Modeling and comparative analysis of spatial distribution of SO 2 concentration using MLR and GWR models: A case study of Karabük, Türkiye

Modeling and comparative analysis of spatial distribution of SO 2 concentration using MLR and GWR models: A case study of Karabük, Türkiye | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Modeling and comparative analysis of spatial distribution of SO 2 concentration using MLR and GWR models: A case study of Karabük, Türkiye Emre Yücer This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8450115/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 04 Mar, 2026 Read the published version in Pure and Applied Geophysics → Version 1 posted 12 You are reading this latest preprint version Abstract In this study, a comparative analysis of Multiple Linear Regression (MLR) and Geographically Weighted Regression (GWR) models was applied within the scope of Land Use Regression (LUR) analysis in order to reveal the spatial distribution of sulfur dioxide (SO 2 ) concentration in Karabük city. 5-fold cross-validation and external data validation methods were used to test the accuracy of the models. The MLR model successfully represented the distribution of SO2 concentrations in general. However, due to the application of fixed coefficients, spatial heterogeneity could not be fully revealed. In the GWR analysis, local differences were revealed more accurately thanks to the coefficient produced separately for each sample point. In the MLR model, the overall accuracy is R2:0.85, the 5-fold cross-validation method accuracy is R2:0.85, and the external data validation method accuracy is R2:0.92. In the GWR model, the overall accuracy is R2:0.95, the 5-fold cross-validation method accuracy is R2:0.95, and the external data validation method accuracy is R2:0.94. In both methods, high SO2 concentrations were observed in the southern part of the city where industrial areas are dense. In the local R2 distribution of the GWR model, high explanation values ​​were again obtained in the southern regions. In the residual analysis of the GWR model, it was observed that the prediction errors of the model were in the range of ± 1 µg/m³ and were randomly distributed except for a few small local regions. According to the results, the GWR model performed better than the MLR model in both predictive accuracy and spatial heterogeneity. Geographically weighted regression Karabük Land use regression Multiple linear regression Spatial modeling Sulfur dioxide Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 1. Introduction With urbanization, industrialization, and the increasing number of vehicles, fossil fuel use is also rapidly increasing. Fossil fuels are used in many areas, from energy production to industrial facilities. Pollutants released into the atmosphere as a result of fossil fuel consumption cause increased sulfur dioxide (SO₂) concentrations. SO 2 causes respiratory irritation, coughing, wheezing, and bronchial infections. Longterm exposure leads to chronic respiratory diseases and cardiovascular diseases. Studies in this area have shown that even low levels of exposure have adverse effects on general health. (Orellano et al., 2021 ; Meo et al., 2024 ). Land Use Regression (LUR) models are used to estimate the spatial distribution of atmospheric pollutants. In the models, concentration values ​​obtained from ground stations are used to estimate areas where measurements cannot be made. In order to obtain the prediction model and coefficients, the values ​​obtained from the measurement stations are associated with spatial data. Spatial data such as traffic density, distribution of industrial facilities, land use, road network and population density are widely used. In addition, meteorological data such as temperature, humidity and pressure are included in the models (Eeftens et al., 2012 ; Ma et al., 2024 ). Different methods such as linear, nonlinear, and machine learning are used in LUR models. Multiple linear regression (MLR) and Geographically Weighted regression (GWR) are the most commonly used linear methods (Guo et al., 2020 ; Shi et al., 2020 ; Xuan et al., 2021 ; Kerckhoffs et al., 2022 ). Correct variable selection, spatial distribution of variables and number of measurement stations are the parameters that affect the success of the model. In addition, the choice of method used in model validation is also important (Aarthi et al., 2020 ; Torres et al., 2024 ). In MLR analysis, there is one dependent variable and multiple independent variables. It is assumed that there is a linear relationship between the dependent variable and the independent variables. Modelling the dispersion of air pollutants can be difficult when the number of ground stations is small or because of their dispersed location. However, successful results have been obtained in LUR studies at different scales for modeling the spatial distribution of air pollutants (Aarthi et al., 2020 ; Wang and Xu 2021 ; Mikeš et al., 2023 ; Siu et al., 2025 ). Huang et al., ( 2017 ) used data from the national monitoring station to model the distribution of PM₂.₅, NO₂ and SO₂ pollutants in Nanjing, China. They used road, land use, meteorological and topographic data for independent variables. In their results, they found an average R² value of 0.83 for SO2. Yücer et al., ( 2023 ) constructed a LUR model using annual averages of SO₂ and PM₁₀ concentrations in the city of Izmit, Türkiye. Indicators such as main road distance, industrial area density, and population density were selected as independent variables in the model. The R² value, representing the explanatory power of the model, was found to be 0.88. The correlation coefficient between observed and predicted values in model validation was found to be 0.90. In a study examining the SO₂ distribution in Saint John, Canada, Siu et al., ( 2025 ) obtained SO₂ concentrations from ground monitoring stations and TROPOMI data. As a result of the 10-fold cross-validation method, they determined the classification accuracy as 83% and the R² value of the daily dataset as 0.46. Global regression models assume that the spatial relationship between the dependent and independent variables is constant across the study area. However, since the city's structure and emission sources do not show the same distribution in every region within the city, this relationship may differ spatially. GWR eliminates spatial stationarity by conducting local regression for each location. It reveals how the variable coefficients change across space (Fotheringham et al., 2009 ; Fotheringham et al., 2017 ; Comber et al., 2020 ). Wei et al., ( 2019 ) examined the relationship between PM 2.5 and SO 2 , NO 2 , PM 10 , CO and O 3 in Heilongjiang, China, using linear mixed models (LMM), GWR and geographically and temporally weighted regression (GTWR) models. It has been found that GWR-based models achieve higher accuracy values ​​than traditional models. Shen et al., ( 2024 ) modeled monthly average air pollutant values ​​according to GWR and MLR analyses in their study on a European scale between 2000 and 2019. Models evaluated according to the 5-fold cross-validation method have yielded successful results. Morshed et al., ( 2024 ) used meteorological indicators as independent variables in the GWR and MLR models. The effect of meteorological variables on pollutant concentration was investigated. According to the results, GWR achieved more successful performance. Liu et al., ( 2025 ) modeled the distribution of PM 2.5 concentration in 284 different cities in China using GWR and Multiscale Geographically Weighted Regression (MGWR) analyses. They found that MGWR was more successful than GWR in revealing the seasonal distribution of PM₂.₅. Studies in the literature have shown that industrial density, proximity to industrial areas, access to the road network, precipitation and wind variables are closely related to SO₂ distribution. GWR has been more successful in revealing local effects on the distribution of air pollutants. Especially in industrial cities, the distribution of SO2 concentration was revealed by the coefficients obtained at the local scale (Yang at al., 2017; Zhou et al., 2019 ). The results obtained from these studies showed that GWR was successful in detecting local heterogeneity in the distribution of pollutant concentration. In this study, MLR and GWR analyses were used to model the spatial distribution of SO₂ in Karabük province. In the regression analyses, the annual average of SO2 concentrations in 2021 was defined as the independent variable. Spatial data such as industrial areas, roads, ground surface temperature, and altitude were selected as independent variables. The results obtained from the analyses were compared with similar MLR and GWR models in the literature. As a result of the analysis, the coefficients of the indicators, the coefficient of determination (R²), and the Root Mean Square Error (RMSE) values were interpreted to explain the relationship between SO₂ and the independent variables. In addition, external data validation and 5-fold cross-validation methods were used to validate the models. The findings of the study will contribute to expanding practices related to SO₂. In addition, it aims to contribute to the development of spatial prediction tools for decision-makers in the management of air quality at the local scale. 2. Materials and methods 2.1. Study area The provincial center of Karabük, an industrial city located in the Western Black Sea Region of Türkiye, was selected as the study area (Fig. 1 ). Karabük is located between 40° 57' and 41° 34' northern latitudes and 32° 04' and 33° 06' eastern longitudes, and it has an area of approximately 4,145 km². The city center is located at an elevation of approximately 280 m above sea level. The topography of the city has a complex structure where valley and plateau areas coexist (Karabük Municipality, 2025 ). Karabük plays a decisive role in the industrial sector in its economic development. Karabük Iron and Steel Works (KARDEMİR), established in 1937, is Türkiye’s first integrated iron and steel production facility. With the establishment of KARDEMİR, the city entered a rapid industrialization process. Iron and steel production and its sub-sectors form the basis of the city's economic structure (Karabük TSO, 2025 ). Most of the industrial facilities are located in areas close to the city center. Pollutants such as SO₂, NOₓ and particulate matter, especially those originating from heavy industrial facilities, negatively affect the air quality of Karabük. Fossil fuels used in iron and steel production, rolling mills and foundries are also among the main sources of SO₂. Therefore, investigating the relationship between industrial facilities, roads, population density, and other spatial variables and air pollutant concentrations in Karabük is important for both environmental management policies and public health. 2.2. Data Sets The dependent variable of the study, SO₂ data for the year 2021, was obtained from four different ground stations of the Ministry of Environment and Urbanization. These data, which are measured hourly by the National Air Quality Monitoring Stations, are published at www.havaizleme.gov.tr . The measurement data were calculated as the annual average of daily values. The lowest number of days measured at the stations is 326. The independent variables used to explain the spatial distribution of SO₂ concentration were created from spatial indicators frequently preferred in similar LUR applications in the literature (Table 1 ). Studies in the literature have found that pollutants, especially SO₂ concentration, exhibit strong relationships with variables such as industrial density and roads (Meng et al., 2015 ; Wang and Xu 2021 ; Mikeš et al., 2023 ; Siu et al., 2025 ). Data on roads, industrial facilities, high-traffic intersections, fuel stations, and neighborhood boundaries were obtained from the city map of Karabük Municipality. Table 1 Independent variables Distance to main roads (m) Distance to secondary roads (m) Total road length within 150 m buffer (m) Total road length within 300 m buffer (m) Total road length within 500 m buffer (m) Elevation (SRTM) (m) Annual average land surface temperature - LST (°C) Population density (persons/hectare) Distance to industrial facilities (m) Industrial density data Total distance to high-traffic intersections (m) Fuel stations density data Distance to main and secondary roads (m): Traffic-related pollutants are among the main causes of urban air pollution, especially in terms of SO 2 and NO x . Studies in the literature have shown that the distances of sampling points to major roads significantly affect pollutant concentrations (Hoek et al., 2008 ; Chen et al., 2020 ; Wong et al., 2021 ; Mikeš et al., 2023 ; Shi et al., 2024 ). Roads within the study area were classified into main and secondary roads. The shortest distance of the sampling points to roads classified into two categories was calculated in meters. Total road length within 150 m, 300 m, and 500 m buffer areas: The length of roads within buffer zones has a determining effect on pollutants such as SO 2 , NO 2 , and PM (Beelen et al., 2013 ; Chen et al., 2020 ; Mikeš et al., 2023 ). To represent the spatial distribution of road-related pollutants, total road lengths in buffer zones at different scales were calculated. Elevation (m): Topography is an influential factor in the transport and accumulation of pollutant gases. It is necessary to include the elevation variable in LUR models, especially because the dispersion of pollutants is limited in valleys. (Wong et al., 2021 ; Mikeš et al., 2023 ). Digital elevation data were downloaded from the Shuttle Radar Topography Mission (SRTM) web page (USGS, 2025 ). Annual average land surface temperature (LST): LST plays a role in pollutant accumulation. Therefore, it has been widely used in remote sensing-based LUR studies to explain the heat island effect and related pollutant distribution (Chen et al., 2020 ; Mikeš et al., 2023 ; Shi et al., 2024 ). Using Landsat 8 data, the annual average land surface temperature of the study area was calculated. Population density (persons/hectare): Population density is an indirect indicator of energy consumption, heating-related emissions, and traffic activities in residential areas. In many studies, population density has shown significant relationships with pollutant concentrations (Beelen et al., 2013 ; Mikees et al., 2023; Shi et al., 2024 ). Neighborhood-based population data were obtained from the Turkish Statistical Institute (TÜİK, 2025 ). Neighborhood boundary data were obtained from Karabük Municipality. Population density was calculated at the neighborhood level using neighborhood boundaries and population data. Distance to industrial facilities (m) and industrial facility density: One of the most important sources of air pollutants in Karabük is industrial facilities. Therefore, the distance to industrial facilities and facility density variables were included in the model. Studies in the literature, especially in heavy industrial regions, have found that proximity to industrial facilities is highly associated with SO 2 and PM concentrations (Beelen et al., 2013 ; Chen et al., 2020 ; Mikeš et al., 2023 ; Yücer et al., 2023 ). The shortest distance of sampling points to industrial areas in vector format was calculated. While creating the density data, the density map of industrial areas converted to point format was obtained in raster format with a resolution of 30 m × 30 m. Total distance to high-traffic intersections: Intersections are areas where emissions are high because vehicles accelerate and decelerate. Intersection density and distance to intersections have been used in LUR models to explain the spatial distribution of NO 2 and CO concentrations (Beelen et al., 2013 ). Nine major intersections with heavy traffic were identified. The total distance of the sample points to these intersections was calculated and included in the model. Fuel stations density data: It causes traffic congestion due to heavy vehicle entry and exit to fuel stations. Pollutant gases released in exhausts due to heavy traffic affect air quality. For this reason, fuel station density data have been included in air quality studies (Ribbean et al., 2016). The density distribution map of the petrol stations was created in raster format with a resolution of 30 m x 30 m. 2.3. Multiple Linear Regression LUR models have been widely used in recent years to model the spatial distribution of air pollutants. In LUR models, pollutant data obtained from a small number of air quality monitoring stations are associated with spatial data. Spatial data consists of data such as traffic density, industrial areas, road networks, population density, land use (Hoek et al., 2008 ). The LUR method has successfully modeled the distribution of air pollutants such as NO 2 , PM 2.5 and SO 2 at different scales (Beelen et al., 2013 ; Meng et al., 2015 ; de Hoogh et al., 2016 ; Liu et al., 2019 ). Different methods such as Linear, Nonlinear and machine learning can be used to create the LUR model. MLR is the most widely used linear method. In this method, pollutant concentration is defined as the dependent variable and land use, roads, meteorological data are defined as the independent variables (Briggs, 1997 ). MLR has several primary objectives. The first is to calculate the coefficient of explanatory power (R 2 ) between the dependent and independent variables. Another purpose is to estimate the values of the dependent variable using the independent variables. It is also determined which independent variable or variables affect the dependent variable and whether this effect is positive or negative. $$\:\text{Y}={{\beta\:}}_{0}+{{\beta\:}}_{1}{\text{X}}_{1}+{{\beta\:}}_{2}{\text{X}}_{2}+...