Mortality and economic burden of PM2.5 and NO2 in Thailand using satellite remote sensing and Random Forest algorithms

dc.contributor.authorKhempunjakul T.
dc.contributor.authorPhosri A.
dc.contributor.authorSangkharat K.
dc.contributor.authorThongphunchung K.
dc.contributor.authorKanchanasuta S.
dc.contributor.authorPatthanaissaranukool W.
dc.contributor.correspondenceKhempunjakul T.
dc.contributor.otherMahidol University
dc.date.accessioned2025-11-16T18:10:47Z
dc.date.available2025-11-16T18:10:47Z
dc.date.issued2025-12-01
dc.description.abstractAir pollution remains a leading environmental health problem in Thailand, where rapid urbanization, biomass burning, and industrial activity contribute to high concentrations of fine particulate matter (PM<inf>2.5</inf>) and nitrogen dioxide (NO<inf>2</inf>). However, most existing studies rely upon data from fixed-site monitoring stations, leaving large areas underrepresented. This study aimed to develop satellite-based Random Forest (RF) models to estimate daily concentrations of PM<inf>2.5</inf> and NO<inf>2</inf> across Thailand from 2018 to 2022 using Aerosol Optical Depth (AOD) data from Terra and Aqua satellites, and NO<inf>2</inf> from TROPOspheric Monitoring Instrument (TROPOMI) aboard the Sentinel-5 Precursor satellite. These estimated concentrations were then linked to health and economic impacts using concentration-response functions and the Value of Statistical Life (VSL). The RF models demonstrated high performance. For PM<inf>2.5</inf>, the model achieved R<sup>2</sup> of 0.94 and RMSE of 7.40 µg m<sup>-3</sup> in training, and R<sup>2</sup> of 0.71 with RMSE of 15.09 µg m<sup>-3</sup> in testing. For NO<inf>2</inf>, R<sup>2</sup> values were 0.94 and 0.73, with RMSE of 2.21 and 4.41 ppb, respectively. Nationwide mean concentrations during the study period were 29.51 ± 2.82 µg m<sup>-3</sup> for PM<inf>2.5</inf> and 4.61 ± 0.81 ppb for NO<inf>2</inf>, with pronounced regional and seasonal contrasts. Specifically, PM<inf>2.5</inf> peaked in northern provinces during the dry season, while NO<inf>2</inf> levels were concentrated in Bangkok and industrial regions. Modeled concentrations were higher than ground-based averages as the model captures unmonitored high-pollution areas, particularly in the north. Long-term exposure to PM<inf>2.5</inf> and NO<inf>2</inf> was associated with 20,487 deaths (95 % CI: 12,833–28,079) and 15,394 deaths (95 % CI: 10,281–20,487), respectively, corresponding to economic losses of 2313 million THB (95 % CI: 1449–3171) and 1738 million THB (95 % CI: 1161–2313) annually, respectively. This study provides a reliable tool for nationwide air quality monitoring and health impact assessment and supports development of sustainable environmental and public health strategies.
dc.identifier.citationEnvironmental Challenges Vol.21 (2025)
dc.identifier.doi10.1016/j.envc.2025.101366
dc.identifier.eissn26670100
dc.identifier.scopus2-s2.0-105020835975
dc.identifier.urihttps://repository.li.mahidol.ac.th/handle/123456789/113007
dc.rights.holderSCOPUS
dc.subjectEnvironmental Science
dc.titleMortality and economic burden of PM2.5 and NO2 in Thailand using satellite remote sensing and Random Forest algorithms
dc.typeArticle
mu.datasource.scopushttps://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=105020835975&origin=inward
oaire.citation.titleEnvironmental Challenges
oaire.citation.volume21
oairecerif.author.affiliationMahidol University
oairecerif.author.affiliationThailand Ministry of Public Health
oairecerif.author.affiliationMinistry of Higher Education, Science, Research and Innovation

Files

Collections