Next Article in Journal
Beyond Top-Down Narratives: Thick Mapping and Participatory Spatial Development in Coastal Colombia
Previous Article in Journal
Urban Expansion Simulation for the Low-Carbon Goal: A Focus on Urban Form Optimization
Previous Article in Special Issue
Marginalized Living and Disabling Spaces: A Bio-Cognitive Perspective
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Land
Volume 15
Issue 3
10.3390/land15030456
land-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Linking Urban Land Use Change and Tropospheric Ozone Dynamics in a Mid-Sized City

by
Ceren Yağcı
Faculty of Engineering and Natural Sciences, Osmaniye Korkut Ata University, Osmaniye 80000, Türkiye
Land 2026, 15(3), 456; https://doi.org/10.3390/land15030456
Submission received: 30 January 2026 / Revised: 27 February 2026 / Accepted: 10 March 2026 / Published: 12 March 2026
(This article belongs to the Special Issue Urban Land Use Change and Its Spatial Planning)
Browse Figures
Versions Notes

Abstract

This study develops an integrated geospatial framework to examine the spatial-temporal relationship between urban land-use change and tropospheric ozone dynamics within a mid-sized functional urban system, using Bolu, Türkiye, as a case study. Mid-sized urban systems remain underrepresented in air-quality and land-use research despite increasing environmental pressures under ongoing urbanization. The spatial framework was defined to encompass the central urban area and its surrounding peri-urban and transportation-influenced transition zones. Future land-use patterns were estimated to 2030 using the MOLUSCE model, while tropospheric ozone indicators were derived from Sentinel-5P observations for the 2020–2024 period and descriptively extended to 2030 using the Theil–Sen slope estimator. A fishnet-based spatial regionalization approach enabled consistent comparison between ozone trends and urban expansion intensity, quantified using the Urban Expansion Intensity Index (UEII). The integrated framework provides a spatially coherent basis for understanding land–atmosphere interactions in mid-sized urban systems.

1. Introduction

Rapid urbanization has significantly altered the spatial configuration of cities by augmenting population density and exacerbating the interplay between land-use change and environmental quality. Horizontal expansion, vertical densification, and growth of transportation networks change the shape of cities, change the way air pollutants spread in the atmosphere, and affect how they build up and move about [1,2,3]. Specifically, emissions of nitrogen dioxide (NO2) and carbon monoxide (CO) from transportation are key to photochemical reactions that create tropospheric ozone (O3) when the weather is right [4,5]. Negative effects on human health and ecosystem health are associated with high levels of ozone. The World Health Organization has named them a major environmental concern [6]. While much research has investigated air-quality dynamics in big metropolitan areas, there exists a relative paucity of studies focusing on mid-sized urban systems and their peri-urban transition zones. In these contexts, fragmented settlement patterns, infrastructure development, and specific topographic limitations can produce intricate and geographically diverse atmospheric reactions [7,8]. Furthermore, evaluations of urban environmental change sometimes depend on administrative boundaries, which may not sufficiently reflect functional urban interactions or emission pathways. Recent progress in remote sensing and geographic information systems has opened up new ways to combine land-use changes with atmospheric indicators. Google Earth Engine and other cloud-based platforms make it easier to work with satellite data from different times. Sentinel-5P/TROPOMI observations enable monitoring of tropospheric pollution at regional scales [9,10]. Simultaneously, land-use/land-cover (LULC) simulation methodologies utilizing Cellular Automata–Markov models and artificial neural networks have been extensively employed to produce spatially detailed urban expansion scenarios [11,12]. The MOLUSCE framework is one of these tools. It uses transition probabilities, neighborhood effects, and various driving forces to model how cities grow. When looking at atmospheric trends, non-parametric estimators like the Theil–Sen slope are often used to look at changes in the direction of environmental time series because they are not affected by outliers or non-normal distributions [13,14]. Nevertheless, research that concurrently analyzes satellite-derived atmospheric indicators and spatially detailed land-use simulations within a unified analytical framework is still scarce, especially for mid-sized cities and urban–rural transitional areas [15,16]. In this regard, Bolu serves as a pertinent instance for analyzing land-atmosphere interactions in a mid-sized metropolitan system. The basin-shaped terrain of the city makes it harder for the boundary layer to ventilate, which could lead to more pollutants building up in certain weather circumstances [17]. Simultaneously, transportation routes like the TEM motorway and the Bolu Tunnel serve as concentrated emission paths, whereas peri-urban growth induces spatially heterogeneous land-use shifts [18]. This study combines the Urban Expansion Intensity Index (UEII), MOLUSCE-based land-use forecasts, Sentinel-5P-derived tropospheric ozone indicators, and fishnet-based spatial regionalization into one analytical framework. This study employs an integrated methodology within a functionally delineated urban boundary to investigate the correlation between geographical patterns of urban expansion and observable atmospheric trends in a mid-sized urban-rural transitional context.

2. Materials and Methods

2.1. Study Area

Bolu is a mid-sized city located in the western Black Sea region of Türkiye, positioned between İstanbul and Ankara along the D-100 Highway and the TEM Motorway (E-80) (Figure 1). This strategic location makes it an important transportation corridor for both passenger and freight traffic [19]. Administratively, Bolu Province covers approximately 8323 km2 and hosts a population of around 300,000 [20]. However, the present study does not analyze the entire provincial territory. Instead, the spatial framework was defined to encompass the central urban area and its functionally connected peri-urban and transportation-influenced transition zones. This delimitation ensures that the analysis reflects the functional urban structure of Bolu rather than the entirety of its administrative boundary.
The city is characterized by a bowl-shaped topography, which restricts air circulation and enhances pollutant accumulation under stagnant atmospheric conditions. The average elevation of the plateau is 1000 m, while the city center lies at 725 m. Bolu has a humid Black Sea climate with precipitation distributed throughout the year. The mean annual temperature is 12 °C, and the average annual precipitation is 388 mm [21]. The prevailing wind direction is west-southwest (WSW), and the mixing layer height decreases to around 150 m in winter, intensifying pollution episodes. Urban development in Bolu follows both contiguous (“oil stain”) and leapfrog patterns. Industrial zones are concentrated in the east, and tourism and recreation areas are in the south and west, while residential districts are primarily located in the north and west [19]. Unplanned and high-density construction in the city center reduces natural ventilation corridors, leading to higher levels of air pollutant accumulation. The main sources of air pollution include vehicular emissions, domestic heating, and small-scale industrial facilities such as cement, metal, woodworking, poultry, and glass manufacturing. Although natural gas was introduced in 2009, coal remains widely used in suburban and rural areas [21]. In addition, long-range transport of pollutants from industrialized regions of Türkiye and Eastern Europe further deteriorates local air quality, leading to elevated concentrations of ozone (O3), nitrogen dioxide (NO2), and volatile organic compounds (VOCs). These VOCs, originating from both anthropogenic activities and natural vegetation, interact with nitrogen oxides under favorable meteorological conditions, thereby enhancing tropospheric ozone formation. Overall, Bolu represents an urban system where dispersed settlement patterns, traffic-related emissions, and microclimatic conditions jointly exacerbate air pollution risks. This context provides a suitable setting for examining the interactions between urban growth, atmospheric conditions, and air quality dynamics. This study employs an integrated methodological framework that combines land-use change modeling with satellite-based analysis of air quality trends derived from Sentinel-5P TROPOMI satellite data (European Space Agency, Paris, France), as summarized in Figure 2. To ensure spatial comparability with urban growth outputs, both land use/land cover (LULC) projections and O3 trends were aggregated within a common fishnet grid structure generated in ArcGIS 10.8 (Esri, Redlands, CA, USA). Finally, outputs from both modules were jointly interpreted to assess the spatiotemporal relationship between projected urban expansion and ozone dynamics, providing an integrated basis for environmental management and urban planning decisions.