+{{\beta\:}}_{\text{n}}{\text{X}}_{\text{n}}+{\epsilon\:}$$ 1 MLR regression is calculated according to the equation in Eq. 1 . In the equation, Y represents the dependent variable, X represents the independent variable, β represents the regression coefficients related to the independent variables, and β₀ is called the intercept, which shows the point where the regression line intersects the y-axis, or the constant term (Alpar 2014 ). 2.4. Geographically Weighted Regression GWR is a spatial regression model developed to overcome the constant coefficient limitation of classical multiple regression models. GWR generates separate coefficient values for each observation point. In this way, the spatial effect of the variables is reflected in the model. This approach is effective in eliminating spatial stationarity in the spatial modeling of air pollutants (Fotheringham et al., 2009 ). In air pollution modeling studies using GWR, parameter coefficients are not dependent on a single global value. It produces different coefficient values according to the spatial location of each sampling point (Morshed et al., 2024 ; Shen et al., 2024 ; Liu et al., 2025 ). In the GWR model, different regression coefficients are obtained for each prediction point using Eq. 2 . $$\:{\text{Y}}_{\text{i}}={{\beta\:}}_{0}({\text{u}}_{\text{i}},{\text{v}}_{\text{i}})+{\Sigma\:}{{\beta\:}}_{\text{k}}({\text{u}}_{\text{i}},{\text{v}}_{\text{i}})\:{\text{X}}_{\text{i}\text{k}}+{{\epsilon\:}}_{\text{i}}$$ 2 Yi: The observed value of each dependent variable in the study area, β₀: constant value, βk: parameter estimate of the k-th variable, Xik: covariance value of the k-th variable, (ui, vi): coordinates of the i-th point, ε: error term (Fotheringham et al., 2009 ). In GWR models, weighting of sampling points relative to each other is carried out using kernel models. Among the common kernel functions are Gaussian and bisquare. In the Gaussian kernel model, there are two different methods: fixed and adaptive. Bandwidth is another important parameter in the construction of GWR models. The bandwidth determines the number of neighboring points or the distance. The bandwidth is usually automatically selected according to the CV (cross-validation) or AICc (Akaike Information Criterion corrected) criteria (Fotheringham et al., 2009 ). 2.5. Assessment of model performances 5-fold cross-validation and validation with external data were used to validate the models. The results of both methods were evaluated using R², RMSE, and MAE validation criteria. 80% of the dataset was used in creating the models and 20% was used in the external data validation method. The predicted values obtained using the coefficients of the models were compared with the observed values and scatter plots were created. Cross-validation method is widely used to evaluate the results of MLR and GWR models. The cross-validation method is useful for testing the suitability of the created models to the training data. In addition, it measures the applicability of the model and tests its performance on different data subsets (Kuhn & Johnson, 2013 ). The cross-validation method is widely preferred to test the accuracy of models, especially in study areas where the number of stations is limited (Vienneau et al., 2013 ; Wang et al., 2013 ; Li et al., 2021 ). 3. Results and Discussions In this study, MLR and GWR models were applied to determine the spatial factors affecting the distribution of SO 2 concentration in Karabük province. The annual average of measurements obtained from four different ground stations is the dependent variable of the models. The independent variables are the twelve spatial indicators specified in Table 1 . Annual average values ​​obtained from ground stations were interpolated according to the study boundaries using the Kriging method. Thus, a raster format SO2 distribution map of the study area with a resolution of 30 m x 30 m was obtained (Fig. 2 ). ArcGIS 10.1 software was used to create the maps in the study. This approach was preferred to ensure the continuity of the pollutant within the study area. For the application of the models, 100 sampling points were randomly distributed within the study area. Eighty of the sampling points were used to construct the models, and 20 of them were used to test the accuracy of the models with external data. SO₂ distribution map interpolated using the Kriging method (µg/m³) In the MLR analysis, the main purpose is to identify the independent variables that explain the dependent variable at the highest rate. SPSS software was used for the implementation of the analysis. It was determined that 6 of the independent variables had a significant relationship with SO 2 concentration (Table 2 ). Sig. and VIF values should be examined for the significance of the relationship between SO 2 and independent variables. The Sig. value is expected to be less than 0.05 at the 95% confidence interval. In addition, a VIF value greater than 10 is an indication of multicollinearity among independent variables. The results obtained showed that there is no multicollinearity, meaning no high correlation among the independent variables. Table 2 Result of the MLR model Independent Variables Coefficent t value Sig. VIF Intercept 17.70 38.01 0.000 Distance to main road -4.20x10 − 4 -1.74 0.050 3.36 Buffer 500 9.62x10 − 5 4.74 0.000 1.80 Distance to industrial facility -6.92x10 − 4 -3.61 0.001 3.89 Industrial Density 7.93 9.07 0.000 5.28 Total distance to high-traffic intersections -9.44 -3.93 0.000 5.34 Fuel station density -1.42x10 − 3 -5.67 0.000 2.43 R² 0.85 Adjusted R² 0.83 According to the results of the MLR model, R² = 0.85 and adjusted R² = 0.83 values were obtained. This indicates that the independent variables explain approximately 83% of the change in SO₂ concentration. The R² values obtained reveal that the created model has a statistically strong prediction performance. When the coefficients of the independent variables are examined, it is seen that the industrial density (β = 7.93, p < 0.000) variable has the highest positive effect on SO 2 concentrations. This result shows that SO 2 concentration increases in regions where industrial facilities are spatially concentrated. This demonstrates that pollutants arising from the fuels used in industrial facilities and from steel production processes significantly negatively affect air quality. The distance to the industrial facility (β = -6.92×10⁻⁴, p < 0.001) variable has a negative effect on the SO 2 variable. This result showed that as the distance to industrial facilities increases, the concentration decreases, and as the distance decreases, the concentration increases. Proximity to main roads and road length within the 500 m buffer are also among important variables affecting SO 2 levels. The shortest distance to the main road variable (β = -4.20×10⁻⁴, p < 0.050) has a negative effect on SO 2 concentration. It is concluded that as the distance to the main roads increases, the SO 2 concentration decreases, and as the distance decreases, the SO 2 concentration increases. This situation shows the effect of vehicles on the main roads on SO 2 concentration. There is a positive relationship between total road length and SO 2 concentration in the 500 m circular buffer area (β = 9.62×10⁻⁵, p < 0.000). This situation causes the SO 2 concentration to increase as the total path length in the buffer zone increases. The total distance indicator to high-traffic intersections (β = -9.44, p < 0.01) has a negative effect on SO 2 concentration. As the total distance to the intersections increases, SO 2 concentration decreases. In the other case, as the total distance to the intersections decreases, the SO 2 concentration increases. This situation shows the effect of traffic density on SO 2 concentration. Fuel station density (β = -1.42×10⁻³, p < 0.001) was determined to have a negative coefficient. The findings obtained as a result of the MLR analysis revealed that industrial activities and the road network in Karabük are decisive on SO 2 concentrations. This situation indicates that industrial emissions have a strong impact on urban air quality. Table 3 Result of the GWR model Independent Variables Mean Std Min Max Distance to main road -8.1x10 − 4 4x10 − 4 -1.26x10 − 3 -1.8x10 − 4 Buffer 500 8.5x10 − 5 2.4x10 − 5 4.1x10 − 5 1.16x10 − 4 Distance to industrial facility -7.2x10 − 4 4.6x10 − 4 -1.28x10 − 3 0 Industrial Density 8.17 0.61 6.60 10.12 Total distance to high-traffic intersections -8.84 2.15 -14.05 -3.67 Fuel station density -1.49x10 − 3 6.7x10 − 4 -2.17x10 − 3 -4.5x10 − 4 R² 0.95 Adjusted R² 0.93 In applying the GWR analysis, variables that were significant in the MLR were included in the analysis. In the model parameter selection, the adaptive model from the Gaussian Kernel models was chosen. In the selection of bandwidth, the AICc (Akaike Information Criterion corrected) criteria, which are most compatible with the adaptive Kernel model, were applied. In the GWR analysis, separate coefficient values are produced for each sampling point. Table 3 presents the minimum, maximum, average, and standard deviation values of the coefficients for the variables. In the GWR analysis, the overall R² value of the model was obtained as 0.95 and the adjusted R² value as 0.93. These values indicate that the overall explanatory power of the model is high. This situation shows the high agreement between the observed and predicted SO₂ concentrations in the GWR model. According to the coefficient values, the buffer 500 and industrial density indicators have a positive effect on SO₂ concentration. It was determined that the variables distance to the main road, distance to the industrial facility, total distance to traffic-intensive intersections, and fuel station density have negative effects on SO₂ concentration. These results are similar to the MLR results. The negative effect of the shortest distance to main roads on SO₂ concentration indicates that as the distance from the main road increases, the concentration decreases, and as the distance to the main road decreases, the concentration increases. A similar situation is observed in the variable of the shortest distance to industrial facilities. It was concluded that as the distance to industrial facilities shortens, the SO 2 concentration increases, while the concentration decreases as the distance increases. The Buffer 500 variable has a positive effect on concentration. This means that increasing the path length in the buffer zone causes an increase in SO 2 concentration. The industrial density variable is another indicator that has a positive effect on SO 2 concentration. This means that high concentrations are observed in dense industrial areas. Total distance to high-traffic intersections has a negative effect on SO 2 concentration. As the total distance of the sampling points to the intersections with high traffic increases, the SO 2 concentration tends to decrease. The concentration increased as the total distance to the intersections decreased. In terms of standard deviation values, industrial density and total distance to high-traffic intersections are the variables with the highest variance. This means that the coefficients of the two indicators vary more between regions. 3.1. Performance assessment Testing the accuracy of the models created in LUR models is an important step. To test the accuracy of the MLR and GWR models, 5-fold cross-validation and external data validation methods were used (Table 4 ). Table 4 Validation results of MLR and GWR models Model Validation Method R² RMSE MAE MLR 5-fold Cross Validation 0.85 0.67 0.53 MLR External data validation 0.92 0.54 0.47 GWR 5-fold Cross Validation 0.95 0.40 0.34 GWR External data validation 0.94 0.45 0.39 The results of the MLR model according to the 5-fold cross-validation method are R² = 0.85, RMSE = 0.67 and MAE = 0.53. Validation results using external data showed improvement and yielded R² = 0.92, RMSE = 0.54 and MAE = 0.47. According to both validation methods, the MLR model achieved successful results. The higher values ​​in the validation with external data indicate that the model may have been affected by variances during training. The results of the GWR model according to the 5-fold cross-validation method are R² = 0.95, RMSE = 0.40 and MAE = 0.34. The obtained values ​​show that the GWR model explains the distribution of SO2 concentration with high accuracy. The values ​​of R² = 0.94, RMSE = 0.45 and MAE = 0.39 obtained in the external data validation method show that the model is also successful according to this validation method. According to the 5-fold cross-validation and external data validation methods, both models produced statistically strong results. The GWR method produces different coefficients for each sample point. As a result, the spatial effects of independent variables affecting SO 2 concentration are reflected more accurately. The use of a constant coefficient for the entire study area in the MLR model caused limitations in explaining local variations. As a result, considering spatial dependency in estimating SO 2 concentration significantly improved model performance. Scatter plot of observed and predicted SO 2 concentrations. The scatter plot showing the relationship between the observed and predicted values according to two different validation methods of the MLR and GWR models is given in Fig. 3 . The scatter plot demonstrates the strength of the relationship between the observed and predicted SO 2 concentrations. In the MLR model, although the 5-fold cross-validation results generally showed a good distribution, a tendency of underestimation was observed between 14–16 µg/m³ values and at some higher values. In the external data validation, it is observed that the values were partially clustered around the axis line. In the GWR model, both the 5-fold cross-validation and external validation plots produced slightly better results compared to the MLR model. This indicates that the GWR model exhibited more consistent results according to both validation methods. It was also revealed in Table 4 that numerically, the GWR model showed better performance than the MLR model in terms of R², RMSE, and MAE values. Spatial distribution map of observed (a) and predicted (b: MLR, c: GWR) SO 2 concentrations In Fig. 4 , in the distribution of observed (a) SO₂ concentration, high pollution was observed in the south and low values in the north. In the MLR (b) prediction map, although the general distribution of SO₂ concentration is similar, the model estimated the areas with the highest concentration in the south at lower levels (Maximum: 19.36 µg/m³). This situation arises due to the spatial stationarity in the MLR model. Therefore, MLR could not fully represent local differences. The GWR (c) prediction map shows high spatial similarity with the observed distribution. The high pollution area in the south and the low pollution areas in the north overlap with the observed distribution in terms of spatial intensity level. This result shows that the GWR model successfully reflects the spatial heterogeneity in the study area. Local R² distribution map in the GWR analysis The Local R² map in Fig. 5 shows how the spatial fit of the GWR model varies across the regions within the study area. Since the Local R² value is calculated by taking into account the observations around each sampling point, it indicates the local level fit of the model. The Local R² value was obtained with maximum 0.92, minimum 0.82, and average 0.87, indicating that the overall explanatory power of the model is high. This result is expected in the GWR analysis and shows that the explanatory power of the model varies spatially. In places with high Local R² (0.92) values, the model explains the distribution of SO₂ concentration very well, while in places where the Local R² (0.82) value is lower, the fit is lower. When the map showing the Local R² distribution is examined, it is understood that the model fit is higher in the south of the study area. A partial decrease in model fit was observed towards the north. This situation may be due to the lower number of stations in the northern region. Residual distribution map of the GWR analysis The Residual map in Fig. 6 shows the spatial error distribution between the observed and predicted values of the GWR model. The positive values on the map indicate the places where the model underestimates, while the negative values indicate the places where the model overestimates. It is seen that the residual values are generally in the range of ± 1 µg/m³. This shows that the predicted values are very close to the observed values. The random spatial distribution of errors is evidence that there is no spatial autocorrelation. In addition, it is seen that there are a few clustered points on the map. The blue areas in the northeast indicate that the model slightly overestimated in these regions. Similarly, in the red areas in the southeast, the model underestimated the SO₂ concentration. However, this situation is not at a level that would affect the overall representation of the model. 