free coverage, (2) night time light data with no further filtering, and (3) night time stable light (NSL) data. Night time stable light (NSL) Stable light image data has been used (1992, 2001, 2011) for this study which was obtained from National Oceanic and Atmospheric Administration, USA captured by F10 (1992), F15 (2001), F18 (2011) DMSP/OLS satellites. Night time stable light (NSL) image refers to a temporal composite build-up of cloud-free images of the Earth at night over a six-month (October–March) period. Data from each DMSP satellite’s OLS are cloud-free annual composites with 1-km resolution.

This NSL data were corrected from sunlight, moonlight, glare, clouds and lighting features and from auroras. An event like fire affected pixels also excluded from the composite image. The annual composite shows the average value of the whole year of different urban areas. The spatial resolution of this NSL image is 30 arc second with a DN (Digital Number) value ranges between 0 and 63. In cases where two satellites had collected data, two separate composites were produced [10]. After obtaining the NSL data administrative boundaries for four Indian megacities and its surrounding districts were extracted to obtain the urban extension and dynamics of urban spatial patterns. Extracted DMSP/OLS data were then projected to UTM projection.

City lights can be easily identified because they exhibit high-intensity light pixels and are relatively more stable, represented as small areas of saturated DN (Digital Number) values decreasing outward from the city core towards the periphery. Therefore, DMSP data provided enough information for mapping the large city footprints and its dynamics of spatial expansion over the years. In India, the urban centers along with industrial, agricultural and domestic sectors, consumes most of the commercial energy.

Statistical data of urban areas (2011) for four metropolitan cities of India (Kolkata, Mumbai, Chennai and Delhi) and its surrounding districts were collected from the census of India, district census handbook, 2011.

Google earth image is also used to examine the accuracy of extracted urban areas from DMPS/OLS night time image.

3 Methodology

3.1 Data pre-processing

Extracted data sets for four Indian megacities and its surrounding district for years 1991, 2001, 2011 are projected to UTM projection for evaluating the urban expansion in surrounding areas of these cities (Fig. 1). As a result of the absence of the onboard calibration of the OLS, this data cannot be used directly for monitoring dynamics of urban growth and its evolution [9, 11]. On the other hand, sensor degradation and a difference in the orbital path can make difference in magnitude and number of lit pixels for same ground areas covered by the particular one pixel [11]. The DMSP/OLS data also have some shortcomings due to some unstable Source of light (such as light emitted from fishing boat, rural areas, wildfire etc.) Outside the actual urban area. In order to reduce these discrepancies and extracting urban areas in DMSP/OLS data must be preprocessed. Most commonly utilized methods to explore the dynamics of urban expansion using DMSP/OLS data are global fixed and local-optimized thresholding [12, 13]. or inter calibration method [14, 15]. In this study, as it is used data for three selected years this study uses threshold method instead of an inter-calibration method.

Fig. 1
figure 1figure 1

DMSP/OLS night time imageries for Chennai, Delhi, Kolkata and Mumbai for 1992, 2001 and 2011

Full size image

A uniform threshold value is needed for comparing between different cities and for different year under evaluation. According to the census of India 2011 handbook, urban areas for four Indian megacities are calculated, and then a sum of the maximum and the minimum pixel value is divided by two for getting initial threshold value of the lit pixels. If it is found that the census-derived urban area is less than the digital number (DN) minimum derived pixel value, minimum DN value should increase by one. When DNmin pixel value derived urban area become less then the census-derived urban area and DNmax pixel value derived urban area become larger than the census derived urban area, then this set of DNmin and DNmax pixel value will be considered for calculating DNT (threshold pixel value). The threshold value is then used to extract the actual urban areas for 4 megacities for the year of 1992, 2001, 2011 (Fig. 2).

Fig. 2
figure 2

Flowchart shows the methodology for determining threshold value for urban light area extraction

Full size image

Where, DNT represents the threshold value, DNmax and DNmin is the maximum and minimum DN value of the image, area (DNT−1) is the area of lit pixels of threshold value − 1 value and higher pixels, area (DN+1) is the area covering by threshold pixel + 1 and greater pixels.

