Satellite Imaging of Global Urbanicity relates to Brain and Behavior in Young People

Urbanicity is a growing environmental challenge for mental-health. While the impact of urban life on brain and behavior might be distinct in different sociocultural conditions and geographies, there might exist features shared between regions. To investigate correlations of urbanicity with brain structure and function, neuropsychology and mental illness symptoms in young people from China and Europe, we developed a remote-sensing satellite-measure termed ‘UrbanSat’ quantifying population-density, a general measure of urbanicity. UrbanSat is correlated with brain volume, surface area and brain-network-connectivity in the medial prefrontal cortex and cerebellum, which mediate its effect on perspective-taking and depression- symptoms. Susceptibility to high population-density is greatest during childhood for the cerebellum and from childhood to adolescence for the prefrontal cortex. As UrbanSat can be generalized to different geographies, it will enable assessing the impact of urbanicity on mental illness and resilience globally, especially in young people where prevention and early interventions are most effective. of high population-density on brain structure and function well on social-cognition and mental-health both CHIMGEN and IMAGEN, their generalization other and geographies. The living on development of structural functional . Satellite data are applied to map urbanization, poverty, climate change and as well as spread of infectious disease 10 . Our study extends the application of remote-sensing satellite data and provides a method to characterize and monitor spatial and temporal patterns of risk for mental disorders. In the optimized CFA models, the four satellite- measures contributing to UrbanSat showed different factor-loadings. Night-time-light with the highest factor-loading can capture the physical environmental features of urbanicity, such as patterns of human settlements 42 , urban expansion 43 and population counts 44 , as well as information about social-environmental features of urbanicity, such as economic activity 45 . Built-up% and cropland% with medium factor-loadings mainly reect physical-environmental features of urbanicity. NDVI with the smallest factor-loading measures residential greenness and has been used extensively to record distribution of green spaces in urban settings 46 . UrbanSat mainly reects the physical- environmental features and indicates its social-environmental features of urbanicity only indirectly.


Introduction
Mental disorders account for 28% of disease burden among non-communicable diseases 1 . Environmental factors account for up to 50% of the attributable risk for mental disorders 2 . The environmental measures investigated in mental-health research include not only individual life events 3 , such as trauma, abuse, neglect, or psychosocial stress, but also, albeit to a lesser extent, individual physical environments 4 .
Urbanicity, the living conditions particular to urban areas, is among the most important environmental challenges globally 5 . While the physical environments are hallmarks of a city, urbanicity also includes the social environment and access to health and social services 5 . The physical, social and service dimensions of urbanicity form a complex relation with each other that has hitherto prevented the development of a unifying concept and measurement of urbanicity 5 .
In 1950, less than 30% of the world's population lived in urban areas, but this number has increased to presently 55% and is expected to rise to 68% in 2050 6 . While Europe is among the most stable urbanized regions, Asia is home to 54% of the world's urban population and subject to massive demographic changes: for example, by 2050 China will have added 255 million urban dwellers 7 . The increasing global urban population emphasizes the importance of investigating how the living conditions particular to urban areas affect human brain and behavior.
We were interested to investigate the relation of urbanicity with brain and behavior in different sociocultural conditions and geographies and further to identify possible susceptibility periods across the life span in young people, enabling targeted preventions when the developing brain may bene t most from environmental modi cation. Whereas there might be distinct in uences of increased population-density in urban settings in different sociocultural conditions and geographies, there are likely to exist common associations with brain and behavior shared in different areas of the globe.
Several studies have focused on the relation of individual physical environments linked to urban living with brain and behavior, such as green space, air and noise pollution 4,8 . However, a more general measure of urbanicity that can objectively assess urban environment with high spatiotemporal resolution and coverage is still lacking. Such a general measure is important, as it registers the overall and susceptibility-period effects of urbanicity on brain and behavior, and may in a subsequent step enable the identi cation and ranking of the individual features of the physical, social and service environment, and their interactions, that contribute most to the observed relation.
