Introduction

The world is being metamorphosed into urbanisable area progressively where 3.5 billion people inhabits in the cities today which is going to increase 60% by 2030 (Sustainable Development Goal 2021). According to World Urbanization Prospects (2018), 95% of urban extension will occur in the developing world in the next decades. Cities are the centres of positive force for technological innovation, human development, poverty reduction and development of economic growth which shares about 60% of global gross domestic product (Sustainable Development Goal 2021; UN 2019). However, these urbanisable areas also account for 70% global carbon emissions, 60% of resource use and 60–70% of energy consumption (UN-Habitat 2019).

Urban growth is a dynamic and complex socio-economic process that changes cities in terms of both size and pattern, i.e. population demography (Kafi et al. 2014; UN 2019). The urban transition and economic growth have been linked closely to modernization, industrialization and globalization that fuels the urbanization (Mallupattu and Reddy 2013; Hasan et al. 2020). People migrate from rural to urban areas due to availability of increased opportunities and facilities such as employment, education, health care and other recreational activities (Hasan et al. 2020; Liping et al. 2018; Sarkar and Chouhan 2019). Consequently, this leads to rapid transformation and alteration of land use and land cover features that are mainly induced by anthropogenic activities than natural phenomenon (Hamad et al. 2018; Mahmoud and Alazba 2016). There are several studies that have shown only few remote locations are still remaining in their natural and pristine state (Alawamy et al. 2020).

Rapid urbanization and urban growth in combination with unsustainable way of using natural resources resulted in the development of slums, insufficient infrastructure, air pollution, increased pressure on fresh water supplies, forest deterioration, biodiversity loss, global warming, transformation of arable land to urban construction, increase in natural disaster events, climate change, urban heat island and desertification (Kumar et al. 2019; Reis 2008; Aliani et al. 2019; Koko et al. 2020). Increase in the urban clusters with increase in the urban population has significantly changed the land surface which is the most important natural resource at the expense of destruction of natural landscape, decline in open spaces and reduction in the capacity of providing ecosystem services and goods (Gharbia et al. 2016; Hamad et al. 2018).

As a capital, Delhi has experienced congestion, problem in providing civic amenities adequately and increased land rent values with increasing population pressure (NCRPB 2021; Naikoo et al. 2020). Therefore, the Master Plan of Delhi (1962) proposed the concept of National Capital Region (NCR) to disperse the pressure of population towards the surrounding areas (DDA 2021). The NCR is a rural–urban planning region, also known as Delhi Extended Urban Agglomeration and its major cities are Delhi, Faridabad, Ghaziabad, Gurugram and Noida (NCRPB 2021). People started moving out from Delhi to look for comparatively cheaper land for residential or commercial purposes and creation of special economic zones (SEZ) also attracted them (Naikoo et al. 2020; Hasnine and Rukhsana 2020). It resulted in a mixed land use pattern and gave rise to many small urban centres which have now become big cities.

The uncontrolled rate of changes in the land use land cover (LULC) of the city especially the transformation of natural to artificial land cover in the cities has become the major concern around the world at global level (Gharbia et al. 2016; Keshtkar and Voigt 2016). These changes need to be continuously monitored and assessed at different spatial and temporal scales in order to understand what, where and when the changes occurred and at which rate they occurred (Mengistu and Salami 2007; Kumar et al. 2016). Regional mapping of LULC is essential to consider positive and negative changes, to evaluate the impacts of “from-to”, to detect driving forces for those changes and to understand the interconnection between people and environment (Kafi et al. 2014; Alipbeki et al. 2020; Mengistu and Salami 2007). Remote sensing and geographic information systems (GIS) are powerful tools and advance means to derive LULC information because they are real time, highly reliable and accurate, cost-effective, and time saving with wide spatial coverage (Maina et al. 2020; Lu et al. 2019). Remote sensing data from Landsat series of satellites have been broadly used in LULC studies as data is easily and freely accessible with comparatively better spatial resolution (Lu et al. 2019).

Urban growth dynamics has become a matter of interest and concern as it changes the basic structure and function of ecosystems (Keshtkar and Voigt 2016). Thus, simulation of urban growth is necessary to examine the future trends in land use and landscape structure (Al‑Shaar et al. 2021). Land change modelling is an important tool that allows space–time predictions, i.e. in what way it is likely to change in the coming years (Lu et al. 2019; Al-sharif and Pradhan 2014). It enables city planners, policy makers and environmentalists to make better planning for future sustainable urbanization and mitigation strategy in order to achieve Sustainable Development Goal 11, i.e. sustainable cities and communities (Sustainable Development Goal 2021; Sarkar and Chouhan 2019). Recent advancements in the geospatial technology enable researchers to predict future LULC changes more realistically and effectively (Devendran and Lakshmanan 2019). In this study, Land Change Modeler (LCM) of TerrSet software is used, which comprises of cellular automata — Markov chain process (CA–Markov) to predict the future urbanization (Kumar et al. 2016; Hasan et al. 2020). This model is widely accepted as it integrates historical LULC data with driving factors of urbanization and produces improved spatial patterns of LULC (Hamad et al. 2018; Liping et al. 2018).