2.2. Data Obtained and Pre-Processing

In this study, RS- and GIS-based datasets were integrated to analyze land-use changes and air-quality dynamics within the predefined analytical boundary (central urban area plus peri-urban and transportation-influenced transition zones). The datasets, together with their temporal coverage, spatial resolution, and analytical purposes, are summarized in Table 1.
Sentinel-5P is equipped with the Tropospheric Monitoring Instrument (TROPOMI), which provides high-accuracy measurements of key atmospheric pollutants with daily global coverage, making it a robust data source for spatiotemporal air quality analysis [10]. In parallel, the original nine Sentinel-2 land cover classes were reclassified into four dominant categories: crops, trees, rangelands, and built-up (Table 2). Considering the relatively homogeneous topography and limited land-cover diversity of Bolu, the original Sentinel-2 land-cover classes were reclassified into four dominant categories [22,23].
  • Driving Factors
Digital Elevation Model (DEM): The Digital Elevation Model (DEM) was derived from the ASTER Global DEM (ASTGDEM), which has a spatial resolution of 1 arc-second (approximately 30 m) [24]. The dataset was downloaded from the NASA Earth Data Search Gateway [25]. Topographic variables such as elevation are critical in influencing urban development patterns by constraining land-use transitions and directing spatial growth trajectories.
Population Density: Population density data were obtained from the World Pop database [26]. The dataset was resampled and spatially aligned to match the resolution and extent of the Sentinel-2 raster products.
This variable represents the spatial pressure exerted by human activities and serves as an important driver of urban expansion.
Distance to Built-up Areas: The distance to built-up areas was calculated using the Euclidean Distance tool in ArcGIS 10.8 Built-up areas were identified based on the “urban” class derived from the reclassified LULC map generated from Sentinel-2 imagery using the Sentinel-2 Land Cover Explorer. The resulting distance raster quantifies the proximity of each pixel to existing urban areas, thereby enabling the identification of zones with high urban expansion potential.
Road Network: Road network data were obtained from the OpenStreetMap (OSM) platform [27]. The vector road data were converted into a raster format using the Line Density and Euclidean Distance tools in ArcGIS. This variable captures the influence of transportation accessibility on urban growth dynamics, as areas closer to road infrastructure typically exhibit higher development potential. All driving factors (DEM, population density, road network, and distance to built-up areas) were normalized to a range of 0–1. This normalization ensured that the influence of each factor operated on a comparable scale, eliminating weight discrepancies during model training. The spatial distribution of the normalized layers is presented in Figure 3. All spatial data sets were standardized to a common projection system (WGS_1984_UTM_Zone_36N) and unified spatial resolution. This standardization was performed to ensure comparability between datasets. Furthermore, variables representing urban dynamics were normalized to achieve a consistent analytical scale throughout the study.

2.3. Air Quality Modeling (2020, 2022, 2024)

Bolu’s ground-based air quality monitoring system, especially the Atatürk Boulevard Kızılay Park Station, provides limited spatial representativeness for broader regional assessments, as it was primarily designed to measure traffic-related pollutants [18,28]. To capture spatially continuous atmospheric patterns, this study used satellite-derived Sentinel-5P TROPOMI tropospheric pollutant products. Daily tropospheric column indicators of CO, NO2, and O3 were processed on the Google Earth Engine (GEE) platform. A quality assurance threshold (qa_value > 0.5) was applied to all products. Valid pixels were composited into monthly means and subsequently aggregated into annual mean maps for 2020, 2022, and 2024. Before aggregation and fishnet-based zonal analysis, all atmospheric datasets were clipped to the predefined analytical boundary encompassing the central urban area and its connected peri-urban and transportation-influenced transition zones. Although all three pollutants (CO, NO2, and O3) were analyzed descriptively, trend estimation and forward extension analyses were conducted exclusively for tropospheric ozone (O3). This selection was based on atmospheric chemistry considerations and the temporal characteristics. As a secondary pollutant formed through photochemical reactions involving NOX, CO, and VOCs, O3 typically exhibits comparatively smoother temporal variability than primary pollutants such as NO2 and CO [29]. Additionally, Sentinel-5P TROPOMI retrievals of O3 are comparatively more stable over multiple years than NO2 and CO, which have shorter atmospheric lifetimes and greater interannual variability [30]. O3 dynamics are further affected by land-use and land-cover features through their influence on precursor emissions, biogenic VOC production, and urban thermal conditions [15,30,31,32]. During the 2020–2024 period, O3 column indicators showed an overall upward tendency across the fishnet grid, while NO2 and CO displayed less interannual variation. Since the analysis is based on three discrete annual observations (2020, 2022, and 2024) rather than a continuous time series, the estimated trend should be interpreted descriptively rather than as definitive evidence of a long-term monotonic increase. In such a short series, interannual anomalies (e.g., meteorological variability or atypical emission conditions) may exert a proportionally stronger influence on slope estimation. Consequently, any forward extension derived from these observations is treated as a scenario-based descriptive extrapolation rather than a deterministic forecast and should be interpreted within the temporal and methodological limits of the study dataset.

Fishnet-Based O3 Estimation

From 2020 to 2024, satellite-derived tropospheric ozone (O3) column indicators in Bolu generally increased over the observation period, especially in the western and central subregions. To reduce spatial variation and allow for consistent subregional comparisons, the study area was divided into four grid cells (A–D), representing the western transportation route, central urban core, and northern and southern rural zones. For each grid cell, annual mean tropospheric O3 column indicators were calculated. Temporal change was quantified using the non-parametric Theil–Sen slope estimator [13,14]. Although the Theil–Sen method is commonly used in environmental trend analyses because of its robustness to outliers, it was applied here to a short time series with only three annual observations (2020, 2022, and 2024). As a result, the statistical power for formal trend detection is limited. Furthermore, due to the very small sample size (n = 3), the estimated slope confidence intervals are inherently wide, indicating substantial uncertainty in the magnitude of the trend despite its directional consistency. The Mann–Kendall test was therefore used in a descriptive way to aid interpretation rather than as definitive proof of statistically significant monotonic behavior. Theil–Sen slope was computed as the median of pairwise slopes between observation years (Equation (1)):
s1 = (O2022 − O2020)/2
s2 = (O2024 − O2022)/2
s3 = (O2024 − O2020)/4
slope = median (s1, s2, s3)
To explore potential future ozone conditions, the estimated slope was linearly extended from the observed 2024 value to 2030 as a scenario-based descriptive extrapolation (Equation (2)):
O2030 = O2024 + slope × 6
Given the uncertainty associated with slope estimation from three discrete annual observations, the 2030 values should be interpreted as exploratory scenario-based projections rather than statistically robust forecasts. This extension assumes linear continuity of the observed tendency and does not account for potential nonlinear atmospheric responses, emission policy changes, or anomalous meteorological conditions. Summary statistics of the 2030 scenario-based O3 column indicators for each fishnet grid cell are presented in Table 3.

2.4. Urban Growth Simulation (MOLUSCE)

Urban growth and land-use/land-cover (LULC) dynamics were simulated using the MOLUSCE plugin in QGIS. MOLUSCE is a GIS-based framework that integrates multiple machine-learning algorithms for spatial transition modeling and land-use change prediction, including artificial neural networks, logistic regression, and weights of evidence [33,34]. In this study, a Multi-Layer Perceptron (MLP) artificial neural network was employed due to its ability to capture nonlinear spatial relationships associated with urban expansion processes [35]. Model training was performed using a backpropagation algorithm, and three sampling strategies (all, random, and stratified) were tested. The “all” sampling mode was selected as it provided the highest classification accuracy, despite longer computation time [36]. Explanatory variables included land-use maps for 2020 (initial) and 2022 (final), together with key urbanization drivers such as elevation, population density, distance to roads, and existing built-up areas. All spatial layers were harmonized to a common projection, resolution, and spatial extent before model implementation. The modelling extent corresponds to the same predefined analytical boundary used in the atmospheric analysis to ensure full spatial consistency across datasets. To capture spatial dependency and neighborhood effects, MOLUSCE’s Cellular Automata (CA) module was used to allocate predicted land-use transitions. This mechanism reflects the spatial contiguity and self-organizing characteristics of urban growth [37,38]. The combined ANN–CA framework enables the generation of spatially coherent urban expansion patterns [39,40]. Transition matrices and class-wise area change statistics were generated to quantify land-use conversions between 2020 and 2022 (Table 4).
Transition potentials were subsequently derived using the trained ANN model, which showed stable learning behavior and achieved a validation Kappa coefficient of 0.87, indicating effective learning without overfitting (Figure 4).
The calibrated transition potentials were then used to drive the CA simulation from 2024 to 2030 using two-year time steps. Model validation was conducted by comparing the simulated 2024 LULC map with the reference dataset, yielding an overall accuracy of 97.4% and a Kappa coefficient of 0.96. These results confirm that the ANN–CA framework implemented in MOLUSCE provides a reliable and spatially consistent basis for projecting future LULC dynamics (Figure 5).
To ensure consistency with the atmospheric analysis, the MOLUSCE-derived LULC projections were subsequently integrated into a four-cell fishnet grid (A–D). Applying the same spatial regionalization to both urban growth simulations and tropospheric ozone trend analysis enabled direct, zone-specific comparison of land-use change intensity and air quality dynamics, supporting a coherent interpretation of urban atmospheric interactions.

