Introduction

Urbanisation is one of the four demographic mega trends along with population growth, ageing, and migration (United Nations, 2019). Urbanisation has different forms. It has often been defined as a phenomenon involving the conversion of rural landscapes to urban forms through a continuous expansion of urban boundaries (Han et al., 2009). Similarly, it is also defined as a process that involves rural–urban migration, resulting in the transformation of land (Abbas, 2016). Urbanisation is multifarious as it involves changes in demographics, land covers and ecology (Tian et al., 2022). With just 30% of the population residing in urban areas in 1950, today's world is mainly urban, with more than 55% of the population across the globe living in areas with urban character (United Nations, 2019). It is anticipated that the share of the urban population will be two third (66%) by 2050, resulting in the urban population exceeding 6 billion (Angel et al., 2016). Lesser developed Asian and African countries are expected to house 90% of the additional 2.5 billion urban dwellers between 2018 and 2050 in their ever-expanding metropolitan areas, often marked by unplanned and haphazard development. A big challenge lies in managing the urban growth resulting from this multifarious phenomenon of urbanisation. The unplanned urbanisation in these Asian and African countries has resulted in socioeconomic problems such as poor urban quality of life, urban poverty, and lack of accessibility to basic infrastructural facilities (public health, sanitation, education, etc.) (Hove et al., 2013; Patil & Sharma, 2022; Sharma & Patil, 2022). It has also resulted in environmental degradation, such as reduced green spaces, loss of fertile lands, habitat fragmentation, and increased pollution (Li & Yeh, 2004; Taubenböck et al., 2009). Understanding the phenomenon of urbanisation in terms of neighbourhood densities, growth typologies, and subsequent management of urban growth is vital to ensure that the evolution of human society happens in line with the United Nation’s Sustainable Development Goals (SDG) (e.g. SDG-11) (United Nations, 2020).

India's urban agglomerations are witnessing unprecedented urbanisation, resulting in rapid and unplanned urban expansion (Gómez et al., 2021; Shafizadeh Moghadam & Helbich, 2013). Indian population tripled between 1950 and 2019 to 1.3 billion. During this period, there was a fivefold increase in the urban population, with 34% living in urban areas as of 2018 (Census, 2011). The Indian urban population share is likely to exceed 50% mark by 2050. It would mean ensuring timely infrastructural facilities for an additional 500 million people to be housed in Indian cities by 2050. Although rapid urbanisation and urban growth of cities in India are facts that need not be proved, the methods used for urban planning are highly criticised as being abortive (Munshi et al., 2014). The same is evident from the negative impacts of urbanisation experienced by various cities in India (Kumar et al., 2007, 2014; Munshi et al., 2014; Vinayak et al., 2021).

The poor planning can be attributed to a meagre cognizance of the urban growth process and spatiotemporal dynamics. Another major challenge for urban planners in India is that rapid urban expansion often outstrips the planning exercise (Kantakumar et al., 2016). Urban planning depends on an exhaustive data set often collected through arduous traditional techniques. Thus, the absence of updated data and suitable tools to measure and monitor urban growth adds to the menace of poor planning. Since cities are considered the drivers of economic growth and contribute significantly to the nation's Gross Domestic Product, formulated policies must prevent haphazard development rather than slowing down the development process. Urban planners need timely information on the current state of urban growth, its past trends, and the probable future (Taubenböck et al., 2009).

The present study uses satellite remote sensing derived neighborhood densities to study urban growth at suitable administrative levels over two decades (1999–2019). The understanding of the urban growth process hypothesised to oscillate between diffusion and coalescence phase is enhanced by identifying urban cores and the relative distribution of medium and low-density developments over time. The haphazard nature of urban growth and its affinity to major road segments is determined by identifying ribbon development occurring along major road corridors in the Mumbai Metropolitan Region. Finally, landscape categories derived through the developed methodology are used to identify the urban growth mode and its directionality. The study proposes an essential tool for planners that can be used to quickly evaluate urban expansion and formulate suitable policies to contain haphazard development.

Literature Review

The satellite remote sensing (RS) data is a suitable alternative to conventional surveying techniques and can be effectively used for studying urban growth dynamics (Herold et al., 2003). When used with the Geographic Information System (GIS), RS is a capable and cost-effective tool for urban studies (Xiao et al., 2006). The historical coverage provided by the Landsat satellite data makes it widely used for urban applications. The geospatial techniques combining RS and GIS have been used increasingly to study the multiple dimensions of urban dynamics, such as the land use/land cover (LULC) alterations (Moisa & Gemeda, 2021; Rojas et al., 2020), monitoring the urban growths (Getu & Bhat, 2021), mapping the land suitability for urban development (Mallick et al., 2024), and modelling future urban growth trends (Hinkosa et al., 2023). Monitoring urban growth helps gather accurate information about past trends that help reveal the dynamics, patterns, and structure of urban growth. Such information is very useful in managing the complex phenomenon of rapid urban growth.