Background model description

Markov chain

In 1906, Andrey Andreyevich Markov a Russian mathematician developed a technique known as Markov chain that gives new direction to predictive probability theory (Seneta 1996; Mohamed and El-Raey 2019). Markov chain is a random process used to model the change over a specified time period from one state to another, i.e. from one LULC class to another and is expressed as (Gidey et al. 2017; Mohamed and El-Raey 2019)

$${X}_{n}=i$$
(1)

where i = state, n = time.

In this model, changes from one state to another are according to a given set of probabilistic rules. The predicted future states depend only on the present states and not upon the past states. Probability of going from present state (i) to future state (j) is given by following mathematical expression (Shawul and Chakma 2019):

$${P}_{ij}=P\left\{\left.{X}_{n+1}=j\right|{X}_{n}=i\right\}\to ,i.e.P=\left\{\left.\mathrm{future}\right|\mathrm{present}\right\}$$
(2)

It produces a transition probability matrix having information of probability of conversion.

$${P}_{ij}=\left[\begin{array}{cccc}{P}_{11}& {P}_{12}& \cdots & {P}_{1m}\\ {P}_{21}& {P}_{22}& \cdots & {P}_{2m}\\ \vdots & \vdots & \vdots & \vdots \\ {P}_{m1}& {P}_{m2}& \cdots & {P}_{mm}\end{array}\right]$$
(3)

where Pij is the transition probability matrix, and m is class number.

Pij must satisfy two conditions:

$$\hspace{0.17em}\le \hspace{0.17em}{P}_{ij}\hspace{0.17em}<\hspace{0.17em}1\mathrm{\ for\ all\ }i\mathrm{\ and\ }j;\ i,\ j\hspace{0.17em}=\mathrm{0,1},2\cdots m$$
$$\sum {P}_{ij}=1\left(\mathrm{row wise}\right)$$

It results in probabilities of the future state for decision making. The following formula is used to predict the state (Lu et al. 2019):

$$P\left(\mathrm{Future period}\right)=P\left(\mathrm{Present period}\right)\times \mathrm{Transition Probability}\left({P}_{ij}\right)$$
(4)

where P is land use land cover conditions.

The basic principle of the Markov chain in Terrset software is that it forecasts the future LULC change from present time (n) to future time (n + 1) (Gidey et al. 2017). In order to model LULC changes using Markov chain, first, the two input images of different dates are compared on a class-by-class basis and then estimate the replacement rate of one class to other for predicting the third date (Yagoub and Bizreh 2014). Markov chain modelling is appropriate to predict the transition trends of LULC system.

Markov chain is a useful tool for simulating LULC dynamics which predicts the future dimensions and trends (Muller and Middleton 1994; Al-sharif and Pradhan 2014). However, it does not provide scientific explanation of projected changes and spatial distribution of LULC categories (Koko et al. 2020; Gidey et al. 2017). The spatial distribution is highly significant in understanding the patterns and potential impact of certain factors such as socio-economic, political and biophysical that changes over the time (Lambin et al. 2000; Kamusoko 2012). To address this problem, the cellular automata is used to give spatial dimension to LULC (Wang et al. 2019).

Cellular automata

The CA is based on the principle that repeats the interaction of simple rules may results in complex phenomena with time. It is a bottom-up, abstract, spatially and temporally discrete dynamic model (Sarkar and Chouhan 2019; Gidey et al. 2017) and composed of following four elements:

  1. (i)

    Grid of cells – cell is a fundamental and smallest unit. Each cell carries data of only one state at any one time from a defined set of states (Kumar et al. 2016). Cell size represents the land area that each cell covers. In this study, Landsat data having 30 m × 30-m cell size has been used.

  2. (ii)

    State – it allocates possibilities to the cell which is typically finite (Virtriana et al. 2014), e.g. 0 and 1 or urban and non-urban for a given time step. These states either change into a new state or remain in their present state at every discrete time step (Keshtkar and Voigt, 2016).

  3. (iii)

    Neighbourhood – each cell has a local neighbourhood, representing a set of cells or adjacent cells that have direct interaction/relationship with the centrally located cell (Kumar et al., 2016). At every iteration, the state of each cell is decided by the states of its local neighbourhood (Gidey et al., 2017). The effect of neighbourhood cells is inversely related to the distance from the central cell (Gharbia et al. 2016; Wolfram 2002; Barredo et al. 2003).

  4. (iv)

    Transition rules – the changes occurred in the cell state is due to simple and logical rules applied on the states of neighbourhood cells iteratively through a number of discrete time steps defined prior to the execution (Keshtkar and Voigt 2016; Al‑Shaar et al. 2021). The formula for calculating cell’s state at any given time t + 1 is as follows (Gidey et al. 2017):

    $${C}_{t+1}=R\left({C}_{t},{O}_{t}\right)$$
    (5)

where C is the cell state at time t and t + 1, O is the state of cells in its neighbourhood and R is the transition rules.

Transition rules are the most essential element in order to simulate real world LULC changes (Gharbia et al. 2016). In relation with urban growth, rules consider various driving factors such as socio-economic, biophysical, accessibility and disturbances which have significant impact on LULC (Arsanjani et al. 2013; White and Engelen 1993; Liu et al. 2008). So, the probability of transitions to forecast LULC changes in the state of neighbourhood cells is not completely random decision (Liping et al. 2018). CA has a strong capability for modelling LULC dynamics because it considers time–space interactions and urban expansion as it uses proximity concept, i.e. cells that are nearer to current class have higher probability of changing into another class (Yang et al. 2014; Nguyen et al. 2019).