3.2 Extraction of urban areas

The threshold value determines using dichotomy method for comparing between different cities and different years has been used to differentiate non-urban areas from urban areas by the reclassify tools of Arc GIS 10.3.1. The lit pixel of DN (Digital Number) value 31 and more is assigned as urban areas and less than 31 DN (Digital Number) value pixels are assigned as non-urban areas.

3.3 Calculation of urban areas from census of India data

Census of India, 2011 provides data about the extent of urban areas which includes Municipal Corporation, municipality, census town and outgrowth. The urban areas of Kolkata (2005.02 km2), Mumbai (1967.31 km2), Chennai (1377.25 km2) and Delhi (2654.74 km2) and surrounding districts of both of this four cities for the census year 2011 are sum up to get the actual urban extent of cities and its surrounding districts, this census of India urban extent area and urban extent derived from DMSP/OLS data is compared to find out the differences between urban areas because there are some loopholes of census of India such as there is a time gap between census data collection and publication of collected data, so there is a chance of lower estimation of urban areas of 2011 as this areas are rapidly increasing its spatial extension. The time gap between census data collection and publication can increase the urban extent. So, it is necessary to compare this two data extracted urban areas for find out the accuracy of both of this two data.

3.4 Validation of result with Google earth

The urban extent map of the four metropolitan cities extracted from DMSP/OLS data for the year 1992, 2001 and 2011 was validated with Google earth images for the years 1992, 2001 and 2011. Validation of extracted urban area from DMSP/OLS night time imagery was conducted by generating 100 random points for whole areas and 20 points are selected which is placed on the isolated urban areas or edge of the extracted urban areas to justify whether any other sources of light out of the actual urban areas is considered as urban areas or not. 20 points are selected for each of the four megacities and its surrounding areas. 1992, 2001 and 2011 urban area image was validated with the Google earth image of the respective years (historical Google earth image of 1992, 2001 and 2011). The points are then exported as kml file to Google earth. The footprints of 20 random points are classified into two categories that are urban and non-urban areas.

4 Results

4.1 Threshold calculation and urban area extraction

Area of four Indian megacities and its surrounding districts has been determined and extracted using the administrative boundaries which excludes the lit pixels of a fishing boat near the coastal area because it only demarcates the territorial area. But it has some difficulties for using this night time image directly because it includes the effect of some unstable light pixels. A threshold value is determined to get the exact urban area of a year for excluding those unstable light sources. Threshold value calculated for four Indian megacities are Chennai (31), Delhi (31), Kolkata (34) and Mumbai (32), In this study 31 DN (Digital Number) value is determined as the threshold value and extract the actual urban area for four mega cities of India for the year of 1992, 2001, 2011 which shows a dynamic nature of urban growth for both of this four cities (Fig. 3).

Fig. 3
figure 3figure 3

Urban areas of four mega cities of India from DMSP/OLS night time image using threshold value for the year 1992, 2001, 2011

Full size image

4.2 Rate of urban growth

Figure 4 shows overlay analysis of derived areas of urban lands uses from DMSP/OLS night time image for the year 1992, 2001, 2011. The overlay analysis exhibits remarkable spatial expansion of Mumbai, Delhi, Kolkata and Chennai for the last three decades (1992–2011) but the rate of lateral spread is not uniform in all direction. The megacities, like Delhi, Kolkata and Mumbai and the incipient mega city of Chennai clearly shows an increase in light intensity over time outward from the core of the city is indicative of the fast expansion of built-up density in the peripheries. DMSP/OLS night time image analysis indicates peripheral rural and transitional lands are rapidly converted into urban uses in Mumbai, Delhi, Kolkata and Chennai metropolitan region. The pattern of lateral expansion of Mumbai and Chennai is similar with growth has slowed in the last decade over the previous one.