4. Conclusion In this study, MLR and GWR methods were used to model the spatial distribution of SO₂ concentrations in Karabük, and the results were compared. Roads, industrial areas, temperature, elevation, intersections, and fuel station data were selected as independent variables for the models. Both models were evaluated using 5-fold cross-validation and external data validation methods. Although MLR achieved a generally good result, it could not fully reveal spatial heterogeneity due to the fixed coefficient assumption. MLR demonstrated good predictive power in both the 5-fold cross-validation (R² = 0.85, RMSE = 0.67, MAE = 0.53) and external data validation methods (R² = 0.92, RMSE = 0.54, MAE = 0.47). In MLR-based LUR studies in the literature, R² values have generally been obtained in the range of 0.70–0.85 (Mikeš et al., 2023 ; Torres et al., 2024 ; Siu et al., 2025 ). In this regard, the findings obtained are consistent with international studies. The GWR model showed better performance than the MLR model in both the 5-fold cross-validation (R² = 0.95, RMSE = 0.40, MAE = 0.34) and external data validation methods (R² = 0.94, RMSE = 0.45, MAE = 0.39). This situation reveals that models based on the fixed coefficient assumption are insufficient in urban areas where industrial activities are intense (Fotheringham et al., 2009 ; Morshed et al., 2024 ; Liu et al., 2025 ). In studies conducted especially in industrial regions related to the GWR model, high explanatory values have been obtained. Cheng et al., ( 2024 ) determined the R² value of the GWR model as 0.93 in industrial cities of China. Morshed et al., ( 2024 ) obtained 10–15% higher accuracy in SO₂ GWR results compared to MLR in Bangladesh. Distribution of industrial areas within the study area. GWR prediction map (a) and residual distribution map (b) In the GWR prediction map (a) in Fig. 7 , it is clearly observed that SO₂ concentration is higher in areas where industrial facilities are concentrated. In the study area where iron and steel plants are concentrated, it is observed that the maximum SO 2 concentration is around these industrial facilities. This situation shows that pollutants emitted from industrial facilities play a decisive role in air quality. This finding obtained in the study is consistent with the high levels of SO 2 concentrations detected in industrial areas in the literature (Shen et al., 2024 ; Liu et al., 2025 ). The residual map (b) in Fig. 7 shows the spatial distribution of prediction errors of the GWR model. The residual values showed a distribution of approximately ± 1 µg/m³. The lower errors in the areas where industrial zones are located indicate that the model made accurate predictions in these regions. On the other hand, there are negative residual values in the north. It is thought that this situation arises as a result of the low number of stations in the region. From the residual values, it was seen that the model could not fully reflect some local effects. This situation has been reported in the studies of Morshed et al., ( 2024 ), Cheng et al., ( 2024 ) and Hu et al., ( 2025 ) in which similar uncertainties were reported in areas with low station density in GWR analyses. The GWR model achieved better results than the MLR in both overall accuracy and modeling the spatial distribution of SO 2 concentration. However, the GWR model revealed the spatial distribution of SO 2 concentration in detail, revealing changes at the local scale. Declarations Competing Interests The authors have no relevant financial or non-financial interests to disclose. Funding The authors declare that no funds, grants, or other support were received during the preparation of this manuscript. Author Contribution All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by Emre Yücer. Data Availability The data sources used in this study can be downloaded from publicly available websites. References Aarthi, A., Gayathri, P., Gomathi, N. R., Kalaiselvi, S., & Gomathi, V. (2020). Air quality prediction through regression model. International Journal of Scientific and Technology Research , 9 (3), 923–928. Alpar. R. (2014). Uygulamalı İstatistik ve Geçerlilik-Güvenirlik: SPSS’de Çözümleme Adımları İle Birlikte. 3rd edition. Detay Yayıncılık Ankara. Beelen, R., Hoek, G., Vienneau, D., et al. (2013). Development of NO₂ and NOx land-use regression models for estimating air pollution exposure in 36 study areas in Europe-The ESCAPE project. Atmospheric Environment , 72, 10–23. https://doi.org/10.1016/j.atmosenv.2013.02.037 . Briggs, D. J. (1997). Mapping urban air pollution using GIS: a regression-based approach. International Journal of Geographical Information Science , 11 (7), 699–718. https://doi.org/10.1080/136588197242158 . Chen, T. H., Hsu, Y. C., Zeng, Y. T., et al. (2020). A hybrid kriging/land-use regression model with Asian culture-specific sources to assess NO2 spatial-temporal variations. Environmental Pollution , 259, 113875. https://doi.org/10.1016/j.envpol.2019.113875 . Cheng, S., Zhang, G., Yang, X., & Lei, B. (2024). A multiscale geographically weighted regression kriging method for spatial downscaling of satellite-based ozone datasets. Frontiers in Environmental Science , 11, 1267752. https://doi.org/10.3389/fenvs.2023.1267752 . Comber, A., Brunsdon, C., Charlton. M., et al. (2020). The GWR route map: a guide to the informed application of Geographically Weighted Regression. arXiv preprint arXi , 2004.06070. de Hoogh, K., Gulliver, J., van Donkelaar, A., et al. (2016). Development of West-European PM2.5 and NO2 land-use regression models incorporating satellite-derived and chemical transport modelling data. Environmental Research , 151, 1–10. https://doi.org/10.1016/j.envres.2016.07.00 . Eeftens, M., Beelen, R., De Hoogh, K., et al. (2012). Development of land use regression models for PM2. 5, PM2. 5 absorbance, PM10 and PMcoarse in 20 European study areas; results of the ESCAPE project. Environmental science & technology , 46(20), 11195–11205. https://doi.org/10.1021/es301948k . Fotheringham, A. S., Brunsdon, C., & Charlton, M. (2009). Geographically weighted regression. The Sage handbook of spatial analysis , 1, 243–254. Fotheringham, A. S., Yang, W., & Kang, W. (2017). Multiscale geographically weighted regression (MGWR). Annals of the American Association of Geographers , 107(6), 1247–1265. https://doi.org/10.1080/24694452.2017.1352480 . Guo, B., Wang, X., Zhang, D., Pei, L., Zhang, D., & Wang, X. (2020). A land use regression application into simulating spatial distribution characteristics of particulate matter (PM 2.5 ) concentration in city of Xi'an, China. Polish Journal of Environmental Studies , 29(6), 4065–4076. https://doi.org/10.15244/pjoes/118426 Hoek, G., Beelen, R., de Hoogh, K., Vienneau, D., Gulliver, J., Fischer, P., & Briggs, D. (2008). A review of land-use regression models to assess spatial variation of outdoor air pollution. Atmospheric Environment , 42(33), 7561–7578. https://doi.org/10.1016/j.atmosenv.2008.05.057 Hu, K., Zhang, X., Yang, X., & Yu, M. (2025). A study on the spatial distribution of life expectancy and its air pollution factors in China based on geographically weighted regression. Frontiers in Public Health , 13, 1565744. https://doi.org/10.3389/fpubh.2025.1565744 Huang, L., Zhang, C., & Bi, J. (2017). Development of land use regression models for PM2. 5, SO2, NO2 and O3 in Nanjing, China. Environmental research , 158, 542–552. https://doi.org/10.1016/j.envres.2017.07.010 Karabük Municipality, (2025). Retrived September 25, 2025, from https://www.karabuk.bel.tr/icerik.asp?i_id=39 . Karabük, TSO. (2025). Karabük Chamber of Commerce And Industry, Retrived September 25, 2025, from https://karabuktso.org.tr/blog-makale/kardemirin-ve-karabukun-dogusu-3-nisan Kerckhoffs, J., Khan, J., Hoek, G., Yuan, Z., Ellermann, T., Hertel, O. et al. (2022). Mixed-effects modeling framework for Amsterdam and Copenhagen for outdoor NO2 concentrations using measurements sampled with google street view cars. Environmental science & technology , 56(11), 7174–7184. https://doi.org/10.1021/acs.est.1c05806 Kuhn, M., & Johnson, K. (2013). Applied predictive modeling. Springer, New York. Li, Z., Ho, K. F., Chuang, H. C., & Yim, S. H. L. (2021). Development and intercity transferability of land-use regression models for predicting ambient PM 10, PM 2.5, NO 2 and O 3 concentrations in northern Taiwan. Atmospheric Chemistry and Physics , 21(6), 5063–5078. https://doi.org/10.5194/acp-21-5063-2021 Liu, Z., Guan, Q., Luo, H., Wang, N., Pan, N., Yang, L., Xiao, S., & Lin, J. (2019). Development of land-use regression model and health risk assessment for NO₂ in different functional areas: a case study of Xi’an, China. Atmospheric Environment , 213, 515–525. https://doi.org/10.1016/j.atmosenv.2019.06.044 Liu, X., Wu, W., Zou, H., Pan, C., Liu, M., & Li, H. (2025). Unveiling seasonal disparities in urban form: Exploring scale and intensity effects on PM2.5 concentrations. Urban Climate , 59, 102319. https://doi.org/10.1016/j.uclim.2025.102319 Ma, X., Zou, B., Deng, J., Gao, J., Longley, I., Xiao, S. et al. (2024). A comprehensive review of the development of land use regression approaches for modeling spatiotemporal variations of ambient air pollution: A perspective from 2011 to 2023. Environment international , 183, 08430. https://doi.org/10.1016/j.envint.2024.108430 Meng, X., Chen, L., Cai, J., Zou, B., Wu, C. F., Fu, Q., Zhang, Y., Liu, Y., & Kan, H. (2015). A land-use regression model for estimating the NO₂ concentration in Shanghai, China. Environmental Research , 137, 308–315. https://doi.org/10.1016/j.envres.2015.01.003 Meo, S. A., Shaikh, N., Alotaibi, M., AlWabel, A. A., & Alqumaidi, H. (2024). Effect of air pollutants particulate matter (PM2. 5, PM10), sulfur dioxide (SO2) and ozone (O3) on cognitive health. Scientific Reports , 14(1), 19616. https://doi.org/10.1038/s41598-024-70646-6 Mikeš, O., Sáňka, O., Rafajová, A., Vlaanderen, J., Chen, J., Hoek, G. et al. (2023). Development of historic monthly land use regression models of SO2, NOx and suspended particulate matter for birth cohort ELSPAC. Atmospheric Environment , 30, 119688. https://doi.org/10.1016/j.atmosenv.2023.119688 Morshed, S. R., Fattah, M. A., Kafy, A. A., Alsulamy, S., Almulhim, A. I., Shohan A. A. A., Khedher, K. M. (2024). Decoding seasonal variability of air pollutants with climate factors: A geostatistical approach using multimodal regression models for informed climate change mitigation. Environmental Pollution , 345, 123463. https://doi.org/10.1016/j.envpol.2024.123463 Orellano, P., Reynoso, J., & Quaranta, N. (2021). Short-term exposure to sulphur dioxide (SO2) and all-cause and respiratory mortality: A systematic review and meta-analysis. Environment international , 150, 106434. https://doi.org/10.1016/j.envint.2021.106434 Ribeiro, H., Almeida, S. M., Martins, E. M., & Silva, A. V. (2016). Petrol stations: An environmental risk factor. Environmental Monitoring and Assessment , 188, 183. https://doi.org/10.1007/s10661-016-5172-0 Shen, Y., de Hoogh, K., Schmitz, O., Clinton, N., Tuxen-Bettman, K., Brandt, J. et al. (2024). Monthly average air pollution models using geographically weighted regression in Europe from 2000 to 2019. Science of the Total Environment , 918, 170550. https://doi.org/10.1016/j.scitotenv.2024.170550 Shi, T., Dirienzo, N., Requia, W. J., Hatzopoulou, M. & Adams, M. D. (2020). Neighbourhood scale nitrogen dioxide land use regression modelling with regression kriging in an urban transportation corridor. Atmospheric Environment , 223, 117218. https://doi.org/10.1016/j.atmosenv.2019.117218 Shi, T., Zhang, Y., Yuan, X., Li, F., & Yan, S. (2024). Spatial Patterns and Determinants of PM2. 5 Concentrations: A Land Use Regression Analysis in Shenyang Metropolitan Area, China. Sustainability , 16(12), 5119. https://doi.org/10.3390/su16125119 Siu, T. K., Greene, C. S., & Fong, K. C. (2025). Identifying Surface Sulphur Dioxide (SO2) Monitoring Gaps in Saint John, Canada with Land Use Regression and Hot Spot Mapping. Atmospheric Environment , 121238. https://doi.org/10.1016/j.atmosenv.2025.121238 Torres, R. A. B., Ramos, R. V., Recto, B. A. B., & Tamondong, A. M. (2024). DEVELOPMENT OF LAND-USE REGRESSION MODELS FOR PARTICULATE MATTER ESTIMATION IN NATIONAL CAPITAL REGION, PHILIPPINES. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences , 48, 437–444. https://doi.org/10.5194/isprs-archives-XLVIII-4-W8-2023-437-2024 TÜİK. (2025). Turkish Statistical Institute. Retrieved July 24, 2025 form https://data.tuik.gov.tr/Kategori/GetKategori?p=Nufus-ve-Demografi-109 USGS. (2025). The United States Geological Survey. Retrieved July 24, 2025 from https://earthexplorer.usgs.gov/ Vienneau, D., de Hoogh, K., Bechle, M. J., Beelen, R., van Donkelaar, A., Martin, R. V. et al. (2013). Western European land-use regression incorporating satellite- and ground-based measurements of NO₂ and PM10. Environmental Science & Technology , 47(23), 13555–13564. https://doi.org/10.1021/es403089q Wang, M., Beelen, R., Basagaña, X., Becker, T., Cesaroni, G., de Hoogh, K. et al. (2013). Evaluation of land-use regression models for NO₂ and particulate matter in 20 European study areas: the ESCAPE project. Environmental Science & Technology , 47(9), 4357–4364. https://doi.org/10.1021/es305129t Wang, J., &Xu, H. (2021). A novel hybrid spatiotemporal land use regression model system at the megacity scale. Atmospheric Environment , 244, 117971. https://doi.org/10.1016/j.atmosenv.2020.117971 Wei, Q., Zhang, L., Duan, W., & Zhen, Z. (2019). Global and geographically and temporally weighted regression models for modeling PM2. 5 in Heilongjiang, China from 2015 to 2018. International journal of environmental research and public health , 16(24), 5107. https://doi.org/10.3390/ijerph16245107 Wong, P. Y., Lee, H. Y., Chen, Y. C., Zeng, Y. T., Chern, Y. R., Chen, N. T. et al. (2021). Using a land use regression model with machine learning to estimate ground level PM2.5. Environmental Pollution , 277, 116846. https://doi.org/10.1016/j.envpol.2021.116846 Xuan, W., Zhang, F., Zhou, H., Du, Z., & Liu, R. (2021). Improving geographically weighted regression considering directional nonstationary for ground-Level PM2.5 estimation. ISPRS International Journal of Geo-Information , 10(6), 413. https://doi.org/10.3390/ijgi10060413 Yang, X., Wang, S., Zhang, W., Zhan, D., & Li, J. (2017). The impact of anthropogenic emissions and meteorological conditions on the spatial variation of ambient SO2 concentrations: A panel study of 113 Chinese cities. Science of the Total Environment , 584, 318–328. https://doi.org/10.1016/j.scitotenv.2016.12.145 Yücer, E., Erener, A., & Sarp, G. (2023). A land use regression model to estimate ambient concentrations of PM10 and SO2 in İzmit, Turkey. Journal of the Indian Society of Remote Sensing , 51(6), 1329–1341. http://dx.doi.org/10.1007/s12524-023-01704-1 Zhou, Q., Wang, C., & Fang, S. (2019). Application of geographically weighted regression (GWR) in the analysis of the cause of haze pollution in China. Atmospheric Pollution Research , 10(3), 835–846. https://doi.org/10.1016/j.apr.2018.12.012 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Published Journal Publication published 04 Mar, 2026 Read the published version in Pure and Applied Geophysics → Version 1 posted Editorial decision: Revision requested 28 Jan, 2026 Reviews received at journal 27 Jan, 2026 Reviewers agreed at journal 09 Jan, 2026 Reviewers agreed at journal 08 Jan, 2026 Reviewers agreed at journal 07 Jan, 2026 Reviews received at journal 06 Jan, 2026 Reviewers agreed at journal 05 Jan, 2026 Reviewers agreed at journal 05 Jan, 2026 Reviewers invited by journal 05 Jan, 2026 Editor assigned by journal 29 Dec, 2025 Submission checks completed at journal 26 Dec, 2025 First submitted to journal 25 Dec, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8450115","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":570178081,"identity":"168a118b-1505-41da-af3b-f7074579d096","order_by":0,"name":"Emre Yücer","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABAUlEQVRIiWNgGAWjYFAD9jYQeYBBAkhKEKeF5xiKFgMitEikEamFX+zsw8e8bXZy5jOfJX6uqLmTOLOB+eBtHoY/+bi0SM5ONzbmbUs2lrmddljyzLFnibMZ2JKteRgMLBtwaDG4ncYmzdt2IHGGdHqDZAPb4cR5DDxm0kAtOF1mfzuN/TdQS/0MyePNPxv+gbTwf8OrxUA6jY0ZqCVBQoLtmGRj22Ggw3jY8GqRuJ3GLDnnXLLhDJ60NMvGvmfGM5vZjC3nGBjj1MI/O43xw5syO3kJ9mPGNxu+3ZGdcbz54Y03FXJ4I4aJB4XLDHYwPg0MDIw/8MuPglEwCkbBSAcAmt9PtLLF2aUAAAAASUVORK5CYII=","orcid":"","institution":"Karabük University","correspondingAuthor":true,"prefix":"","firstName":"Emre","middleName":"","lastName":"Yücer","suffix":""}],"badges":[],"createdAt":"2025-12-25 16:23:08","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8450115/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8450115/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s00024-026-03950-z","type":"published","date":"2026-03-04T15:58:08+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":99814008,"identity":"f7b126f6-cdf8-4b0e-9f0a-37bc428c0788","added_by":"auto","created_at":"2026-01-08 