Urban Growth Monitoring and Prediction Using Remote Sensing Urban Monitoring Indices Approach and Integrating CA-Markov Model: A Case Study of Lagos City, Nigeria

Abstract

As one of the swiftly advancing megacities globally, Lagos faces significant challenges in managing its urban expansion. Mainly, this study focuses on monitoring and predicting urban growth using a comprehensive approach incorporating Global Land 30 (GL30), satellite-based nighttime light observations, and built-up and population density data. The application of remote sensing techniques, combined with utilizing the GL30 dataset, provides an effective means to monitor and predict urban growth trends and patterns. The major patterns occurred from 2000 to 2020, including increased cultivated land; reductions in grasslands, shrublands, and wetlands; and major urbanization. Predictive models indicate that urbanization will continue. Furthermore, employing the Cellular Automata (CA) Markov model in land-use and land-cover (LULC) change prediction. The findings revealed significant changes in LULC over the two decades. Particularly, the percentage of artificial terrain increased from 17.016% to 25.208%, and the area under cultivation increased significantly, rising from 46,771 km² (1.238%) in 2000 to 75,283 km² (1.993%) in 2020. Grasslands fell from 7.839% to 1.875%, while forest cover somewhat increased, climbing from 39.319% to 43.081%. Additionally, marshes fell from 9.788% to 5.646%, while shrublands decreased from 4.421% to 2.640%. Surprisingly, bare ground decreased sharply from 0.677% to 0.003%. To forecast future LULC changes, the study also used a Markov Chain Transition Matrix. According to the data, there is a 3.54% chance that agricultural land will become urban, converting it from being used for agriculture to urban development. On the other hand, just 1.05% of forested regions were likely to become municipal areas. This study offers foundations for the upcoming research to enhance urban growth models and sustainability strategies in the face of rising urbanization and environmental concerns in the region, as well as laying the groundwork for informed decision-making in the region.

1. Introduction

Global cities have undergone significant growth and transformation in recent decades [1,2], accompanied by substantial changes in land use. Urban growth monitoring and prediction play pivotal roles in understanding the evolving dynamics of urban landscapes, offering essential insights for sustainable urban development [3,4,5].

These measures evaluate human activity and socioeconomic advancement, encompassing qualitative and quantitative aspects across different temporal and spatial dimensions [4]. Advances in satellite sensors and related technologies, through remote sensing approaches, have facilitated urban geospatial monitoring and visual cognition research [6]. Utilizing remote sensing data from various sources allows for an objective and scientific analysis of the spatio-temporal dimensions of urban development and changes in land use [7].

Different studies have underscored the insufficiency and lack of reliability in the existing data and information regarding urban growth in Nigerian cities, presenting a hurdle for well-informed decision-making [8,9,10,11]. Metropolitan areas such as Lagos face a shortage of accurate and current information concerning the scale of urban expansion, impeding comprehensive urban planning efforts. Therefore, regularly updated datasets, like GlobeLand30, satellite-based nighttime light (NTL) observations, built-up data, and population density data, have become indispensable for monitoring and forecasting urban growth in developing nations.

Urban growth is being monitored and predicted using GlobeLand30, a product developed by China’s National Geomatics Centre [12,13,14]. Taking the timeframe from 2000 to 2020 and presenting high-resolution images at a 30 m scale, this dataset shows considerable potential for delineating regions with sparse land-cover details, particularly in developing nations [14].

Observations derived from satellite-based nighttime light (NTL) have demonstrated efficacy in tracking human-generated sources of light in urban areas, towns, and various locations, including traffic patterns. This capability facilitates the assessment of spatial-temporal shifts in socioeconomic activities and the progression of urbanization [15]. Unlike conventional remotely sensed data, which characterize landscape morphology and texture, NTL data detail light connections and spatial-temporal distribution [16]. Incorporating nighttime light data for delineating and mapping metropolitan regions offers the advantages of global mapping of urban entity expansion with increased temporal frequencies and reduced spatial resolution [17]. Although past research endeavors have sought to extract urban entities from NTL data, various challenges persist, underscoring the need for additional investigation. Hence, it is essential to employ impervious surfaces or urbanized areas as validated references [18]. Urban regions are intricate assortments that mirror the spatial clustering of human activities; hence, the examination and validation of urban entities derived from NTL data are crucial for credibility and applicability. Researchers mainly apply diverse indices, like the normalized difference built-up index (NDBI), to delineate urbanized areas [19].