Fig. 4
figure 4

Growth of urban areas of Chennai, Delhi, Kolkata and Mumbai for 1992, 2001, 2011

Full size image

The lateral spread of Chennai was much faster in between 1992 and 2001 (growth rate was 84.4% over the existing urban land) but the rate of growth has slowed down in the last decade (grown at a rate of 60.25% in between 2001 and 2011). The lateral spread of the largest urban agglomeration in India (i.e. Mumbai) has also slowed down in between 2001 and 2011 (23.44%) over the previous decade (57.3%). While, in Kolkata and Delhi, the rate of outward expansion is continuously increasing. The rate of lateral expansion of Kolkata was 27.48% over the existing urban area in between 1992 and 2001 which has increased to 41.74% in for the period of 2001–2011. The national capital, Delhi, is continuously expanding in the last two decades (The rate of spatial expansion was 44.56% in between which is increased to 59.78 during the period of 2001–2011 (Table 1). Therefore, it can be said that urban growth trend is shifting from Mumbai and Chennai to Kolkata and Delhi but till now the fastest growing megacity of India is Chennai in terms of 1992, 2001 and 2011 DMSP/OLS night time image analysis.

Table 1 Actual and percentage growth of four Indian megacities for 1992–2001 and 2001–2011 time span
Full size table

The megacities of Delhi, Kolkata and Mumbai show a unique morphological pattern of growth with high built-up density at the core and growing periphery development (spatial extension and lateral spread of the city). It is found that urban area of Chennai is growing towards the surrounding district of Kancheepuram than any other adjoining districts in an elongated pattern along transportation lines. The spatial extents of Delhi exhibit a circular pattern of growth and lateral spread of the city is faster towards adjoining districts of Gurgaon and Gautam Buddha Nagar than other surrounding districts. 2001 and 2011 DMSP/OLS data shows the elongated growth of urban area of Kolkata mega city toward South and North 24 Parganas. Mumbai is the least growing megacities of India in respect to 2001 and 2011 data analysis (23.44%) and growth rate is almost same for both of two surrounding districts of Mumbai. Mumbai is also growing in an elongated pattern.

4.3 Validation with census data

Table 2 shows comparison between census identified urban areas extend data and DMSP/OLS derived data, census calculated urban area data is calculated which includes Municipal Corporation, municipality, census town and outgrowth area. It shows that in case of Kolkata and Mumbai the difference between urban area extracted from census of India and DMSP/OLS night time image data is 7.12 and 1.74% which is acceptable DMSP/OLS night time image extracted urban area extract in comparison to census-defined urban areas. But in case of Chennai and Delhi difference between extracted urban areas from census data and DMSP/OLS night time data is 15 and 13.81% which shows a moderately high difference between two data sets. But census data collection procedure starts more or less 2 years before the data published in 2011 but DMSP/OLS data was captured in 2011 so this 2 years gap can be a cause of this difference between extracted urban areas from census and DMSP/OLS night time image. So, the difference between two data sets will reduce as the urban area is continuously growing day by day if the both data used for 2011.

Table 2 Comparison between census of India calculated urban area and DMSP/OLS deriver urban area for 2011
Full size table

4.4 Validation with Google earth

For validating the extracted urban area (2011) to test whether the threshold value can identify the actual urban area or not, 20 random points have been selected from randomly generated 100 points (only those 20 points are selected which placed on isolated extracted urban areas or placed at the edge of the extracted urban areas to test whether another source of emitted light is considered as urban areas or not) urban areas of each urban city then this points are overlying on Google earth map that shows a very high rate of accuracy of the extracted urban areas from DMSP/OLS night time image. Out of the 20 random points, all points are overlying on urban area in Google earth image for Delhi and Mumbai and 19 points out of 20 (95%) overlying on urban areas in Google earth image (Table 3). So, it can be said that calculated threshold value is appropriate for urban area extraction for four Indian megacities.

Table 3 Validation result of DMSP/OLS derived urban areas with Google earth for the year 2011
Full size table