In the literature, several techniques are used to quantitatively analyse urban growth through spatial metrics using remote sensing data and GIS. The metrics range from landscape ecology-based indices (Mcgarigal, 2015), information theory-based indices (Shukla et al., 2021), and fractal analysis-based methods (Bosch et al., 2020). Feng et al. (2016) use Shannon’s entropy (SE) to analyse the spatial patterns of urban sprawl in Jiangning District, China. Getu and Bhat (2021) and Xu et al. (2007) use the largest patch index (LPI), number of patches (NP) and area-weighted mean fractal dimension (AWMFD) to study the patterns of Bahir Dar in Ethiopia and Nanjing in China. Das and Angadi (2021) use landscape metrics such as class area, aggregation index, percentage landscape (PLAND), landscape shape index (LSI), LPI, edge density (ED), area-weighted mean shape index (AWMSI), and SE to decipher the process of urban development of Barrackpore town in West Bengal, India. Rastogi and Jain (2018) use SE and fractal dimensions to study the dispersion of urban areas of Tiruchirappalli city in India. Dutta and Das (2019) use NP, LPI, ED, LSI, clumpiness index and perimeter area fractal dimension to study the urban sprawl of English Bazar agglomeration. Similarly, urban sprawl metrics (USM) developed by Angel, Civco, and Parent (2007) have also been used for Kolkata (Sahana et al., 2018), Delhi (Sharma & Joshi, 2013) and Pune (Kantakumar et al., 2016) to reveal the spatiotemporal patterns in these cities. Although these metrics help understand urban growth's spatial patterns, they do not directly reveal the urban structures.

Understanding a city’s spatial structures, including its extent and important urban centers, is crucial for planning suitable urban infrastructure. The urban core classification is an effective tool that can help identify the extent of the city and people’s activity centers. Several studies thus attempt to understand urban structures quantitatively, with the more recent ones including using nighttime light intensity (NTI) data (Chen et al., 2017; Kii et al., 2023). Chen et al. (2017) proposed a thresholding-based NTI contour tree to identify the urban centers of Shanghai city. However, the method helps identify only the major centers and ignores the smaller ones due to issues like averaging the smaller light intensity peaks. On the other hand, the method proposed by Kii et al. (2023) uses the trip behaviour data combined with the NTI. Although effective, the methods are primarily limited by the availability of such extensive travel-behaviour data that are difficult to obtain for cities in developing countries like India. Further, using NTI data to capture urban development patterns in developing countries such as India faces additional complexities. This is because some development patterns, such as the ribbon developments or informal settlements, come up without authorization from the competent authorities and have limited electricity supplies. Using NTI data in such cases may lead to underrepresenting such developments. The present study overcomes these limitations using the neighborhood densities identified directly through the satellite remote sensing data.

Along with the urban density-based spatial metrics that provide quantitative measures of urban growth, the qualitative measures that describe the underlying process involved are also essential. Qualitatively, the spatiotemporal patterns have been traditionally described as a transition between the diffusion and coalescence patterns. Generally, urbanization begins with the diffusion phase marked by a shoot-up in the number of urban patches. In the later phase, these multiple patches merge to form larger continuous urban patches, termed the coalescence phase (Dietzel et al., 2005). Understanding such patterns quantitatively and qualitatively for an urban agglomeration at the individual administrative units like the urban local body (ULB) may help urban planners strategize urban development. However, the studies that combine the quantitative and qualitative aspects of urban growth are limited. Instead of using the growth modes as an indicator of growth phases, the present study uses the urban built-up density variation and evolution of cores as the measure of growth patterns to associate it with the popular diffusion-coalescence hypothesis.

These patterns of urban areas are often a consequence of various geophysical, accessibility and socio-economic factors (Li et al., 2018). The geophysical factors include slopes (Mondal et al., 2020), elevation (Vinayak et al., 2022), accessibility parameters include proximity to roads (Mustafa et al., 2018), and socioeconomic factors include population (Osumanu & Ayamdoo, 2022). However, among all other factors, roads serve as important corridors of urban expansion. Affinity to the road networks often results in a ribbon development pattern of urban growth wherein the city sprawls along the road corridors. Identifying the ribbon development patterns is vital since such developments along the roads are considered resource-intensive and represent haphazard/unplanned settlements (Ilyassova et al., 2021).

However, only a few past studies have attempted to identify ribbon development. Kumar and Sharma (2023) studied the industrialisation and urban growth along the Delhi-Mumbai Industrial Corridor by taking a buffer of 10 km along the highway and analysing the LULC variations within the buffer. Sharma and Joshi (2013) used a 500 m buffer in the Delhi region, while Chettry (2022) used a 100 m buffer along major highways in Thiruvananthapuram to analyse the ribbon development. Shukla and Jain (2019) used the characteristics of urban sprawl to identify ribbon development as a separate class. Although these studies have theoretically explained the ribbon development concepts, they use buffer analysis to identify them. Such a simplistic measure fails to establish a systematic quantification approach for correctly identifying ribbon development and may result in an overestimation of ribbon development. As in the present study, combining urban densities and buffers helps develop a more reliable measure of ribbon development.

Although a few studies in the past have attempted to analyse the spatiotemporal patterns of Mumbai City and its suburbs, limited works have been done to study the urban growth of the entire Mumbai Metropolitan Region (MMR), and none have studied them at individual administrative units most relevant to the urban planners. The present study addresses these gaps by carrying out a spatiotemporal analysis of the Mumbai Metropolitan Region, one of India's largest and fastest-growing urban agglomerations for two decades; Interval-I (1999–2009) and Interval-II (2009–2019) using an urban density-based approach. The administrative setup of MMR consists of eight municipal corporations, nine municipal councils, 35 census towns, and over 900 villages, making it a unique urban agglomeration. The study optimises the urban sprawl metrics initially developed by Angel et al. (2007) to segregate the built-up areas using densities of walkable neighbourhoods. The derived urban typologies are used to identify the primary and secondary urban cores, map the urban growth directionality, extract the ribbon development and understand the urban growth typologies. The objectives of the study are:

  1. 1.

