Introduction

Environmental noise pollution (ENP) has been recognized as a leading risk factor, threatening the health of millions globally1,2,3,4. Although research in this field is still developing, a growing body of robust studies indicates that ENP, particularly in cities originating from transportation, construction, and social activities, may impact the central nervous system1,4,5,6. These noise-related impacts increase the risk of mental health issues, such as depression, anxiety, suicidal tendencies, and behavioral disorders, particularly in children and adolescents4,6,7,8,9. For example, long-term exposure to transportation noise has led to approximately 18 million people experiencing significant annoyance and an additional 5 million suffering from severe sleep disturbances6. Furthermore, the World Health Organization reports that ENP contributes to a loss of over 1.6 million healthy life years annually in Western European countries6.

A sobering reality is that a growing body of evidence indicates children are particularly vulnerable to ENP due to their limited ability to manage noise-related stressors10,11. Additionally, research highlights concern about the cumulative lifetime effects of environmental risk factors (ERFs) on children, given the rapid and ongoing development of their organ systems10,12. Prolonged exposure to such ERF can impede cognitive development10,13. According to Jean Piaget’s theory of cognitive development, children aged 7 to 12 are in a crucial stage of cognitive growth, making them especially susceptible to the long-term effects of excessive noise, which may affect their future development10,11. Numerous studies have confirmed the detrimental impact of ENP on children’s cognitive processes, including literacy, attention, mathematics, and memory10,11.

Children spend approximately 20–30% of their daily time in schools and kindergartens10,14. Therefore, investigating ERFs such as ENP in these settings is crucial for enhancing their health in response to environmental exposure14. Such studies can offer valuable insights for policymakers to develop and implement effective interventions to reduce ENP around schools and kindergartens. As the most robust approach, in-situ measurements are often impractical and unaffordable, particularly for covering all places of interest. Consequently, modeling-based approaches, such as land use regression (LUR), are more feasible and valid, as reported numerous times in previous studies on air pollution15,16,17,18,19,20,21 and, recently, in the context of ENP5,22,23,24. This approach is used to map various ERFs of interests at non-monitored locations and illustrate its spatial variability16,17,20,22,25. In fact, this regression-based approach utilizes spatial land use data, including green spaces; residential, commercial, and industrial areas; different types of roads; terminals; bus and taxi stations; public parking lots; and population density15,16,17,18,19,20,21. Therefore, we designed this study to map ENP around all elementary schools and kindergartens in Tehran using LUR, aiming to prompt action to address the cumulative lifetime effects of ENP on children in Tehran.

Methods

Study domain and spatial distribution of elementary schools and kindergartens in Tehran

This study focused on Tehran, the capital and most populous city of Iran, with a resident population approaching 10 million, which increases to over 12 million during the daytime26. Geospatial data on the locations of elementary schools and kindergartens were obtained from Tehran’s General Department of Education and Training. This data was subsequently processed and visualized using ArcGIS software (version 10.8). Our analysis covered 895 elementary schools (Fig. 1A) and 998 kindergartens (Fig. 1B), collectively serving around 800,000 students in Tehran.

Fig. 1
figure 1

Spatial distribution of element ary schools (A), and kindergartens (B) across Tehran.

Full size image
Fig. 2
figure 2

The points where ENP was measured across Tehran (A) and the points used for creating and validating the model (B).