2.5. Urban Expansion Intensity Index (UEII)

Urban expansion dynamics were further quantified using the Urban Expansion Intensity Index (UEII), originally proposed by Ren et al. [41]. UEII provides a standardized measure of urban growth by relating changes in built-up land areas to the total spatial extent of a given unit over a defined time interval. Compared with traditional landscape metrics, UEII enables a clearer interpretation of both spatial differences in urban expansion intensity and temporal development trends [42]. In this study, UEII was calculated by relating the increase in the built-up area between two time points to the total area of each fishnet cell and the corresponding time interval. The index represents the average annual rate of urban expansion and has been widely applied in both national and international studies to assess urban growth dynamics and to support land-use planning and sustainability-oriented decision-making [42,43]. The Urban Expansion Intensity Index (UEII) was calculated using Equation (3):
UEII = (ULA(i, t2) − ULA(i, t1))/(TLAi × Δt) × 100
where ULA(i, t2) and ULA(i, t1) represent the built-up land area (expressed as the number of urban pixels) within fishnet cell i at the later and earlier time points (e.g., 2024 and 2022), respectively. TLAi denotes the total area of fishnet cell i (in pixels), and Δt represents the time interval between the two observation years (in years). To facilitate interpretation and interregional comparison, UEII values were classified into qualitative growth categories ranging from stable to very fast expansion, following the threshold scheme proposed by Al-Sharif and Pradhan [44] (Table 5).
Based on the methodological framework and indices described above, the following section presents the empirical results of the study, focusing on the spatiotemporal dynamics of atmospheric pollutants and land-use change in Bolu within the defined analytical boundary.

3. Results

3.1. Spatial and Temporal Dynamics of CO and NO2 (2020–2024)

Figure 6 and Figure 7 illustrate pollutant maps for CO and NO2. Both pollutants show west-northwest clustering, aligned with the Bolu Tunnel, TEM corridor, and central urban areas.
Average changes over the 2020–2024 period indicate modest yet consistent increases in primary traffic-related pollutants. Tropospheric NO2 column indicators increased from approximately 8.27 × 10−5 to 8.54 × 10−5 mol m−2, while tropospheric CO column indicators rose slightly from about 0.0301 to 0.0313 mol m−2. These gradual upward trends indicate persistent spatial patterns aligned with major transportation corridors and densely built-up zones. Despite their moderate magnitude, the spatial coherence of these increases suggests sustained anthropogenic pressure rather than episodic pollution events.

3.2. O3 Trends and 2030 Projections

The estimated trend is derived from three temporally discrete annual composites (2020, 2022, and 2024) rather than from a continuous multi-year time series. Tropospheric ozone exhibited a directionally consistent upward tendency across the study area during the 2020–2024 period (Figure 8). Satellite-derived tropospheric O3 column indicators increased from approximately 0.141–0.142 mol m−2 in 2020 to 0.150–0.152 mol m−2 in 2024. Given the limited number of observations, the identified tendency should be interpreted as indicative rather than statistically conclusive. Spatially, the strongest relative increases were observed in sub-regions A and B, corresponding to the western transportation corridor and the central urban area, respectively. Sub-region C had average O3 levels that were rising slowly, while sub-region D had lower levels, but there was still a slight upward trend over the available time period.
Based on the linear extension of the Theil–Sen slope, ozone distributions for 2030 were generated as a descriptive continuation of the observed trends to support spatial interpretation of potential future tendencies (Figure 9).
Under this linear extension assumption, projected 2030 values ranged between 0.164 and 0.167 mol m−2, indicating a continued increase relative to the observed period. Spatially, the strongest increases were observed in sub-regions A and B, corresponding to the western transportation corridor and the central urban area, respectively. Subregion C exhibited moderate O3 levels with a weaker increasing tendency, while subregion D showed comparatively lower values, although a gradual upward trend remained evident throughout the study period. Overall, the fishnet-based analysis revealed distinct subregional differences in both the magnitude and spatial distribution of tropospheric O3 changes.

3.3. LULC Changes

3.3.1. Accuracy Assessment

To ensure the reliability of the multi-temporal LULC maps (2020, 2022, and 2024), a comprehensive accuracy assessment was conducted. A stratified random sampling approach was implemented in Google Earth Engine (GEE) to generate 380 validation points for each year (Crops: 60; Trees: 60; Rangeland: 60; Built-up: 200). The sampling design and class allocation were kept consistent across all years to ensure methodological uniformity and inter-annual comparability. Reference class labels were assigned through manual visual interpretation in Google Earth Pro using high-resolution satellite imagery (0.3–1.2 m spatial resolution), primarily sourced from Maxar Technologies (Figure 10). The selection of Maxar imagery as ground-truth data was guided by its established credibility and widespread use in remote sensing validation studies. Owing to its sub-meter spatial resolution and superior spatial detail, Maxar imagery enables precise discrimination of land cover types. It reduces uncertainties associated with mixed pixels and indistinct class boundaries. Previous research has demonstrated the effectiveness of high-resolution imagery within the Google Earth Engine framework for validating land-cover classifications and related remote sensing applications [45].
For each year, confusion matrices were constructed by comparing the classified maps with the interpreted reference data. Overall Accuracy (OA), Producer’s Accuracy (PA), User’s Accuracy (UA), and the Kappa coefficient (κ) were calculated to quantify classification performance and agreement beyond chance. The accuracy assessment results are summarized in Table 6 while the detailed confusion matrices for each year are provided in Appendix A. The findings indicate relatively stable classification performance across years. OA values were 80.0% for 2020, 77.1% for 2022, and 78.2% for 2024. The corresponding Kappa coefficients ranged between 0.66 and 0.73, indicating substantial agreement across all years. The harmonized validation framework enhances the robustness of the multi-temporal comparison and strengthens the credibility of the detected land-use change dynamics.
Figure 11 illustrates the spatial and temporal evolution of the four major land-use/land cover (LULC) classes—Crops, Trees, Rangeland, and Built-up—across Bolu from 2020 to 2030. The figure highlights the spatial distribution and expansion patterns of each class over time. Complementing the spatial analysis, Figure 12 presents a quantitative assessment of class-specific area changes (km2) for the years 2020, 2022, 2024, and the 2030 land-use estimates, enabling a direct comparison of temporal trends and the magnitude of land-use transitions.
Between 2020 and 2024, land-use/land cover (LULC) trajectories in Bolu within the defined analytical boundary reveal a moderate yet systematic redistribution of land resources (Figure 11 and Figure 12). Crops and rangeland areas exhibit a gradual decline, whereas trees and built-up classes show concurrent increases:
  • Crops: 214.36 → 209.83 km2 (−4.53 km2; −0.66%)
  • Trees: 322.46 → 326.92 km2 (+4.46 km2; +0.64%)
  • Rangeland: 68.89 → 64.45 km2 (−4.44 km2; −0.65%)
  • Built-up: 77.80 → 82.37 km2 (+4.57 km2; +0.66%)
These patterns indicate a transition away from crops and rangeland uses toward both urban expansion and tree recovery. Among all classes, the built-up category exhibits the steepest and most consistent upward trajectory, reflecting intensified urban development, particularly along the western and central corridors of Bolu. Tree cover demonstrates a mild but steady recovery, while rangeland shows the highest temporal variability, characterized by an initial decline followed by partial stabilization.