    To segregate the urban built-up areas based on urban densities as high, medium, and low-density urban areas.

  2. 2.

    To utilise density-based urban typologies to study the evolution of spatial structures of urban growth by identifying primary and secondary urban cores within suitable administrative units.

  3. 3.

    To develop a technique for identifying the ribbon development along major roads.

  4. 4.

    To determine the growth directionality by carrying out the octant analysis.

  5. 5.

    To study the urban growth typologies for each administrative unit.

Study Area and Data

Study Area

Mumbai Metropolitan Region (MMR) is located in the western part of Maharashtra state of India. It is one of the largest urban agglomerations in India, with a geographic area of 4355 sq. km. It consists of eight municipal corporations, nine municipal councils, 35 census towns, and 994 villages with a total population of 22.8 million (Census, 2011). Apart from this, the MMR also has areas that come under the special planning authority such as Bhiwandi Special Notified Areas (BSNA), Ambernath-Kulgaon-Badlapur-Special Notified Area (AKBSNA), CIDCO New Town, and Khopata. Out of the total population in MMR, about 90% resides in eight municipal corporations, and overall, 94% of the population of MMR resides in urban areas. Hence, the population of MMR is mainly urban. Greater Mumbai, which consists of the island city of Mumbai and its eastern and western suburbs, forms the largest municipal corporation, while Ambernath forms the largest municipal council. Greater Mumbai is also a significant contributor to the economic growth of Maharashtra and India. The availability of infrastructure coupled with government policies has resulted in the city's economic and financial prosperity, significantly contributing to urban growth. Being an island city, Mumbai's lack of land space has resulted in the dispersal of economic activities and substantial urban growth in its suburbs. Greater Mumbai's population share in MMR has decreased from 77% in 1971 to 55% in 2011, indicating rapid growth of other municipal corporations and councils in MMR (MMRDA, 2016). These peri-urban regions are witnessing haphazard and unplanned development and are subject to lesser governance. The present study aims to understand urban growth dynamics by segregating the urban areas of MMR into high, medium, and low-density regions based on the urbaneness of walkable neighbourhoods. MMR's future urban growth patterns by using various geospatial models. Figure 1 shows the location of the study area and the urban local bodies (ULBs) considered for analysis in the present study, while Table 1 shows its demographics.

Fig. 1
figure 1

Study Area

Full size image
Table 1 Demographics of Mumbai Metropolitan Region
Full size table

Data Used

Level-II archives of Landsat satellite images of 1999, 2009 and 2019 are used to classify LULC. The required cloud-free images are procured from the United States Geological Survey (USGS), and the study area is masked using administrative boundaries. The data downloaded is from the same months (April–May), with nearby dates each year to ensure comparability of images and capture permanent changes. The images are co-registered to ensure that a point represents the exact location in all three images. The study also uses the road network with all significant corridors, such as highways, expressways, freeways, and other arterials, obtained from the Open Street Maps (OSM) and checked using the Comprehensive Transportation Study (CTS) Report for MMR (MMRDA, 2019). Table 2 shows the summary of the data used.

Table 2 Data Used
Full size table

Methodology

Classification of LULC

Figure 2 shows the methodological flowchart. A maximum likelihood (ML) supervised algorithm that considers the class conditional probability to be multivariate Gaussian is used to produce classified thematic maps. Based on regional reports such as MMR development plans (MMRDA, 2016), dominant land cover classes are identified. Five classes considered are Other land, Vegetation, Built-up, Waterbody, and Wetland. Training pixels for each selected class are obtained based on Google Earth Imagery. Spectral signatures for each class obtained through the training samples are then used to generate the classified maps for 1999, 2009, and 2019. The generated LULC maps are visually inspected by comparing them with Google Earth imagery. Also, a post-classification refinement is performed by emulating the classified image in synchronisation with the Google Earth Imagery and manually re-coding the wrongly classified pixels. The entire process of post-classification improvements helps improve classification accuracy.

Fig. 2
figure 2

Methodological Flowchart

Full size image

The thematic maps are assessed for accuracy using measures suggested by Congalton & Green (2019), including determining the overall and class-level accuracy. Seven hundred sixty stratified random points are taken independent of the training data, and classified vs. actual classes are compared using Google Earth hybrid maps. The thematic maps are reclassified into three classes: Built-up, Waterbodies, and Others (combining Wetlands, Other Land and Vegetation) after ascertaining the classification accuracy.

Density-based Spatial Landscape Typologies

The built-up areas classified using the Landsat images are segregated into three types: high-density urban areas or urban built-up, medium-density urban areas or suburban built-up, and low-density areas or rural built-up, termed urban spatial typologies. These typologies are conceptualised based on attributes characterising urban sprawl, including continuous outward expansion of city boundaries beyond walkable distances, persistent decreases in urban densities, and increased suburbanisation. Each cell is centered within a circular neighbourhood to determine its urbaneness value. A circular moving neighbourhood of 1 sq. km (radius = 584 m), corresponding to roughly an 8 to 10-min walking distance around each built-up cell of 30 m × 30 m, is considered for this purpose. The distance of 584 m is in line with the generally defined walkable distance between 400 and 800 m (Cracu et al., 2024; Daniels & Mulley, 2013; Shin & Woo, 2024). The percentage of urbaneness of the cell is determined as the proportion of the neighbourhood covered by built-up areas. Once the cell urbaneness is calculated, the center of the neighbourhood is shifted to an adjacent cell, and the percentage urbaneness is recalculated. Similarly, the urbaneness is determined for each built-up cell, and it is reclassified into one of the following three categories:

  1. 1.