Full size image

Characteristics of schools and kindergartens included and ENP data

The data used in this study were measured and collected by the Tehran Air Quality Control Company (TAQCC) at 308 locations near selected primary schools and kindergartens (Fig. 2A) as ENP (10-minute equivalent noise levels) across the 22 districts of Tehran (Table 1). It should be noted that the measurements were carried out from Saturday to Wednesday—the official working days in the Iranian calendar—during the hours when children typically spend approximately seven hours (from 8:00 a.m. to 3:00 p.m.) in schools or kindergartens. This schedule was deliberately chosen to align with children’s occupancy periods in schools and kindergartens. Based on accessibility and feasibility, three 10-minute samples were taken at each site (school and kindergarten) from Saturday to Wednesday between 8:00 a.m. and 3:00 p.m. For the schools and kindergartens located along streets/highways on two sides, three measurements were conducted on each side, resulting in a total of six 10-minute samples. Since no significant difference was observed between measurements at a given location, their average was calculated and reported and we applied these averages in subsequent analyses. Of the 308 points measured, 11 were excluded from the study because they were located outside the 22 districts of Tehran. The equivalent noise levels were measured using a B&K2230 sound level meter. To meet standard measurement conditions, the device was positioned 1.5 m above the ground and within a 1.5-3-meter radius free of any obstacles, including vehicles, trees, buildings, walls, etc. Additionally, a foam windscreen was placed on the sensor to minimize the effects of wind and air disturbances. Pedestrian movement near the sensors was strictly prohibited to avoid interference.

Table 1 Summary of characteristics of schools and kindergartens included in the study for measuring ENP around them.
Full size table

Variables and LUR model development

To predict noise pollution levels around all elementary schools and kindergartens, we employed established methods for developing our LUR models16,25. It should be noted that our variable selection process combined statistical criteria, methodological considerations, and subject expertise, in line with established practices in LUR modeling15,16,18,21,25,30. Notably, among the 288 total measurements, 50 measurements - geographically distributed across all districts of Tehran - were used for model development, while the remaining 238 measurements were utilized for model fitting according to the Bland-Altman plot analysis (Fig. 2A). We generated eight buffer zones with radii of 25, 50, 75, 100, 150, 200, 250, 300, 350, 400 m around each Noise Monitoring Station and extracted 135 variables across seven categories (Table S1). These categories included green spaces, highways, roads, residential, commercial, industrial areas, fuel stations, bus and taxi stations, public parking lots, and population density, in accordance with methodologies used in previous studies21,27,28,29.

Before model construction, we evaluated the normality of all 138 variables using the Shapiro-Wilk test. For variables, we then applied Pearson correlation for normally distributed variables and Spearman correlation for non-normal variables, analyzing their relationship with the response. Variables showing a significant correlation with the response (p-value < 0.1) were identified as potential predictors for inclusion in the forward multiple regression analysis15,21,31. Third, we developed models with enhanced predictive power by utilizing refined algorithms. We did not impose constraints on the initially identified potential predictor variables. Beyond the variables that showed a significant correlation with the response variable (p-value < 0.1) in the initial models, we re-examined all 138 variables to potentially improve the model’s accuracy. As an acceptable approach applied by the previous studies15,18,21,25,30, we manually added and removed variables to determine whether those previously excluded due to high p-values (low correlation) could contribute to the new models. To clarify, some predictors (e.g., green space or length of primary roads etc.) were calculated across several buffer distances (e.g., 25 m, 50 m, 75 m, 100 m, 150 m, 200 m, 250 m 300 m, 350 m, and 400 m). Including multiple correlated versions of the same variable in the model could lead to multicollinearity issues. To address this, we first tested the distribution of the dependent and independent variables using the Shapiro–Wilk test. For normally distributed independent variables, we used Pearson correlation with the outcome, and for non-normally distributed variables, we applied Spearman correlation. For variables with multiple buffer sizes (e.g., green space across 10 buffers), we selected the buffer with the strongest correlation (or the lowest p-value) with the outcome as the representative predictor. This approach reduced redundancy while retaining the most informative buffer size. Our findings indicated that relying solely on predictor variables with a p-value < 0.1 might not be sufficient. We reported R² and adjusted R² for the models and leave-one-out cross-validation (LOOCV) for validation. Additionally, we assessed the variance inflation factor (VIF) to confirm that the variables included in the models were not collinear and that all potential predictor variables were appropriately selected15,17,21,31. Finally, the models were deemed acceptable if the R², adjusted R², LOOCV R² demonstrated minimal differences, and if the LOOCV R² showed an improvement of at least 1% compared to the previously acceptable models. All predictor variables included in our models had a VIF < 2, confirming that they were appropriately selected21. After finalizing the models, we used them to estimate noise level around elementary schools and kindergartens in Tehran.