3.3.2. Transition Matrix Area Consistency Analysis (2020–2024)

To assess the internal consistency of observed LULC changes, the transition matrix derived from the MOLUSCE model for the 2020–2024 period was cross-validated against empirical area differences. The matrix identifies the dominant land-use transformation pathways:
  • Rangeland → Trees: 7.98%
  • Crops → Built-up: 3.58%
  • Rangeland → Built-up: 1.57%
  • Built-up stability: 94.95%
When applied to the 2020 baseline and converted to area units, these transition probabilities closely reproduce the observed 2024 LULC configuration (Table 4). Minor deviations remain within acceptable limits, confirming strong consistency between matrix-derived estimates and observed class areas.

3.4. LULC Projection (2020–2030)

The ANN–CA-based MOLUSCE simulation provides a spatially explicit prediction of land-use dynamics within the defined analytical boundary by 2030. The model integrates transition potentials, neighborhood effects, and class-level change tendencies calibrated over the 2020–2024 period. High validation performance (overall accuracy: 97.4%; κ = 0.9604) supports the reliability of the model in predicting near-term land-use patterns.

3.4.1. Overall Projected Changes (2020–2030)

Based on the MOLUSCE-based ANN–CA simulation, land-use change trajectories indicate the continuation of urban growth in Bolu within the defined analytical boundary toward 2030, although at a reduced rate compared to the calibration period (2020–2024). Built-up areas are projected to increase by approximately +6.15 km2 (+0.79%) over the full 2020–2030 period, confirming urban expansion as the dominant land-use transition. In contrast, cropland and rangeland areas exhibit gradual declines, with projected net changes of −1.43 km2 and −0.19 km2, respectively, between 2024 and 2030. Areas classified as trees remain largely stable throughout the projection horizon, maintaining an aerial extent of approximately 326.92 km2, which indicates a high level of forest-cover persistence in the modeled land-use transitions. Overall, the projections reveal a gradual redistribution of land-use classes characterized by incremental urban expansion and limited conversion of agricultural and rangeland areas, while tree-covered areas exhibit strong spatial stability.

3.4.2. Spatial Patterns of Projected Change

The projected 2030 LULC map (Figure 11) shows that urban growth is happening in a way that makes sense spatially. The main way this is happening is through corridor-oriented expansion along the western transportation axis, which is linked to the TEM motorway and the Bolu Tunnel. This linear growth configuration reflects transport-oriented development under relatively low topographic constraints. In contrast, the central urban core exhibits predominantly infill-dominated growth, indicating consolidation rather than extensive outward sprawl. At mid- and high-elevation zones, tree-covered areas remain largely stable, while rangeland losses primarily concentrate within peri-urban and foothill transition zones.

3.4.3. Transition Dynamics Driving the 2030 Land-Use Prediction

The projected 2030 land-use configuration is governed by a limited number of dominant transition pathways identified during the calibration period. Corridor-oriented urban expansion is primarily driven by conversions from cropland and rangeland to built-up land. Minor increases in tree cover are sustained through rangeland-to-trees transitions. High persistence within the built-up and trees classes reinforces spatial continuity and enables the cellular automata (CA) mechanism to generate realistic long-term patch dynamics.

3.4.4. Model–Projection Consistency

The ANN–CA framework demonstrates strong consistency between observed land-use changes and projected outcomes. Validation results confirm that future urban growth follows road-dependent and topographically constrained patterns, tree cover remains largely stable, and rangeland represents the most dynamic land-cover class. Overall, the predicted 2030 land-use pattern indicates controlled yet persistent urban expansion, accompanied by stable tree cover and gradual reductions in cropland and rangeland areas.

3.5. UEII (2020–2030)

UEII values calculated for the periods 2020–2022, 2022–2024, 2020–2024, and 2024–2030 (Table 7) indicate that urban expansion in Bolu remains low in intensity but spatially heterogeneous.
Across all sub-regions, UEII values predominantly fall within the moderate to slow growth categories, confirming that urban expansion in Bolu remained limited in magnitude throughout the study period. During the years 2020 to 2022, expansion intensity was at its lowest point, with UEII values close to zero or negative in some areas. This means that urban growth was almost stagnant. A moderate increase was observed during the 2022–2024 period, particularly in sub-regions A and B, followed by a slight decline or stabilization during the 2024–2030 assessment period. There are clear differences between sub-regions when it comes to space. Sub-regions A and B consistently exhibit higher UEII values relative to the other areas, reflecting more pronounced expansion tendencies along the main urban core and transportation corridors. Sub-region C shows a gradual but persistent increase in UEII values over time, indicating incremental expansion in the rural–suburban transition zone. In contrast, sub-region D maintains consistently low UEII values across all periods, suggesting a stable development pattern with minimal spatial expansion. Overall, the UEII results indicate that urban growth in Bolu is characterized by low intensity and spatially uneven expansion, with localized increases occurring in specific sub-regions rather than uniform outward growth across the study area.

3.6. Relationship Between Urban Expansion Intensity and Tropospheric O3 Column Indicators

Figure 13 illustrates the relationship between projected 2030 tropospheric ozone (O3) column indicators and the Urban Expansion Intensity Index (UEII). A negative correlation was observed between UEII and O3 (r ≈ −0.72, n = 4). However, given the very limited number of spatial units included in the analysis, this coefficient should be interpreted with caution. The small sample size substantially limits statistical robustness and does not allow strong inference regarding causality or generalizable relationships.
Within the 2030 projection framework, UEII values and Sentinel-5P-derived tropospheric ozone column indicators were jointly evaluated. Summary statistics for the four subregions (A–D) are presented in Table 3. Projected mean tropospheric O3 column indicators for 2030 ranged between approximately 0.165 and 0.167 mol m−2, with interregional differences of less than 0.002 mol m−2, indicating a relatively homogeneous spatial distribution across the study area. In contrast, UEII values varied between 0.11 and 0.14, corresponding to the “very slow expansion” class across all subregions. Subregions A and B exhibited the highest projected tropospheric O3 column indicators despite lower UEII values, whereas subregion C recorded the highest UEII value but only moderate O3 levels. Subregion D displayed both relatively low UEII and lower projected O3 values compared to the other subregions.
Overall, the observed association suggests weak spatial correspondence between projected urban expansion intensity and tropospheric ozone column indicators within the study area. Given the limited number of spatial units and the scenario-based nature of the 2030 ozone extension, the correlation result should be considered exploratory rather than confirmatory.

4. Discussion

This study examines the relationship between urban growth and tropospheric ozone dynamics in a mid-sized Anatolian city by integrating remote-sensing-based land-use modeling with long-term atmospheric trend analysis. Although the analysis is based on a single case study, the methodological structure developed here may be adaptable to other mid-sized urban systems with comparable topographic and transportation characteristics. Rather than implying direct generalizability of the numerical results, the contribution of this study lies in demonstrating how land-use simulation, satellite-derived atmospheric indicators, and zone-based spatial regionalization can be jointly interpreted within a coherent analytical framework. While relationships between urbanization and air quality have been widely examined, existing studies have predominantly focused on megacities, often treating urban scale as the primary determinant of pollution dynamics [46,47]. In contrast, recent research indicates that urban form, land-use configuration, and developmental stage can be equally influential, particularly in small and mid-sized cities [48]. In this context, the persistent increase in tropospheric ozone observed in this study, despite relatively modest changes in primary pollutants (NO2 and CO), is consistent with VOC-limited photochemical regimes dominated by secondary formation processes [49,50]. Spatially, ozone intensification along the western transportation corridor reflects the combined influence of transport-related precursor accumulation and basin-shaped topographic confinement. The inverse relationship observed between urban expansion intensity (UEII) and tropospheric ozone across subregions further indicates that urban expansion intensity alone is insufficient to explain ozone accumulation. From a planning and policy perspective, these findings indicate that air-quality risks in mid-sized cities cannot be effectively addressed through land-use control or compactness-oriented growth strategies alone. Even when cities grow very slowly, tropospheric ozone levels may keep going up because of the combined effects of transport-related precursor accumulation, basin-like topographic confinement, and biogenic VOC contributions from nearby forested areas. This suggests that conventional mitigation approaches focusing primarily on reducing the spatial extent of urban growth may be insufficient in VOC-limited environments. Instead, targeted emission management along transportation corridors, preservation of urban ventilation pathways, and differentiated ozone control strategies tailored to mid-sized cities could be considered. In this context, the observed decoupling between urban expansion intensity and ozone accumulation highlights the need for integrated planning frameworks that explicitly incorporate atmospheric chemistry regimes alongside spatial development indicators. Similar regime-dependent ozone responses, characterized by inverse relationships between urban development intensity and ozone levels under VOC-limited photochemical environments, have been reported in regional-scale studies [18,29]. Findings reported in the literature further support this interpretation. Analyses based on Sentinel-5P-derived NO2 and HCHO observations demonstrate that sustained increases in ozone concentrations may occur under VOC-limited photochemical regimes even when trends in primary pollutants remain limited [51]. This highlights the critical role of remote sensing data in revealing scale-dependent urban ozone responses in contexts where land-use change indicators alone are insufficient to explain observed air-quality dynamics.