The absence of prompt, accurate, and reliable data about the position, spatial scope, pace, and underlying influences of urban development in expansive cities within developing nations, exemplified by Lagos, has impeded the formulation of effective planning strategies and municipal governance for a prolonged duration [9,11,20].

Therefore, our study aims to bridge this critical data gap by leveraging remote sensing technologies and comprehensive data analysis techniques. It intends to provide valuable insights into the location and prediction of future urban growth using the Integrating CA-Markov Model. As a result, our research will extensively use available sustained data to conduct a long-term analysis of urban area evolution, along with projecting future urban growth data using the newly updated Global Land 30 data. This research on urban growth monitoring and prediction in Lagos City, Nigeria, using Global Land 30 data and remote sensing urban monitoring indices, contributes to the broader understanding of urbanization dynamics in rapidly developing regions. By integrating comprehensive datasets and advanced analytical tools, the study offers a nuanced examination of the factors influencing urban growth, providing valuable information for informed decision-making and sustainable urban development strategies.

2. Literature Review

Urban growth monitoring and prediction have emerged as critical elements in contemporary urban planning strategies, applying advanced methodologies and technologies to ensure accurate assessments [21]. Extensive research has delved into understanding the dynamics of urban expansion, particularly focusing on harnessing remote sensing technologies and datasets like Global Land 30 (GL30) to augment the monitoring process [8,12,22,23,24]. They conducted a comprehensive study incorporating GL30 data and a remote sensing urban monitoring indices approach to forecast urban growth. Their findings highlighted the significance of employing advanced data processing techniques for precise and prompt analysis of urban enlargement and land-use changes.

Studies have underscored challenges associated with insufficient and unreliable urban development data within developing countries such as Nigeria [25]. These investigations emphasized the pivotal role of continuous datasets like GL30 in bridging substantial data gaps, which are essential for effective long-term urban planning. Similarly, Wang and Maduako [11] shed light on the scarcity of up-to-date data in cities like Lagos, emphasizing the pivotal role of remote sensing technologies in addressing these challenges.

Various studies have emphasized the effectiveness of remote sensing technologies, especially when combined with GL30 data, in monitoring and predicting urban growth [15,26], and the significance of nighttime light observations in monitoring human-induced light sources and processes of urbanization. These observations provide crucial insights into light distribution and socio-economic activity changes, facilitating a deeper understanding of urban growth dynamics.

The collective body of literature underscores the pivotal role of remote sensing technologies and datasets, such as GL30 and NTL (nighttime light observations), in urban growth monitoring and prediction [12,15,16,26,27,28]. It stresses the need for continuous and reliable data sources to address critical gaps in rapidly urbanizing regions. Integrating advanced data-processing techniques and urban monitoring indices provides a comprehensive framework for comprehending the dynamics of urban expansion for facilitating sustainable urban planning and development. Remote sensing continuously captures valuable data crucial for predictive studies [29]. The accuracy of analyzing historical and present conditions significantly influences the reliability of forecasted changes [30].

Past research extensively delved into simulating land-use changes using the CA-Markov model. Parsa et al. [31] effectively applied this model to Iran’s Arasbaran biosphere reserve to forecast future LULC patterns, aiding decision makers in addressing forthcoming land-use challenges. They highlighted its usefulness in shaping land-use policies and its potential as an early warning system. However, Ozturk [32] compared the performance of the CA-Markov and MLP-MC models in predicting LULC changes in Atakum, Samsun, Turkey, discovering that the MLP-MC model yielded superior projections compared with the CA-Markov model. Conversely, Regmi et al. [33] compared CA-Markov and GEOMOD models when analyzing LULC dynamics in Nepal’s Phewa Lake watershed, concluding that CA-Markov chains effectively projected future scenarios. These studies integrated diverse driving forces such as infrastructure, socio-economic factors, and terrain characteristics, demonstrating their influence on the spatial patterns of the watershed’s LULC.

In summary, these researchers collectively underscore the complexity of LULC and emphasize that Markov-based cellular automata models offer extensive insights into the intricate aspects of spatial systems.