Results

Spatial distribution of ENP around included elementary schools and kindergartens

A total of 297 schools and kindergartens across the 22 districts of Tehran were included in our study to assess ENP around these educational institutions (Table 1). Among these, 121 (approximately 40.7%) were located along streets or highways on two sides, potentially exposing them to higher levels of traffic noise. Only 30 institutions (10.1%) were equipped with double-glazed windows, which are often used as a noise reduction intervention. Notably, district 2 had the highest number of institutions included in our study (38), followed by districts 5 and 7 with 26 and 25, respectively. In contrast, district 13 had only 2 institutions represented. The prevalence of institutions with double-glazed windows varied widely, with some districts like 3, 4, 8, 12, 13, 20, 21, and 22 having no such installations at all (Table 1).

ENP showed significant variation across 22 districts of Tehran (Table 1). The lowest average of ENP was found in district 22 with 58.7 dB(A), ranging from 49.5 to 64.1 dB(A), while the highest average was in district 12 with 75.4 dB(A), ranging from 72.3 to 78.1 dB(A). Other districts with high average ENP included districts 13 (74.8 dB(A)), 14 (74.4 dB(A)), 15 (74.5 dB(A)), 16 (75.1 dB(A)), and 20 (75.1 dB(A)), indicating a potential environmental concern in these areas. In 19 out of 22 districts, the average of ENP around the schools and kindergartens exceeded even the commercial area standard of 65 dB(A). For instance, in district 12, the average of ENP is 20.4 dB(A) above the residential standard, 15.4 dB(A) above the residential-commercial standard, and 10.4 dB(A) above the commercial threshold. Similarly, districts 13, 14, 15, 16, and 20 also recorded mean noise levels exceeding the standard of ambient noise level for residential, residential-commercial, commercial, and residential-industrial areas. Even in district 22—the only district with an average Leq10min below 60 dB(A)—the recorded value of 58.7 dB(A) still surpasses the residential noise limit of 55 dB(A), and falls just within the acceptable range for residential-commercial areas (Table 1).

ENP around elementary schools and kindergartens based on LUR model

Figure 3 provides the spatial distribution of ENP around elementary schools and kindergartens in Tehran and Fig. 4 gives detailed information regarding the number and percentage of both elementary schools and kindergartens in each category of ENP in Tehran. Given Fig. 4, the results reveal considerable variation in ENP, ranging from below 55 dB(A) to above 75 dB(A) across different districts of the city. In fact, there is a clear spatial inequity in ENP around elementary schools and kindergartens, specifically by comparing those located in the north with center and south districts of the city. With respect to all elementary schools and kindergartens, the ENP around all of them exceeded 50.1 dB(A). The highest ENP—ranging from 65.1 dB(A) to 85 dB(A)—are predominantly concentrated in the central, southern, and southeastern districts of the city. However, nearly all of the elementary schools and kindergartens situated in the north experienced lower ENP, often in the range of 50.1–65 dB(A).

On the other hand (Fig. 4), the largest proportion of elementary schools and kindergartens—approximately 36% (690 out of 1893)—are exposed to ENP in the range of 70.1–75 dB(A), suggesting that over one-third of all institutions are situated in noisy environments. The category of 65–70 dB(A) includes almost 30% (575 out of 1893) of elementary schools and kindergartens, highlighting that more than a quarter of educational institutions for children in Tehran are located in areas with considerably elevated noise. Of greater concern, almost 13% of educational institutions for children in Tehran (260 elementary schools and kindergartens) was located in the category of ENP more than 75 dB(A), meaning almost one in every eight educational institutions for children in Tehran is under high ENP. Finally, only 4% (80 institutions) of schools and kindergartens are located in areas with ENP less than 60 dB(A).

Fig. 3
figure 3

Spatial distribution of environmental noise level around elementary schools and kindergartens in Tehran.