5. Restrictions and Ambiguities

This study proposes an integrated and spatially coherent analytical framework for examining land–atmosphere interactions in a mid-sized urban system. Nevertheless, several methodological and interpretative limitations should be acknowledged.
First, the atmospheric analysis relies on satellite-derived tropospheric ozone (O3) column indicators obtained from Sentinel-5P/TROPOMI. These indicators represent vertically integrated atmospheric concentrations rather than ground-level surface measurements. Consequently, they do not directly quantify human exposure at the surface. Instead, they should be interpreted as proxies of regional-scale atmospheric variability and directional tendencies rather than as equivalents of surface ozone concentrations. Second, the ozone dataset covers a relatively short temporal window (2020–2024), which limits the capacity to detect long-term climatic or structural atmospheric shifts. Although annual aggregation reduces short-term variability, tropospheric ozone formation is influenced by seasonal cycles, temperature fluctuations, solar radiation intensity, and large-scale circulation dynamics. Because explicit meteorological normalization or chemical transport modeling was not implemented, a portion of the observed interannual variability may reflect meteorological influences in addition to land-use-related processes. Nevertheless, the directional consistency observed across the study period suggests the presence of a spatial tendency within the defined analytical boundary. Third, spatial aggregation into four fishnet subregions introduces potential Modifiable Areal Unit Problem (MAUP) effects. Grid-based regionalization may smooth intra-urban heterogeneity and influence the strength of statistical relationships between the Urban Expansion Intensity Index (UEII) and ozone indicators. However, the consistent application of the same grid structure across all datasets ensures internal comparability and methodological coherence at the selected spatial scale. Accordingly, findings should be interpreted at the grid-cell level rather than as fine-scale intra-urban gradients. Fourth, uncertainties are also associated with land-use classification and ANN–CA-based simulation. The reclassification of Sentinel-2 land-cover categories into four aggregated classes may affect boundary delineation, particularly in transitional peri-urban zones. Nonetheless, validation results demonstrate satisfactory overall accuracy and Kappa coefficients, supporting the reliability of the detected land-use patterns. Residual classification errors may influence the precise magnitude of area changes, but they are unlikely to fundamentally alter the identified spatial growth tendencies. Finally, the 2030 ozone values are derived from a linear extension of Theil–Sen slopes estimated from only three discrete annual observations (2020, 2022, and 2024). This approach assumes linear continuity and does not account for potential nonlinear atmospheric responses, emission-policy shifts, or anomalous meteorological conditions. Therefore, the 2030 estimates should be interpreted as scenario-based descriptive extensions intended to explore possible future conditions rather than as deterministic forecasts.
Overall, the identification of spatial variability in ozone indicators and a directionally consistent tendency during the 2020–2024 period can be considered robust within the defined spatial and temporal framework. In contrast, the 2030 extensions and the UEII–ozone correlations should be regarded as exploratory, scenario-based insights that provide an initial analytical foundation for spatial planning discussions and future hypothesis-driven investigations.

6. Conclusions

This research investigated the interconnected evolution of urban land-use change and tropospheric ozone dynamics in Bolu within a spatially integrated analytical framework. The methodology facilitated a coherent comparison between urban expansion intensity and atmospheric trends inside a specified functional urban border by integrating MOLUSCE-based land-use simulations, the Theil–Sen slope estimate, and fishnet-based spatial regionalization. The findings demonstrate a directional rise in tropospheric ozone column indicators over the 2020–2024 observation period. The limited temporal coverage constrains formal statistical inference; however, the observed consistency of the trend across subregions indicates a quantifiable alteration in regional atmospheric conditions during the study period. The scenario-based extension to 2030 suggests the potential continuance of this trend under linear extrapolation assumptions; nonetheless, this prediction should be regarded as exploratory rather than predictive. At the applied grid scale, ozone patterns did not display a straightforward proportionate correlation with the intensity of urban growth. The found inverse geographical correlation between UEII and tropospheric ozone indicates that air quality dynamics in mid-sized urban systems may be influenced not only by land conversion rates but also by transport-related emissions, topographical limitations, and localized photochemical processes. The results are contingent upon the established analytical boundaries and the spatial and temporal resolution utilized in this research. Rather than offering prescriptive policy recommendations, the contribution of this research lies in demonstrating how spatially explicit land-use simulations and satellite-derived atmospheric indicators can be jointly interpreted to explore land–atmosphere interactions in mid-sized urban contexts. A subsequent study that integrates extended time series, enhanced geographical resolution, and meteorological normalization would fortify attribution and augment the resilience of scenario-based extensions.

Funding

This research received no external funding.

Data Availability Statement

Sentinel-2 MSI and Sentinel-5P/TROPOMI data were obtained from the ESA Copernicus Programme. ASTER Global DEM (ASTER GDEM v3) data were accessed via the NASA Earthdata platform, population density data were obtained from WorldPop, and road network data were acquired from OpenStreetMap. The data used in this study were derived from publicly available resources. Processed data and derived datasets supporting the findings of this study are available from the corresponding author upon reasonable request.

Conflicts of Interest

The author declares no conflicts of interest.

Abbreviations

The following abbreviations are used in this manuscript:
AbbreviationDefinition
GISGeographic Information Systems
RSRemote Sensing
LULCLand-Use/Land Cover
O3Tropospheric Ozone
NO2Nitrogen Dioxide
COCarbon Monoxide
DEMDigital Elevation Model
MSIMultispectral Instrument
UEIIUrban Expansion Intensity Index
GEEGoogle Earth Engine

Appendix A. Confusion Matrices for LULC Classification (2020–2024)

Appendix A.1. Confusion Matrix for 2020

Table A1. Confusion matrix for the 2020 LULC classification (n = 380).
Table A1. Confusion matrix for the 2020 LULC classification (n = 380).
Reference\ClassifiedCropsTreesRangelandBuilt-upRow Total
Crops (60)4844460
Trees (60)4484460
Rangeland (60)4448460
Built-up (200)101020160200
Column Total666676172380

Appendix A.2. Confusion Matrix for 2022

Table A2. Confusion matrix for the 2022 LULC classification (n = 380).
Table A2. Confusion matrix for the 2022 LULC classification (n = 380).
Reference\ClassifiedCropsTreesRangelandBuilt-UpRow Total
Crops (60)4654560
Trees (60)4474560
Rangeland (60)5544660
Built-up (200)151217156200
Column Total706969172380

Appendix A.3. Confusion Matrix for 2024

Table A3. Confusion matrix for the 2024 LULC classification (n = 380).
Table A3. Confusion matrix for the 2024 LULC classification (n = 380).
Reference\ClassifiedCropsTreesRangelandBuilt-UpRow Total
Crops (60)4744560
Trees (60)4483560
Rangeland (60)5545560
Built-up (200)141316157200
Column Total707068172380