Understanding the drivers behind urban expansion involves exploring both biophysical and socioeconomic factors [34]. Various mathematical models simulate and predict urban expansion; however, the Combined Cellular Automata-Markov chain model can produce a more refined spatio-temporal pattern of LULC change [35]. Hence, the CA-Markov model emerges as a fitting choice for the present study, given that the Markovian model gauges the magnitude of changes, while a CA model assesses spatial alterations. Therefore, the model is suitable for predicting Lagos’ future urban growth changes.

Finally, this study aims to contribute further to the growing urban growth monitoring and prediction field. It applies Global Land 30 data (GL30 data) and remote sensing urban monitoring indices better to understand urban growth in Lagos City, Nigeria.

3. Materials and Methods

3.1. Study Area Description

Lagos lies in the coastal lowlands of southwest Nigeria, specifically between latitudes 6°22′–6°42′ N and longitudes 2°42′–4°20′ E, respectively (Figure 1) [22]. The Republic of Benin bounds Lagos State on its western side, while the Nigerian state of Ogun bounds it on its northern and eastern sides. While the Atlantic Ocean borders it on its southern side. Lagos, with 9,113,605 inhabitants in 2006, is Nigeria’s most populous state, accounting for roughly 6.44% of the country’s total population. It also has the continent’s most populated city, with 12 million [20]. Furthermore, among the 28 present and future megacities, Lagos was ranked ninth in a recent United Nations study on World Urbanization Prospects [11].

3.2. Data and Methodology

3.2.1. GlobeLand30 Data

The importance of employing the GlobeLand30 dataset becomes apparent in tracking urban expansion. This dataset’s exceptional 30 m resolution perfectly meets the requirement to objectively and scientifically examine the spatio-temporal aspects of municipal growth and alterations in land use [7]. Developing countries find it especially significant as it offers crucial data to delineate and oversee regions with incomplete information on land cover [8,14].

The expansion of urban areas, which significantly affects environmental sustainability and urban planning, requires accurate and thorough data sources. The 30 m fine resolution provided by the GlobeLand30 dataset emerges as a transformative factor in this context. This level of precision enables researchers and planners to explore the subtleties of urban growth dynamics in intricate detail. Whether urbanized regions, natural characteristics, or vegetation, the dataset facilitates precise mapping of various land-cover types, providing valuable insights into the continuously growing urban landscape (Figure 2).

In monitoring urban expansion, particularly in developing nations where data gaps are often notable, the GlobeLand30 dataset assumes increased importance. It effectively addresses a crucial challenge by enabling the mapping and monitoring of areas with incomplete land-cover data. This detailed data source sheds light on previously overlooked or poorly understood regions. As a result, decision makers, urban planners, and researchers gain the ability to make informed choices for sustainable development, considering areas that were previously underrepresented or not adequately understood.

In the broader context of urban development and land-use planning, the GlobeLand30 dataset stands as a crucial instrument. Its 30 m resolution presents an unparalleled opportunity for thoroughly monitoring urban sprawl and its consequences. In pursuing sustainable urbanization, this dataset provides the precision and richness essential for shaping policies, making informed decisions, and ensuring that future cities evolve in a balanced and environmentally conscious manner.

3.2.2. Satellite-Based Nighttime Light

This study utilized nighttime light data maps covering the period from 1992 to 2013, sourced from two providers: the DMSP/OLS and the NPP-VIRS. The DMSP/OLS data span from 1992 to 2013 with a resolution of 1 km by 1 km. On the other hand, the NPP-VIRS data cover the years 2012 to 2020 and offer a higher resolution of 500 m by 500 m. The data from 1992 to 2013 originated from the global nighttime light data captured by the Operational Line Scan Sensor aboard the Defense Meteorological Satellite Program (DMSP-OLS data) satellites launched in 1976. Over the period from 1992 to 2013, the DMSP-OLS dataset illustrates a representation of stable lights, with each image of consistent lights constructed from all cloud-free DMSP-OLS data for that respective year. These stable lights correspond to nightlights, with DN numbers ranging from 0 to 63 (Figure 3).