Objectives

The objectives of the study are to (i) evaluate the changes in LULC using satellite imagery for the period from 1989 to 2020, and (ii) model and predict future LULC changes, particularly growth in urban area for the year 2030. The findings will help policy makers in improving the land use policies that will safeguard the local environment and sustainable development.

Study area

The study area covers Delhi (capital of India) and neighbouring four important NCR districts namely Ghaziabad, Faridabad, Gurugram and Gautam Buddha Nagar lies between longitudes 77°21′12.36"E to 77°25′16.78"E and 28°14′15.59"N to 28°55′52.05"N latitudes (Fig. 1). It spreads over an area of 5955 km2. The climatic conditions of the area are characterized by hot semi-arid summers with maximum temperature 48 °C and fairly cold subtropical winters with minimum temperature 4 °C. As per the Indian Meteorological Department, the average annual rainfall is approximately 886 mm.

Fig. 1
figure 1

Location map of the study area

Full size image

Materials and methods

Satellite data

Satellite imageries are helpful in mapping and representing the spatial distribution of LULC information (Rawat and Kumar 2015). The LULC data used in this study were prepared from the Landsat images for the years 1989, 2000, 2010 and 2020. These satellite images were obtained from the US Geological Survey (USGS) data portal (http://earthexplorer.usgs.gov/). The images were selected from the same season to reduce the effect of seasonal variation. The details of satellite data are given in Table 1.

Table 1 Details of satellite data used in the study
Full size table

Ancillary data

This study considered the biophysical and socioeconomic driving factors that exert direct effect on the urban growth pattern. The following five factors were selected for running the model to predict the urban expansion:

  1. (i)

    Proximity from roads – road network is an important factor in expanding as well as stimulating urban growth. It increases the accessibility of land and decreases the cost of getting common facilities as all the urban land use categories such as residential, commercial and institutional places are interconnected to each other (Abdullahi and Pradhan, 2018). Road network data was extracted from the Open Street Map (https://www.openstreetmap.org/) dated 15 March 2021. In this study, it was assumed that the road network data remained unchanged. Therefore, this data was entered for the year 2010 assuming that network was similar during that period. Both major roads and minor roads were considered in this study. The Euclidean distance tool was applied to calculate the distance from roads based on the closest cell (Gharbia et al. 2016).

  2. (ii)

    Elevation – elevation map was generated from Digital Elevation Model (DEM) to represent topography. DEM was obtained from Shuttle Radar Topographic Mission (SRTM) which was downloaded from USGS Earth Explorer.

  3. (iii)

    Slope – slope map was derived from the DEM using slope tool.

  4. (iv)

    Land use and land cover – LULC maps are the most crucial input for urban growth projection. These data can be used to extract the trend of changes in the landscape.

  5. (v)

    Distance from urban disturbance – it is most likely that the beginning of new urban areas usually spread close to the existing urban areas, i.e. there are high chances for urban growth in nearby places due to neighbourhood effect (Gharbia et al. 2016). To estimate distance from urban disturbance, Euclidean distance tool was used.

Method

The workflow that was carried out in this study consists of the following: (1) image pre-processing, (2) image classification and accuracy assessment, (3) change detection and (4) urban growth prediction using LCM for the year 2030. Flowchart illustrating the methodology used in the study is given in Fig. 2.

Fig. 2
figure 2

Flowchart of Methodology adopted in the study for change detection and prediction of urban growth

Full size image

Image pre-processing

Pre-processing of satellite images is a vital step as it enhances the quality and visibility of images by minimizing the errors (Alawamy et al. 2020). The foremost step is layer stacking, i.e. combining different bands into a standard False colour composite (FCC) image used to determine the colour tone of a specific class (Mohamed and El-Raey 2019). The area of interest (AOI) was masked and extracted from the FCC image.

Image classification and accuracy assessment

The goal of the image classification procedure is to classify image’s pixels into several LULC classes based on their different spectral signatures or reflectance values (Alawamy et al. 2020). To categorize the images, supervised classification technique was performed with the maximum likelihood classifier algorithm which aggregates the pixels having similar spectral properties of the image (Mahmoud and Alazba 2016; Maina et al. 2020). For this study, NRSC classification scheme was adopted and images were classified into 12 LULC classes, namely forest, scrub, grass, urban, rural, crop land, fallow land, barren area, open area, waterbody, riverbed and mining. The LULC maps for the years 1989, 2000, 2010 and 2020 have been prepared. Description of these LULC classes is presented in Table 2.

Table 2 Description of the land use land covers classes identified in the study area
Full size table

Due to the high spatial heterogeneity in the study area, recoding was done on the basis of local knowledge, familiarity with the study area, field observations and Google earth images for improving the classification results considerably (Sarkar and Chouhan 2019). Accuracy assessment was performed on the classified images as it quantitatively evaluates how accurately pixels are classified into different LULC classes and examines the reliability of classified image (Alipbeki et al. 2020). High-resolution imagery from Google Earth Pro and field visits were used as reference data of corresponding time periods (Luo et al. 2015) and 610 randomly distributed ground control points (GCPs) were generated across each across each LULC categories. After comparing the classified images with reference, it gives final result in the form of overall accuracy and kappa coefficient for the years 1989, 2000, 2010 and 2020 (Table 3). The overall accuracy is the ratio of total number of pixels classified correctly to the total number of pixels (Aliani et al. 2019). Value of kappa coefficient ranges from + 1 to –1 which indicates the accuracy between the ground reality classes and the classified classes (Duric et al. 2014; Lillesand et al. 2015).