Traditionally, the characterization to urbanicity was carried out using census data, which are ascertained infrequently in different ways and at different times in different countries 5 . Thus, census data are less useful for comparing urbanicity across different countries. More recently, the Global Human Settlement Layer (GHSL) has provided globally standardised human settlements, including urbanisation and urbanicity 9 . GHSL data, however, are only available at large and infrequent intervals, namely 1975, 1990, 2000 and 2015 9 .
To facilitate global comparative analyses of the overall effects of urbanicity on brain and behavior and to identify potential susceptibility-periods, dense quantitative and longitudinal environmental measures that can be obtained from different geographies are required. Remotely-sensed satellite-data provide globally standardized quantitative environmental measures enabling the tracing of environmental features going back nearly 50 years 10 . Population-density is a well-established and quanti able general measure of urbanicity, frequently applied for neighbourhood classi cation and used around the globe 9 .
Here, we aimed to use population-density as a general measure of urbanicity to investigate if the urban environment is correlated with brain and behavior, and if these correlations are comparable in China and Europe. Speci cally, we developed a satellite-based measure of population-density termed 'UrbanSat', and applied UrbanSat in China and Europe to investigate the relation of population-density, as a proxy of urbanicity, with brain structure, function and behavior in two neuroimaging datasets of young people: as exploration dataset, we used the Chinese CHIMGEN cohort (www.chimgen.tmu.edu.cn) 11 and as replication dataset, we used the European longitudinal IMAGEN-cohort (www.imagen-europe.com) 12 . While we did not have any a priori hypotheses, we were interested in investigating if: (i) UrbanSat is associated with brain structure, function and behavior; (ii) Brain features associated with UrbanSat mediate the association between UrbanSat and behavior; (iii) Correlations of UrbanSat with brain and behavior are similar in Chinese and Europeans; Furthermore, we were interested in (iv) identifying susceptibilityperiods for the effects of UrbanSat during child and adolescent development on brain and behavior. A schematic summary is shown in Fig. 1.

Demographics
We recruited young-adult participants with lifetime-residential geographies from CHIMGEN (n=3306) and IMAGEN second follow-up (FU2) (n=561). Detailed inclusion and exclusion criteria are presented in Supplementary Tables 1 and 2. Demographics of samples used in statistical analyses and sample attrition are described in Supplementary Table 3  UrbanSat: a satellite-based measure of urbanicity To develop a satellite-based measure of urbanicity, we selected information from nine types of satellite registrations relevant for detecting and characterising urban settlements, including night-time light (NL), normalized difference built-up index (NDBI), normalized difference water index (NDWI), normalized difference vegetation index (NDVI), and ve measures derived from land cover mapping (Built-up%, cropland%, grassland%, forest% and water body%) (Online and Supplementary Methods, Supplementary Table 7). After imputing the nine annual satellite registrations for the participants with missing values using Bayesian data augmentation (Supplementary Table 8 and Online Methods), we carried out a ten-fold cross validation strati ed by spatio temporality to optimize the con rmatory factor analysis (CFA) models, and to predict annual UrbanSat scores of each participant from birth to the age of recruitment (Online Methods). UrbanSat was generated by the optimized CFA model consisting of NL, built-up%, cropland% and NDVI, which best captured variation of urban features while maximizing goodnessof-t. UrbanSat in CHIMGEN (Supplementary Tables 9-10) and IMAGEN-FU2 (Supplementary Tables 11-12) had a Tucker-Lewis-Index (TLI) and comparative-fit-index (CFI)>0.95, root mean-square-error of approximation (RMSEA)<0.06 and standard root mean-square-residual (SRMR)<0.08, indicating excellent model fit (Online Methods). UrbanSat was robust across time and geographies, as validated by its correlations with ground-level population-density from GHSL-POP 9 for China and Europe for the years 1990, 2000 and 2015 (Fig.2). Histograms of the distribution of UrbanSat score in each center are shown in Supplementary  Fig.3. UrbanSat showed higher correlations with population-density in different residential categories (rural, town, city and overall), countries (Asia and Europe) and years (1990, 2000 and 2015) than any individual satellite-measures (Fig. 2).