Using nighttime light data from satellites is crucial for observing and characterizing urban expansion processes, providing valuable insights into the spatiotemporal dynamics of urbanization [4,22]. Integrating land-use data with nighttime light data yields valuable fresh perspectives on the connections between land-use alterations and variations in the brightness of municipal nighttime lights [36]. This combined investigation improves our comprehension of the dynamics of urbanization, a critical aspect of making informed decisions related to infrastructure development, sustainable management, and urban planning. Given that human activity predominantly generates artificial light, it is a reliable indicator of assessing economic activity and the progression of urbanization [37]. Higher levels of artificial light are a sign of more urbanization and affluence in the economy [15,16,22].

3.2.3. Other Data

The analysis of Lagos City’s urban expansion also involved the examination of built-up data in Figure 4, along with the utilization of various indices (NDBI, NDVI, BUI). Various image features, spanning from 2000 to 2020, were extracted from Landsat TM multi-temporal satellite data accessible on the website (https://earthexplorer.usgs.gov/ (accessed on 1 June 2023)), and applied to map and oversee the expansion of urban areas in Lagos city (

Table 1. Satellite image information.

Sensor	Landsat 5	Landsat 7	Landsat 8
Band	Resolution	Resolution	Resolution
(m)	(m)	(m)
Band 1	Blue—30	Blue-Green—30	Coastal—30
Band 2	Green—30	Green—30	Blue—30
Band 3	Red—30	Red—30	Green—30
Band 4	Near IR—30	Near IR—30	Red—30
Band 5	SW IR—30	SW IR—30	Near IR—30
Band 6	LW IR—120	LW IR—60	SW IR1—30
Band 7	SW IR—30	SW IR—30	SW IR2—30
Band 8	Pan—15	Pan—15	N/A
Band 9	Cirrus—30	N/A	N/A

). This study discriminates between non-urban and urban built-up regions using the Building Density Index (BUI) and the Nighttime Brightness Index (NBI). Non-urban vegetation encompasses parks, gardens, open spaces, water bodies, and agricultural vegetation. The urban areas were categorized into closely and scattered urban zones. NDVI, NDBI, and BUI images were generated by employing the following formulas [19]:

NDVI = NIR − Red/NIR + Red

(1)

NDBI = SWIR − NIR/SWIR + NIR

(2)

BUI = NDBI − NDVI

(3)

Supplementary data, including the digital elevation model (DEM), road networks, and artificial surfaces, produced forecasting elements such as elevation, incline, proximity to water bodies, and distance from significant roadways. We classified these factors into static and dynamic variables. Dynamic variables are adjusted and re-evaluated during prediction, whereas static variables remain constant over time [25,26]. The sub-modeling in this study focused on two main variables: critical LULC transitions and predictive variables.

3.3. Data Processing

The preprocessing of remote sensing images encompassed steps to enhance the quality of the downloaded satellite images. This involved various procedures, such as geometric adjustment, radiometric calibration, atmospheric correction, image extraction, image mosaicking, and image blending. These processes are crucial in refining the raw data from remote sensing satellites.

Geometric correction involved rectifying geometric distortions in the collected satellite images to enhance spatial representation accuracy. Simultaneously, radiometric calibration standardized pixel radiometric measurements, enabling meaningful quantitative analysis. Various variables (Figure 5) are considered in predicting future land-use changes, encompassing two types: static variables like height (m), slope (degree), and proximity to water bodies (m), and changeable variables, including proximity from major roadways and existing artificial surfaces in (m). It is essential to highlight that dynamic variables can be adjusted and recalculated throughout the prediction process, whereas static variables remain constant over time [25,26]. This study’s sub-modeling concentrated on two primary variables: crucial LULC transitions and the propelling forces behind them. Figure 4 presents the standardized variables employed in the current investigation.

Transition maps are created by incorporating changes in LULC and static and dynamic variables into a multilayer perceptron (MLP) neural network, a widely utilized artificial neural network type. The MLP neural network utilizes a supervised “backpropagation” training algorithm to adapt model parameters and reduce errors, ultimately enhancing overall accuracy.