Table 3 Estimation of LULC changes and accuracy assessment for 1989, 2000, 2010 and 2020
Full size table

Change detection

Change detection is a comparative analysis used to identify the changes in the nature and extent of different LULC features between different periods (Alawamy et al. 2020). In order to detect LULC changes over three decades from 1989 to 2020, the classified images were compared which estimates the net changes, gains and losses in the LULC categories and to understand the dynamics of changes (Singh 1989). The magnitude of change is a rate of expansion or reduction in the area of each class. A negative value indicates loss in the area while a positive value indicates gain in the area (Kafi et al. 2014).

Urban growth prediction using LCM

LCM is an innovative and vertical application in the Terrset software used to predict future urban growth with better spatiotemporal pattern (Hyandye and Martz 2017; Keshtkar and Voigt 2016; Sayemuzzaman and Jha 2014). The following steps were involved:

  1. (i)

    Images of both earlier and later LULC maps, i.e. 2000–2010 and 2010–2020 were used as an input into the change analysis panel to estimate the change between two different LULC that help in predicting changes in 2020 and 2030, respectively (Eastman 2016). The images were reclassified into five LULC classes which are important to predict urban growth namely-scrub, urban, agriculture, open area and other lands. There is a spatial trend analysis tool that gives output of pattern of changes. Trends of 3rd and 4th order were calculated.

  2. (ii)

    The transition potential panel displays a list of transition sub models that consist of a group of transitions existed between two LULC classes and filtered out the specific sub models required for the urban growth prediction (Lu et al. 2019). The transition probability matrix was computed based on the later LULC image set as starting point for change simulation by Markov chain which compares all pixels and their transition for each LULC class (Kumar et al. 2016; Eastman 2016). This reflects the probability and quantity of expected LULC change from one class to another (Al-sharif and Pradhan 2014).

  3. (iii)

    Some of the driving factors that act as agents of urbanization are responsible for the cause of dynamics of LULC events (Eastman 2016). Cramer's V coefficient has been used to assess association of each of the driving factor towards urban growth. Higher the Cramer’s V coefficient value indicates that the driver has higher relationship with urban growth (Eastman 2016; Eastman and Toledano 2018). In this study, we have considered major variables based on literature, i.e. elevation, slope, distance from roads, distance from urban disturbances and trends of 3rd or 4th order.

  4. (iv)

    Transitions are modelled using multi-layer perceptron (MLP) neural network method as it allows multiple transitions at one time and equal sample size per class (Devendran and Lakshmanan 2019; Eastman and Toledano 2018). This empirical method operated in automatic mode as it decides on its own how parameters should be used and changed to get better output data (Gidey et al. 2017). The dynamic learning procedure initiates with start learning rate and lowers down successively as it approaches to end learning rate and with this number of iterations completed (Kumar et al. 2016; Eastman and Toledano 2018). During each time step, pixels with higher transition probability changed to new land type class and others remain unchanged to generate a new land map at the end of each iteration (Wang et al. 2019). Once the training process is completed, it gives transition potential images (Eastman 2016).

  5. (v)

    Based on the proximity concept of CA, spatial reallocation of each land class is estimated and overlaid to produce output (Kamusoko 2012). For predicting the future scenario for a specific period, previous rates of change and transition potentials were used. We have chosen the hard prediction mapping which is based on multi-objective land allocation (MOLA) module of LCM (Abuelaish 2018; Kumar et al. 2016).

  6. (vi)

    Model validation is a decisive part as it examines the predictive power of any model and dependability of outcome (Gidey et al. 2017; Kamusoko 2012). For this purpose, a comparison is done between actual and predicted maps of the year 2020 (Singh et al. 2015).

Results and discussion

LULC change detection

LULC change is a continual process influenced by socioeconomic, political, cultural, demographic and environmental factors (Rahman et al. 2012). It needs a comprehensive understanding and monitoring of all the factors which are important for establishing association between the policy decisions, driving forces of the process and to overcome the problems (Mahmoud and Alazba 2016; Mallupattu and Reddy 2013). The change analysis is carried out to understand the land transformation and trends in the LULC. Positive and negative changes are noticed over 31 years in the area under the LULC categories. The coverage percentage of each land use class in the years 1989–2020 is presented in Fig. 3. The LULC classification results which are summarized for the 1989–2020 period are shown in Table 3. The LULC maps for the years 1989, 2000, 2010 and 2020 are depicted in Fig. 4. Results of gains and losses of different LULC classes are presented in Fig. 5.