Voxel-wise multiple-regression of mean UrbanSat before age 18 with brain gray-matter-volume (GMV) was performed in CHIMGEN (n=2176). We controlled for age in all analyses, thus accounting for the older and wider age spread in CHIMGEN (age: 23.54±2.33 years) compared to IMAGEN (age: 18.89±0.66 years). We also controlled throughout for gender, education, site, body-mass-index (BMI), genetic population-strati cation and socioeconomic status (SES) (Supplementary Tables 13-14). Total intracranial volume was controlled in all imaging analyses, except for the analyses of cortical thickness (CT) and surface area (SA), where mean CT and total SA were controlled, respectively. Parental history of mental illness was an exclusion criterion for CHIMGEN and controlled for in IMAGEN. Uncorrected statistical maps of the association of UrbanSat with brain GMV in CHIMGEN adjusting for confounding covariates under parametric testing and non-parametric permutation testing are shown in Fig.3a and Supplementary Fig. 4. We found negative correlation of UrbanSat with medial prefrontal cortex (mPFC) volume (peak MNI-coordinate: x=-7.5, y=30, z=45; 676 voxels; peak t-value=-6.42; Fig.3b) and a positive correlation with cerebellar volume (peak MNI-coordinate: x=10.5, y=-51, z=-18; 978 voxels; peak t-value=6.22; Fig.3b) (Parametric testing P c <0.05, family-wise error (FWE) corrected for voxel numbers, imaging modalities and data categories; see Online Methods). We con rmed the results with non-parametric permutation-testing (TFCE-FWE, Pc<0.05, Supplementary Methods and Supplementary Fig.4). Potential imputation-bias was ruled-out by sensitivity-analyses in 1491 participants with complete satellite and neuroimaging data  Table 15). Using voxel-wise multiple regression of individual satellite-measures with GMV, we found signi cant correlation with mPFC-GMV being driven by NL and built-up%, and correlation with cerebellar-GMV being driven by NL and cropland%. We found no correlation of NDVI with either mPFC-or cerebellar-GMV ( Supplementary Fig.8). We observed similar results in GHSL ground-level population-density data, thus validating the relation of UrbanSat and GMV ( Supplementary Fig.8).
In IMAGEN-FU2 (n=415), we replicated CHIMGEN ndings. The uncorrected statistical correlation map of UrbanSat with brain GMV in CHIMGEN showed signi cant spatial correlation (r=0.40, P<0.001) with that of IMAGEN-FU2 ( Supplementary Fig.9).  Table 16). Thus, the observed correlation between UrbanSat and brain structure is robust across geographies and socio-cultural conditions. We applied distributed lag models (DLMs) to identify susceptibility-periods of lifetime UrbanSat on GMV and SA (Online Methods). We observed a negative association of UrbanSat with mPFC-GMV from age 4 to 15 (Fig.3c) and SA from age 5 to 7 (Fig.3c), indicating a susceptibility-period during childhood and adolescence, driven by NL, built-up%, cropland% and NDVI ( Supplementary Fig.13). Correlation of UrbanSat with cerebellar GMV was signi cant from age 1 to 10 years, indicating a susceptibility-period during childhood (Fig.3c), driven by NL, built-up% and cropland% ( Supplementary Fig.13).
To measure the relation between age of migration and brain structure, we split CHIMGEN participants into who migrated to the city before age 14 years (n=229, mean-age at migration=8.24±4.86 years), after age 14 (n=1385, mean-age at migration=17.17±2.68 years), and life-long city-dwellers (n=562) (Fig.3d). We found that participants born in the city or early migrants showed smaller mPFC-GMV (P=0.040) and SA (P=7.28×10 -9 ) as well as greater cerebellar-GMV (P=5.00×10 -5 ) than those with later exposure (Fig.3e and Supplementary Table 17).

No correlation of UrbanSat with white-matter microstructure
Using tract-based spatial statistics (TBSS) analysis of diffusion-tensor imaging (DTI) data, we did not nd signi cant correlation of UrbanSat with brain fractional anisotropy (FA) in either CHIMGEN or IMAGEN-FU2 (TFCE-FWE P c <0.05).