References

  1. Angel, S.; Parent, J.; Civco, D.L. The fragmentation of urban landscapes: Global evidence of a key attribute of the spatial structure of cities, 1990–2000. Environ. Urban. 2012, 24, 249–283. [Google Scholar] [CrossRef]
  2. Güneralp, B.; Seto, K.C. Futures of global urban expansion: Uncertainties and implications for biodiversity conservation. Environ. Res. Lett. 2013, 8, 014025. [Google Scholar] [CrossRef]
  3. Perveen, S.; Yigitcanlar, T.; Kamruzzaman, M.; Agdas, D. How can transport impacts of urban growth be modelled? An approach to consider spatial and temporal scales. Sustain. Cities Soc. 2020, 55, 102031. [Google Scholar] [CrossRef]
  4. Sun, Y.; Liu, C.; Palm, M.; Vigouroux, C.; Notholt, J.; Hu, Q.; Jones, N.; Wang, W.; Su, W.; Zhang, W.; et al. Ozone seasonal evolution and photochemical production regime in the polluted troposphere in eastern China derived from high-resolution Fourier transform spectrometry (FTS) observations. Atmos. Chem. Phys. 2018, 18, 14569–14583. [Google Scholar] [CrossRef]
  5. Monks, P.S.; Archibald, A.T.; Colette, A.; Cooper, O.; Coyle, M.; Derwent, R.; Fowler, D.; Granier, C.; Law, K.S.; Mills, G.E.; et al. Tropospheric ozone and its precursors from the urban to the global scale from air quality to short-lived climate forcer. Atmos. Chem. Phys. 2015, 15, 8889–8973. [Google Scholar] [CrossRef]
  6. World Health Organization. WHO Global Air Quality Guidelines; WHO: Geneva, Switzerland, 2021. [Google Scholar]
  7. Duranton, G.; Turner, M.A. The fundamental law of road congestion: Evidence from US cities. Am. Econ. Rev. 2011, 101, 2616–2652. [Google Scholar] [CrossRef]
  8. Cummings, L.E.; Stewart, J.D.; Reist, R.; Shakya, K.M.; Kremer, P. Mobile monitoring of air pollution reveals spatial and temporal variation in an urban landscape. Front. Built Environ. 2021, 7, 648620. [Google Scholar] [CrossRef]
  9. Gorelick, N.; Hancher, M.; Dixon, M.; Ilyushchenko, S.; Thau, D.; Moore, R. Google Earth Engine: Planetary-scale geospatial analysis for everyone. Remote Sens. Environ. 2017, 202, 18–27. [Google Scholar] [CrossRef]
  10. Veefkind, J.P.; Aben, I.; McMullan, K.; Förster, H.; de Vries, J.; Otter, G.; Claas, J.; Eskes, H.J.; de Haan, J.F.; Kleipool, Q.; et al. TROPOMI on the ESA Sentinel-5 Precursor: A GMES mission for global observations of the atmospheric composition for climate, air quality and ozone layer applications. Remote Sens. Environ. 2012, 120, 70–83. [Google Scholar] [CrossRef]
  11. Munthali, M.G.; Mustak, S.; Adeola, A. Modelling land use and land cover dynamics of Dedza district of Malawi using hybrid Cellular Automata and Markov model. Remote Sens. Appl. Soc. Environ. 2020, 17, 100276. [Google Scholar] [CrossRef]
  12. Erkek, D.; Yağcı, C.; İşcan, F. Urban growth predictions: Optimization of urbanization strategy for risk mitigation in medium-sized cities. Growth Change 2025, 56, e70054. [Google Scholar] [CrossRef]
  13. Theil, H. A rank-invariant method of regression analysis. Ned. Tijdschr. Natuurk. 1950, 3, 386–392. [Google Scholar]
  14. Sen, P.K. Estimates of the regression coefficient based on Kendall’s tau. J. Am. Stat. Assoc. 1968, 63, 1379–1389. [Google Scholar] [CrossRef]
  15. Fang, Y.; Zhao, L. Assessing the environmental benefits of urban ventilation corridors: A case study in Hefei, China. Build. Environ. 2022, 212, 108810. [Google Scholar] [CrossRef]
  16. Seto, K.C.; Güneralp, B.; Hutyra, L.R. Global forecasts of urban expansion to 2030 and direct impacts on biodiversity and carbon pools. Proc. Natl. Acad. Sci. USA 2012, 109, 16083–16088. [Google Scholar] [CrossRef]
  17. Dörter, M.; Mağat-Türk, E.; Döğeroğlu, T.; Özden-Üzmez, Ö.; Gaga, E.O.; Karakaş, D.; Yenisoy-Karakaş, S. An assessment of spatial distribution and atmospheric concentrations of ozone, nitrogen dioxide, sulfur dioxide, benzene, toluene, ethylbenzene, and xylenes: Ozone formation potential and health risk estimation in Bolu city of Turkey. Environ. Sci. Pollut. Res. 2022, 29, 53569–53583. [Google Scholar] [CrossRef]
  18. Karner, A.A.; Eisinger, D.S.; Niemeier, D.A. Near-roadway air quality: Synthesizing the findings from real-world data. Environ. Sci. Technol. 2010, 44, 5334–5344. [Google Scholar] [CrossRef]
  19. Taner, İ. Kent profili: Bolu. Bolu Abant İzzet Baysal Üniversitesi Sos. Bilim. Enstitüsü Derg. 2016, 16, 159–197. [Google Scholar] [CrossRef]
  20. Bolu Directorate of Environment and Urbanization (BDEU). Bolu Provincial Environmental Status Report 2018; BDEU: Bolu, Türkiye, 2018. [Google Scholar]
  21. Turkish State Meteorological Service. Genel İstatistik Verileri. Available online: https://www.mgm.gov.tr (accessed on 16 June 2025).
  22. Karra, K.; Kontgis, C.; Statman-Weil, Z.; Mazzariello, J.C.; Mathis, M.; Brumby, S.P. Global land use/land cover with Sentinel 2 and deep learning. In Proceedings of the 2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS, Brussels, Belgium, 11–16 July 2021; IEEE: Piscataway, NJ, USA, 2021; Volume 8, pp. 4704–4707. [Google Scholar] [CrossRef]
  23. Hegazy, I.R.; Kaloop, M.R. Monitoring urban growth and land use change detection with GIS and remote sensing techniques in Daqahlia Governorate, Egypt. Int. J. Sustain. Built Environ. 2015, 4, 117–124. [Google Scholar] [CrossRef]
  24. NASA/METI/AIST/Japan Spacesystems; U.S./Japan ASTER Science Team. ASTER Global Digital Elevation Model Version 3; NASA EOSDIS Land Processes DAAC. 2019. [CrossRef]
  25. Earthdata Search. NASA Earthdata. Available online: https://search.earthdata.nasa.gov/ (accessed on 15 July 2025).
  26. WorldPop. WorldPop Global Population Data. Available online: https://www.worldpop.org/ (accessed on 15 July 2025).
  27. OpenStreetMap Contributors. OpenStreetMap. Available online: https://www.openstreetmap.org/ (accessed on 15 July 2025).
  28. Bolu Governorship. Bolu Province 2023 Environmental Status Report; Provincial Directorate of Environment, Urbanization and Climate Change: Bolu, Türkiye, 2024. [Google Scholar]
  29. Seinfeld, J.H.; Pandis, S.N. Atmospheric Chemistry and Physics: From Air Pollution to Climate Change, 3rd ed.; John Wiley & Sons: Hoboken, NJ, USA, 2016. [Google Scholar]
  30. Borsdorff, T.; Andrasec, J.; Hu, H.; Aben, I.; Landgraf, J. Detection of carbon monoxide pollution from cities and wildfires on regional and urban scales: The benefit of CO column retrievals from SCIAMACHY 2.3 µm measurements under cloudy conditions. Atmos. Meas. Tech. 2018, 11, 2553–2565. [Google Scholar] [CrossRef]
  31. Jin, X.; Fiore, A.M.; Boersma, K.F.; De Smedt, I.; Valin, L.C. Inferring changes in summertime surface ozone–NOx–VOC chemistry over U.S. Urban Areas from Two Decades of Satellite and Ground-Based Observations. Environ. Sci. Technol. 2020, 54, 6518–6529. [Google Scholar] [CrossRef]
  32. Çelik, M.A.; Bilik, A.; Akiner, M.E.; Gemeda, D.O. How do land use/cover changes influence air quality in Türkiye? A Satellite-Based Assessment. Land 2025, 14, 1945. [Google Scholar] [CrossRef]
  33. Sood, A.; Smakhtin, V. Global hydrological models: A review. Hydrol. Sci. J. 2015, 60, 549–565. [Google Scholar] [CrossRef]