Full size image
Fig. 4
figure 4

The number (A) and percentage (B) of both elementary schools and kindergartens in each category of environmental noise level in Tehran.

Full size image

Developed LUR model

According to the final regression model presented in Table 2, seven spatial predictor variables (out of 135).

were found to statistically significant predict the ENP around elementary schools and kindergartens in Tehran. The multivariable LUR model demonstrated a strong explanatory power with an R² of 0.70 and an adjusted R² of 0.65, respectively. The LOOCV R² and RMSE values were 0.59 and 3.15, respectively, indicating acceptable predictive performance of the model. The Durbin-Watson statistic was 1.86, suggesting no serious autocorrelation among residuals. Additionally, VIF values below 2 suggest that there is no serious multicollinearity among the predictor variables. To further assess the reliability and validity of the model, three diagnostic plots—namely the Bland–Altman plot, residual plot, and observed-to-predicted ratio plot—were analyzed (Fig. 5). The Bland–Altman plot confirmed the correlation between the observed and predicted values, as no specific patterns were observed (e.g., a peak-shaped pattern (∧) or a directional trend (> or <)), and there was no evidence of systematic over- or underestimation. Additionally, most data points fell within the 95% limits of agreement, with a mean difference close to zero. According to observed-to-predicted ratio plot, most ratios were concentrated around the value of 1, indicating that the predicted values were relatively close to the observed ones.

The variables identified as significant determinants of ENP around elementary schools and kindergartens of Tehran included green space area within a 350-meter buffer (AGS), length of secondary roads within a 100-meter buffer (LoSR), area of commercial parcels within a 100-meter buffer (AoCP), the distance to the nearest military zone (DTNMZ), the distance to the nearest terminal (DTNT), the distance to the nearest primary road (DTNPR), and the distance to the nearest highway (DTNH) (Table 2). Among these variables, four—DTNT, DTNPR, DTNH, and AGS—with coefficients of approximately − 3.2, − 1.6, − 1.5, and − 0.4, respectively, had negative effects on ENP around schools and kindergartens, meaning that an increase in these variables was associated with a reduction in ENP. In contrast, the other variables, including LoSR, AoCP, and DTNMZ, showed positive effects, indicating that their increase contributed to higher noise levels.

Fig. 5
figure 5

Bland–Altman plot (A), residual plot (B), and observed-to-predicted ratio plot (C) for assessing the agreement between observed and predicted ENP.

Full size image
Table 2 The final model developed to predict environmental noise level around elementary schools and kindergartens in Tehran.
Full size table

Discussion

This study mapped ENP around 1,893 schools and kindergartens in Tehran to provide evidence supporting the need for innovative urban planning strategies. A LUR model was employed to evaluate the impact of 135 spatial predictor variables on ENP, using measured noise levels from 308 schools and kindergartens across all 22 districts of Tehran. Among these institutions, nearly 41% were located along streets or highways on at least two sides. Furthermore, the vast majority (90%) of the schools and kindergartens included in the ENP measurements lacked double-glazed windows—an affordable and practical intervention for reducing exposure to ENP. When comparing schools and kindergartens located along streets or highways on at least two sides with those located on only one side, no significant difference in ENP was observed, with average noise levels of 70.7 ± 5.0 dB(A) and 69.1 ± 5.6 dB(A), respectively. Based on the measured noise levels around schools and kindergartens across the districts of Tehran, a clear spatial inequity was observed. The lowest average ENP levels were recorded in districts 22, 4, and 2, with values of 58.7, 66.4, and 67.4 dB(A), respectively. In contrast, the highest levels were found in districts 12, 16, 20, and 13, with average ENP levels exceeding 74 dB(A).

