Ultra-high-resolution modeling of ground-level ozone for long-term exposure risk assessment driven by GTR-transformer

To address the demand for fine-grained, long-term exposure assessment of ground-level ozone pollution in China, this study overcomes the limitations of traditional kilometer-level resolution modeling by developing the geo-temporal transformer with residual enhancement (GTR-Transformer) model. This m...

Full description

Saved in:
Bibliographic Details
Main Authors: Jiahuan Chen, Heng Dong, Zili Zhang, Sicong He
Format: Article
Language:English
Published: Elsevier 2025-08-01
Series:Environment International
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S0160412025004568
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849247464949809152
author Jiahuan Chen
Heng Dong
Zili Zhang
Sicong He
author_facet Jiahuan Chen
Heng Dong
Zili Zhang
Sicong He
author_sort Jiahuan Chen
collection DOAJ
description To address the demand for fine-grained, long-term exposure assessment of ground-level ozone pollution in China, this study overcomes the limitations of traditional kilometer-level resolution modeling by developing the geo-temporal transformer with residual enhancement (GTR-Transformer) model. This model integrates bidirectional temporal attention and spatial encoding techniques to construct a dataset of monthly average ozone concentrations at an ultra-high (100-m) resolution across representative regions of China from 2020 to 2023. The model exhibits outstanding performance in multi-dimensional validation, achieving a coefficient of determination of up to 0.94 and a root mean square error as low as 7.59 μg/m3. Employing SHapley Additive exPlanations (SHAP) analysis, we identify key drivers of ozone, including 2 m dewpoint temperature and surface solar radiation downwards (SHAP: 4.88 and 3.83, respectively), substantiating the photochemical regulation mechanism. Moreover, across scales spanning regions, counties, neighborhoods, and even blocks, our dataset reveals distinct ozone distribution patterns and concentration gradients that strongly correlated with road density. Furthermore, the 100-m resolution ozone maps showed an absolute difference of up to 8.9 percentage points in county-scale exposure assessment compared to 1-km resolution data in 2020. This disparity is further magnified when combined with the same ultra-high-resolution population data. These findings demonstrate that the proposed model and ultra-high-resolution maps provide actionable support for precise ozone control and exposure assessment.
format Article
id doaj-art-41efe7bf51f9448e91ff4d6c7e138878
institution Kabale University
issn 0160-4120
language English
publishDate 2025-08-01
publisher Elsevier
record_format Article
series Environment International
spelling doaj-art-41efe7bf51f9448e91ff4d6c7e1388782025-08-20T03:58:11ZengElsevierEnvironment International0160-41202025-08-0120210970510.1016/j.envint.2025.109705Ultra-high-resolution modeling of ground-level ozone for long-term exposure risk assessment driven by GTR-transformerJiahuan Chen0Heng Dong1Zili Zhang2Sicong He3School of Resource and Environmental Engineering, Wuhan University of Technology, Wuhan 430070, ChinaSchool of Resource and Environmental Engineering, Wuhan University of Technology, Wuhan 430070, ChinaZhejiang Ecological Environment Monitoring Center, Hangzhou 310012, China; Zhejiang Key Laboratory of Ecological Environment Monitoring, Early Warning and Quality Control Research, Hangzhou 310012, ChinaSchool of Resource and Environmental Engineering, Wuhan University of Technology, Wuhan 430070, China; Corresponding author.To address the demand for fine-grained, long-term exposure assessment of ground-level ozone pollution in China, this study overcomes the limitations of traditional kilometer-level resolution modeling by developing the geo-temporal transformer with residual enhancement (GTR-Transformer) model. This model integrates bidirectional temporal attention and spatial encoding techniques to construct a dataset of monthly average ozone concentrations at an ultra-high (100-m) resolution across representative regions of China from 2020 to 2023. The model exhibits outstanding performance in multi-dimensional validation, achieving a coefficient of determination of up to 0.94 and a root mean square error as low as 7.59 μg/m3. Employing SHapley Additive exPlanations (SHAP) analysis, we identify key drivers of ozone, including 2 m dewpoint temperature and surface solar radiation downwards (SHAP: 4.88 and 3.83, respectively), substantiating the photochemical regulation mechanism. Moreover, across scales spanning regions, counties, neighborhoods, and even blocks, our dataset reveals distinct ozone distribution patterns and concentration gradients that strongly correlated with road density. Furthermore, the 100-m resolution ozone maps showed an absolute difference of up to 8.9 percentage points in county-scale exposure assessment compared to 1-km resolution data in 2020. This disparity is further magnified when combined with the same ultra-high-resolution population data. These findings demonstrate that the proposed model and ultra-high-resolution maps provide actionable support for precise ozone control and exposure assessment.http://www.sciencedirect.com/science/article/pii/S0160412025004568Ground-level ozoneSpatiotemporal modelingUltra-high resolutionExposure assessmentSHAPTransformer
spellingShingle Jiahuan Chen
Heng Dong
Zili Zhang
Sicong He
Ultra-high-resolution modeling of ground-level ozone for long-term exposure risk assessment driven by GTR-transformer
Environment International
Ground-level ozone
Spatiotemporal modeling
Ultra-high resolution
Exposure assessment
SHAP
Transformer
title Ultra-high-resolution modeling of ground-level ozone for long-term exposure risk assessment driven by GTR-transformer
title_full Ultra-high-resolution modeling of ground-level ozone for long-term exposure risk assessment driven by GTR-transformer
title_fullStr Ultra-high-resolution modeling of ground-level ozone for long-term exposure risk assessment driven by GTR-transformer
title_full_unstemmed Ultra-high-resolution modeling of ground-level ozone for long-term exposure risk assessment driven by GTR-transformer
title_short Ultra-high-resolution modeling of ground-level ozone for long-term exposure risk assessment driven by GTR-transformer
title_sort ultra high resolution modeling of ground level ozone for long term exposure risk assessment driven by gtr transformer
topic Ground-level ozone
Spatiotemporal modeling
Ultra-high resolution
Exposure assessment
SHAP
Transformer
url http://www.sciencedirect.com/science/article/pii/S0160412025004568
work_keys_str_mv AT jiahuanchen ultrahighresolutionmodelingofgroundlevelozoneforlongtermexposureriskassessmentdrivenbygtrtransformer
AT hengdong ultrahighresolutionmodelingofgroundlevelozoneforlongtermexposureriskassessmentdrivenbygtrtransformer
AT zilizhang ultrahighresolutionmodelingofgroundlevelozoneforlongtermexposureriskassessmentdrivenbygtrtransformer
AT siconghe ultrahighresolutionmodelingofgroundlevelozoneforlongtermexposureriskassessmentdrivenbygtrtransformer