Fig. 3
figure 3

Graphical representation of LULC change for selected time periods in the years 1989–2020

Full size image
Fig. 4
figure 4

Classified maps of the study area of a 1989, b 2000, c 2010 and d 2020

Full size image
Fig. 5
figure 5

Gains and losses of LULC classes for each time period

Full size image

Green cover (forest, scrub and grass)

The green cover areas are sanctuary, ridge, biodiversity parks, botanical gardens, protected and reserved forests, city forest, recreational areas and parks. Green cover area is observed to have declined from 1989 (1644.72 km2) to 2000 (1319.89 km2) and also further significant decline is observed from 2000 (1319.89 km2) to 2010 (1168.88 km2). This is attributed to conversion of forest into scrub and green cover into built-up and crop land classes prominently. The study area comprises of Aravalli range which was an active mining site with stone crushing zone. This left the area highly deteriorated and deprive of green cover. Other anthropogenic activities such as illegal tree cutting, continuous expansion of settlement and roads traverse through green cover reduce the natural vegetation. These human-induced changes caused fragmentation of green cover, tremendous loss of habitat for native flora and fauna, decline in air quality, adversely affect the ecological settings of the region and unfavourable hydrological changes (Mengistu and Salami 2007). This raised the concern at global and national level through the United Nations Environment Programme, Convention on Biological Diversity, the Earth Summits, Biodiversity Act 2002, setting up of protected areas and so on in order to save the natural heritage and maintaining the environmental qualities.

The area of green cover increased by 4.55% in 2020. The transformation of land from scrub to forest and expansion of the green zone can be attributed to the implementation of reforestation, afforestation and restoration programmes initiated by government, academic institution and local communities (CPCB 2020; Sinha 2014). A ban was imposed on mining in 2004 (Iamgurgaon 2020) and conservative as well as innovative concept of biodiversity park and city forest were evolved to rewild the area and revive the nearly extinct biodiversity (DBF 2016). In Delhi, there are 4 ridges (Northern ridge, Central Ridge, South Central ridge and Southern ridge), 7 biodiversity parks (Yamuna Biodiversity park, Aravalli Biodiversity park, Neela Hauz Biodiversity Park, Tilpath Valley Biodiversity Park, Kamla Nehru Ridge, Tughlaqabad Biodiversity Park and Kalindi Biodiversity Park), Asola-Bhatti Wildlife Sanctuary and 26 notified protected and reserved forests such as Sanjay Van, Jahanpana forest, Rajokri forest and Mitraon forest (Forest Department 2015). However, there are only a few notified green areas — 5 in Gurugram (Sultanpur bird sanctuary, Sarbasirpur reserve forest, Eco restoration Chakkarpur Bundh, Badshahpur Forest corridor, Sikanderpur forest area), 5 in Gautam Buddha Nagar (Okhla bird sanctuary, Surajpur bird sanctuary, Night safari, Miyawaki forest, biodiversity park sec-91), 1 in Faridabad (Magar Bani forest) and 1 in Ghaziabad (City forest). So, these areas must focus on increasing the green breathing space which are self-sustained and healthy ecosystems that act as green lung to metropolitan cities. The green cover in the study area needs more protection as they are still under threat and more disturbed due to encroachment, unsustainable development, illegal construction, illegal tree cutting, proximity to densely populated urban centres, construction of highways and expressway bisect them. These green areas help in combating the local climate change and boost the quality of urban environment as they act as sink for urban pollutants, improve local weather conditions and mitigate the urban heating (Singh 2017). The grass area witnessed an increase in its size from the initial 60.07 km2 in 1989 to 122.17 km2 in 2020 as the riverbed, farm house lawns and abandoned land got covered by grass with the time.

Waterbody

The waterbody area shows a trend of decreasing first to 5.58 km2 during the period 1989–2000, then increasing 32.95 km2 (2000–2010) and again decreasing status 42.84 km2 (2010–2020). The rapid decrease in waterbodies is occurred due to their conversion into agricultural areas, filling them up for creating human settlements, dumping of solid waste materials, discharge of untreated sewage causes biomagnification and construction of embankments for preventing flood water entry. This causes shrinking, degrading, vanishing and drying up of waterbodies. In fact, many of waterbodies have become open drains and lost their self-purification abilities due to human disturbances. Presently, the study area depends on the Yamuna river, Hindon river and groundwater sources for its water supply which are contaminated and water table level has been dropping from 1 to 3 m every year at different places (CGWB 2015; Dev et al. 2020). These water-stressed cities are largely urbanized with a huge and growing population required to accomplish the accelerating fresh water demand which is becoming more difficult with every passing day. To address this semi-critical situation, rejuvenation of the city’s water networks becomes necessary. The concept of constructed wetlands is adopted which work similarly to the natural ones; absorb the impurities in water through natural processes at low-cost, ensures the original ecological integrity and recharging the groundwater (Gorgoglione and Torretta 2018). Some successful cases of the restored lakes or wetlands in the study area are Neela hauz, Hauz khas lake, Surajpur wetland, Rajokri lake, Old Fort lake and Najafgarh Jheel (Saxena 2017). The noticeable change observed is that deep pits of mines within the green cover area have filled up with water and converted into healthy functional lakes gradually. But the current rate of population and its demand growing is higher than that of rate of reviving waterbodies which is at slow pace. There is need of quick initiatives and speedy participation with careful balancing between development and environment.