The correlations of UrbanSat with WNFCs and BNFCs were stable in a meta-analysis of all CHIMGEN and IMAGEN sites (Supplementary Fig.12 and Supplementary Table 16). Brain localization ( Fig.4b and Fig.4f) and susceptibility-periods ( Fig.4d and Fig.4h) of WNFCs and BNFCs in CHIMGEN and IMAGEN were consistent with those observed for brain structure ( Supplementary Fig.15-17), except for non-signi cant during adolescence. The WNFCs and BNFCs changes between 14 and 19 years in IMAGEN were correlated with UrbanSat ( Fig.4b and Fig.4f and Supplementary Table 15 and Supplementary Results). In CHIMGEN, WNFCs and BNFCs were correlated with age of migration to the city (Fig. 4c and Fig.4g and Supplementary Results).

Correlations of UrbanSat with behavior
We investigated whether UrbanSat is related to measures of cognition and mental-health, i.e. depression and anxiety. In CHIMGEN (n=2148), the social-cognition measure 'perspective-taking', perceiving a situation from an alternative point of view 16 , was positively correlated with UrbanSat (reaction-time for perspective-taking: rho=-0.14, P c <0.05, Bonferroni-corrected for datacategories and 21 behavioral assessments, see Online Methods) and replicated in IMAGEN-FU2 (rho=0.14, P c <0.05) ( Table 1). A negative correlation between UrbanSat and reaction-time for perspective-taking performance was observed from 12 to 22 years in CHIMGEN (Fig.5a).
In CHIMGEN and IMAGEN, increased NL and built-up%, decreased NDVI and cropland% were signi cantly correlated with enhanced perspective-taking performance and increased depression-symptoms (Table 1). The susceptibility-periods for individual satellite-measures were similar to UrbanSat in CHIMGEN ( Supplementary Fig.18). Although most correlations of UrbanSat with brain and behaviour were consistent between males and females, some correlations, especially with brain development in IMAGEN BL-FU2, were sex-speci c (Supplementary Tables 19-20).

Discussion
Using a remote-sensing satellite-measure, 'UrbanSat', we characterized the relation of population-density, a proxy of urbanicity, with brain structure, function and behavior during childhood and adolescence in large datasets in China and Europe. We provide converging evidence for association of UrbanSat during childhood and adolescence with GMV and SA of the mPFC and aDMN, but not with CT and FA. The mPFC and aDMN mediate the correlation between UrbanSat and improved perspective-taking and increased depression-symptoms. We also found positive correlations of UrbanSat during childhood with cerebellar volume, which mediated the association with perspective-taking and depression-symptoms. We are extending previous observations reporting an association of depression-symptoms with urban settings 17 , by demonstrating the stability of this observation in different geographical and sociocultural regions, and by discovering possible underlying brain mechanisms and susceptibilityperiods during childhood and adolescent development.
Our results suggest urban living has both bene cial and adverse correlations with health: enhanced social cognition (perspective-taking) and increased depression-symptoms, in contrast to previous studies, which mainly reported adverse aspects of urbanicity 18 . The mPFC, the core brain area of the aDMN, has been implicated in a variety of social-cognition and affective functions commonly compromised in psychiatric disorders 19 . The susceptibility of mPFC to urban environment is supported by the greater sensitivity of the mPFC, to urbanicity-related risk-factors, including chronic stress 20 and air-pollution 21 .
While our ndings are consistent with reports of an association between urbanicity and mPFC in smaller European samples 22 , they differ from these studies as we found associations with GMV and SA rather than CT, and an absence of sex speci city.
We found a positive correlation of UrbanSat with cerebellar volume, a mediator for the association of UrbanSat with perspectivetaking and depression-symptoms. The functional network connectivity of the cerebellum also mediates the association of UrbanSat with perspective-taking. Cerebellar lesions cause the 'Cerebellar-Cognitive-Affective-Syndrome', characterized by impairments in executive-function and memory, as well as affect 23 . Animal studies extend these ndings to stress-dependent depressive affect 24 and impairment in social behavior 25 . It is tempting to speculate that these pathways may connect to brain regions involved in perspective-taking and depression-symptoms 26 . Imaging features related to cerebellum showed susceptibility-periods to high population-density at the age of 1-10 years, during which cerebellum and cortex are increasing in volume [27][28][29] .