14:40:11","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1997116,"visible":true,"origin":"","legend":"","description":"","filename":"Manuscript.docx","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/9088ca71308e6d2dd3ae9134.docx"},{"id":99813712,"identity":"8e8904e2-feda-4809-9516-eac3af993545","added_by":"auto","created_at":"2026-01-08 14:39:32","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3628,"visible":true,"origin":"","legend":"","description":"","filename":"cb9a4c327ad84bbaad16f0aebe7eec97.json","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/b2c349826043bc6657409b48.json"},{"id":99814079,"identity":"2cefb2a7-d27e-449a-83c3-f92f22bd52be","added_by":"auto","created_at":"2026-01-08 14:40:16","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":135129,"visible":true,"origin":"","legend":"","description":"","filename":"cb9a4c327ad84bbaad16f0aebe7eec971enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/f4f2811e976c9d7194effa27.xml"},{"id":99813539,"identity":"26d5cfb6-b1f8-473c-92fb-07170b1a9f70","added_by":"auto","created_at":"2026-01-08 14:39:14","extension":"jpeg","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":945185,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/1b2f213014ed34f311424921.jpeg"},{"id":99813768,"identity":"c912a2b2-0a3c-4093-a46e-ba4b94896b90","added_by":"auto","created_at":"2026-01-08 14:39:46","extension":"jpeg","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":136181,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/13ae0ff7dcb2169f1c74cc06.jpeg"},{"id":99814009,"identity":"2219ac84-1798-427f-a066-b4209f3ad1d2","added_by":"auto","created_at":"2026-01-08 14:40:12","extension":"jpeg","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":162186,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/8c429b7e494687891ffd5142.jpeg"},{"id":99813717,"identity":"f313c19d-dac5-4612-a47b-785ad457ea33","added_by":"auto","created_at":"2026-01-08 14:39:33","extension":"jpeg","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":317816,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/9b0a9eeb150e5ad1f93688a7.jpeg"},{"id":99813595,"identity":"d3c92e35-a110-4081-a765-df9049789627","added_by":"auto","created_at":"2026-01-08 14:39:25","extension":"jpeg","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":128026,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/a18125c36f4e56fe4270b12c.jpeg"},{"id":99813715,"identity":"e6956154-dc0b-4889-a976-e19173e2593b","added_by":"auto","created_at":"2026-01-08 14:39:33","extension":"jpeg","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":148457,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/a2d4b686fb21bcafdcd9957c.jpeg"},{"id":99813846,"identity":"c3aad0e4-c2c0-47df-b96d-210db3fa74b0","added_by":"auto","created_at":"2026-01-08 14:39:54","extension":"jpeg","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":246456,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage7.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/4f39a2316164b4443e298b85.jpeg"},{"id":99813762,"identity":"1cb98533-fa2a-4a0d-af71-eb3b2d9a5034","added_by":"auto","created_at":"2026-01-08 14:39:44","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":600328,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/c32c533ac5a1898f17dbc5d7.png"},{"id":99814076,"identity":"5f371ef4-bc53-4944-be09-682cce6206f8","added_by":"auto","created_at":"2026-01-08 14:40:15","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":44307,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/ea1738d3fcd9ade62b5d5b32.png"},{"id":99813993,"identity":"54d0f3c2-964a-4da2-b3d8-35d0a8dd4fd8","added_by":"auto","created_at":"2026-01-08 14:40:07","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":50930,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/6fca27ade5dc7106657e842b.png"},{"id":99813766,"identity":"0326d8fc-7f9a-412b-b2e3-379045fa6257","added_by":"auto","created_at":"2026-01-08 14:39:46","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":74926,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/a7962f1dbf37b208d6bed716.png"},{"id":99813714,"identity":"17412c9f-b5a7-44a9-b979-e92888e0a91e","added_by":"auto","created_at":"2026-01-08 14:39:33","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":39153,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/a429660d3c03c07c8580b01b.png"},{"id":99813847,"identity":"a96dcb81-7ba1-4490-82f0-7c8636ee0003","added_by":"auto","created_at":"2026-01-08 14:39:54","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":52255,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/baf87d1da30355963e346efd.png"},{"id":99813843,"identity":"bf118b04-26a7-442a-8c57-bb7373361bf1","added_by":"auto","created_at":"2026-01-08 14:39:53","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":69077,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/b1fcda72b2eee9636e945894.png"},{"id":99813970,"identity":"4376a0a7-d195-4cf8-b02b-d248f75431f3","added_by":"auto","created_at":"2026-01-08 14:40:05","extension":"xml","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":134336,"visible":true,"origin":"","legend":"","description":"","filename":"cb9a4c327ad84bbaad16f0aebe7eec971structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/6cb9721860f52457630073ba.xml"},{"id":99813590,"identity":"543c26ea-f6d5-4618-829f-ee09dae016f0","added_by":"auto","created_at":"2026-01-08 14:39:22","extension":"html","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":140831,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/de0f6a010dd6b7d720d809fc.html"},{"id":99813981,"identity":"a193da4f-ce59-4700-a5e4-8f40881e3e59","added_by":"auto","created_at":"2026-01-08 14:40:06","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":945185,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eStudy area\u003c/em\u003e\u003c/p\u003e","description":"","filename":"image1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/7ac70ca06710dce543e2ed42.jpeg"},{"id":100374438,"identity":"0f02ed35-da40-4d2d-8f32-0f696001e25a","added_by":"auto","created_at":"2026-01-16 08:28:14","extension":"jpeg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":136181,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eSO₂ distribution map interpolated using the Kriging method (µg/m³)\u003c/em\u003e\u003c/p\u003e","description":"","filename":"image2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/2191c135c933d8f01fbb5345.jpeg"},{"id":99813776,"identity":"5c762c08-f664-4a5d-aea8-e595a065669c","added_by":"auto","created_at":"2026-01-08 14:39:48","extension":"jpeg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":162186,"visible":true,"origin":"","legend":"\u003cp\u003eScatter plot of observed and predicted SO\u003csub\u003e2 \u003c/sub\u003econcentrations.\u003c/p\u003e","description":"","filename":"image3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/379254f45e95e572cd86ed56.jpeg"},{"id":99813906,"identity":"65760783-b0d7-4dd7-8b00-38436495d0f4","added_by":"auto","created_at":"2026-01-08 14:40:01","extension":"jpeg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":317816,"visible":true,"origin":"","legend":"\u003cp\u003eSpatial distribution map of observed (a) and predicted (b: MLR, c: GWR) SO\u003csub\u003e2 \u003c/sub\u003econcentrations\u003c/p\u003e","description":"","filename":"image4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/cce788d534aa0fbcb241a2ae.jpeg"},{"id":99813614,"identity":"cb8cb551-a7ca-498b-83c5-8d66465f91ac","added_by":"auto","created_at":"2026-01-08 14:39:29","extension":"jpeg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":128026,"visible":true,"origin":"","legend":"\u003cp\u003eLocal R² distribution map in the GWR analysis\u003c/p\u003e","description":"","filename":"image5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/15bca766a6f129716224bc1b.jpeg"},{"id":99813601,"identity":"42a79641-28eb-4ff5-b677-0aa7878bacfb","added_by":"auto","created_at":"2026-01-08 14:39:27","extension":"jpeg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":148457,"visible":true,"origin":"","legend":"\u003cp\u003eResidual distribution map of the GWR analysis\u003c/p\u003e","description":"","filename":"image6.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/0ba9f755d29b7548d8d6d2fd.jpeg"},{"id":100356281,"identity":"422320ed-af13-40d1-b366-053c67b033d7","added_by":"auto","created_at":"2026-01-16 06:59:33","extension":"jpeg","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":246456,"visible":true,"origin":"","legend":"\u003cp\u003eDistribution of industrial areas within the study area. GWR prediction map (a) and residual distribution map (b)\u003c/p\u003e","description":"","filename":"image7.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/33118d7dfb90b9706ffaef85.jpeg"},{"id":104250650,"identity":"c9afa3f3-e4fa-4b57-8f69-b26e9fa21e39","added_by":"auto","created_at":"2026-03-09 16:04:15","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2831867,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8450115/v1/76232e10-56ac-4a11-aaca-7ff7ba80a1f4.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Modeling and comparative analysis of spatial distribution of SO 2 concentration using MLR and GWR models: A case study of Karabük, Türkiye","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eWith urbanization, industrialization, and the increasing number of vehicles, fossil fuel use is also rapidly increasing. Fossil fuels are used in many areas, from energy production to industrial facilities. Pollutants released into the atmosphere as a result of fossil fuel consumption cause increased sulfur dioxide (SO₂) concentrations. SO\u003csub\u003e2\u003c/sub\u003e causes respiratory irritation, coughing, wheezing, and bronchial infections. Longterm exposure leads to chronic respiratory diseases and cardiovascular diseases. Studies in this area have shown that even low levels of exposure have adverse effects on general health. (Orellano et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Meo et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eLand Use Regression (LUR) models are used to estimate the spatial distribution of atmospheric pollutants. In the models, concentration values ​​obtained from ground stations are used to estimate areas where measurements cannot be made. In order to obtain the prediction model and coefficients, the values ​​obtained from the measurement stations are associated with spatial data. Spatial data such as traffic density, distribution of industrial facilities, land use, road network and population density are widely used. In addition, meteorological data such as temperature, humidity and pressure are included in the models (Eeftens et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Ma et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Different methods such as linear, nonlinear, and machine learning are used in LUR models. Multiple linear regression (MLR) and Geographically Weighted regression (GWR) are the most commonly used linear methods (Guo et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Shi et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Xuan et al., \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Kerckhoffs et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Correct variable selection, spatial distribution of variables and number of measurement stations are the parameters that affect the success of the model. In addition, the choice of method used in model validation is also important (Aarthi et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Torres et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn MLR analysis, there is one dependent variable and multiple independent variables. It is assumed that there is a linear relationship between the dependent variable and the independent variables. Modelling the dispersion of air pollutants can be difficult when the number of ground stations is small or because of their dispersed location. However, successful results have been obtained in LUR studies at different scales for modeling the spatial distribution of air pollutants (Aarthi et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Wang and Xu \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Mikeš et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Siu et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Huang et al., (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2017\u003c/span\u003e) used data from the national monitoring station to model the distribution of PM₂.₅, NO₂ and SO₂ pollutants in Nanjing, China. They used road, land use, meteorological and topographic data for independent variables. In their results, they found an average R\u0026sup2; value of 0.83 for SO2. Y\u0026uuml;cer et al., (\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) constructed a LUR model using annual averages of SO₂ and PM₁₀ concentrations in the city of Izmit, T\u0026uuml;rkiye. Indicators such as main road distance, industrial area density, and population density were selected as independent variables in the model. The R\u0026sup2; value, representing the explanatory power of the model, was found to be 0.88. The correlation coefficient between observed and predicted values in model validation was found to be 0.90. In a study examining the SO₂ distribution in Saint John, Canada, Siu et al., (\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) obtained SO₂ concentrations from ground monitoring stations and TROPOMI data. As a result of the 10-fold cross-validation method, they determined the classification accuracy as 83% and the R\u0026sup2; value of the daily dataset as 0.46.\u003c/p\u003e \u003cp\u003eGlobal regression models assume that the spatial relationship between the dependent and independent variables is constant across the study area. However, since the city's structure and emission sources do not show the same distribution in every region within the city, this relationship may differ spatially. GWR eliminates spatial stationarity by conducting local regression for each location. It reveals how the variable coefficients change across space (Fotheringham et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Fotheringham et al., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Comber et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eWei et al., (\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) examined the relationship between PM\u003csub\u003e2.5\u003c/sub\u003e and SO\u003csub\u003e2\u003c/sub\u003e, NO\u003csub\u003e2\u003c/sub\u003e, PM\u003csub\u003e10\u003c/sub\u003e, CO and O\u003csub\u003e3\u003c/sub\u003e in Heilongjiang, China, using linear mixed models (LMM), GWR and geographically and temporally weighted regression (GTWR) models. It has been found that GWR-based models achieve higher accuracy values ​​than traditional models. Shen et al., (\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) modeled monthly average air pollutant values ​​according to GWR and MLR analyses in their study on a European scale between 2000 and 2019. Models evaluated according to the 5-fold cross-validation method have yielded successful results. Morshed et al., (\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) used meteorological indicators as independent variables in the GWR and MLR models. The effect of meteorological variables on pollutant concentration was investigated. According to the results, GWR achieved more successful performance. Liu et al., (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) modeled the distribution of PM\u003csub\u003e2.5\u003c/sub\u003e concentration in 284 different cities in China using GWR and Multiscale Geographically Weighted Regression (MGWR) analyses. They found that MGWR was more successful than GWR in revealing the seasonal distribution of PM₂.₅. Studies in the literature have shown that industrial density, proximity to industrial areas, access to the road network, precipitation and wind variables are closely related to SO₂ distribution. GWR has been more successful in revealing local effects on the distribution of air pollutants. Especially in industrial cities, the distribution of SO2 concentration was revealed by the coefficients obtained at the local scale (Yang at al., 2017; Zhou et al., \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). The results obtained from these studies showed that GWR was successful in detecting local heterogeneity in the distribution of pollutant concentration.\u003c/p\u003e \u003cp\u003eIn this study, MLR and GWR analyses were used to model the spatial distribution of SO₂ in Karab\u0026uuml;k province. In the regression analyses, the annual average of SO2 concentrations in 2021 was defined as the independent variable. Spatial data such as industrial areas, roads, ground surface temperature, and altitude were selected as independent variables. The results obtained from the analyses were compared with similar MLR and GWR models in the literature. As a result of the analysis, the coefficients of the indicators, the coefficient of determination (R\u0026sup2;), and the Root Mean Square Error (RMSE) values were interpreted to explain the relationship between SO₂ and the independent variables. In addition, external data validation and 5-fold cross-validation methods were used to validate the models. The findings of the study will contribute to expanding practices related to SO₂. In addition, it aims to contribute to the development of spatial prediction tools for decision-makers in the management of air quality at the local scale.