There are obstacles in conserving and preserving the waterbodies present in the study area. For example, encroachment, construction of skyscrapers looming nearby, large amount of domestic and industrial effluents released into them, appearance of unusual white foam, concretization around waterbodies, and waste dumping have (i) drastically degraded the quality of water, (ii) blocked the rain water absorption, and (iii) affected the aquatic organisms. Some of the practices can be adopted for reviving and improving the health of waterbodies such as Phytorid-Scientific Wetland with Active Biodegradation technology developed by NEERI, establishment of porous pavements, installation of rain water harvesting systems for residential and commercial buildings as well as flyovers and roads, removal of encroachments, creating awareness about the waterbodies and their potential, and active water resources management (Saxena 2017; Ritter 2019).

The waterbodies showed a maximum increment in the area 132.32 km2 in 2010. This happened because of the cloudburst incident taken place at Leh which also affects the study area. The area received highest rainfalls and Yamuna river overflows (Guhathakurta et al., 2020).

Urban area

During the 1989–2020 period, the level of urbanization in the area increased remarkably from 407.06 to 1207.59 km2. It indicates that urban area is experiencing continuous growth in the last 31 years. On the contrary, the crop land, open area and green cover reduced during the same time. At present, the study area is characterized by intensive growth in the population (Figs. 6 and 7) and high inflow of migrants resulting in the accelerated demand for housing and facilities that can be attributed to the change in land use pattern. Table 4 shows total migrants and their socio-economic reasons for migration.

Fig. 6
figure 6

Decadal population growth

Full size image
Fig. 7
figure 7

Decadal population density

Full size image
Table 4 Socio economic factors for migration and total migrants
Full size table

There are certain factors that are responsible for attracting the population and evident for the changes in the urban cluster level are as follows:

  1. (i)

    Transportation infrastructure plays a crucial role in providing mobility to people, goods and services (Aljoufie et al. 2011). It tends to stimulate an active urban growth, land use changes and enhances the level of economic activity through the accessibility it provides to the region (Marovic et al. 2018). In the study area, road network and metro rail are the main drivers of urbanization as well as expansion of fringe zones of Delhi due to the fast and good connectivity providing links to various places and close proximity. Road network serves as a skeleton for urban development and dominant transport asset used on a daily basis by millions of commuters (Shi et al., 2019). Table 5 shows major roads in the study area and their expansion with the time to bring different regions closer together. The increase in the total road length from 36,248 (2012) to 38,716 km (2019) (DES 2019a; DESH 2019; DESUP 2019), aided in linking the cities (NCRPB, 2021). In the study area, these expressways and highways act as catalyst behind the rising of numerous commercial and residential complexes. It attracts both buyers and investors which offers a lot of scope for future townships. Metro rail is a mass rapid transit system connects all five main metropolitan areas of study with each other having 12 operational lines covering total length of 391 km (Metro 2022). The metro rail stretches have been witnessing rapid development. There is a continuous increase in ridership, i.e. average daily ridership was 80,000 (2002), 12.59 lakhs (2010) and 57 lakhs (2020) (ESI 2020).

  2. (ii)

    The 7th 5-year plan (1985–1990) and 8th 5-year plan (1992–1997) of the Government of India proposed the development of small and medium towns around large cities. These towns aimed as key centres of attraction to the migrants by providing availability to economic activities and employment opportunities that could further help in reducing immigration into the large cities (Chachra, 1993). Accordingly, the National Capital Region Planning Board was established in 1985 and new district Gautam Budh Nagar is created in 1997 to minimize the pressure of population in Delhi. In the study area, the diversion of population influx towards Ghaziabad, Noida, Greater Noida, Faridabad and Gurugram is observed.

  3. (iii)

    The New Economic Policy (1991) opened the doors for globalization that gave entry to private and foreign players, lowered the import duties and allowed the international flow of goods, services and technology with few restrictions (Madaan and Madan 1995). This raised the economic growth centres in the study area related to various fields such as automobiles, banking, pharmaceutical, information technology, food processing, telecommunications and pesticide that generates more employment opportunities (Dev et al. 2020). With the entry of multinational companies (MNC’s), the number of industries within the study area had increased from 11,836 in 2011 to 13,038 in 2018 (Table 6). It implies large-scale infrastructure construction with more residential spaces for working people. Therefore, it can be noted that changing pattern of economy had a considerable role in LULC dynamics in the national capital and urban fringe zones. In Delhi, there is a decrease in the industries due to strict environmental norms and sealing of small-scale industries caused the shifting to nearby outskirt area while increase the corporate sector simultaneously (DES 2019a, 2019b).

    The Special Economic Zone Act (2005) created free trade zones that have more liberal economic laws, efficient transport system, uninterrupted power supply and global telecommunication network for enhancing foreign investment. Noida and Gurugram are Special Economic Zones present in the study area. The maximum extension of urban area has been observed from Noida than Gurugram. The government is launching policies time to time for creating an ecosystem for doing business with ease such as Enterprise Promotion Policy 2015 and Invest India that attracts investors and large number of employees towards Delhi-NCR.

  4. (iv)

    The study area is the home to many educational institutions and renowned universities which bring people from all over India. Other infrastructure facilities also are getting available with urban development such as malls, hospitals, community and sports facilities and water supply which further attract people to have a good standard of living (Qadir et al. 2020).