  34. Alqadhi, S.; Mallick, J.; Balha, A.; Bindajam, A.; Singh, C.K.; Hoa, P.V. Spatial and decadal prediction of land use/land cover using multi-layer perceptron-neural network (MLP-NN) algorithm for a semi-arid region of Asir, Saudi Arabia. Earth Sci. Inform. 2021, 14, 1547–1562. [Google Scholar] [CrossRef]
  35. Kondum, F.A.; Rowshon, M.K.; Luqman, C.A.; Hasfalina, C.M.; Zakari, M.D. Change analyses and prediction of land use and land cover changes in the Bernam River Basin, Malaysia. Remote Sens. Appl. Soc. Environ. 2024, 36, 101281. [Google Scholar] [CrossRef]
  36. NextGIS. MOLUSCE—Modules for Land Use Change Evaluation. Available online: https://docs.nextgis.com/docs_ngqgis/source/molusce.html (accessed on 15 July 2025).
  37. Clarke, K.C.; Hoppen, S.; Gaydos, L. A self-modifying cellular automaton model of historical urbanization in the San Francisco Bay Area. Environ. Plan. B Plan. Des. 1997, 24, 247–261. [Google Scholar] [CrossRef]
  38. Li, X.; Yeh, A.G.O. Neural-network-based cellular automata for simulating multiple land use changes using GIS. Int. J. Geogr. Inf. Sci. 2002, 16, 323–343. [Google Scholar] [CrossRef]
  39. Yang, X.; Zheng, X.Q.; Lv, L.N. A spatiotemporal model of land use change based on ant colony optimization, Markov chain and cellular automata. Ecol. Model. 2012, 233, 11–19. [Google Scholar] [CrossRef]
  40. Liu, Y.; Hu, Y.; Long, S.; Liu, L.; Liu, X. Analysis of the effectiveness of urban land-use-change models based on the measurement of spatio-temporal, dynamic urban growth: A cellular automata case study. Sustainability 2017, 9, 796. [Google Scholar] [CrossRef]
  41. Ren, P.; Gan, S.; Yuan, X.; Zong, H.; Xie, X. Spatial expansion and sprawl quantitative analysis of mountain city built-up area. In Geo-Informatics in Resource Management and Sustainable Ecosystem (GRMSE 2013); Communications in Computer and Information Science; Springer: Berlin/Heidelberg, Germany, 2013; Volume 398, pp. 166–176. [Google Scholar] [CrossRef]
  42. Wei, G.; He, B.-J.; Sun, P.; Liu, Y.; Li, R.; Ouyang, X.; Luo, K.; Li, S. Evolutionary trends of urban expansion and its sustainable development: Evidence from 80 representative cities in the belt and road initiative region. Cities 2023, 138, 104353. [Google Scholar] [CrossRef]
  43. Yan, J.; Wang, J.; Su, F.; Liu, B. Morphology changes and expansion of major port cities in the Philippines from 1990 to 2020. Cities 2024, 147, 104818. [Google Scholar] [CrossRef]
  44. Al-Sharif, A.A.; Pradhan, B. Monitoring and predicting land use change in Tripoli using CA–Markov models. Arab. J. Geosci. 2014, 7, 4291–4301. [Google Scholar] [CrossRef]
  45. Gümüş, M.G. Performance analysis of water extraction indices with geospatial and statistical techniques using Google Earth Engine platform: A case study of Ramsar wetlands in Türkiye. J. Indian Soc. Remote Sens. 2025, 53, 2697–2721. [Google Scholar] [CrossRef]
  46. Shukla, V.; Parikh, K. The environmental consequences of urban growth: Cross-national perspectives on economic development, air pollution, and city size. Urban Geogr. 1992, 13, 422–449. [Google Scholar] [CrossRef]
  47. Wu, F.; Zhang, F. Rethinking China’s urban governance: The role of the state in neighbourhoods, cities and regions. Prog. Hum. Geogr. 2022, 46, 775–797. [Google Scholar] [CrossRef]
  48. Liang, L.; Gong, P. Urban and air pollution: A multi-city study of long-term effects of urban landscape patterns on air quality trends. Sci. Rep. 2020, 10, 18618. [Google Scholar] [CrossRef]
  49. Sillman, S. The relation between ozone, NOx and hydrocarbons in urban and polluted rural environments. Atmos. Environ. 1999, 33, 1821–1845. [Google Scholar] [CrossRef]
  50. Ren, J.; Guo, F.; Xie, S. Diagnosing ozone–NOx–VOC sensitivity and revealing causes of ozone increases in China based on 2013–2021 satellite retrievals. Atmos. Chem. Phys. 2022, 22, 15035–15047. [Google Scholar] [CrossRef]
  51. Dong, S.; Ma, P.; Yang, X.; Luo, N.; Chen, L.; Wang, L.; Song, H.; Zhao, S.; Zhao, W. Characteristics and source analysis of ozone pollution in Tianjin from 2013 to 2022. Remote Sens. 2024, 16, 3970. [Google Scholar] [CrossRef]
Land 15 00456 g001
Figure 1. (a) Location of Bolu in Türkiye (b) Land use/land cover (LULC) distribution within Bolu Province. (c) study area showing sub-regions (A–D).
Figure 1. (a) Location of Bolu in Türkiye (b) Land use/land cover (LULC) distribution within Bolu Province. (c) study area showing sub-regions (A–D).
Land 15 00456 g001
Land 15 00456 g002
Figure 2. Flowchart of the study. The left panel represents the land-use change modeling process based on Sentinel-2 LULC data and driving factors (DEM). The right panel shows the satellite-based air quality analysis performed using Sentinel-5P data on the Google Earth Engine platform. Arrows indicate the sequence of analytical steps, diamond shapes represent model validation thresholds, and the dashed vertical line separates the land-use modeling and satellite data analysis modules. Blue dashed boxes highlight the prediction stage used to estimate future conditions.
Figure 2. Flowchart of the study. The left panel represents the land-use change modeling process based on Sentinel-2 LULC data and driving factors (DEM). The right panel shows the satellite-based air quality analysis performed using Sentinel-5P data on the Google Earth Engine platform. Arrows indicate the sequence of analytical steps, diamond shapes represent model validation thresholds, and the dashed vertical line separates the land-use modeling and satellite data analysis modules. Blue dashed boxes highlight the prediction stage used to estimate future conditions.
Land 15 00456 g002
Land 15 00456 g003
Figure 3. Spatial distribution of normalized driving factors.
Figure 3. Spatial distribution of normalized driving factors.
Land 15 00456 g003
Land 15 00456 g004
Figure 4. Neural network learning curve for the 2020–2022 calibration period used in the ANN–CA model.
Figure 4. Neural network learning curve for the 2020–2022 calibration period used in the ANN–CA model.
Land 15 00456 g004
Land 15 00456 g005
Figure 5. Multiple resolution budget validation of the MOLUSCE-based ANN–CA land-use change simulation.
Figure 5. Multiple resolution budget validation of the MOLUSCE-based ANN–CA land-use change simulation.
Land 15 00456 g005
Land 15 00456 g006
Figure 6. Spatial distribution of tropospheric carbon monoxide (CO) over the defined analytical boundary.
Figure 6. Spatial distribution of tropospheric carbon monoxide (CO) over the defined analytical boundary.
Land 15 00456 g006
Land 15 00456 g007
Figure 7. Spatial distribution of tropospheric nitrogen dioxide (NO2) over the defined analytical boundary.
Figure 7. Spatial distribution of tropospheric nitrogen dioxide (NO2) over the defined analytical boundary.
Land 15 00456 g007
Land 15 00456 g008
Figure 8. Temporal variation in satellite-derived tropospheric O3 column indicators across the four fishnet subregions (A–D) during the 2020–2024 period. Values are derived from three discrete annual observations (2020, 2022, and 2024).