3.4. Land-Use/Land-Cover Changes Prediction

In this study, the MLP neural network played a crucial role in forecasting future LULC changes by establishing transition weights for inclusion in the likelihood matrices of the Markov chain. The resulting matrix quantifies the extent of change leading to the predicted end dates for each weighted transition. We performed a detailed quantification on the transition probability matrix concerning LULC variations from 2000 to 2020. Using the CA Markov chain method, a projection for 2040 was conducted, facilitating the identification of potential LULC changes (Figure 6).

The CA-Markov model is a prevalent tool within various LULC modeling methodologies, encompassing spatial and temporal alterations [38,39,40]. This model amalgamates cellular automata and Markov chain techniques, enabling the projection of LULC changes and their respective characteristics over time [41]. Its utility extends to aiding planning processes by supporting the analysis of temporal shifts and spatial distribution patterns within LULC [42]. Its significance extends to land-use policy formulation and planning strategies for sustainable land utilization [42]. Its significance further extends to the realm of land-use policy formulation and planning strategies directed towards sustainable land utilization [42]. Investigating historical LULC changes becomes essential, enabling a comprehensive comprehension of the interplay between human activities and the environment from a long-term standpoint [43].

4. Results

4.1. LULC Statistics: An Evolution from 2000 to 2020

Table 2. Dynamics of LULC change (2000–2020).

Year	2000		2020
LULC	Area (km2)	Area (%)	Area (km2)	Area (%)
Water bodies	744.204	19.702	738.622	19.554
Artificial surfaces	642.749	17.016	952.178	25.208
Grassland	296.083	7.839	70.812	1.875
Cultivated land	46.771	1.238	75.283	1.993
Wetland	369.722	9.788	213.251	5.646
Shrubland	166.991	4.421	99.728	2.640
Forest	1485.171	39.319	1627.301	43.081
Bare land	25.586	0.677	0.102	0.003

and Figure 7 illustrate the dynamics of LULC from 2000 to 2020, presenting changes in proportions across various LULC classes over the two-decade period. The expansion of cultivated areas is evident, increasing significantly from 46.771 km² (1.238%) in 2000 to 75.283 km² (1.993%) in 2020, indicating shifts in agricultural practices. Forest cover has experienced a modest increase, from 1,485.171 km² (39.319%) to 1,627.301 km² (43.081%), suggesting a relatively stable condition. In contrast, grasslands have seen a substantial decline, decreasing from 296.083 km² (7.839%) to 70.812 km² (1.875%), likely attributed to urbanization growth. Shrubland areas have also notably diminished, going from 166.991 km² (4.421%) to 99.728 km² (2.640%), indicating alterations in land management or natural processes. Wetland areas have contracted, reducing from 369.722 km² (9.788%) in 2000 to 213.251 km² (5.646%) in 2020, potentially influenced by changing natural conditions or human activity. Bodies of water have remained relatively stable, experiencing a slight decrease from 744.204 km² (19.702%) to 738.622 km² (19.554%). The most prominent change is observed in artificial surfaces, surging from 642.749 km² (17.016%) in 2000 to 952.178 km² (25.208%) in 2020, attributed to urbanization and infrastructural development. Lastly, bare land has significantly reduced, from 25.586 km² (0.677%) to 0.102 km² (0.003%).

The spatial-temporal analysis unveils noteworthy alterations in LULC categories over the two previous decades. These changes signify shifts in landscape dynamics, with certain categories expanding while others diminish. Particularly, artificial surfaces have witnessed a substantial growth, accounting for 7.192% of the total area, indicative of urbanization and development (Figure 8), while grassland and bare ground have experienced a significant reduction in coverage. The transformation within these LULC classes mirrors the evolving nature of the study area’s environment between 2000 and 2020.

4.2. Change Analysis of LULC with Land-Change Modeler

Table 3. Statistics for gain, loss, and net change.