Table 5 Changes and expansion of major roads in the study area
Full size table
Table 6 Number of industries and number of employees in the study area
Full size table

The study area is subjected to intensive human influence and to meet its daily needs has its direct impacts on the resources of the region. The transformation of land into urban areas leads to increase in impervious surfaces that decrease the percolation of water into the soil as a result decrease in the ground water recharge (GWYB 2019). Land use changes effect the biodiversity of the region and a constant intrusion into green cover. The study area witnessed haphazard growth of construction area which do not follow local laws as a consequence the area suffered from different types of pollution (air, water, noise and light pollutions).

The World Air Quality Report (2019) highlighted the elevated PM 2.5 pollution level in the air and faces the environmental and human health risk issues gained from the rapid urbanization of metropolitan cities. The most polluted cities are from study area — Ghaziabad (110.2 µg/m3), Delhi (98.6 µg/m3), Noida (97.7 µg/m3), Gurugram (93.1 µg/m3) and Greater Noida (91.3 µg/m3). During lockdown period of COVID-19, reduction in the air pollutant concentration is observed significantly from 55 to 67% due to temporary drop down in economic activities (Greenpeace India 2020). Therefore, the government should focus in implementing the sustainable development along with green framework around the urban clusters to protect the human health and ensure a healthy planet for all.

Rural area

There is a gain in rural area from 131.2 (1989) to 218.22 km2 (2020). It is observed that the number of villages and size of villages increases as the number of villagers also increases as well. It occurred because of the economic transformation takes place in the surrounding areas. But simultaneously, some of the rural areas have been occupied for urban development.

Crop land

The crop land which has an area of 2504.74 km2 in 1989 has raised to 2654.87 km2 in 2000. The expansion of crop land due to increased demand for food is the major factor that opens up an opportunity to bring green cover under cultivation.

The study has shown a sudden decrease in the available agricultural land from 2452.3 (2010) to 1856.82 km2 (2020), with net decrease of 595.48 km2 within the study area. The study area witnessed a large amount of crop land acquired by the government authorities for urban settlements and infrastructural development. There are many private real estate developers that take up the crop land for various upcoming and ongoing projects and have changed its land use pattern of the area.

Fallow land

The fallow land is reduced in the study area from 493.2 (1989) to 354.43 km2 (2000) and further decreased to 313.87 km2 in 2010 as cultivation has resumed in some of the parts again and some of the areas has converted to built-up area. The net increase shown in the area is 69.68 km2 in 2020 as they are not cultivated for one or more seasons.

Open area

The open area increased continuously during the 31 years from 355.63 (1989) to 498.3 km2 (2020) as the area acquired by the government and private sector for the upcoming construction projects. The disturbed area of green cover is also represented by open area because of the human influence.

Barren area

The barren area reduced by 21.74 km2 (2010) and moved to open area, crop land and recreation cover. During last one decade, the total barren area increased as the other land cover getting abandoned.

Riverbed

The study has revealed a constant decrease in the riverbed area as some of the portion has been occupied by crop land or build up while some part is naturally covered by grass or shrub with the time.

Mining

The Aravalli range in the study area was adversely affected by the mining activity which was banned in 2004 (iamgurgaon 2020) and area was reforested with the help of eco task force.

Urban growth Prediction

The study area consists of Ghaziabad, Gautam Buddha Nagar, Faridabad and Gurugram which no longer to be considered as satellite towns to Delhi. With the pace of time and accelerated development, they become large urban agglomerations and now can be classified as independent cities. But after the opening of the economic corridors and changes in policies, the study area experienced haphazard, unplanned and unauthorized colonization and development activities. The advancement in remote sensing and GIS technologies and large amount spatiotemporal data availability helps in evaluating changes in the past and forecasting the future urban growth can provide assistance to decision-makers for developing a strategic sustainable plan (Fathizad et al. 2015). The results indicate that 20.28% of the entire study area has been occupied as urban area in 2020 which will be expected to increase by 24.30% in 2030 is explained by the decline in agriculture and open area. The change trajectories between the actual and predicted land use land cover classes for the year 2020–2030 are shown in Fig. 8. Figure 9 illustrates a comparison of the variations in urban area between 2020 and 2030. The changes in the area for years 2020 and 2030 are presented in Table 7. Table 8 represents the potential explanatory power of each driving force, calculated by Cramer’s V.

Fig. 8
figure 8

Graphical representation of actual and predicted LULC classes for 2020–2030

Full size image
Fig. 9
figure 9

Urban growth maps of the study area (a). Actual map 2020 (b). Predicted map 2020 (c). Predicted map 2030

Full size image
Table 7 Variation in LULC area of actual 2020 and predicted 2020 and area statistics of predicted 2030
Full size table
Table 8 Factors considered and their Cramer’s V values for predicting urban growth
Full size table

To a large extent, there are consequences of urban growth as it puts pressure on the natural resources which requires a necessary consideration and effective measures towards them such as shortage of water resources, landscape fragmentation, loss of arable land, deforestation, land degradation, mitigating local climatic conditions and pollution (Gidey et al. 2017; Lu et al. 2019). Therefore, in order to balance the growth of urban areas, economic benefits, protection of arable land and conservation of the natural environment, a holistic approach with harmonizing way is essential (Koko et al. 2020).