\u003c/p\u003e"},{"header":"2. Materials and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n\u003ch2\u003e2.1. Study area\u003c/h2\u003e\n\u003cp\u003eThe provincial center of Karab\u0026uuml;k, an industrial city located in the Western Black Sea Region of T\u0026uuml;rkiye, was selected as the study area (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e). Karab\u0026uuml;k is located between 40\u0026deg; 57' and 41\u0026deg; 34' northern latitudes and 32\u0026deg; 04' and 33\u0026deg; 06' eastern longitudes, and it has an area of approximately 4,145 km\u0026sup2;. The city center is located at an elevation of approximately 280 m above sea level. The topography of the city has a complex structure where valley and plateau areas coexist (Karab\u0026uuml;k Municipality, \u003cspan class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e\n\u003cp\u003eKarab\u0026uuml;k plays a decisive role in the industrial sector in its economic development. Karab\u0026uuml;k Iron and Steel Works (KARDEMİR), established in 1937, is T\u0026uuml;rkiye\u0026rsquo;s first integrated iron and steel production facility. With the establishment of KARDEMİR, the city entered a rapid industrialization process. Iron and steel production and its sub-sectors form the basis of the city's economic structure (Karab\u0026uuml;k TSO, \u003cspan class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e\n\u003cp\u003eMost of the industrial facilities are located in areas close to the city center. Pollutants such as SO₂, NOₓ and particulate matter, especially those originating from heavy industrial facilities, negatively affect the air quality of Karab\u0026uuml;k. Fossil fuels used in iron and steel production, rolling mills and foundries are also among the main sources of SO₂. Therefore, investigating the relationship between industrial facilities, roads, population density, and other spatial variables and air pollutant concentrations in Karab\u0026uuml;k is important for both environmental management policies and public health.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e2.2. Data Sets\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe dependent variable of the study, SO₂ data for the year 2021, was obtained from four different ground stations of the Ministry of Environment and Urbanization. These data, which are measured hourly by the National Air Quality Monitoring Stations, are published at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ewww.havaizleme.gov.tr\u003c/span\u003e\u003c/span\u003e. The measurement data were calculated as the annual average of daily values. The lowest number of days measured at the stations is 326.\u003c/p\u003e\n\u003cp\u003eThe independent variables used to explain the spatial distribution of SO₂ concentration were created from spatial indicators frequently preferred in similar LUR applications in the literature (Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e). Studies in the literature have found that pollutants, especially SO₂ concentration, exhibit strong relationships with variables such as industrial density and roads (Meng et al., \u003cspan class=\"CitationRef\"\u003e2015\u003c/span\u003e; Wang and Xu \u003cspan class=\"CitationRef\"\u003e2021\u003c/span\u003e; Mike\u0026scaron; et al., \u003cspan class=\"CitationRef\"\u003e2023\u003c/span\u003e; Siu et al., \u003cspan class=\"CitationRef\"\u003e2025\u003c/span\u003e). Data on roads, industrial facilities, high-traffic intersections, fuel stations, and neighborhood boundaries were obtained from the city map of Karab\u0026uuml;k Municipality.\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n\u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n\u003ctable id=\"Tab1\" border=\"1\"\u003e\u003ccaption\u003e\n\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n\u003cdiv class=\"CaptionContent\"\u003e\n\u003cp\u003e\u003cem\u003eIndependent variables\u003c/em\u003e\u003c/p\u003e\n\u003c/div\u003e\n\u003c/caption\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eDistance to main roads (m)\u003c/p\u003e\n\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eDistance to secondary roads (m)\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eTotal road length within 150 m buffer (m)\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eTotal road length within 300 m buffer (m)\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eTotal road length within 500 m buffer (m)\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eElevation (SRTM) (m)\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAnnual average land surface temperature - LST (\u0026deg;C)\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003ePopulation density (persons/hectare)\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eDistance to industrial facilities (m)\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eIndustrial density data\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eTotal distance to high-traffic intersections (m)\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eFuel stations density data\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003eDistance to main and secondary roads (m): Traffic-related pollutants are among the main causes of urban air pollution, especially in terms of SO\u003csub\u003e2\u003c/sub\u003e and NO\u003csub\u003ex\u003c/sub\u003e. Studies in the literature have shown that the distances of sampling points to major roads significantly affect pollutant concentrations (Hoek et al., \u003cspan class=\"CitationRef\"\u003e2008\u003c/span\u003e; Chen et al., \u003cspan class=\"CitationRef\"\u003e2020\u003c/span\u003e; Wong et al., \u003cspan class=\"CitationRef\"\u003e2021\u003c/span\u003e; Mike\u0026scaron; et al., \u003cspan class=\"CitationRef\"\u003e2023\u003c/span\u003e; Shi et al., \u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003e). Roads within the study area were classified into main and secondary roads. The shortest distance of the sampling points to roads classified into two categories was calculated in meters.\u003c/p\u003e\n\u003cp\u003eTotal road length within 150 m, 300 m, and 500 m buffer areas: The length of roads within buffer zones has a determining effect on pollutants such as SO\u003csub\u003e2\u003c/sub\u003e, NO\u003csub\u003e2\u003c/sub\u003e, and PM (Beelen et al., \u003cspan class=\"CitationRef\"\u003e2013\u003c/span\u003e; Chen et al., \u003cspan class=\"CitationRef\"\u003e2020\u003c/span\u003e; Mike\u0026scaron; et al., \u003cspan class=\"CitationRef\"\u003e2023\u003c/span\u003e). To represent the spatial distribution of road-related pollutants, total road lengths in buffer zones at different scales were calculated.\u003c/p\u003e\n\u003cp\u003eElevation (m): Topography is an influential factor in the transport and accumulation of pollutant gases. It is necessary to include the elevation variable in LUR models, especially because the dispersion of pollutants is limited in valleys. (Wong et al., \u003cspan class=\"CitationRef\"\u003e2021\u003c/span\u003e; Mike\u0026scaron; et al., \u003cspan class=\"CitationRef\"\u003e2023\u003c/span\u003e). Digital elevation data were downloaded from the Shuttle Radar Topography Mission (SRTM) web page (USGS, \u003cspan class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e\n\u003cp\u003eAnnual average land surface temperature (LST): LST plays a role in pollutant accumulation. Therefore, it has been widely used in remote sensing-based LUR studies to explain the heat island effect and related pollutant distribution (Chen et al., \u003cspan class=\"CitationRef\"\u003e2020\u003c/span\u003e; Mike\u0026scaron; et al., \u003cspan class=\"CitationRef\"\u003e2023\u003c/span\u003e; Shi et al., \u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003e). Using Landsat 8 data, the annual average land surface temperature of the study area was calculated.\u003c/p\u003e\n\u003cp\u003ePopulation density (persons/hectare): Population density is an indirect indicator of energy consumption, heating-related emissions, and traffic activities in residential areas. In many studies, population density has shown significant relationships with pollutant concentrations (Beelen et al., \u003cspan class=\"CitationRef\"\u003e2013\u003c/span\u003e; Mikees et al., 2023; Shi et al., \u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003e). Neighborhood-based population data were obtained from the Turkish Statistical Institute (T\u0026Uuml;İK, \u003cspan class=\"CitationRef\"\u003e2025\u003c/span\u003e). Neighborhood boundary data were obtained from Karab\u0026uuml;k Municipality. Population density was calculated at the neighborhood level using neighborhood boundaries and population data.\u003c/p\u003e\n\u003cp\u003eDistance to industrial facilities (m) and industrial facility density: One of the most important sources of air pollutants in Karab\u0026uuml;k is industrial facilities. Therefore, the distance to industrial facilities and facility density variables were included in the model. Studies in the literature, especially in heavy industrial regions, have found that proximity to industrial facilities is highly associated with SO\u003csub\u003e2\u003c/sub\u003e and PM concentrations (Beelen et al., \u003cspan class=\"CitationRef\"\u003e2013\u003c/span\u003e; Chen et al., \u003cspan class=\"CitationRef\"\u003e2020\u003c/span\u003e; Mike\u0026scaron; et al., \u003cspan class=\"CitationRef\"\u003e2023\u003c/span\u003e; Y\u0026uuml;cer et al., \u003cspan class=\"CitationRef\"\u003e2023\u003c/span\u003e). The shortest distance of sampling points to industrial areas in vector format was calculated. While creating the density data, the density map of industrial areas converted to point format was obtained in raster format with a resolution of 30 m \u0026times; 30 m.\u003c/p\u003e\n\u003cp\u003eTotal distance to high-traffic intersections: Intersections are areas where emissions are high because vehicles accelerate and decelerate. Intersection density and distance to intersections have been used in LUR models to explain the spatial distribution of NO\u003csub\u003e2\u003c/sub\u003e and CO concentrations (Beelen et al., \u003cspan class=\"CitationRef\"\u003e2013\u003c/span\u003e). Nine major intersections with heavy traffic were identified. The total distance of the sample points to these intersections was calculated and included in the model.\u003c/p\u003e\n\u003cp\u003eFuel stations density data: It causes traffic congestion due to heavy vehicle entry and exit to fuel stations. Pollutant gases released in exhausts due to heavy traffic affect air quality. For this reason, fuel station density data have been included in air quality studies (Ribbean et al., 2016). The density distribution map of the petrol stations was created in raster format with a resolution of 30 m x 30 m.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\n\u003ch2\u003e2.3. Multiple Linear Regression\u003c/h2\u003e\n\u003cp\u003eLUR models have been widely used in recent years to model the spatial distribution of air pollutants. In LUR models, pollutant data obtained from a small number of air quality monitoring stations are associated with spatial data. Spatial data consists of data such as traffic density, industrial areas, road networks, population density, land use (Hoek et al., \u003cspan class=\"CitationRef\"\u003e2008\u003c/span\u003e). The LUR method has successfully modeled the distribution of air pollutants such as NO\u003csub\u003e2\u003c/sub\u003e, PM\u003csub\u003e2.5\u003c/sub\u003e and SO\u003csub\u003e2\u003c/sub\u003e at different scales (Beelen et al., \u003cspan class=\"CitationRef\"\u003e2013\u003c/span\u003e; Meng et al., \u003cspan class=\"CitationRef\"\u003e2015\u003c/span\u003e; de Hoogh et al., \u003cspan class=\"CitationRef\"\u003e2016\u003c/span\u003e; Liu et al., \u003cspan class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e\n\u003cp\u003eDifferent methods such as Linear, Nonlinear and machine learning can be used to create the LUR model. MLR is the most widely used linear method. In this method, pollutant concentration is defined as the dependent variable and land use, roads, meteorological data are defined as the independent variables (Briggs, \u003cspan class=\"CitationRef\"\u003e1997\u003c/span\u003e). MLR has several primary objectives. The first is to calculate the coefficient of explanatory power (R\u003csup\u003e2\u003c/sup\u003e) between the dependent and independent variables. Another purpose is to estimate the values of the dependent variable using the independent variables. It is also determined which independent variable or variables affect the dependent variable and whether this effect is positive or negative.\u003c/p\u003e\n\u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\n\u003cdiv id=\"FileID_Equ1\" class=\"mathdisplay\"\u003e$$\\:\\text{Y}={{\\beta\\:}}_{0}+{{\\beta\\:}}_{1}{\\text{X}}_{1}+{{\\beta\\:}}_{2}{\\text{X}}_{2}+...+{{\\beta\\:}}_{\\text{n}}{\\text{X}}_{\\text{n}}+{\\epsilon\\:}$$\u003c/div\u003e\n\u003cdiv class=\"EquationNumber\"\u003e1\u003c/div\u003e\n\u003c/div\u003e\n\u003cp\u003eMLR regression is calculated according to the equation in Eq.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e. In the equation, Y represents the dependent variable, X represents the independent variable, \u0026beta; represents the regression coefficients related to the independent variables, and \u0026beta;₀ is called the intercept, which shows the point where the regression line intersects the y-axis, or the constant term (Alpar \u003cspan class=\"CitationRef\"\u003e2014\u003c/span\u003e).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\n\u003ch2\u003e2.4. Geographically Weighted Regression\u003c/h2\u003e\n\u003cp\u003eGWR is a spatial regression model developed to overcome the constant coefficient limitation of classical multiple regression models. GWR generates separate coefficient values for each observation point. In this way, the spatial effect of the variables is reflected in the model. This approach is effective in eliminating spatial stationarity in the spatial modeling of air pollutants (Fotheringham et al., \u003cspan class=\"CitationRef\"\u003e2009\u003c/span\u003e). In air pollution modeling studies using GWR, parameter coefficients are not dependent on a single global value. It produces different coefficient values according to the spatial location of each sampling point (Morshed et al., \u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003e; Shen et al., \u003cspan class=\"CitationRef\"\u003e2024\u003c/span\u003e; Liu et al., \u003cspan class=\"CitationRef\"\u003e2025\u003c/span\u003e). In the GWR model, different regression coefficients are obtained for each prediction point using Eq.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\n\u003cdiv id=\"Equ2\" class=\"Equation\"\u003e\n\u003cdiv id=\"FileID_Equ2\" class=\"mathdisplay\"\u003e$$\\:{\\text{Y}}_{\\text{i}}={{\\beta\\:}}_{0}({\\text{u}}_{\\text{i}},{\\text{v}}_{\\text{i}})+{\\Sigma\\:}{{\\beta\\:}}_{\\text{k}}({\\text{u}}_{\\text{i}},{\\text{v}}_{\\text{i}})\\:{\\text{X}}_{\\text{i}\\text{k}}+{{\\epsilon\\:}}_{\\text{i}}$$\u003c/div\u003e\n\u003cdiv class=\"EquationNumber\"\u003e2\u003c/div\u003e\n\u003c/div\u003e\n\u003cp\u003eYi: The observed value of each dependent variable in the study area, \u0026beta;₀: constant value, \u0026beta;k: parameter estimate of the k-th variable, Xik: covariance value of the k-th variable, (ui, vi): coordinates of the i-th point, \u0026epsilon;: error term (Fotheringham et al., \u003cspan class=\"CitationRef\"\u003e2009\u003c/span\u003e).\u003c/p\u003e\n\u003cp\u003eIn GWR models, weighting of sampling points relative to each other is carried out using kernel models. Among the common kernel functions are Gaussian and bisquare. In the Gaussian kernel model, there are two different methods: fixed and adaptive. Bandwidth is another important parameter in the construction of GWR models. The bandwidth determines the number of neighboring points or the distance. The bandwidth is usually automatically selected according to the CV (cross-validation) or AICc (Akaike Information Criterion corrected) criteria (Fotheringham et al., \u003cspan class=\"CitationRef\"\u003e2009\u003c/span\u003e).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\n\u003ch2\u003e2.5. Assessment of model performances\u003c/h2\u003e\n\u003cp\u003e5-fold cross-validation and validation with external data were used to validate the models. The results of both methods were evaluated using R\u0026sup2;, RMSE, and MAE validation criteria. 80% of the dataset was used in creating the models and 20% was used in the external data validation method. The predicted values obtained using the coefficients of the models were compared with the observed values and scatter plots were created.\u003c/p\u003e\n\u003cp\u003eCross-validation method is widely used to evaluate the results of MLR and GWR models. The cross-validation method is useful for testing the suitability of the created models to the training data. In addition, it measures the applicability of the model and tests its performance on different data subsets (Kuhn \u0026amp; Johnson, \u003cspan class=\"CitationRef\"\u003e2013\u003c/span\u003e). The cross-validation method is widely preferred to test the accuracy of models, especially in study areas where the number of stations is limited (Vienneau et al., \u003cspan class=\"CitationRef\"\u003e2013\u003c/span\u003e; Wang et al., \u003cspan class=\"CitationRef\"\u003e2013\u003c/span\u003e; Li et al., \u003cspan class=\"CitationRef\"\u003e2021\u003c/span\u003e).