LULC Category	Gain (km2)	Loss (km2)	Net Change (km2)	% Change
Cultivated land	68.9679	40.4559	28.512	0.43
Waterbodies	27.1485	32.7303	−5.582	−0.08
Wetland	28.8135	185.2848	−156.471	−2.39
Shrubland	75.5451	142.8075	−67.263	−1.03
Grassland	30.7125	255.9834	−225.271	−3.48
Artificial surfaces	356.2821	46.8531	309.429	8.19
Forest	391.0347	248.9049	142.13	2.32
Bare land	0.1017	25.5861	−25.484	−0.38

and Figure 9 provide a comprehensive breakdown of changes in LULC categories, encompassing gains, losses, net changes, and percentage changes. Notably, the category of artificial surfaces exhibited substantial growth, primarily attributed to urbanization and infrastructure development, with a significant net increase of 309.429 km² or 8.19%. Conversely, the waterbody category experienced both losses and gains, resulting in a modest decline of −5.582 km² or −0.08%, potentially influenced by various factors affecting water features. Wetland areas underwent the most significant net change, encountering a loss of 156.471 km² or −2.39%, likely due to conversions to alternative land uses. Both shrubland and grassland recorded substantial losses, with net decreases of −67.263 km² (−1.03%) and −225.271 km² (−3.48%), respectively, potentially associated with alterations in land management or human activities. Cultivated land witnessed an increase of 28.512 km² or 0.43%, possibly resulting from conversions from other land categories. Forest exhibited gains and a net change of 142.13 km² or 2.32%, likely influenced by reforestation initiatives. Bare land experienced a notable decrease, with a net loss of −25.484 km² or −0.38%, indicating changes in land use. These dynamics are spatially illustrated in Figure 9 and Figure 10, offering valuable insights into the evolving landscape over the study period from 2000 to 2010.

4.3. The Markov Chain Transition Matrix and Prediction

Table 4. Matrix of transition probability.

LULC Class	Cultivated Land	Forest	Grassland	Shrubland	Wetland	Waterbodies	Artificial Surfaces/Urban Areas	Bare Land
Bare land	0.0966	0.4082	0.0059	0.1282	0.0782	0.0051	0.2777	0.0000
Artificial surfaces	0.0003	0.0513	0.006	0.0119	0.0008	0.0026	0.9271	0.0000
Grassland	0.0299	0.4253	0.1354	0.1005	0.002	0.0053	0.3016	0.0000
Shrubland	0.0083	0.3596	0.0427	0.1448	0.0301	0.0245	0.3898	0.0000
Wetland	0.0008	0.3677	0.0059	0.0135	0.4989	0.0347	0.0786	0.0000
Water bodies	0.001	0.0218	0.0048	0.0017	0.0069	0.956	0.0077	0.0001
Forest	0.0371	0.8324	0.0048	0.0173	0.0103	0.0044	0.0938	0.0000
Cultivated land	0.135	0.2025	0.1433	0.063	0.0054	0.0075	0.4433	0.0000

displays a sequential probability matrix, offering a quantitative understanding of how various LULC classes anticipated undergoing changes over a specific period using Markov chain analysis, as detailed in the referenced book. This matrix proves valuable for predicting future landscape dynamics, especially within the context of urbanization and land-use planning. The provided probabilities in the table offer specific information about the likelihood of particular LULC transitions. For instance, the probability of cultivated land transitioning into artificial surfaces/urban zones is 3.54%, which shows a small yet noticeable possibility of agricultural land transforming into urban projects. Conversely, the probability of forest land changing into artificial surfaces is lower at 1.05%, suggesting that converting forested areas into urban zones is less probable.

The findings also indicate various potential transitions. Grassland, shrubland, wetland, water bodies, and barren land all have distinct probabilities of transforming into artificial surfaces in the future, providing insights into likely land-use changes. For instance, the relatively elevated probability of shrubland (17.54%) transitioning to artificial surfaces signifies a notable risk of urbanization expanding into areas previously covered by shrubs. Likewise, the relatively higher probabilities of Grassland (4.62%) and cultivated land (3.54%) becoming artificial surfaces suggest significant potential conversions of these areas into urban landscapes.

The projection of LULC for 2040 (Figure 11) illustrates substantial shifts in the distribution of land-cover categories compared to data from 2000 and 2020. Notably, the trends in artificial surfaces, or urban areas, and forests hold particular significance. In 2000, urban areas encompassed 642,749 km², constituting 17.016% of the total land area. By 2020, this figure had markedly increased to 952,178 km², representing 25.208% of the total land area. The most noteworthy transformation is anticipated in 2040, with urban areas projected to cover 1,094,151 km², accounting for nearly 28.967% of the total land area. This pattern underscores a consistent and substantial rise in urbanization (Figure 10). Meanwhile, forests, covering 1,485,171 km² (39.319%) in 2000, are expected to occupy 1,627,301 km² (43.081%) by 2020. In 2040, forests are projected to cover 1,551,632 km² (41.078%) of the land area. Despite this slight decrease, forests remain a crucial component of the landscape. Overall, the statistics highlight a significant trend toward increasing urbanization (artificial surfaces), emphasizing the imperative need for sustainable land-use planning and conservation initiatives to balance urban growth with the preservation of natural ecosystems.

4.4. Urban Sprawl Indicators

The map in Figure 12 provides data for the years 2000 and 2020 concerning three distinct urban sprawl indicators: “NTL” (nighttime light), “Built-up”, and “Artificial Surfaces”, all measured in square kilometers (km²). Notably, all three indicators have exhibited significant increases over the past two decades. For instance, “NTL” witnessed a substantial rise from 116.4055 km² in 2000 to 426.0804 km² in 2020. Similarly, “Built-up” areas expanded from 913.5737 km² to 1138.694 km², while “Artificial Surfaces” surged from 642.749 km² to 952.178 km².

The data hold substantial implications for analyses related to LUL and urban growth. The notable expansion observed in these variables over the past two decades signifies swift urbanization and land development in the studied region. This urbanization carries various consequences, including heightened requirements for infrastructure, housing, and services, alongside alterations in local ecosystems, land fragmentation, and environmental considerations.

According to [44], the rise in population and the expanding trend of urbanization, especially in developing nations, are anticipated to impose additional pressure on the productivity of agricultural land and intensify the challenges faced by farmers in their agricultural practices. Moreover, the combination of urban expansion and a dense population, often associated with poverty, is commonly linked to increased occurrences of environmental degradation, impacting diverse land uses, such as forests, wetlands, and natural resources [45].

These data are essential for urban planners, politicians, and researchers as they underscore the significance of employing sustainable urban development approaches, implementing effective land-use policies, and engaging in conservation initiatives.

5. Discussions

5.1. NTL (Nighttime Light) Data

NTL data and land-use data are distinct but complementary datasets used in urban and environmental analysis. Nighttime light data, often derived from satellite imagery, capture the artificial illumination present during the night, offering insights into the spatial distribution and intensity of anthropological activities [17,29]. On the other hand, land-use data categorize the different types of activities or covers on the Earth’s surface, such as residential, commercial, industrial, or green spaces. Integrating nighttime light data with land-use data enhances our understanding of urban dynamics, allowing researchers and policymakers to examine the patterns of development, identify areas of high human activity, and make informed decisions about urban planning, resource allocation, and sustainable growth strategies [17,29]. The outcomes of the current study demonstrate a strong correlation between changes in nighttime brightness and particular types of land-use transitions at the individual pixel level (Figure 13).

Additionally, a notable location-based correlation discerned between nighttime light luminosity transitions and municipal areas characterized by high levels of human activity.

The nighttime light data, illustrated in Figure 3, quantify the artificial light observed during the night through satellite imagery. This information is frequently utilized as an indicator for assessing urbanization and economic activity, given the tendency for areas with heightened artificial light levels to reflect increased urban development. Notably, in the year 2000, the NTL value recorded at 116.4055 km², and by 2020, it had undergone a significant surge to 426.0804 km². This substantial expansion over a two-decade span signifies robust urbanization and notable economic growth. The analysis emphasizes the valuable insights gathered by integrating nighttime light data with land-use data, offering a deeper understanding of urbanization dynamics. Such insights play a crucial role in informed decision-making for urban planning, infrastructure development, and sustainable management.

5.2. Built-Up Areas

In 2000, the “Built-up” area covered 913.5737 square kilometers, while the “LULC” area (artificial surfaces) encompassed 642.749 square kilometers. By 2020, the “Built-up” area expanded to 1138.694 square kilometers, and the “LULC” area increased to 952.178 square kilometers. The high correlation coefficients for both 2000 and 2020 signify a robust positive correlation between the “Built-up” and “LULC” areas.

This implies that the augmentation of the “Built-up” area corresponds to a significant increase in the artificial surfaces over the two decades. This strong correlation is anticipated, given that “Built-up” typically represents areas with human development and urbanization [46,47,48,49]. This frequently aligns with “LULC” areas like artificial surfaces (urban areas). As urbanization progresses, the scope of artificial surfaces expands, as evidenced by the “LULC” data. The correlation analysis reconfirms the connection between urban development, quantified by the “Built-up” area, and the growth of artificial surfaces over time.

5.3. Markov Prediction 2040

The Markov model has been commonly utilized in LUC change forecasts in recent years [50]. In the context of time-series data or dynamic systems, Markov prediction models assess the probabilities of transitioning from one state to another over successive intervals [29]. This matrix is a crucial tool for predicting changes in LULC. The model incorporates a wide range of influencing driver factors, which are classified as independent and dependent variables [8,11,36]. These factors stem from both natural characteristics and human activities. Examples of dependent variables encompass the DEM, surface terrain, road network and distance, proximity to municipal areas, an interval from rivers, and populace density. Conversely, explanatory variables involve transitions from one land-use or land-cover type to another, such as the transformation of barren land (BL), water bodies (WB), artificial surfaces (AL), or all land-use types to cultivated land (CL). The Land Change Modeler (LCM) employed in this research heavily relies on these explanatory variables as its primary driving forces.

Drawing upon the outcomes of this investigation, notable trends in LULC alterations manifest across three distinct time spans (2000, 2020, and 2040) within the realm of urban sprawl prognosis. Specifically, the proportion of artificial surfaces has markedly risen from 17.016% in 2000 to 28.967% in 2040 (Figure 14), indicative of the expansion of urban areas—a clear signal of urban sprawl. Although there is a marginal decrease in forested areas, possibly attributed to some conversion for urban expansion, cultivated land exhibits fluctuations, possibly influenced by shifts in agricultural practices or land-use regulations. Waterbodies generally remain stable, yet grassland, shrubland, and wetland areas exhibit negligible fluctuations, suggesting a certain resilience to urban growth in this region. The percentages of bare land continue to remain at low levels.

6. Conclusions

This study mainly focuses on urban growth monitoring and prediction within the unique context of Lagos City, Nigeria, using a combined approach of Global Land 30 data and remote sensing urban monitoring indices to examine Lagos as a rapidly growing megacity. Substantial alterations in LULC data have observed between 2000 and 2020, indicating a dynamic transformation in landscape over the two-decade period. Prominent trends encompass substantial expansions in cultivated land, a modest uptick in forest cover, noteworthy declines in grasslands and shrublands, reductions in wetland areas, and a relatively stable status for water bodies. The most conspicuous transformation, however, is the remarkable surge in artificial surfaces, attributable to urbanization and infrastructure development. Additionally, a noteworthy decrease in bare land is evident, marking another significant change in the landscape during this period.

The outcomes indicate the notable increase in cultivated areas and artificial surfaces while wetland, shrubland and grassland decreased significantly. The proliferation of artificial surfaces underscores the rapid urbanization and infrastructure development within the study area. The decline in grasslands and shrublands could indicate shifts in land management or human activities affecting natural ecosystems. Additionally, the reduction in wetland areas emphasizes the urgency of conservation initiatives and monitoring changing environmental conditions.

Moreover, the insights gleaned from the Markov Chain Transition Matrix and prediction analysis offer valuable foresight into future landscape dynamics, particularly in the realms of urbanization and land use planning. The matrix’s probabilities give detailed information regarding the likelihood of specific LULC transitions, aiding in the forecasting of changes of land use. These forecasts can serve as guiding principles for urban development and environmental preservation initiatives.

The escalation in urban indicators like “NTL”, “Built-up”, and “Artificial Surfaces” over the past two decades implies rapid urbanization and land development analysis. To mitigate the adverse effects of urban expansion, the study underscores the critical importance of implementing sustainable urban development strategies, effective land-use policies, and conservation activities, particularly in developing nations. Ultimately, the study’s findings furnish crucial insights into evolving landscape dynamics, urbanization patterns, and potential environmental ramifications. These insights are indispensable for informed decision-making, long-term urban planning, and conservation endeavors within the research area. Finally, this study recommends future researchers refine predictive models, incorporating real-time data and considering socio-economic factors in the monitoring and prediction framework of urban growth.