We found that the proximity of educational institutions to urban infrastructures—such as green spaces, primary and secondary roads, highways, commercial parcels, military zones, and terminals—significantly influences ENP around them. Regarding the military areas, it should be noted that they are typically located on the outskirts of the city or in low-density regions due to security and safety considerations, with limited public access. Therefore, proximity to a military zone does not necessarily imply exposure to urban sources of noise pollution—such as heavy traffic, commercial centers, or routine urban activities. On the contrary, as one moves farther from these military zones, they are more likely to approach central and high-traffic urban areas where the presence of multiple noise-generating sources (e.g., traffic, economic activities, and population density) is greater. Thus, DTNMZ may indirectly serve as a proxy indicator for proximity to noisier and more congested urban environments.

Our study reveals that the northern districts of Tehran experience lower ENP compared to the central and southern districts. This difference can be attributed to variations in urban morphology and socio-economic status (SES)21,32,33. Most ministries and administrative buildings are concentrated in the central and southern areas of Tehran, attracting approximately three million commuters daily for business and administrative purposes21,32,33,34. This substantial influx of traffic significantly contributes to increased ENP in these areas21,35. Regarding SES, Tehran is generally divided into four categories, ranging from the northern districts with the highest SES to the southern districts with the lowest32,35. In higher SES areas, residents tend to use higher-quality vehicles, which emit less noise, thereby resulting in lower ENP levels21,35.

Our findings were consistent with previous studies regarding the spatial variations of ENP in Tehran36,37,38,39. For example, in a burden of disease estimation attributable to ENP conducted by Shamsipour and colleagues (2022), the authors showed that there was a clear spatial inequity in exposure to ENP across Tehran’s districts with more than 55 dB(A) in the central parts of the city where traffic density is higher36. This study, also, estimated 697 disability-adjusted life year per 100,000 population for ENP in Tehran. Another study conducted by Biglari and colleagues revealed that Tehran citizens were exposed to high levels of ENP, ranging from 57.4 dB(A) to 84.5 dB(A)37.

Although the application of LUR based on measurement campaigns to predict ENP at high-resolution spatial scales is increasingly recognized around the world as a well-documented approach22,23,24,40,41,42,43, there is limited evidence from Tehran and other megacities in Iran5. In fact, a study conducted by Gharehchahi and colleagues mapped ENP in Shiraz5. Therefore, we only compared our spatial predictor variables with this study and those identified in studies on ambient air pollution in Tehran. Additionally, we compared our spatial predictor variables for mapping ENP with those used in studies conducted in other countries. In the study by Gharehchahi and colleagues5, road-related variables reported as influential factors on ENP were largely similar to those in our study. These included the length of residential and secondary roads within a 20 m buffer, proximity to trunk and residential roads, the residential area within a 60 m buffer, and the length of trunk roads within a 100 m buffer5. Our findings were largely consistent with the findings of previous LUR studies on air pollution in Tehran, as they revealed that green spaces positively attenuate ambient air pollution, while proximity to secondary roads, terminals, and highways increases air pollution15,18,21.

Similarly, the study by Xu and colleagues (2022) revealed that road-related variables, urban areas, residential areas, and commercial-related variables—such as the number of restaurants—were the most significant predictors of increased ENP in Shanghai, China22. In contrast, they found that the greenery area (specifically farmland) was the only negative predictor in their model22. In São Paulo, Brazil, Raess and colleagues (2021) revealed that residential land use and green space were predictor variables with positive effects on ENP, meaning that these spatial predictors contributed to a decrease in ENP44. In contrast, they found that road-related variables, such as medium roads, had a negative effect on ENP44. Another study in Dalian Municipality, Northwest China, by Xie and colleagues (2011) revealed that green space was the only negative predictor in their model, while industrial land use and urban roads were positive spatial predictor variables30.

Implications based on our findings and previous studies

The following strategies can help reduce ENP around schools and kindergartens, thereby improving children’s health and even enhancing their educational performance.