\u003c/p\u003e\n\u003c/div\u003e"},{"header":"3. Results and Discussions","content":"\u003cp\u003eIn this study, MLR and GWR models were applied to determine the spatial factors affecting the distribution of SO\u003csub\u003e2\u003c/sub\u003e concentration in Karab\u0026uuml;k province. The annual average of measurements obtained from four different ground stations is the dependent variable of the models. The independent variables are the twelve spatial indicators specified in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. Annual average values ​​obtained from ground stations were interpolated according to the study boundaries using the Kriging method. Thus, a raster format SO2 distribution map of the study area with a resolution of 30 m x 30 m was obtained (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). ArcGIS 10.1 software was used to create the maps in the study. This approach was preferred to ensure the continuity of the pollutant within the study area. For the application of the models, 100 sampling points were randomly distributed within the study area. Eighty of the sampling points were used to construct the models, and 20 of them were used to test the accuracy of the models with external data.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cem\u003eSO₂ distribution map interpolated using the Kriging method (\u0026micro;g/m\u0026sup3;)\u003c/em\u003e \u003c/p\u003e \u003cp\u003eIn the MLR analysis, the main purpose is to identify the independent variables that explain the dependent variable at the highest rate. SPSS software was used for the implementation of the analysis. It was determined that 6 of the independent variables had a significant relationship with SO\u003csub\u003e2\u003c/sub\u003e concentration (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Sig. and VIF values should be examined for the significance of the relationship between SO\u003csub\u003e2\u003c/sub\u003e and independent variables. The Sig. value is expected to be less than 0.05 at the 95% confidence interval. In addition, a VIF value greater than 10 is an indication of multicollinearity among independent variables. The results obtained showed that there is no multicollinearity, meaning no high correlation among the independent variables.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003e\u003cem\u003eResult of the MLR model\u003c/em\u003e\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIndependent Variables\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCoefficent\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003et value\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSig.\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eVIF\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIntercept\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e17.70\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e38.01\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDistance to main road\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-4.20x10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e-1.74\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.050\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e3.36\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBuffer 500\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e9.62x10\u003csup\u003e\u0026minus;\u0026thinsp;5\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e4.74\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e1.80\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDistance to industrial facility\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-6.92x10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e-3.61\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e3.89\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIndustrial Density\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e7.93\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e9.07\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e5.28\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTotal distance to high-traffic intersections\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-9.44\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e-3.93\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e5.34\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFuel station density\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-1.42x10\u003csup\u003e\u0026minus;\u0026thinsp;3\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e-5.67\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e2.43\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eR\u0026sup2;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAdjusted R\u0026sup2;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.83\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eAccording to the results of the MLR model, R\u0026sup2; = 0.85 and adjusted R\u0026sup2; = 0.83 values were obtained. This indicates that the independent variables explain approximately 83% of the change in SO₂ concentration. The R\u0026sup2; values obtained reveal that the created model has a statistically strong prediction performance.\u003c/p\u003e \u003cp\u003eWhen the coefficients of the independent variables are examined, it is seen that the industrial density (β\u0026thinsp;=\u0026thinsp;7.93, p\u0026thinsp;\u0026lt;\u0026thinsp;0.000) variable has the highest positive effect on SO\u003csub\u003e2\u003c/sub\u003e concentrations. This result shows that SO\u003csub\u003e2\u003c/sub\u003e concentration increases in regions where industrial facilities are spatially concentrated. This demonstrates that pollutants arising from the fuels used in industrial facilities and from steel production processes significantly negatively affect air quality. The distance to the industrial facility (β = -6.92\u0026times;10⁻⁴, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) variable has a negative effect on the SO\u003csub\u003e2\u003c/sub\u003e variable. This result showed that as the distance to industrial facilities increases, the concentration decreases, and as the distance decreases, the concentration increases.\u003c/p\u003e \u003cp\u003eProximity to main roads and road length within the 500 m buffer are also among important variables affecting SO\u003csub\u003e2\u003c/sub\u003e levels. The shortest distance to the main road variable (β = -4.20\u0026times;10⁻⁴, p\u0026thinsp;\u0026lt;\u0026thinsp;0.050) has a negative effect on SO\u003csub\u003e2\u003c/sub\u003e concentration. It is concluded that as the distance to the main roads increases, the SO\u003csub\u003e2\u003c/sub\u003e concentration decreases, and as the distance decreases, the SO\u003csub\u003e2\u003c/sub\u003e concentration increases. This situation shows the effect of vehicles on the main roads on SO\u003csub\u003e2\u003c/sub\u003e concentration. There is a positive relationship between total road length and SO\u003csub\u003e2\u003c/sub\u003e concentration in the 500 m circular buffer area (β\u0026thinsp;=\u0026thinsp;9.62\u0026times;10⁻⁵, p\u0026thinsp;\u0026lt;\u0026thinsp;0.000). This situation causes the SO\u003csub\u003e2\u003c/sub\u003e concentration to increase as the total path length in the buffer zone increases.\u003c/p\u003e \u003cp\u003eThe total distance indicator to high-traffic intersections (β = -9.44, p\u0026thinsp;\u0026lt;\u0026thinsp;0.01) has a negative effect on SO\u003csub\u003e2\u003c/sub\u003e concentration. As the total distance to the intersections increases, SO\u003csub\u003e2\u003c/sub\u003e concentration decreases. In the other case, as the total distance to the intersections decreases, the SO\u003csub\u003e2\u003c/sub\u003e concentration increases. This situation shows the effect of traffic density on SO\u003csub\u003e2\u003c/sub\u003e concentration. Fuel station density (β = -1.42\u0026times;10⁻\u0026sup3;, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001) was determined to have a negative coefficient. The findings obtained as a result of the MLR analysis revealed that industrial activities and the road network in Karab\u0026uuml;k are decisive on SO\u003csub\u003e2\u003c/sub\u003e concentrations. This situation indicates that industrial emissions have a strong impact on urban air quality.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003e\u003cem\u003eResult of the GWR model\u003c/em\u003e\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIndependent Variables\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMean\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eStd\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMin\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMax\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDistance to main road\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-8.1x10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4x10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-1.26x10\u003csup\u003e\u0026minus;\u0026thinsp;3\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-1.8x10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBuffer 500\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e8.5x10\u003csup\u003e\u0026minus;\u0026thinsp;5\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2.4x10\u003csup\u003e\u0026minus;\u0026thinsp;5\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e4.1x10\u003csup\u003e\u0026minus;\u0026thinsp;5\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.16x10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDistance to industrial facility\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-7.2x10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4.6x10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-1.28x10\u003csup\u003e\u0026minus;\u0026thinsp;3\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIndustrial Density\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e8.17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.61\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e6.60\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e10.12\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTotal distance to high-traffic intersections\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-8.84\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2.15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-14.05\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-3.67\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFuel station density\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-1.49x10\u003csup\u003e\u0026minus;\u0026thinsp;3\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6.7x10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-2.17x10\u003csup\u003e\u0026minus;\u0026thinsp;3\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-4.5x10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eR\u0026sup2;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.95\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAdjusted R\u0026sup2;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.93\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eIn applying the GWR analysis, variables that were significant in the MLR were included in the analysis. In the model parameter selection, the adaptive model from the Gaussian Kernel models was chosen. In the selection of bandwidth, the AICc (Akaike Information Criterion corrected) criteria, which are most compatible with the adaptive Kernel model, were applied. In the GWR analysis, separate coefficient values are produced for each sampling point. Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e presents the minimum, maximum, average, and standard deviation values of the coefficients for the variables.\u003c/p\u003e \u003cp\u003eIn the GWR analysis, the overall R\u0026sup2; value of the model was obtained as 0.95 and the adjusted R\u0026sup2; value as 0.93. These values indicate that the overall explanatory power of the model is high. This situation shows the high agreement between the observed and predicted SO₂ concentrations in the GWR model. According to the coefficient values, the buffer 500 and industrial density indicators have a positive effect on SO₂ concentration. It was determined that the variables distance to the main road, distance to the industrial facility, total distance to traffic-intensive intersections, and fuel station density have negative effects on SO₂ concentration. These results are similar to the MLR results. The negative effect of the shortest distance to main roads on SO₂ concentration indicates that as the distance from the main road increases, the concentration decreases, and as the distance to the main road decreases, the concentration increases. A similar situation is observed in the variable of the shortest distance to industrial facilities. It was concluded that as the distance to industrial facilities shortens, the SO\u003csub\u003e2\u003c/sub\u003e concentration increases, while the concentration decreases as the distance increases. The Buffer 500 variable has a positive effect on concentration. This means that increasing the path length in the buffer zone causes an increase in SO\u003csub\u003e2\u003c/sub\u003e concentration. The industrial density variable is another indicator that has a positive effect on SO\u003csub\u003e2\u003c/sub\u003e concentration. This means that high concentrations are observed in dense industrial areas. Total distance to high-traffic intersections has a negative effect on SO\u003csub\u003e2\u003c/sub\u003e concentration. As the total distance of the sampling points to the intersections with high traffic increases, the SO\u003csub\u003e2\u003c/sub\u003e concentration tends to decrease. The concentration increased as the total distance to the intersections decreased. In terms of standard deviation values, industrial density and total distance to high-traffic intersections are the variables with the highest variance. This means that the coefficients of the two indicators vary more between regions.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e3.1. Performance assessment\u003c/h2\u003e \u003cp\u003eTesting the accuracy of the models created in LUR models is an important step. To test the accuracy of the MLR and GWR models, 5-fold cross-validation and external data validation methods were used (Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003e\u003cem\u003eValidation results of MLR and GWR models\u003c/em\u003e\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eValidation Method\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eR\u0026sup2;\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eRMSE\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMAE\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e5-fold Cross Validation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.67\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.53\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eExternal data validation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.54\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.47\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGWR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e5-fold Cross Validation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.95\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.40\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.34\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGWR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eExternal data validation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.94\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.45\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.39\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe results of the MLR model according to the 5-fold cross-validation method are R\u0026sup2; = 0.85, RMSE\u0026thinsp;=\u0026thinsp;0.67 and MAE\u0026thinsp;=\u0026thinsp;0.53. Validation results using external data showed improvement and yielded R\u0026sup2; = 0.92, RMSE\u0026thinsp;=\u0026thinsp;0.54 and MAE\u0026thinsp;=\u0026thinsp;0.47. According to both validation methods, the MLR model achieved successful results. The higher values ​​in the validation with external data indicate that the model may have been affected by variances during training.\u003c/p\u003e \u003cp\u003eThe results of the GWR model according to the 5-fold cross-validation method are R\u0026sup2; = 0.95, RMSE\u0026thinsp;=\u0026thinsp;0.40 and MAE\u0026thinsp;=\u0026thinsp;0.34. The obtained values ​​show that the GWR model explains the distribution of SO2 concentration with high accuracy. The values ​​of R\u0026sup2; = 0.94, RMSE\u0026thinsp;=\u0026thinsp;0.45 and MAE\u0026thinsp;=\u0026thinsp;0.39 obtained in the external data validation method show that the model is also successful according to this validation method.\u003c/p\u003e \u003cp\u003eAccording to the 5-fold cross-validation and external data validation methods, both models produced statistically strong results. The GWR method produces different coefficients for each sample point. As a result, the spatial effects of independent variables affecting SO\u003csub\u003e2\u003c/sub\u003e concentration are reflected more accurately. The use of a constant coefficient for the entire study area in the MLR model caused limitations in explaining local variations. As a result, considering spatial dependency in estimating SO\u003csub\u003e2\u003c/sub\u003e concentration significantly improved model performance.