Model validation

Validation of model was performed by comparing the predicted map of 2020 with the actual map of 2020 (Gidey et al. 2017). The overall accuracy and kappa coefficient of the predicted output of 2020 are found to be 98.63% and 0.96 respectively. This reveals that LCM model is capable of forecasting the future changes successfully. Although the model used in this study is an effective technique for performing the simulation, there are several uncertainties in the prediction of land type classes in the future, which are described as follows: First, the future changes made by the government in the land development policies and master plans as well as natural processes that effect the land use pattern cannot be estimated through model. As the model predicts, the future changes on the basis of present LULC patterns identified (Keshtkar and Voigt 2016). Second, the driving factors keep on changing with the time. For example, change in the transportation system directly influences the direction and shape of growth in the city as settlement locations and density shifts. Thus, the selection of most favourable set of driving forces is important to improve the model prediction accuracy; otherwise, it may result in error in estimating changes correctly (Arsanjani et al. 2013; Wang et al. 2019).

Land use and planning

There is a complex relationship between land use land cover changes, human activities and sustainable use. The gaps in the present land use planning (Master Plans) of the study area that needs to be filled are as follows.

  1. (i)

    We should construct regularly updated databases that include population economics, population density and environmental issues which help in framing strategic plans. The lower the density per square kilometre of a region makes things to be managed in a better way. For example, the Ghaziabad Municipal Corporation is not taking the necessary steps to address solid waste remediation (NGT 2020) as its the population density is high, i.e. 3971 persons per sq.km (Census 2011).

  2. (ii)

    To mitigate immigration and excessive load on natural resources, the government should take decisions and efforts for sustainable development by uniform distribution of economic zones, services and facilities throughout the country instead of creating infrastructure at particular places.

  3. (iii)

    The master plans of different areas in the study excluding Delhi have mentioned the recreational or open space land use category which is generally used in the form of public parks, playgrounds, golf course and amusement park but this will not cope up the air pollution problem. Proper planning is required to deal with pollution and provide healthy life to the resident people. There should be conscious efforts for increasing green cover with the creation of healthy dense green areas or biodiversity parks to mitigate local environmental problems.

  4. (iv)

    After the establishment of Delhi Development Authority (DDA), it has prepared only three master plans for the years 1962, 2001 and 2021, Ghaziabad Development Authority (GDA) master plans 2001 and 2021 (interval of 20 years), and NOIDA master plan 2021 and 2031 (interval of 10 years). This prolonged time interval between the master plans could be the origin of haphazard growth in the study area. There is a need to amend the duration of master plan updation to meet with the pace of fast-growing urban cities. This will help the planners and authorities effectively monitor and manage land uses and also formulation of integrated systematic master plans which can be modified according to the ground feedback.

  5. (v)

    During the preparation process of Master Plan, there is a lack of public and stakeholder’s participation which helps in understanding and resolving local level issues and improving social infrastructure. In this direction, DDA is taking suggestions from the public for preparing master plan for 2041. Other regions may also follow bottom-up approach.

  6. (vi)

    To adopt nature-based solutions and sustainable urban designs such as walkability principle. To make the city walkable, it should meet certain conditions — safety, continuity, comfort, accessibility and attractiveness. It would help in health, environment and socio-economic benefits (Shafray and Kim 2017).

  7. (vii)

    Now a days, plantation of ornamental plants is more on roadsides, sidewalks and parks. It should be mixed type of plantation which includes native species, medicinal species, air purifying species, ornamental species and fruit species. It could help in reconnecting people with local vegetation, improvement in quality of environment, increases attractiveness, provide fruits to homeless people and welcoming back the birds to cities.

Conclusion

The study demonstrates that the integration of remote sensing, GIS and LCM with multi-temporal data helps in understanding and assessing the LULC dynamics and future trends of urban growth. LULC changes serves as a scientific basis to understand the implications of cause–effect relationship between humans and environment that studied area is experiencing, such as poor drainage, waterlogging, imperious surfaces, ecological problems, over exploitation of natural resources and unauthorized constructions. This is the result of weaker policy implementation and less responsive behaviour.

The study revealed a significant expansion of urban area from 1989 to 2020 due to government policies, population growth, infrastructure facilities and also close proximity to capital. There is an improvement in the green cover area which may benefit the local climatic conditions, provide valuable ecosystem services and enhance the quality of life. The decrease in the waterbody may contribute to irregularity in water supply and water scarcity. It is required to maintain equilibrium with the nature by creating sustainable cities as land is limited and its utilization should be judicious and well managed. In the view of tomorrow, sustainable cities should offer healthy environment, more green spaces, zero waste, carbon neutrality with progressive and green economy. To strengthen this process, it is imperative to actively engage local communities in planning and monitoring activities.

The study shows that there are number of drivers that challenges the future development of sustainable cities. Therefore, urban simulation studies can help in understanding the effects of future changes and drawbacks of current planning policies to safeguard the adverse effects. The prediction results indicate that the model has good potential and accuracy to stimulate the future trends. Such database is very useful for multi-criteria decision making, environmentally sound management and future urban planning to harmonize the demand and reduce the impacts arises from land use pattern.

Furthermore, the study offers the policy-makers and planners to improve the policies and formulate new master plans with integrated multidisciplinary framework. Considering the difference between current situation and projected changes in urban growth, the study area needs to take short-term action plans with long-term vision to achieve urban green–blue network. There must be strict building by-laws on all construction activities which restricts haphazard growth within the cities, introduction of urban regeneration projects at all levels (economic, socio-cultural and environmental) and redevelopment of urban riverfronts of Yamuna and Hindon.