While previous studies focused on the effect of mean exposure to urban-living on brain and mental-health 30 , we identi ed neurodevelopmental periods with increased susceptibility to urban-living. Consistent with observations of susceptibility-periods of non-affective psychosis to residential mobility during childhood and adolescence 31 , we found that structure and function of the mPFC, as well as depression-symptoms have pronounced susceptibility to high population-density during childhood and adolescence, a period more sensitive to social stress 32 . The correlation of UrbanSat with mPFC-GMV, and change rate, was driven by SA rather than CT, indicating that SA may be more sensitive to environmental factors than CT. Perspective-taking was more sensitive to high population-density during adolescence and young adulthood, implying a time-window for neurobehavioural interventions targeting social-cognition.
Our results are suggestive of a cumulative effect of urbanicity on brain and behavior, whereby participants born or migrating to the city at an earlier age had more pronounced effects than those who become city dwellers later. Given that CHIMGEN participants were students who moved to cities for their studies, we do not have any data on people who after spending some years in the city moved back to the country side. We also do not have data to distinguish possible short but extreme exposure to urban life, in utero or during susceptibility periods from a moderate continuous exposure.
We found several shared effects of high population-density in urban settings on brain structure and function as well as on social-cognition and mental-health in both CHIMGEN and IMAGEN, indicating their generalization to other sociocultural conditions and geographies. The effect of urban living on brain development during adolescence was con rmed by exploring the correlation of UrbanSat with brain structural and functional changes from age 14 to 19 in IMAGEN. Taking into account normative references [27][28][29] , our observations are consistent with an accelerated development in densely populated urban areas of cerebellum during childhood and mPFC during childhood and adolescence. We also found inconsistent results between CHIMGEN and IMAGEN: only 4/49 BNFCs correlating with UrbanSat in CHIMGEN were replicated in IMAGEN. The more extensive effects of urban living on BNFCs in CHIMGEN may re ect the more drastic changes in urbanization in China compared to Europe 6 , but may also relate to confounding factors beyond the covariates controlled in our study 33 .
UrbanSat was correlated with GMV, SA and functional connectivity, but not with FA and CT, indicating different sensitivities of brain properties to residential environments. UrbanSat showed positive correlation with cerebellar volume, negative correlation with mPFC volume, but non-signi cant correlation with volumes of other regions, suggesting different spatial sensitivities to residential environments; the cerebellum was sensitive to urban residential environments during childhood, but the mPFC sensitive during both childhood and adolescence, indicating different temporal sensitivities to residential environments. This framework of different spatial and temporal sensitivities to urban residential environments may help to understand the association of urban living with brain and mental-health.
High population-density a general measure of urbanicity, can cause increased social stress and air pollution, both of which affect brain structure in young people 34,35 . A recent study has observed an association of urbanicity with brain activity in regions linked to social stress processing 30 . Such brain changes may mediate the well-established impact of urbanicity on mental-health, including on mood disorders 36 and social-cognition 37 . Stress in childhood can accelerate brain development and lead to faster maturation of certain brain regions during adolescence, including cerebellum and the mPFC 38 . Faster brain maturation results in enhanced cognitive development 39 and may account in part for the positive correlation of urbanicity and perspective-taking observed in our study. However, faster maturation of the mPFC and cerebellum may come at a cost of decreased plasticity, including of fear extinction mechanisms (mPFC), which may contribute to increased vulnerability for anxiety and depression 40 . Air pollution induces neuroin ammation in the brain, leading to the damage and loss of neural tissue in the prefrontal cortex 35 and may provoke depression-symptoms 41 . Thus, urban upbringing may cause affective and anxiety symptoms by way of both, increased social stress and pollution.