Figure 8. Temporal variation in satellite-derived tropospheric O3 column indicators across the four fishnet subregions (A–D) during the 2020–2024 period. Values are derived from three discrete annual observations (2020, 2022, and 2024).
Land 15 00456 g008
Land 15 00456 g009
Figure 9. Spatiotemporal patterns of tropospheric O3 over the defined analytical boundary (2020–2024) and scenario-based 2030 extension derived from three discrete annual observations (2020, 2022, and 2024).
Figure 9. Spatiotemporal patterns of tropospheric O3 over the defined analytical boundary (2020–2024) and scenario-based 2030 extension derived from three discrete annual observations (2020, 2022, and 2024).
Land 15 00456 g009
Land 15 00456 g010
Figure 10. Illustration of random points generated for accuracy analysis.
Figure 10. Illustration of random points generated for accuracy analysis.
Land 15 00456 g010
Land 15 00456 g011
Figure 11. Temporal Evolution of The LULC Classes in Bolu.
Figure 11. Temporal Evolution of The LULC Classes in Bolu.
Land 15 00456 g011
Land 15 00456 g012
Figure 12. Area changes (km2) of LULC classes in Bolu for 2020, 2022, 2024, and projected 2030.
Figure 12. Area changes (km2) of LULC classes in Bolu for 2020, 2022, 2024, and projected 2030.
Land 15 00456 g012
Land 15 00456 g013
Figure 13. Relationship between projected tropospheric ozone (O3) column indicators and UEII for the 2030. Points A–D correspond to the spatial zones defined in this figure: A—Western urban corridor around the Bolu Tunnel; B—Central dense urban area; C—Northern rural zone; D—Southern rural zone.
Figure 13. Relationship between projected tropospheric ozone (O3) column indicators and UEII for the 2030. Points A–D correspond to the spatial zones defined in this figure: A—Western urban corridor around the Bolu Tunnel; B—Central dense urban area; C—Northern rural zone; D—Southern rural zone.
Land 15 00456 g013
Table 1. Datasets used in the study.
Table 1. Datasets used in the study.
Data CategoryDataset/SourceTemporal CoverageSpatial
Resolution		Main Variables/ClassesPurpose of the Study
LULCSentinel-2 MSI (ESA Copernicus)2020, 2022, 202410 mCrops, trees, rangeland, and built-upBaseline and transition maps for LULC change analysis and MOLUSCE simulations
SimulationMOLUSCE (QGIS plugin)2020–2030 (projection)10 m (input-based)Cellular automata–Markov transition probabilitiesSimulation of future LULC urban expansion patterns
O3Sentinel-5P/TROPO-MI (ESA Copernicus)2020–2024 (observed); 2030 (projected)~7 × 3.5 kmTropospheric ozone column indicatorTrend analysis and 2030 projection of ozone dynamics
NO2Sentinel-5P/TROPO-MI2020–2024 (observed only)~7 × 3.5 kmTropospheric NO2 column indicatorSupporting observational analysis of traffic-related emissions
COSentinel-5P/TROPO-MI2020–2024 (observed only)~7 × 3.5 kmTropospheric CO column indicatorSupporting observational analysis of combustion-related emissions
Trend analysis (O3 only)Theil–Sen slope
estimator		2020–2030 (projection)~7 × 3.5 kmMedian-based non-parametric slope estimate (Theil–Sen)Estimation and extrapolation of tropospheric ozone trends
Spatial aggregationRegular fishnet gridStaticZone-based (custom)Mean zonal statisticsHarmonization of multi-resolution datasets
Urban expansion metricUrban Expansion Intensity Index (UEII)2020–2030Zone-basedExpansion intensity classesQuantification of the spatial intensity of urban growth
Table 2. Description of the LULC classification used in this study.
Table 2. Description of the LULC classification used in this study.
IndexLULCDescription
1CropsHumans planted/plotted cereals, grasses, and crops not at tree height.
2TreesAny significant clustering of tall (~15 feet or higher) dense vegetation, typically with a closed or dense canopy.
3RangelandOpen areas covered in homogenous grasses with little to no taller vegetation; wild cereals and grasses with no obvious human plotting
4Built-upHuman-made structures: major road and rail networks; large, homogenous, impervious surfaces
Table 3. Summary statistics of scenario-based 2030 tropospheric ozone (O3) column indicators derived from three discrete annual observations (2020, 2022, 2024).
Table 3. Summary statistics of scenario-based 2030 tropospheric ozone (O3) column indicators derived from three discrete annual observations (2020, 2022, 2024).
IDMINMAXMEANSTD
A0.16490.16720.165920.00044
B0.16560.16740.166460.00040
C0.16430.16730.165320.00063
D0.16480.16700.165770.00033
Table 4. Land use/land cover change between 2020 and 2022.
Table 4. Land use/land cover change between 2020 and 2022.
Class2020 (km2)2022 (km2)Change (km2)2020 (%)2022 (%)Change (%)
Crops214.41216.48+2.0731.3631.66+0.30
Trees322.60327.36+4.7647.1847.88+0.70
Rangeland68.8960.40−8.4910.088.83−1.25
Built Area77.8279.47+1.6511.3811.62+0.24
Table 5. Classification of UEII Values.
Table 5. Classification of UEII Values.
UEII RangeClassDescription
>1.92Very FastIntense urban expansion
1.05–1.92FastHigh growth tendency
0.59–1.05ModerateBalanced development
0.28–0.59SlowLimited growth
0.08–0.28Very SlowLow-intensity expansion
<0.08StableNo significant growth
Table 6. Classification accuracy assessment results for LULC maps (2020–2024).
Table 6. Classification accuracy assessment results for LULC maps (2020–2024).
YearOA (%)Kappa
κ		Crops (PA/UA)Trees (PA/UA)Rangeland (PA/UA)Built-Up (PA/UA)
202080.00.7380.0/72.780.0/72.780.0/63.280.0/93.0
202277.10.6676.7/65.778.3/68.173.3/63.878.0/90.7
202478.20.6778.3/67.180.0/68.675.0/66.278.5/91.3
Table 7. UEII values, growth classes, and interpretative assessment for sub-regions across multiple times (2020–2030).
Table 7. UEII values, growth classes, and interpretative assessment for sub-regions across multiple times (2020–2030).
RegionUEII (2020–2022)Class UEII (2022–2024)Class UEII (2020–2024)Class UEII (2024–2030)Class Interpretation
A0.148 Very slow0.419 Very slow0.284 Slow0.112 Very slowWestern corridor—spatial intensification
B0.296 Very slow0.274 Slow0.285 Slow0.119 Very slowCentral area—stable, relatively intensive
C−0.014 Stable0.046 Very slow0.031 Very slow0.142 Very slowNorthern rural zone—gradual increase
D0.055 Very slow0.112 Very slow0.083 Very slow0.133 Very slowSouthern belt—limited expansion
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Yağcı, C. Linking Urban Land Use Change and Tropospheric Ozone Dynamics in a Mid-Sized City. Land 2026, 15, 456. https://doi.org/10.3390/land15030456

AMA Style

Yağcı C. Linking Urban Land Use Change and Tropospheric Ozone Dynamics in a Mid-Sized City. Land. 2026; 15(3):456. https://doi.org/10.3390/land15030456

Chicago/Turabian Style

Yağcı, Ceren. 2026. "Linking Urban Land Use Change and Tropospheric Ozone Dynamics in a Mid-Sized City" Land 15, no. 3: 456. https://doi.org/10.3390/land15030456

APA Style

Yağcı, C. (2026). Linking Urban Land Use Change and Tropospheric Ozone Dynamics in a Mid-Sized City. Land, 15(3), 456. https://doi.org/10.3390/land15030456

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop