Urban Expansion Projections in Maricá/Rio De Janeiro—RJ: Modeling with Cellular Automata and Sentinel Images for 2030 and 2040 ^†

Abstract

Maricá, located on the eastern coast of Rio de Janeiro, experiences rapid urban growth driven by infrastructure and economic development. This study presents the first high-resolution projection of Maricá’s urban expansion (2030–2040), integrating oil industry impacts and protected area constraints. Using Sentinel-2 MSI data (10–20 m resolution) classified via Random Forest on Google Earth Engine (90% accuracy) and a Dinamica EGO Cellular Automata model (5 × 5 Moore neighborhood, calibrated on 2015–2020 transitions), results indicate 18.4% urban growth by 2030 (129 km²), expanding to 151 km² (+38.5% total) by 2040, with 72% replacing pastures. This supports sustainable urban management strategies.

1. Introduction

Urbanization in the Global South has intensified in recent decades, especially in medium-sized cities located on the peripheries of major metropolitan areas [1]. This expansion often occurs in contexts marked by insufficient infrastructure and limited regulatory capacity, resulting in fragmented spatial configurations and increased socio-environmental vulnerability [2]. In this scenario, spatially explicit modeling techniques, such as Cellular Automata (CA), have become essential tools for understanding land cover change and forecasting urban expansion patterns [3]. The integration of CA-based models with high-resolution satellite data, particularly Sentinel-2 imagery, has significantly enhanced the ability to produce detailed spatial projections and to inform anticipatory planning strategies in fast-changing territories [4,5].

Cellular models in urban modeling began to be implicitly utilized in early computational models of the 1960s, such as the works of Chapin and Weiss for North Carolina and Waldo Tobler’s model for Detroit [6,7]. Cellular Automata (CA) have become the main approach in spatial models, being defined by Stephen Wolfram as mathematical representations of physical systems in which time and space are discrete [8]. These models utilize uniform grids where the variables of each cell are updated simultaneously, grounded in local rules and the values of neighboring cells [9,10]. CA are widely applied in the simulation of land cover changes, enabling the creation of high-resolution dynamic models and their integration with Geographic Information Systems (GISs) for the analysis of urban dynamics [11,12,13]. Brazil has experienced a marked increase in urbanization beyond its major metropolitan centers, with medium-sized cities playing a key role in the reconfiguration of urban networks and land use patterns [14]. Cities like Maricá, positioned at the urban–rural interface of metropolitan regions, are particularly susceptible to unregulated expansion, environmental degradation, and infrastructure deficits [1G]. Urban growth in these cities is often characterized by dispersed occupation and encroachment on environmentally sensitive areas, such as wetlands and forest fragments [14,15]. This study aligns with this methodological framework by simulating urban expansion scenarios for Maricá, a rapidly growing municipality in southeastern Brazil. Maricá serves as a representative case of medium-sized coastal cities in the Global South, experiencing rapid urbanization due to oil-related economic activities and proximity to metropolitan cores. Between 2010 and 2022, Maricá’s population grew by 54.87%, the highest in the state of Rio de Janeiro and the ninth in Brazil [16]. Its dynamics reflect broader global trends in peripheral urban growth and environmental conflicts.

1.1. Research Objectives

This study projects urban expansion in Maricá for the years 2030 and 2040 using Cellular Automata and Sentinel-2 imagery, with the following objectives: to identify areas most susceptible to urban growth, to quantify land cover changes between the baseline and projected scenarios, and to assess the potential of CA-derived results to inform urban planning and environmental policies.

1.2. Study Area

Maricá, located on the eastern coast of the state of Rio de Janeiro, Brazil, is part of the Metropolitan Region of Rio de Janeiro (MRRJ) and has undergone notable demographic expansion in recent decades. The municipality covers an area of 362.5 km² and includes approximately 34 km of coastline, incorporating environmentally sensitive zones such as restinga formations, coastal ridges, lagoons, and forest fragments. According to data from the Brazilian Institute of Geography and Statistics (IBGE) [16,17], Maricá’s population increased from approximately 70,000 inhabitants in 1991 to 91,000 in 2000, 127,461 in 2010, and 171,000 in 2022 [18]. This growth is largely driven by oil royalty revenues, public housing programs, and infrastructure investments, including the expansion of road networks and real estate developments [19,20] (Figure 1).

Between 2003 and 2013, the municipality registered the fourth highest population growth rate among Brazilian cities [9,10], reflecting broader regional dynamics of urban expansion and socio-environmental transformation.

2. Materials and Methods

This study integrates remote sensing and dynamic modeling through Cellular Automata (CA), using the Dinamica EGO 7.4 platform. The methodological workflow is structured into three main phases—data input, modeling, and simulation—as illustrated in Figure 2.

The Dinamica EGO platform and CA model were selected not only for their computational efficiency but also for their accessibility, reproducibility, and adaptability to mid-sized urban areas with limited data availability—characteristics common in cities of the Global South.

2.1. Data Input

The input database consists of land cover maps derived from Sentinel-2 MSI images (Level-1C, 10–20 m resolution) for the years 2015, 2020, and 2023. The images were processed on the Google Earth Engine (GEE) platform. Land cover classification was performed using the Random Forest algorithm with 100 decision trees and a stratified random sampling strategy for training data. The classification achieved 90% overall accuracy and a Kappa index of 0.87. Driving forces such as slope, proximity to roads, hydrography, and protected areas were extracted and standardized. Transitions between classes were analyzed via cross-tabulation of classified maps. Corrections were applied to remove illogical transitions, and classes were reclassified to improve model interpretability.

2.2. Modeling

The modeling phase involved defining transition potentials and spatial dependencies. A transition matrix was generated using land cover maps from 2015 to 2020. Continuous variables were categorized to support statistical modeling. Spatial dependency was modeled using the Weights of Evidence (WoE) method, which calculates the spatial association between land use change and driving forces grounded in Bayesian inference. The WoE coefficients were calculated, and regional adjustments were applied to improve local sensitivity through regionalized simulation.

2.3. Simulation

The CA simulation followed these steps: (i) Sentinel-2 preprocessing, (ii) land cover classification, (iii) transition matrix calculation, (iv) WoE coefficient estimation, (v) transition potential computation, (vi) CA calibration, (vii) fuzzy validation, and (viii) scenario generation for 2030–2040. Model calibration was conducted using a 5 × 5 Moore neighborhood and a comparison between the simulated and observed 2023 land cover. Validation employed fuzzy similarity metrics, allowing for assessment of spatial patterns with tolerance to minor positional differences. Once calibrated, the model was used to simulate urban expansion scenarios for 2030 and 2040. These future scenarios provide spatialized insights into potential land use dynamics under the influence of existing physical and regulatory constraints. After classification and validation through fuzzy analysis, future scenarios were generated for the years 2030 and 2040.

3. Results

Figure 3 presents the results obtained from the modeling for the year 2023, as well as the projected land cover classes for 2030 and 2040, providing a detailed analysis of the spatial evolution of Maricá.

The points 1–6 highlight urban expansion over pasture areas and forest fragments and along main roads. Growth concentrated in strategic corridors reinforces the role of road infrastructure, requiring monitoring to mitigate impacts on environmentally sensitive areas (Figure 4).

Between 2023 and 2030, the urbanized area of Maricá is expected to grow from 109 km² to 129 km², representing an 18.4% increase. Subsequently, by 2040, a further expansion to 151 km² is estimated, corresponding to an additional 17.5% growth. This expansion is not evenly distributed across the territory, with distinct spatial patterns emerging, such as urban sprawl over areas previously occupied by pastures and forest fragments, as well as increased development near major transportation routes.

Moreover, detailed spatial analysis reveals changes in protected areas and a densification of buildings, highlighting challenges for territorial and environmental management in the municipality. The presence of growth vectors in environmentally sensitive regions underscores the need for public policies focused on sustainable urban planning, balancing environmental preservation with land use dynamics.

4. Discussion

The modeling results reveal a consistent pattern of urban expansion in Maricá, concentrated along transportation corridors and over areas previously occupied by pastures and forest fragments. This trend is commonly observed in coastal and peri-urban regions, where infrastructure and accessibility drive land use change [2,3,4]. Maricá’s dynamics resemble urban trajectories in other Global South cities such as Gurgaon (India), Lagos (Nigeria), and Durban (South Africa), where urban growth often advances over ecologically sensitive areas with limited institutional regulation [21]. In Maricá, projected urban growth between 2023 and 2040 is particularly concentrated in its northern and central zones, overlapping with environmental protection areas, restinga vegetation, and lagoon systems. This pressure on coastal ecosystems raises concerns about biodiversity loss, water quality degradation, and ecosystem fragmentation—challenges also reported in rapidly growing cities in Latin America and Africa [3]. The forecasted 40% increase in urbanized areas by 2040 emphasizes the urgent need for integrated land use planning. Without effective regulatory measures, urban sprawl may lead to irreversible impacts on environmentally strategic zones. In this context, Cellular Automata (CA) modeling proves to be a valuable tool for anticipating spatial conflicts and supporting sustainable development strategies, as shown in studies conducted in Mexico, Ghana, Ethiopia, and China [21].

This study contributes by applying CA modeling to a fast-growing mid-sized city embedded within a transitioning metropolitan region. Maricá’s specific context—marked by strong real estate pressure, rapid demographic growth, and economic dependence on oil royalties—makes it a unique case for predictive spatial modeling. Comparisons with other Global South cities reinforce the need to integrate social, environmental, and economic variables into urban planning practices. One limitation of this study is the reduced number of land cover classes and the absence of socioeconomic variables in the transition modeling. Future research should incorporate demographic data, housing typologies, and infrastructure layers, as well as validate simulation outcomes with local planning documents and stakeholder engagement. While not introducing a novel algorithm, this study aligns with best practices in land change modeling, applying established methods to a relevant and underexplored planning scenario. Similar approaches have been validated in recent studies focused on emerging urban regions [12,21].

5. Conclusions

It is concluded that Cellular Automata with Sentinel imagery enable coherent simulation of urban growth and support sustainable urban planning. The results highlight environmental impacts and infrastructure deficits, reinforcing the need for effective public policies. Future work will include additional modeling variables.