5 Discussion

The study has demonstrated pattern of lateral spread of the megacities and its surrounding areas of Mumbai, Delhi and Kolkata and incipient mega city of Chennai for years of 1992, 2001 and 2011. DMSP–OLS NTL data facilitated in identifying the specific patterns of growth of Mumbai, Delhi, Kolkata and Chennai. Studies using DMSP/OLS data was of national or global scale but this study shows that it is also relevant for city-specific studies at small scale. In the post-reform era as a result of intensification of city-specific economic activities the megacities, incipient megacities and the larger urban agglomeration (agglomerations with more than 2.5 million inhabitants) started to grow outward leads to rapid peripheral growth and development of city region. The growth of city region around mega cities, incipient megacities and the larger urban agglomeration can be monitor through DMSP/OLS night time image. DMSP/OLS have shown encouraging results to map lateral spread of all megacities and Chennai (incipient megacities). The dataset adequately provides the detailed pattern and specific feature of growth of Mumbai, Delhi, Kolkata and Chennai. It has been found that the growth rate of Chennai and Mumbai have slowed down over the years as the main center is saturated with the present population, while Kolkata and Delhi continue to grow and affects its surrounding areas rapidly for providing shelters for the surplus population of the city areas. The spatial extension of these four cities has acquired specific morphological forms. Mumbai, Kolkata and Chennai showed an elongated and disaggregated pattern of growth unlike Delhi, which had grown radially. These patterns are indicative of the spatial and economic linkages of these cities with its adjoining districts. The rapid development of periphery (spatial extension and lateral spread of the city) in Mumbai is based on the newly established industries (IT industry and Pharmaceuticals chains). Chennai is mainly growing towards the Vellore city and coastal areas (major drivers are automobile industries and IT industries). Kolkata is growing toward adjoining the North and South 24 Parganas district due to the shifting of different economic centers outside the Kolkata and rapid development real estate and new township (New Town Project). Delhi is radially expanding towards its adjoining districts of National Capital Region. So, it is found that all this four megacities are expanding towards the economic centres and following the main road communication for better development of the city’s economic activities.

A study on Indian urban growth tries to find out the influencing factors for urban growth in Indian cities especially for the Indian megacities where he found that manufacturing industries, different types of secondary and tertiary services, proximity to big cities and coverage of primary school areas are the main factors of urban growth in Indian big cities [16]. Whereas another study identified that increasing educational level become a pull for migration from rural areas to urban areas, there are some other factors which they identified are non-agricultural employment and development of small and medium town for urban development of Indian megacities [17].

Another factor that is found from previous studies and also from his studies that main core of the city is saturated earlier 1992 but the population of these cities are continuously increasing for providing shelter of the surplus population of the main city surrounding areas are converting to urban areas creating sprawl conditions. Sprawl areas also impacted the bio-physical environment of the city surrounding as well as city cores. One of the main problems of city increase is slum development which is also same for these cities. These factors are responsible for urban lateral expansion of four Indian megacities.

The growth of Kolkata agglomeration results in conversion of East Kolkata wetland, which becomes international importance for its service to the people and diverse ecosystem. It also converted agricultural lands into urban areas and reduces vegetation cover of that area which is also true for other megacities of India considered for this study. Some largely affected eco-environmental problems related to urban growth in four Indian megacities are air quality depletion, vegetation cover reduction, water shortage (groundwater depletion), and increasing stormwater flood in urban areas due to the reduction of perviousness etc. which need to be considered for sustainable urban planning.

So, it can be said that the study successfully demonstrated the dynamics of urban growth of all mega cities, and one incipient megacities using DMSP/OLS night time image (1992, 2001 and 2011). Mapping and monitoring of growth of these cities are important to provide dynamics of urban footprints and reduce the discrepancies other datasets (census of India or MODIS or LANDSAT TM).

6 Conclusion

This study explores the growth dynamics of four Indian big cities which is very important in Indian economy as well as on the basis of population carrying capacity, so continuous monitoring of urban spatial growth is essential for sustainable development and planning formulation. This study proves that it is better to use DMSP/OLS night time image for continuous urban monitoring and extract exact urban areas by using the capacity of capturing lights from the urban areas. Although this sensor captures some other lights emitted from some other sources rather than urban areas, a threshold based approach is used to eliminate light emitted from other sources and eliminate the impact of overglowing characteristics of DMSP/OLS night time image. Analysis indicates that Chennai is the most rapidly growing city out of these four Indian cities followed by Delhi, Kolkata and Mumbai, but Chennai and Mumbai shows decreasing trend of urban growth in 2011 from 2001 and Delhi and Kolkata shows increasing trend of urbanization in 2011 from 2001. On the other hand, Kolkata and Mumbai show less difference between census of India data and DMSP/OLS night time image deriver urban areas but Chennai and Delhi shows higher level of difference between this two data sets compare to other two cities. So, it can be said that DMSP/OLS night time data is applicable for studying urban growth phenomena; also it is helpful to extract actual urban areas using its light capturing capability for continuous monitoring of urban growth.