  • We found that the proximity of educational institutions to urban infrastructures—such as green spaces, primary and secondary roads, highways, commercial parcels, military zones, and terminals—significantly influences ENP around them. Our study, consistent with previous research on promoting environmentally healthy cities, highlights the need to modify existing infrastructure. Such changes could substantially reduce the burden of disease associated with risk factors like ENP, especially among children, who represent a particularly vulnerable group45,46,47,48,49. Cities and their infrastructures not only affect citizens’ exposures but also shape their interactions with ERFs—such as ENP, air pollution, heat island, and light pollution—through various pathways, including green and walkable infrastructure45,46,47,48,49. As a result, several innovative urban models—such as the Superblock, the Compact Cities, the Car-Free Cities, the 15-Minute City, or a combination of these—have been introduced in various cities around the world to effectively address the aforementioned ERFs, particularly ENP. These models commonly lead to a reduction in private car use and an increase in both public and active/green transportation—such as walking, cycling, e-scooters, and shared mobility services—thereby promoting urban environments as more environmentally healthy cities45,46,47,48,49. Several studies have evaluated the environmental and health benefits of the superblock model (Fig. 6) implemented in the Poblenou, Sant Antoni, and Horta neighborhoods of Barcelona, Spain50,51,52,53,54. These studies have reported that Superblocks in these neighborhoods have improved citizens’ well-being and emotional health due to reduced ENP, lower ambient air pollution, and increased socialization50,51,52,53,54. For example, Mueller et al. (2019) estimated that the superblock model in Barcelona could prevent 163 deaths annually (95% CI: 83–246) due to reductions in ENP51,53,54.

  • Another implication of our findings, supported by previous research on other environmental hazards such as air pollution, is that future educational institutions for children should be sited farther from major sources of ENP—including terminals, primary roads, military zones, and commercial areas—and instead be planned in areas with greater green space15,21. For existing schools and kindergartens, it is important to highlight that, where relocation is feasible, transferring these institutions to greener and quieter areas could be considered as part of long-term urban planning strategies.

  • Given that only 10% of these institutions were equipped with double-glazed windows, the installation of such windows could be considered a practical and cost-effective intervention to reduce children’s exposure to ENP65,66.

  • In addition to the aforementioned long-term interventions, children’s exposure to ENP can also be individually mitigated through the use of earplugs or noise-cancelling headphones during outdoor activities in schoolyards52 as the effectiveness of these devices in reducing ENP exposure has been demonstrated in the general population65.

Fig. 6
figure 6

The Super block model before (baseline) and after implementation54.

Full size image

Strengths and limitations

The current study has both strengths and limitations. The best of our knowledge, this is the first study to predict ENP around all elementary schools and kindergartens in Tehran. Additionally, we enhanced the accuracy of the developed model by incorporating 135 variables across seven categories representing all land uses such as green spaces, highways, roads, residential, commercial, industrial areas, fuel stations, bus and taxi stations, public parking lots, and even population density. Hence, our findings can be valuable for policy-makers and public health bodies, aiding in the improvement of schools and kindergartens against ENP. Regarding the limitations, we did not measure ENP within the yard of these educational institutions, although we made an effort to measure this leading risk factor closest to these institutions. As mentioned earlier in the method section, our variable selection process during the production of the models was mainly manual, based on prior evidence. This procedure can lead to an increased risk of omitting counterintuitive but potentially meaningful findings.

Conclusion

We designed this modeling study to map ENP—as an ERF affecting children’s health and educational performance—around 1,893 schools and kindergartens, utilizing 135 spatial predictor variables within buffers of less than 400 m. This study aims to provide evidence supporting the need for new urban design strategies. Seven predictor variables were found to have considerable effects on ENP around these educational institutions for children. Based on the measured and predicted noise levels, all schools and kindergartens experienced ENP exceeding national standards for residential and commercial areas, as noise levels in all districts of Tehran were higher than 55 dB(A). A spatial inequity in ENP was observed across the districts of Tehran, with elementary schools and kindergartens located in the northern parts of the city experiencing lower noise levels compared to those in the central and southern districts. The predictor variables for ENP included proximity to green space areas, primary and secondary roads, commercial parcels, military zones, terminals, and highways. Based on these variables, our study highlights the need to adopt innovative urban planning strategies to address and reduce exposure to this and other ERFs.