    High-Density Urban Areas or Urban Built-up (UBU): Urbaneness greater than 50%.

  2. 2.

    Medium-Density Urban Areas or Suburban Built-up (SUBU): Urbaneness between 20 to 50%.

  3. 3.

    Low-Density Urban Areas or Rural Built-up (RBU): Urbaneness less than 20%.

The cut-off set for percentage urbaneness is selected best to suit the delineation of the urban areas in MMR. Various studies in the past have used different cut-offs; for example, Sharma and Joshi (2013) use cut-offs as < 10%, 10–50%, and > 50% for defining RBU, SUBU, and UBU. Similarly, Sahana et al. (2018) use 20%, Kantakumar et al. (2016) use 30%, and Angel et al. (2016) use 25% instead of 10% for defining RBU. Cut-off values of 10%, 20%, and 30% are evaluated in this study to determine the landscape typologies. The results obtained using each cut-off value are visually inspected using the Google Earth Hybrid Images for the recent year, and since the 20% value leads to better identification of urban densities, it is used for further analysis. Using the above definitions, the UBU, SUBU, and RBU share is identified for each ULB across the entire MMR.

The non-built-up areas are also reclassified into three categories as follows.

  1. 4.

    Urban Open Space (UOS): The non-built-up area with urbaneness greater than 50%. It also includes any secluded patch of less than 200 ha of the undeveloped area entirely surrounded by UBU, SUBU or RBU.

  2. 5.

    Fringe Open Land (FOL): A developable area within 100 m of a built-up area that does not belong to the UOS category.

  3. 6.

    Rural Open Land (ROL): An undeveloped area not belonging to UOS, FOL and ROL.

The UOS and FOL represent important categories of open land with high susceptibility to land degradation. In addition to three classes of developable lands, non-developable or restricted zones are classified under the Restricted category. Classifying undeveloped lands into the above three categories helps identify the urban growth typologies. The density-based landscape typologies help us develop a methodology for a detailed analysis of urban areas in MMR; the methodology is discussed in the next section.

Analysis of Urban Core

The threefold categorisation of urban growth densities is further used to derive valuable insights into understanding spatiotemporal patterns of urban growth. It includes analysing the urban cores, extracting ribbon development, and understanding the directionality of urban development. The urban cores are defined as the high-density urban areas (having > 50% urbaneness) located within each urban local body. Urban cores are further classified as Primary (PC) and Secondary urban cores (SC). The PC is the largest patch of high-density built-up areas and may be considered a driving point of urban development within that ULB. However, there can be other urban clusters within the ULB apart from the primary urban cluster. These smaller, high-density urban clusters are classified as SC. Thus, any ULB may have one or more SCs but has one and only one PC. Setting the criteria to define the contiguity of urban clusters is also essential because these criteria form the basis for deciding whether the other smaller clusters are a part of the PC or otherwise. Although no hard and fast rules are defined for the same, we use a 3 × 3 neighbourhood with an eight-neighbour rule along with the criteria for the minimum size of the urban cluster. Thus, if the adjoining urban cluster is not within the 3 × 3 neighbour and is of the size > 5 ha, it is considered an independent cluster.

Directionality of Growth

The PC identified for the initial year (1999) can be associated with the inceptive phases of urban development within each ULB. A directional analysis is performed to study the evolution of urban structures for major municipal corporations in the study areas. The analysis begins with determining the centroid of high-density urban areas within each ULB for the inceptive year (here, 1999). Once the UBU centroid is determined, the ULB is divided into eight sectors, each of 45 degrees, representing North, South, West, and East and their four sub-directions: Northeast, Southeast, Southwest, and Northwest. The percentage of land area covered by built-up areas in each direction is extracted and plotted for each year to determine the directionality of urban growth.

Identification of Ribbon Development

The urban core and directional analysis help us understand the variations in patterns of high-density urban areas. On the other hand, the variations in the low and medium-density urban areas help us identify and extract the ribbon development. The ribbon development is defined as low-density development along the major road segments. To identify the ribbon development, 250 m buffer is considered along the major road network consisting of national and state highways, freeways, and express highways.

Once the buffers are ready, the next step is to extract the built-up areas within the buffers. The buffer is overlaid on the low and medium-density built-up areas, and the built-up area lying between the buffer is extracted. While extracting the low and medium-density built-up areas, it may be possible that a contiguous group of the built-up area lies partly within and partly outside the buffer. However, for such a case, if only the built-up area lying within the buffer is considered and the part of the contiguous urban cluster lying outside is removed, it may lead to erroneous results. Thus, to cater to this problem and correctly calculate the ribbon development, the contiguous built-up clusters are identified using the region group tool within the GIS environment with an eight-neighbour rule. Thus, all contiguous built-up areas lying entirely or partially within the buffer are extracted and plotted for each study year.