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eScatter plot of observed and predicted SO\u003csub\u003e2\u003c/sub\u003e concentrations.\u003c/p\u003e \u003cp\u003eThe scatter plot showing the relationship between the observed and predicted values according to two different validation methods of the MLR and GWR models is given in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e. The scatter plot demonstrates the strength of the relationship between the observed and predicted SO\u003csub\u003e2\u003c/sub\u003e concentrations. In the MLR model, although the 5-fold cross-validation results generally showed a good distribution, a tendency of underestimation was observed between 14\u0026ndash;16 \u0026micro;g/m\u0026sup3; values and at some higher values. In the external data validation, it is observed that the values were partially clustered around the axis line. In the GWR model, both the 5-fold cross-validation and external validation plots produced slightly better results compared to the MLR model. This indicates that the GWR model exhibited more consistent results according to both validation methods. It was also revealed in Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e that numerically, the GWR model showed better performance than the MLR model in terms of R\u0026sup2;, RMSE, and MAE values.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eSpatial distribution map of observed (a) and predicted (b: MLR, c: GWR) SO\u003csub\u003e2\u003c/sub\u003e concentrations\u003c/p\u003e \u003cp\u003eIn Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, in the distribution of observed (a) SO₂ concentration, high pollution was observed in the south and low values in the north. In the MLR (b) prediction map, although the general distribution of SO₂ concentration is similar, the model estimated the areas with the highest concentration in the south at lower levels (Maximum: 19.36 \u0026micro;g/m\u0026sup3;). This situation arises due to the spatial stationarity in the MLR model. Therefore, MLR could not fully represent local differences. The GWR (c) prediction map shows high spatial similarity with the observed distribution. The high pollution area in the south and the low pollution areas in the north overlap with the observed distribution in terms of spatial intensity level. This result shows that the GWR model successfully reflects the spatial heterogeneity in the study area.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eLocal R\u0026sup2; distribution map in the GWR analysis\u003c/p\u003e \u003cp\u003eThe Local R\u0026sup2; map in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e shows how the spatial fit of the GWR model varies across the regions within the study area. Since the Local R\u0026sup2; value is calculated by taking into account the observations around each sampling point, it indicates the local level fit of the model. The Local R\u0026sup2; value was obtained with maximum 0.92, minimum 0.82, and average 0.87, indicating that the overall explanatory power of the model is high. This result is expected in the GWR analysis and shows that the explanatory power of the model varies spatially. In places with high Local R\u0026sup2; (0.92) values, the model explains the distribution of SO₂ concentration very well, while in places where the Local R\u0026sup2; (0.82) value is lower, the fit is lower. When the map showing the Local R\u0026sup2; distribution is examined, it is understood that the model fit is higher in the south of the study area. A partial decrease in model fit was observed towards the north. This situation may be due to the lower number of stations in the northern region.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eResidual distribution map of the GWR analysis\u003c/p\u003e \u003cp\u003eThe Residual map in Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e shows the spatial error distribution between the observed and predicted values of the GWR model. The positive values on the map indicate the places where the model underestimates, while the negative values indicate the places where the model overestimates. It is seen that the residual values are generally in the range of \u0026plusmn;\u0026thinsp;1 \u0026micro;g/m\u0026sup3;. This shows that the predicted values are very close to the observed values. The random spatial distribution of errors is evidence that there is no spatial autocorrelation. In addition, it is seen that there are a few clustered points on the map. The blue areas in the northeast indicate that the model slightly overestimated in these regions. Similarly, in the red areas in the southeast, the model underestimated the SO₂ concentration. However, this situation is not at a level that would affect the overall representation of the model.\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Conclusion","content":"\u003cp\u003eIn this study, MLR and GWR methods were used to model the spatial distribution of SO₂ concentrations in Karab\u0026uuml;k, and the results were compared. Roads, industrial areas, temperature, elevation, intersections, and fuel station data were selected as independent variables for the models. Both models were evaluated using 5-fold cross-validation and external data validation methods. Although MLR achieved a generally good result, it could not fully reveal spatial heterogeneity due to the fixed coefficient assumption.\u003c/p\u003e \u003cp\u003eMLR demonstrated good predictive power in both the 5-fold cross-validation (R\u0026sup2; = 0.85, RMSE\u0026thinsp;=\u0026thinsp;0.67, MAE\u0026thinsp;=\u0026thinsp;0.53) and external data validation methods (R\u0026sup2; = 0.92, RMSE\u0026thinsp;=\u0026thinsp;0.54, MAE\u0026thinsp;=\u0026thinsp;0.47). In MLR-based LUR studies in the literature, R\u0026sup2; values have generally been obtained in the range of 0.70\u0026ndash;0.85 (Mikeš et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Torres et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Siu et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). In this regard, the findings obtained are consistent with international studies.\u003c/p\u003e \u003cp\u003eThe GWR model showed better performance than the MLR model in both the 5-fold cross-validation (R\u0026sup2; = 0.95, RMSE\u0026thinsp;=\u0026thinsp;0.40, MAE\u0026thinsp;=\u0026thinsp;0.34) and external data validation methods (R\u0026sup2; = 0.94, RMSE\u0026thinsp;=\u0026thinsp;0.45, MAE\u0026thinsp;=\u0026thinsp;0.39). This situation reveals that models based on the fixed coefficient assumption are insufficient in urban areas where industrial activities are intense (Fotheringham et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Morshed et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Liu et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). In studies conducted especially in industrial regions related to the GWR model, high explanatory values have been obtained. Cheng et al., (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) determined the R\u0026sup2; value of the GWR model as 0.93 in industrial cities of China. Morshed et al., (\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) obtained 10\u0026ndash;15% higher accuracy in SO₂ GWR results compared to MLR in Bangladesh.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eDistribution of industrial areas within the study area. GWR prediction map (a) and residual distribution map (b)\u003c/p\u003e \u003cp\u003eIn the GWR prediction map (a) in Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e, it is clearly observed that SO₂ concentration is higher in areas where industrial facilities are concentrated. In the study area where iron and steel plants are concentrated, it is observed that the maximum SO\u003csub\u003e2\u003c/sub\u003e concentration is around these industrial facilities. This situation shows that pollutants emitted from industrial facilities play a decisive role in air quality. This finding obtained in the study is consistent with the high levels of SO\u003csub\u003e2\u003c/sub\u003e concentrations detected in industrial areas in the literature (Shen et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Liu et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe residual map (b) in Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e shows the spatial distribution of prediction errors of the GWR model. The residual values showed a distribution of approximately\u0026thinsp;\u0026plusmn;\u0026thinsp;1 \u0026micro;g/m\u0026sup3;. The lower errors in the areas where industrial zones are located indicate that the model made accurate predictions in these regions. On the other hand, there are negative residual values in the north. It is thought that this situation arises as a result of the low number of stations in the region. From the residual values, it was seen that the model could not fully reflect some local effects. This situation has been reported in the studies of Morshed et al., (\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), Cheng et al., (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) and Hu et al., (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) in which similar uncertainties were reported in areas with low station density in GWR analyses.\u003c/p\u003e \u003cp\u003eThe GWR model achieved better results than the MLR in both overall accuracy and modeling the spatial distribution of SO\u003csub\u003e2\u003c/sub\u003e concentration. However, the GWR model revealed the spatial distribution of SO\u003csub\u003e2\u003c/sub\u003e concentration in detail, revealing changes at the local scale.\u003c/p\u003e "},{"header":"Declarations","content":"\u003cp\u003e \u003ch2\u003eCompeting Interests\u003c/h2\u003e \u003cp\u003eThe authors have no relevant financial or non-financial interests to disclose.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eFunding\u003c/h2\u003e \u003cp\u003eThe authors declare that no funds, grants, or other support were received during the preparation of this manuscript.\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eAll authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by Emre Y\u0026uuml;cer.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe data sources used in this study can be downloaded from publicly available websites.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAarthi, A., Gayathri, P., Gomathi, N. R., Kalaiselvi, S., \u0026amp; Gomathi, V. (2020). Air quality prediction through regression model. \u003cem\u003eInternational Journal of Scientific and Technology Research\u003c/em\u003e, 9 (3), 923\u0026ndash;928.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlpar. R. (2014). Uygulamalı İstatistik ve Ge\u0026ccedil;erlilik-G\u0026uuml;venirlik: SPSS\u0026rsquo;de \u0026Ccedil;\u0026ouml;z\u0026uuml;mleme Adımları İle Birlikte. 3rd edition. \u003cem\u003eDetay Yayıncılık\u003c/em\u003e Ankara.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBeelen, R., Hoek, G., Vienneau, D., et al. (2013). Development of NO₂ and NOx land-use regression models for estimating air pollution exposure in 36 study areas in Europe-The ESCAPE project. \u003cem\u003eAtmospheric Environment\u003c/em\u003e, 72, 10\u0026ndash;23. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.atmosenv.2013.02.037\u003c/span\u003e\u003cspan address=\"10.1016/j.atmosenv.2013.02.037\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBriggs, D. J. (1997). Mapping urban air pollution using GIS: a regression-based approach. \u003cem\u003eInternational Journal of Geographical Information Science\u003c/em\u003e, 11 (7), 699\u0026ndash;718. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/136588197242158\u003c/span\u003e\u003cspan address=\"10.1080/136588197242158\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen, T. H., Hsu, Y. C., Zeng, Y. T., et al. (2020). A hybrid kriging/land-use regression model with Asian culture-specific sources to assess NO2 spatial-temporal variations. \u003cem\u003eEnvironmental Pollution\u003c/em\u003e, 259, 113875. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.envpol.2019.113875\u003c/span\u003e\u003cspan address=\"10.1016/j.envpol.2019.113875\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCheng, S., Zhang, G., Yang, X., \u0026amp; Lei, B. (2024). A multiscale geographically weighted regression kriging method for spatial downscaling of satellite-based ozone datasets. \u003cem\u003eFrontiers in Environmental Science\u003c/em\u003e, 11, 1267752. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fenvs.2023.1267752\u003c/span\u003e\u003cspan address=\"10.3389/fenvs.2023.1267752\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eComber, A., Brunsdon, C., Charlton. M., et al. (2020). The GWR route map: a guide to the informed application of Geographically Weighted Regression. \u003cem\u003earXiv preprint arXi\u003c/em\u003e, 2004.06070.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ede Hoogh, K., Gulliver, J., van Donkelaar, A., et al. (2016). Development of West-European PM2.5 and NO2 land-use regression models incorporating satellite-derived and chemical transport modelling data. \u003cem\u003eEnvironmental Research\u003c/em\u003e, 151, 1\u0026ndash;10. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.envres.2016.07.00\u003c/span\u003e\u003cspan address=\"10.1016/j.envres.2016.07.00\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEeftens, M., Beelen, R., De Hoogh, K., et al. (2012). Development of land use regression models for PM2. 5, PM2. 5 absorbance, PM10 and PMcoarse in 20 European study areas; results of the ESCAPE project. \u003cem\u003eEnvironmental science \u0026amp; technology\u003c/em\u003e, 46(20), 11195\u0026ndash;11205. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/es301948k\u003c/span\u003e\u003cspan address=\"10.1021/es301948k\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFotheringham, A. S., Brunsdon, C., \u0026amp; Charlton, M. (2009). Geographically weighted regression. \u003cem\u003eThe Sage handbook of spatial analysis\u003c/em\u003e, 1, 243\u0026ndash;254.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFotheringham, A. S., Yang, W., \u0026amp; Kang, W. (2017). Multiscale geographically weighted regression (MGWR). \u003cem\u003eAnnals of the American Association of Geographers\u003c/em\u003e, 107(6), 1247\u0026ndash;1265. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/24694452.2017.1352480\u003c/span\u003e\u003cspan address=\"10.1080/24694452.2017.1352480\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGuo, B., Wang, X., Zhang, D., Pei, L., Zhang, D., \u0026amp; Wang, X. (2020). A land use regression application into simulating spatial distribution characteristics of particulate matter (PM\u003csub\u003e2.5\u003c/sub\u003e) concentration in city of Xi'an, China. \u003cem\u003ePolish Journal of Environmental Studies\u003c/em\u003e, 29(6), 4065\u0026ndash;4076. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.15244/pjoes/118426\u003c/span\u003e\u003cspan address=\"10.15244/pjoes/118426\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHoek, G., Beelen, R., de Hoogh, K., Vienneau, D., Gulliver, J., Fischer, P., \u0026amp; Briggs, D. (2008). A review of land-use regression models to assess spatial variation of outdoor air pollution. \u003cem\u003eAtmospheric Environment\u003c/em\u003e, 42(33), 7561\u0026ndash;7578. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.atmosenv.2008.05.057\u003c/span\u003e\u003cspan address=\"10.1016/j.atmosenv.2008.05.057\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHu, K., Zhang, X., Yang, X., \u0026amp; Yu, M. (2025). A study on the spatial distribution of life expectancy and its air pollution factors in China based on geographically weighted regression. \u003cem\u003eFrontiers in Public Health\u003c/em\u003e, 13, 1565744. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fpubh.2025.1565744\u003c/span\u003e\u003cspan address=\"10.3389/fpubh.2025.1565744\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuang, L., Zhang, C., \u0026amp; Bi, J. (2017). Development of land use regression models for PM2. 5, SO2, NO2 and O3 in Nanjing, China. \u003cem\u003eEnvironmental research\u003c/em\u003e, 158, 542\u0026ndash;552. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.envres.2017.07.010\u003c/span\u003e\u003cspan address=\"10.1016/j.envres.2017.07.010\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKarab\u0026uuml;k Municipality, (2025). Retrived September 25, 2025, from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.karabuk.bel.tr/icerik.asp?i_id=39\u003c/span\u003e\u003cspan address=\"https://www.karabuk.bel.tr/icerik.asp?i_id=39\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKarab\u0026uuml;k, TSO. (2025). Karab\u0026uuml;k Chamber of Commerce And Industry, Retrived September 25, 2025, from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://karabuktso.org.tr/blog-makale/kardemirin-ve-karabukun-dogusu-3-nisan\u003c/span\u003e\u003cspan address=\"https://karabuktso.org.tr/blog-makale/kardemirin-ve-karabukun-dogusu-3-nisan\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKerckhoffs, J., Khan, J., Hoek, G., Yuan, Z., Ellermann, T., Hertel, O. et al. (2022). Mixed-effects modeling framework for Amsterdam and Copenhagen for outdoor NO2 concentrations using measurements sampled with google street view cars. \u003cem\u003eEnvironmental science \u0026amp; technology\u003c/em\u003e, 56(11), 7174\u0026ndash;7184. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/acs.est.1c05806\u003c/span\u003e\u003cspan address=\"10.1021/acs.est.1c05806\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKuhn, M., \u0026amp; Johnson, K. (2013). Applied predictive modeling. Springer, New York.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi, Z., Ho, K. F., Chuang, H. C., \u0026amp; Yim, S. H. L. (2021). Development and intercity transferability of land-use regression models for predicting ambient PM 10, PM 2.5, NO 2 and O 3 concentrations in northern Taiwan. \u003cem\u003eAtmospheric Chemistry and Physics\u003c/em\u003e, 21(6), 5063\u0026ndash;5078. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.5194/acp-21-5063-2021\u003c/span\u003e\u003cspan address=\"10.5194/acp-21-5063-2021\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu, Z., Guan, Q., Luo, H., Wang, N., Pan, N., Yang, L., Xiao, S., \u0026amp; Lin, J. (2019). Development of land-use regression model and health risk assessment for NO₂ in different functional areas: a case study of Xi\u0026rsquo;an, China. \u003cem\u003eAtmospheric Environment\u003c/em\u003e, 213, 515\u0026ndash;525. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.atmosenv.2019.06.044\u003c/span\u003e\u003cspan address=\"10.1016/j.atmosenv.2019.06.044\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu, X., Wu, W., Zou, H., Pan, C., Liu, M., \u0026amp; Li, H. (2025). Unveiling seasonal disparities in urban form: Exploring scale and intensity effects on PM2.5 concentrations. \u003cem\u003eUrban Climate\u003c/em\u003e, 59, 102319. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.uclim.2025.102319\u003c/span\u003e\u003cspan address=\"10.1016/j.uclim.2025.102319\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMa, X., Zou, B., Deng, J., Gao, J., Longley, I., Xiao, S. et al. (2024). A comprehensive review of the development of land use regression approaches for modeling spatiotemporal variations of ambient air pollution: A perspective from 2011 to 2023. \u003cem\u003eEnvironment international\u003c/em\u003e, 183, 08430. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.envint.2024.108430\u003c/span\u003e\u003cspan address=\"10.1016/j.envint.2024.108430\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMeng, X., Chen, L., Cai, J., Zou, B., Wu, C. F., Fu, Q., Zhang, Y., Liu, Y., \u0026amp; Kan, H. (2015). A land-use regression model for estimating the NO₂ concentration in Shanghai, \u003cem\u003eChina. Environmental Research\u003c/em\u003e, 137, 308\u0026ndash;315. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.envres.2015.01.003\u003c/span\u003e\u003cspan address=\"10.1016/j.envres.2015.01.003\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMeo, S. A., Shaikh, N., Alotaibi, M., AlWabel, A. A., \u0026amp; Alqumaidi, H. (2024). Effect of air pollutants particulate matter (PM2. 5, PM10), sulfur dioxide (SO2) and ozone (O3) on cognitive health. \u003cem\u003eScientific Reports\u003c/em\u003e, 14(1), 19616. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41598-024-70646-6\u003c/span\u003e\u003cspan address=\"10.1038/s41598-024-70646-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMikeš, O., S\u0026aacute;ňka, O., Rafajov\u0026aacute;, A., Vlaanderen, J., Chen, J., Hoek, G. et al. (2023). Development of historic monthly land use regression models of SO2, NOx and suspended particulate matter for birth cohort ELSPAC. \u003cem\u003eAtmospheric Environment\u003c/em\u003e, 30, 119688. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.atmosenv.2023.119688\u003c/span\u003e\u003cspan address=\"10.1016/j.atmosenv.2023.119688\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMorshed, S. R., Fattah, M. A., Kafy, A. A., Alsulamy, S., Almulhim, A. I., Shohan A. A. A., Khedher, K. M. (2024). Decoding seasonal variability of air pollutants with climate factors: A geostatistical approach using multimodal regression models for informed climate change mitigation. \u003cem\u003eEnvironmental Pollution\u003c/em\u003e, 345, 123463. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.envpol.2024.123463\u003c/span\u003e\u003cspan address=\"10.1016/j.envpol.2024.123463\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOrellano, P., Reynoso, J., \u0026amp; Quaranta, N. (2021). Short-term exposure to sulphur dioxide (SO2) and all-cause and respiratory mortality: A systematic review and meta-analysis. \u003cem\u003eEnvironment international\u003c/em\u003e, 150, 106434. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.envint.2021.106434\u003c/span\u003e\u003cspan address=\"10.1016/j.envint.2021.106434\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRibeiro, H., Almeida, S. M., Martins, E. M., \u0026amp; Silva, A. V. (2016). Petrol stations: An environmental risk factor. \u003cem\u003eEnvironmental Monitoring and Assessment\u003c/em\u003e, 188, 183. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10661-016-5172-0\u003c/span\u003e\u003cspan address=\"10.1007/s10661-016-5172-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShen, Y., de Hoogh, K., Schmitz, O., Clinton, N., Tuxen-Bettman, K., Brandt, J. et al. (2024). Monthly average air pollution models using geographically weighted regression in Europe from 2000 to 2019. \u003cem\u003eScience of the Total Environment\u003c/em\u003e, 918, 170550. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.scitotenv.2024.170550\u003c/span\u003e\u003cspan address=\"10.1016/j.scitotenv.2024.170550\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShi, T., Dirienzo, N., Requia, W. J., Hatzopoulou, M. \u0026amp; Adams, M. D. (2020). Neighbourhood scale nitrogen dioxide land use regression modelling with regression kriging in an urban transportation corridor. \u003cem\u003eAtmospheric Environment\u003c/em\u003e, 223, 117218. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.atmosenv.2019.117218\u003c/span\u003e\u003cspan address=\"10.1016/j.atmosenv.2019.117218\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShi, T., Zhang, Y., Yuan, X., Li, F., \u0026amp; Yan, S. (2024). Spatial Patterns and Determinants of PM2. 5 Concentrations: A Land Use Regression Analysis in Shenyang Metropolitan Area, China. \u003cem\u003eSustainability\u003c/em\u003e, 16(12), 5119. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/su16125119\u003c/span\u003e\u003cspan address=\"10.3390/su16125119\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSiu, T. K., Greene, C. S., \u0026amp; Fong, K. C. (2025). Identifying Surface Sulphur Dioxide (SO2) Monitoring Gaps in Saint John, Canada with Land Use Regression and Hot Spot Mapping. \u003cem\u003eAtmospheric Environment\u003c/em\u003e, 121238. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.atmosenv.2025.121238\u003c/span\u003e\u003cspan address=\"10.1016/j.atmosenv.2025.121238\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTorres, R. A. B., Ramos, R. V., Recto, B. A. B., \u0026amp; Tamondong, A. M. (2024). DEVELOPMENT OF LAND-USE REGRESSION MODELS FOR PARTICULATE MATTER ESTIMATION IN NATIONAL CAPITAL REGION, PHILIPPINES. The International Archives of the Photogrammetry, \u003cem\u003eRemote Sensing and Spatial Information Sciences\u003c/em\u003e, 48, 437\u0026ndash;444. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.5194/isprs-archives-XLVIII-4-W8-2023-437-2024\u003c/span\u003e\u003cspan address=\"10.5194/isprs-archives-XLVIII-4-W8-2023-437-2024\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eT\u0026Uuml;İK. (2025). Turkish Statistical Institute. Retrieved July 24, 2025 form \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://data.tuik.gov.tr/Kategori/GetKategori?p=Nufus-ve-Demografi-109\u003c/span\u003e\u003cspan address=\"https://data.tuik.gov.tr/Kategori/GetKategori?p=Nufus-ve-Demografi-109\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUSGS. (2025). The United States Geological Survey. Retrieved July 24, 2025 from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://earthexplorer.usgs.gov/\u003c/span\u003e\u003cspan address=\"https://earthexplorer.usgs.gov/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVienneau, D., de Hoogh, K., Bechle, M. J., Beelen, R., van Donkelaar, A., Martin, R. V. et al. (2013). Western European land-use regression incorporating satellite- and ground-based measurements of NO₂ and PM10. \u003cem\u003eEnvironmental Science \u0026amp; Technology\u003c/em\u003e, 47(23), 13555\u0026ndash;13564. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/es403089q\u003c/span\u003e\u003cspan address=\"10.1021/es403089q\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, M., Beelen, R., Basaga\u0026ntilde;a, X., Becker, T., Cesaroni, G., de Hoogh, K. et al. (2013). Evaluation of land-use regression models for NO₂ and particulate matter in 20 European study areas: the ESCAPE project. \u003cem\u003eEnvironmental Science \u0026amp; Technology\u003c/em\u003e, 47(9), 4357\u0026ndash;4364. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/es305129t\u003c/span\u003e\u003cspan address=\"10.1021/es305129t\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, J., \u0026amp;Xu, H. (2021). A novel hybrid spatiotemporal land use regression model system at the megacity scale. \u003cem\u003eAtmospheric Environment\u003c/em\u003e, 244, 117971. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.atmosenv.2020.117971\u003c/span\u003e\u003cspan address=\"10.1016/j.atmosenv.2020.117971\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWei, Q., Zhang, L., Duan, W., \u0026amp; Zhen, Z. (2019). Global and geographically and temporally weighted regression models for modeling PM2. 5 in Heilongjiang, China from 2015 to 2018. \u003cem\u003eInternational journal of environmental research and public health\u003c/em\u003e, 16(24), 5107. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/ijerph16245107\u003c/span\u003e\u003cspan address=\"10.3390/ijerph16245107\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWong, P. Y., Lee, H. Y., Chen, Y. C., Zeng, Y. T., Chern, Y. R., Chen, N. T. et al. (2021). Using a land use regression model with machine learning to estimate ground level PM2.5. \u003cem\u003eEnvironmental Pollution\u003c/em\u003e, 277, 116846. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.envpol.2021.116846\u003c/span\u003e\u003cspan address=\"10.1016/j.envpol.2021.116846\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXuan, W., Zhang, F., Zhou, H., Du, Z., \u0026amp; Liu, R. (2021). Improving geographically weighted regression considering directional nonstationary for ground-Level PM2.5 estimation. \u003cem\u003eISPRS International Journal of Geo-Information\u003c/em\u003e, 10(6), 413. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/ijgi10060413\u003c/span\u003e\u003cspan address=\"10.3390/ijgi10060413\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang, X., Wang, S., Zhang, W., Zhan, D., \u0026amp; Li, J. (2017). The impact of anthropogenic emissions and meteorological conditions on the spatial variation of ambient SO2 concentrations: A panel study of 113 Chinese cities. \u003cem\u003eScience of the Total Environment\u003c/em\u003e, 584, 318\u0026ndash;328. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.scitotenv.2016.12.145\u003c/span\u003e\u003cspan address=\"10.1016/j.scitotenv.2016.12.145\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eY\u0026uuml;cer, E., Erener, A., \u0026amp; Sarp, G. (2023). A land use regression model to estimate ambient concentrations of PM10 and SO2 in İzmit, Turkey. \u003cem\u003eJournal of the Indian Society of Remote Sensing\u003c/em\u003e, 51(6), 1329\u0026ndash;1341. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1007/s12524-023-01704-1\u003c/span\u003e\u003cspan address=\"10.1007/s12524-023-01704-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou, Q., Wang, C., \u0026amp; Fang, S. (2019). Application of geographically weighted regression (GWR) in the analysis of the cause of haze pollution in China. \u003cem\u003eAtmospheric Pollution Research\u003c/em\u003e, 10(3), 835\u0026ndash;846. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.apr.2018.12.012\u003c/span\u003e\u003cspan address=\"10.1016/j.apr.2018.12.012\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"pure-and-applied-geophysics","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"paag","sideBox":"Learn more about [Pure and Applied Geophysics](https://www.springer.com/journal/24)","snPcode":"24","submissionUrl":"https://submission.nature.com/new-submission/24/3","title":"Pure and Applied Geophysics","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Geographically weighted regression, Karabük, Land use regression, Multiple linear regression, Spatial modeling, Sulfur dioxide","lastPublishedDoi":"10.21203/rs.3.rs-8450115/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8450115/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eIn this study, a comparative analysis of Multiple Linear Regression (MLR) and Geographically Weighted Regression (GWR) models was applied within the scope of Land Use Regression (LUR) analysis in order to reveal the spatial distribution of sulfur dioxide (SO\u003csub\u003e2\u003c/sub\u003e) concentration in Karab\u0026uuml;k city. 5-fold cross-validation and external data validation methods were used to test the accuracy of the models. The MLR model successfully represented the distribution of \u003csub\u003eSO2\u003c/sub\u003e concentrations in general. However, due to the application of fixed coefficients, spatial heterogeneity could not be fully revealed. In the GWR analysis, local differences were revealed more accurately thanks to the coefficient produced separately for each sample point. In the MLR model, the overall accuracy is R2:0.85, the 5-fold cross-validation method accuracy is R2:0.85, and the external data validation method accuracy is R2:0.92. In the GWR model, the overall accuracy is R2:0.95, the 5-fold cross-validation method accuracy is R2:0.95, and the external data validation method accuracy is R2:0.94. In both methods, high \u003csub\u003eSO2\u003c/sub\u003e concentrations were observed in the southern part of the city where industrial areas are dense. In the local R2 distribution of the GWR model, high explanation values ​​were again obtained in the southern regions. In the residual analysis of the GWR model, it was observed that the prediction errors of the model were in the range of \u0026plusmn;\u0026thinsp;1 \u0026micro;g/m\u0026sup3; and were randomly distributed except for a few small local regions. According to the results, the GWR model performed better than the MLR model in both predictive accuracy and spatial heterogeneity.\u003c/p\u003e","manuscriptTitle":"Modeling and comparative analysis of spatial distribution of SO 2 concentration using MLR and GWR models: A case study of Karabük, Türkiye","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-08 14:07:16","doi":"10.21203/rs.3.rs-8450115/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-01-28T08:49:39+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-01-27T18:03:30+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"206080076615500310485402136481089551654","date":"2026-01-09T14:29:34+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"23826711186797646440455980811490403243","date":"2026-01-08T06:25:44+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"198513258574284235462743773759866174552","date":"2026-01-07T13:41:14+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-01-06T11:49:33+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"56802588960021628078409890701474261274","date":"2026-01-05T15:48:43+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"65787447245162634659668625565245443961","date":"2026-01-05T13:32:39+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-01-05T12:01:01+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-12-29T07:29:40+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-12-26T11:40:34+00:00","index":"","fulltext":""},{"type":"submitted","content":"Pure and Applied Geophysics","date":"2025-12-25T16:07:05+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"pure-and-applied-geophysics","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"paag","sideBox":"Learn more about [Pure and Applied Geophysics](https://www.springer.com/journal/24)","snPcode":"24","submissionUrl":"https://submission.nature.com/new-submission/24/3","title":"Pure and Applied Geophysics","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"6c00f5a2-df2f-4e7f-ad1a-63dbc1c9b2b1","owner":[],"postedDate":"January 8th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2026-03-09T16:01:32+00:00","versionOfRecord":{"articleIdentity":"rs-8450115","link":"https://doi.org/10.1007/s00024-026-03950-z","journal":{"identity":"pure-and-applied-geophysics","isVorOnly":false,"title":"Pure and Applied Geophysics"},"publishedOn":"2026-03-04 15:58:08","publishedOnDateReadable":"March 4th, 2026"},"versionCreatedAt":"2026-01-08 14:07:16","video":"","vorDoi":"10.1007/s00024-026-03950-z","vorDoiUrl":"https://doi.org/10.1007/s00024-026-03950-z","workflowStages":[]},"version":"v1","identity":"rs-8450115","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8450115","identity":"rs-8450115","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}