Remotely-sense satellite data play a critical role in monitoring the Earth's surface to track environmental conditions that are intimately related to human health 10 . Satellite data are applied to map urbanization, poverty, climate change and pollution, as well as spread of infectious disease 10 . Our study extends the application of remote-sensing satellite data and provides a method to characterize and monitor spatial and temporal patterns of risk for mental disorders. In the optimized CFA models, the four satellite-measures contributing to UrbanSat showed different factor-loadings. Night-time-light with the highest factorloading can capture the physical environmental features of urbanicity, such as patterns of human settlements 42 , urban expansion 43 and population counts 44 , as well as information about social-environmental features of urbanicity, such as economic activity 45 . Built-up% and cropland% with medium factor-loadings mainly re ect physical-environmental features of urbanicity. NDVI with the smallest factor-loading measures residential greenness and has been used extensively to record distribution of green spaces in urban settings 46 . UrbanSat mainly re ects the physical-environmental features and indicates its social-environmental features of urbanicity only indirectly.
For privacy reasons, our satellite-measures were obfuscated to a spatial resolution of one kilometer, preventing the capture of important aspects of urban life, such as daily mobility paths. Future studies will investigate the integrated effect of urban physical and social environment, their interaction with genetics and relation to brain and behavior. This study is not epidemiological but neurobiological, aiming to identify brain mechanisms by which urbanicity in uences behavior. How representative the identi ed mechanisms are among the general population is a different task for future epidemiological studies.
Our ndings were made possible due to recent advancements in remote-sensing satellite technologies which were leveraged to measure the relation of urbanicity with brain and behavior. We were able to (1) apply a general measure of urbanicity, population-density, which is not dependent on census de nitions of urban areas that might be con ated by densely populated rural areas, or sparsely populated areas within urban settlements, or may vary between nations 5 ; (2) obtain a high spatial and temporal resolution 10 ; and (3) use a measure applicable anywhere on earth from 1970s to the present day. Thus, UrbanSat provides a unique opportunity to identify the cumulative effects and susceptibility-periods of urbanicity on brain and behavior. However, we note that UrbanSat cannot unravel environmental pathways and their interactions that cause the aversive effects of urban living. This is a task for subsequent studies with access to su cient ground level data for comprehensive characterization of causal environmental pathways that underlie the observed correlations.
In the current work, we have provided proof of principle establishing the use of satellite-data to inform the relation between urban environment, brain and behavior. As our approach can be extended and generalized to other geographies and is easy to implement even in the absence of detailed or directly comparable ground level data, it may be relevant for public health, policy and urban planning globally. 2.26×10 -11 (-0.14) 6.09×10 -11 (-0.14) 3.53×10 -9 (-0.13) 2.35×10 -10 (0.14) 4.62×10 -10 (0.14) RT  index (NDWI), normalized difference vegetation index (NDVI) and ve measures derived from land cover mapping (Built-up%, cropland%, grassland%, forest% and water body%). b. Ten-fold cross validation of con rmatory factor analysis (CFA) strati ed by spatiotemporality was applied to predict annual UrbanSat score for each participant. The optimized CFA model includes NL, built-up%, cropland% and NDVI. The mean UrbanSat scores before 18 years showed higher correlation with ground level population-density from global human settlement layers (GHSL) than any individual satellite-measures both in CHIMGEN and IMAGEN-FU2. c. Investigation of the cumulative effects of UrbanSat on brain and behavior. d. Identi cation of susceptibility periods of lifetime UrbanSat on brain and behavior using distributed lag models. S, satellite-measures of urbanicity; Sub, subjects; Y, years old. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.  Correlations of UrbanSat with brain structure. a. Uncorrected statistical maps in the voxel-wise multiple regression of mean UrbanSat before 18 years with brain GMV under parametric testing in CHIMGEN (n=2176). b. In CHIMGEN, UrbanSat is negatively (blue) correlated with mPFC GMV (left) and positively correlated with cerebellar GMV (right) (FWE Pc<0.05). The correlation of UrbanSat with mPFC GMV is driven by SA rather than CT and these correlations are replicated in IMAGEN-FU2 (n=415); UrbanSat is correlated with brain volumetric change in the mPFC (n=340) between 14 and 19 years in IMAGEN BL-FU2,