Identification of Modes of Urban Growth

The urban growth typologies or the modes of urban growth are usually classified based on the spatial relationship between the existing built-up areas and new developments. Although there have been many techniques to identify urban growth typologies, there has been a consensus about the three categories of urban spatial typologies:

  1. 1.

    Densification or Infill

  2. 2.

    Extension or Edge Expansion

  3. 3.

    Spontaneous or Outlying

In the present study, the urban growth typologies are classified based on the three-fold classification of developable land. The location of the new built-up is identified, and the new development is classified as infill or densification if it occurs in the UOS. The new development is classified as an extension if it occurs within 100 m of any built-up area, and it is classified as outlying if the development occurs in the ROL. The growth mode is identified for each administrative unit and compared between the municipal corporations, municipal councils, and rural/special notified areas.

Results

Urban Growth Quantification

The overall accuracy of 90%, 92%, and 89% with a kappa coefficient of 0.87, 0.90, and 0.87 is obtained for 1999, 2009, and 2019. The individual class-level accuracy calculations show that no class has an accuracy of less than 85%. The overall accuracy and individual class accuracies are satisfactory; therefore, the images are further analysed.

The classification results show that the built-up area is 414.8, 579.90, and 785.80 sq. km in 1999, 2009, and 2019. The built-up area increased at an alarming rate of 89.44% over the past two decades. During the same period, the wetlands reduced by 13.5%, waterbodies decreased by 8.5%, and other land reduced by 16%. However, the vegetative cover saw a 20% increase in MMR. The built-up areas are further analysed by combining all other classes. Figure 3 shows the spatial distribution of built-up areas.

Fig. 3
figure 3

LULC Classification for the Year (a) 1999 (b) 2009 and (c) 2019

Full size image

Greater Mumbai contributes about 34% of the region's built-up area. It shows a 16.7% growth in the first interval (1999–2009), but the urban growth of this region has slowed down with just 8.6% growth in the next interval (2009–2019). Like Greater Mumbai, Ulhasnagar Municipal Corporation also shows nominal growth (8.9% in the first decade and 5.4% in the second decade). Contradicting the area's overall high growth, these regions lack growth mainly because of the space constraints for further development within their administrative boundaries. In fact, except Vasai-Virar, all other municipal corporations show a slowed-down urban growth over time.

On the other hand, most municipal councils show an increasing urban growth trend over time. Among others, the urban areas in Vasai Virar, Ambernath-Kulgaon-Badlapur Special Notified Area, Kulgaon-Badlapur, CIDCO New Town, Karjat, Outer MMR, and Rural MMR doubled between 2009 and 2019. The MMR has recently seen a rapid increase in sub/peri-urban areas. One primary reason is affordable housing in these regions and good road-rail connectivity with Mumbai city. Industrial expansion in Kalyan-Dombivli, the development of Badlapur as a logistic hub, the construction of a new port in CIDCO New Town, and the planned new airport near Navi Mumbai have also contributed to the large-scale urban expansion in these regions.

Identification of Spatial Landscape Typologies

The spatiotemporal analysis of spatial landscape typologies reveals significant variations in urban development patterns over different ULBs and within the same ULBs over time. The six categories of landscape, consisting of three categories of developed and three categories of undeveloped lands, based on densities, are shown in Fig. 4. Table 3 gives the changes in the landscape that occurred in Interval-I and Interval-II. As shown in Table 3, the urban built-up indicating the high-density built-up areas increased by about 50% in the first period, while the growth of both medium and low-density built-up areas stood at about 16%. Due to increased high-density development, the UOS representing the developable spaces within the urban landscape decreased by about 7.5%, while overall rural spaces were reduced by 5%. In the next period, the growth of high-density built-up areas slowed to 39%, while a steep increase in suburban fringes was observed. The increase in suburban fringes resulted in an increase in urban open spaces by 21%; however, the rural open lands decreased by 8%. The classified landscapes are further analysed by classifying the urban built-up areas as primary and secondary cores, while the suburban and rural built-up areas are used to identify the ribbon development.

Fig. 4
figure 4

Spatial Landscape Typologies of MMR for the Year (a) 1999, (b) 2009, and (c) 2019

Full size image
Table 3 Extracted Density-based Spatial Landscape Typologies
Full size table

The spatial urban typologies obtained using the neighborhood densities are verified using Google Earth imagery. This verification is conducted by overlaying the urban typology boundaries on high-resolution Google Earth imagery and comparing whether the density-wise segregation obtained through neighbourhood density coincides with the actual boundaries observed on the ground. The details on the procedure adopted for the verification of results can be referred to in Appendix A.

Urban Core Analysis

The high-density urban areas are further classified as the primary urban core and the secondary urban core. For the entire MMR, the area under the PC increased from 15,215 hectares in 1999 to 21,921 hectares in 2009 and then to 24,153 hectares (Table 4). The SC also showed a significant increase during this period. It increased from 11,917 ha in 1999 to 18,864 ha in 2009 and 32,241 ha in 2019. The analysis of growth rates shows a balanced development in the first interval, where the PC and SC increased rapidly (by 44% and 58%, respectively). However, in the next interval (2009–2019), the PC increased by just 12%, while the SC increased by 71%, indicating an overall diffusion phase for MMR. Thus, between 2009 and 2019, many new urban centers flourished, while the PC did not increase. The suburban built-up has also shown a tremendous increase, while the rural built-up areas showed steady growth over the two development periods. The trends of urban spatial typologies are extracted at individual ULB levels (Table 5).

Table 4 Spatial Urban Typologies of Entire MMR
Full size table
Table 5 Spatial Urban Typologies including Ribbon Development at ULB Level
Full size table

In 1999 regions such as Alibag, Karjat, Khopate, Matheran, Pen, and Rural MMR had no PC, mainly due to their locations being far away from major urban centers. With improved rail-road connectivity between various ULBs, by 2019, all ULBs except two, Khopate and Matheran, had developed PC. During the first interval (1999–2009), ULBs Bhiwandi, Kalyan, Greater Mumbai, Mira Bhayandar, and Navi Mumbai were already developed and showed tremendous PC expansion. The increase in PC resulted mainly from the merger of SC with the PC, as evidenced by the negative growth rates of SC during this period. The regions Outer MMR, Thane, Uran, and CIDCO showed much higher SC expansion rates than PC expansion. It indicated the development of new urban clusters in these regions and PC development. A few regions, such as Vasai Virar and Rural MMR, showed a balanced growth of both PC and SC.

During the second interval (2009–2019), the patterns of urban cores in these regions changed significantly. The regions Bhiwandi, Greater Mumbai, Thane, and Kalyan showed slowed PC growth mainly due to saturation of spaces for further expansion, while the SC expanded rapidly. These SCs could be expected to merge with the PC, resulting in a single large PC in these regions. It may mark the final stages of growth, especially for Greater Mumbai, which has exhausted most of the spaces available within the administrative boundaries. During the second period, the newly developing areas like Outer MMR continued to disperse by forming a new SC. The regions Mira Bhayandar, Navi Mumbai, Vasai Virar, and Rural MMR showed rapid PC expansion, and the SC developed at a slower pace, while the Uran and CIDCO showed a balanced development.

Extraction of Ribbon Development

The urban spatial typologies- medium and low-density urban areas- help extract the Ribbon Development using the process discussed in the Methodology section. Figure 5 shows the extracted Ribbon Development for each study year. In 1999, 6237.4 ha of built-up area was found along the major road networks passing through MMR, constituting about 15.50% of the total built-up area. Thus, a significant proportion of built-up area was constructed near road networks. In 2009, the total built-up area was found to be 7035.4 ha, which increased to 9490.91 ha in 2019. Although the total built-up areas classified as Ribbon Development increased, its contribution to the total built-up area reduced to 12.6% in 2009 and 12.4% in 2019. The reduction in the overall percentage of the built-up area classified as Ribbon Development indicates that although new patches of Ribbon Development emerged at various locations in MMR, many were densified and converted to high-density built-up areas. The statistics of Ribbon Development are extracted at ULB levels to carry out a detailed analysis. Table 5 shows the distribution of Ribbon Development over different ULBs during the study interval.

Fig. 5
figure 5

Ribbon Development Extracted for the Year (a) 1999 (b) 2009 and (c) 2019

Full size image

Regions Greater Mumbai, Navi Mumbai, and Panvel showed a continuous reduction in ribbon development in both intervals. Greater Mumbai and Navi Mumbai saw massive development in the early 2000s and have become relatively densified. With the saturation of land spaces within these administrative boundaries, the regions show meagre amounts of ribbon development with continuously reducing trends. On the contrary, regions Bhiwandi and its exterior, Rural MMR, Outer MMR, and Vasai-Virar, which are in a developing stage, show a continued increase in Ribbon Development. Urban local bodies Kalyan and Contiguous areas and Mira Bhayandar showed a decrease in Ribbon Development in the first period (1999–2009); however, Ribbon Development increased in the second period. Contrary to this type of development, Thane and CIDCO's new town showed an increase in ribbon development during the initial period, but later, with increased urban densities, the ribbon development showed a reduction.

Growth Typologies

The modes of urban growth are identified for each administrative unit and compared between the municipal corporations, councils, and special notified areas. The proportion of urban growth types within each administrative unit helps understand the changes in urban patterns. Figure 6 (a, c, e) shows the distribution of urban growth typologies (infill, extension and spontaneous) in Interval-I, and Fig. 6 (b, d, f) shows the distribution in Interval- II. Figure 7 shows the spatial distribution of urban growth. In general, edge expansion is the dominant mode of urban growth for both periods. Greater Mumbai and Ulhasnagar municipal corporations show infill as the dominant mode for the Interval-I. Besides these two municipal corporations, Navi Mumbai and Bhiwandi transitioned from edge expansion-dominated urban growth mode to infill-dominated.

figure 6

Relative Shares of Three Growth Modes in Interval I and II

Fig. 7
figure 7

Spatial Distribution of Growth Modes for the Year (a) 1999–2009 and (b) 2009–2019

Full size image

On the other hand, all other municipal corporations show a continuous edge expansion in both intervals. Also, all the municipal councils show edge expansion with proportions higher than 69% in both intervals. While edge expansion remains the dominant urban growth mode in municipal corporations (except a few) and municipal councils, the infilling mode of urban growth comes next. The proportion of spontaneous growth remains very low in these administrative units.

Similar to most ULBs in MMR, the special notified areas and rural areas also show edge expansion as the dominant growth type; however, the proportion of spontaneous development within these lesser governed areas is also significant.

Directional Analysis

Figure 8 gives the directional distribution of urban footprints at the ULB level. Regions Alibag, Greater Mumbai, Thane, Mira-Bhayandar, and Karjat show unidirectional growth, i.e., they have one dominant direction of growth. Regions Vasai Virar, Navi Mumbai, Panvel, Kalyan Dombivli, and Pen show multi-directional growth patterns (having more than one dominant direction). For Greater Mumbai, by 2019, six out of eight sectors show more than 50% of the land area covered by built-up area, indicating saturation of the biggest patch of MMR. Mira-Bhayandar has shown more than 200% growth in the Northeast and East direction in the last two decades, while the North and Northwest direction shows minimal urban growth due to marshy lands. Thane, a neighbouring city of Greater Mumbai, is surrounded by a creak in the Northeast and a national park in the West, resulting in limited growth opportunities. It shows a growth of 255% over two decades in the North direction. Vasai-Virar has grown more than 150% in the last two decades in all directions except in the South because of the hilly terrain. Due to hilly terrain on the Northeast side and a creek on the Southwest side, Navi Mumbai showed less than 20% growth in the last decade, which may indicate a lack of land availability for new growth. Bhiwandi shows saturation of land spaces, with 49% of land in the Southwest and 32% in the Southeast direction covered by built-up. The growth in the Southwest direction spilled over Bhiwandi urban boundaries and continued along with State Highway 35. Kalyan-Dombivli showed minimal growth between the North and West directions because of constrained topography, with the river passing along Northwest and Northeast directions.

Fig. 8
figure 8

Directional Distribution of urban footprint at ULB Level

Full size image

Among others, the road network is assumed to be one of the significant factors shaping urban growth direction. The road network for 2019 is used to evaluate this hypothesis. Table 6 gives Pearson's Correlation between the Percentage of Land Covered by Built-up and Road Density. It can be observed that while Greater Mumbai and Thane show a slight positive correlation between road density and the percentage of land covered by built-up, all other developing ULBs show a significant positive correlation. It may imply that urban growth is more likely to occur in directions with higher road densities.

Table 6 Correlation coefficient between the percentage of land covered by built-up and road density
Full size table

Discussion

Several studies in the past highlighted the rapid urban expansion of Indian cities, often characterised by unplanned and haphazard patterns (Chakraborty et al., 2022). The urban expansion surrounding the metropolitan regions of India consists of mixed patterns created by mosaics of urban–rural transitions surrounding the parent city, making the physical distinction of urban structures highly challenging (Sahana et al., 2023). The present study uses the concept of urban densities to determine urban structures and analyses the spatial patterns of Mumbai and its surrounding areas within the Mumbai Metropolitan Region (MMR). The result indicates a rapid urban expansion characterised by mixed-density patterns, with the built-up areas increasing by 89% in the past two decades. The increased built-up areas come at the cost of a substantial decrease in the other natural land covers, such as the other land getting reduced by 16%, waterbodies by 13.5%, and wetlands reduced by 8.5%.

Analysing urban spatial typologies reveals that MMR became denser between 1999 and 2009, with the high-density built-up areas increasing by an additional half. However, a significant increase in suburban fringes is also observed during the second interval (2009–2019), along with the high-density areas. It indicates a shift in urban development dynamics over time with rapid growth in the areas surrounding the city centers in MMR. Additional details are obtained by classifying high-density urban areas as the Primary Core (PC) and Secondary Core (SC) and analysing them at the urban local body level. Three development patterns in a particular period are identified – Diffusion (PC expands slower than SC), Coalescence (PC expands faster than SC) and Balanced (both PC and SC expand). The analysis reveals the emergence of urban cores in Rural and Outer MMR, Special Planning Areas such as CIDO and within the city outskirts such as the exterior of Bhiwandi and Kalyan. The lack of sufficient land for expanding its urban footprint within Greater Mumbai, Navi Mumbai, Bhiwandi, and Ulhasnagar municipal boundaries has infilled urban open spaces in these dense municipal corporations.

On the other hand, municipal councils in the developing phase show a rapid outward expansion of urban footprint with edge expansion proportions exceeding 70%. While high-density urban growth may result in intense pressure on public services and increase the land surface temperatures significantly, the rapid expansions in the fringes, too, may have severe implications on the sustainability of the city’s growth. A high proportion of spontaneous development is also observed in the special planning and rural areas, suggesting new growth hot spots. Suitable planning strategies must be implemented to ensure a balanced development of urban cores and surrounding areas. Such an arrangement may help preserve the green spaces within the urban cores under pressure from the constant infilling of urban open spaces. Further, stringent regulations to govern the spontaneous growth and enhancement of the infrastructure in the emerging suburban/peri-urban areas are required to prevent haphazard development in these lesser-governed areas.

The findings suggest an increase in the ribbon development for the ULBs in the developing phase, while a reduction is observed for ULBs in the later stages of development. Significant developments along the roads in Bhiwandi and its exterior are observed along NH-160. The ribbon development here is mainly due to the construction of warehouses in Bhiwandi that serve businesses throughout the entire MMR. Development in Mira Bhayandar is observed along the Western Express Highway (one of the busiest roads in MMR) near the Ghodbunder-Vasai bridge connecting Mumbai city to the north of MMR and serving as an important entry point to Mumbai along NH-48. Within Kalyan and contiguous areas, extensive development is observed along SH-80 in the form of small roadside eateries and retail shops adjoined by haphazard informal settlements. The ribbon development observed in rural areas of MMR, are along the major transportation corridors that provide inter-city movements such as Mumbai-Pune Highway, Mumbai-Pune Expressway, and Mumbai-Nashik Highway. Ribbon development, which is often overlooked in literature, is one of the initial stages of development. It takes place due to excellent accessibility to the roads, but at the same time, such developments often obstruct traffic movements, resulting in increased traffic congestion problems. Since ribbon development is resource-intensive and may lead to haphazard urbanisation, capturing it at relevant administrative scales is essential. Such high rates of ribbon development in rural areas of MMR require continuous monitoring mainly due to insufficient planning and governance in these areas. Proposed policies may include proposing development zones at strategic intervals to confine the highway utilities and commercial establishments within designated locations within the MMR.

The directional analysis reveals the importance of road densities in shaping urban growth. Apart from the road densities, the topographical features play a significant role. Notably, low growths are observed towards the north and northwest of Mira Bhayandar due to marshy lands and towards the north and northeast of Thane due to creeks. Similarly, hilly terrain towards the south of Virar and east of Navi Mumbai and the river in Kalyan’s north and west directions restrict the development. Implementing policies and intensive regulations in these places is imperative to minimise the environmental impact. This will ensure that the ongoing pressure from urban growth and population expansion does not result in encroachment of these natural features.

The patterns of urban growth obtained can be effectively integrated with various urban growth prediction models like Cellular Automata and Machine Learning Algorithms. Many studies that predict future urban growth patterns are mainly based on incorporating the proximity to various infrastructure facilities (Bhatia et al., 2024a). Huang et al. (2019) studied the effects of urban expansion on urban heat islands using parameters such as proximity to roads, slope and population density. Bhatia et al. (2024b) used the slope, proximity to infrastructures and transportation networks with the Markov Chain-Logistic Regression-Cellular Automata hybrid model to predict the future urban growth of MMR. Similarly, Zhou et al. (2019) use the slopes and road networks in the SLEUTH model to predict future urban growth patterns. As in the present study, the classification of urban areas into high, medium, and low densities can be incorporated into future growth models to simulate future patterns better. It can allow planners to set specific density targets for each administrative unit (based on its current state) and create scenarios incorporating a mix of densification, outward expansion, and development of new urban centers.

Further, a few studies, such as Chen et al. (2020) and Li and Zhou (2019), used the concept of shared socioeconomic pathways (SSPs) to create multiple scenarios for urban growth prediction. The SSP framework can be integrated with urban densities. For example, the SSP1 (Sustainability Scenario) can be modelled to focus on high-density growth with transit-oriented developments while maintaining urban open spaces. In contrast, the SSP3 (Regional Rivalry Scenarios) can be modelled with uneven growth patterns through low-density spontaneous growth and new growth hotspots in outer areas.

Despite the comprehensive nature of the analysis, a few limitations still exist. The study primarily relies on the use of static administrative boundaries. However, the re-demarcation of administrative boundaries often lags behind urban expansion, making it essential to use dynamic boundary demarcations along with administrative setup to obtain a more accurate picture of urban growth patterns. Further, while the present study highlights the importance of planning and regulatory measures, implementing these measures can involve substantial inter-agency coordination. Addressing these limitations with a continuous data update can be crucial for ensuring the applicability of future urban growth studies.

Conclusion

The present study employs an urban density-based approach to evaluate the spatiotemporal dynamics of urban expansion in a fast-growing Mumbai Metropolitan Region using three Landsat satellite images of 1999, 2009, and 2019. The built-up areas are classified as high, medium, and low-density urban areas, while the non-built-up developable areas are classified as urban open space, fringe open land and rural open land. The high-density areas define primary and secondary urban cores within each administrative unit, and the ribbon development along the major road segments is identified based on the low and medium-density urban areas. New built-up analysed relative to previous developable lands are used to identify the mode of urban growth. The urban growth centroids are computed to understand the directionality of the urban footprint.

The urban core analysis reveals that primary and secondary cores developed quickly during the initial development period (1999–2009), indicating a balanced development. In the second period, the expansion of the primary core was marginal; however, the secondary cores expanded rapidly, indicating the dispersion pattern of development. Larger municipal corporations like Kalyan and Bhiwandi showed a coalescence development pattern during the first period. However, with the saturation of spaces for further infilling, these areas exhibited the development of new urban centers in the second period and, thereby, a dispersion pattern. Three development patterns are identified within the urban local bodies: diffusion, coalescence, and balanced growth.

Overall ribbon development reduced from 15.5% in 1999 to 12.4% in 2019 but increased in outgrown villages and suburban fringes, mainly across the highways providing intercity movements such as Mumbai-Ahmedabad, Mumbai-Nashik, and Mumbai-Pune. The outward expansions dominate the urban growth in MMR, though the municipal corporations show significant infilling while municipal councils show extension with lesser infilling. The rural and special notified areas show a significant proportion of spontaneous development, indicating the establishment of new development centers. The growth directions also varied significantly, with a few regions showing unidirectional expansion while others showing multi-directional expansion. A strong positive correlation between road densities and growth directions is found, emphasising the role of roads and natural barriers in shaping urban footprints.

The methodology used in the present study provides a suitable tool for planners to understand urban growth patterns. The notions of the urban core, ribbon development, and growth directions of urban footprint are directly relevant to ensure environmental sustainability and understanding urban growth patterns at administrative unit levels may help monitor and make effective and sustainable growth policies.