{"id":601,"date":"2021-02-03T09:07:38","date_gmt":"2021-02-03T01:07:38","guid":{"rendered":"http:\/\/tapdata.org\/?page_id=601"},"modified":"2026-02-27T17:43:59","modified_gmt":"2026-02-27T09:43:59","slug":"methods","status":"publish","type":"page","link":"http:\/\/tap-dev.makenv.com\/?page_id=601&lang=en","title":{"rendered":"Methods"},"content":{"rendered":"\n<style>\n#content-core\n{\nbackground:#fff;\nbox-shadow: 0px 3px 3px rgba(35, 35, 35, 0.15);\npadding:30px;\nmargin-top:30px;\n}\n#content-core .wp-block-column:not(:first-child) {\n   margin-left: 0px;\n}\n#content-core h3{\ncolor:#3EB457;\n}\n.ffx-menuOne {\n  width:100%;\n  padding: 5px;\n  font-size: 16px;\n  font-weight: 600;\n  line-height: 20px;\n  color: #333333;\n  background: #F7F7F7;\n  margin-bottom:10px;\n}\n.ffx-submenu {\n  display: flex;\n  flex-direction: row;\n  align-items: center;\n  padding: 0px 0px 0px 15px;\n  gap: 15px;\n  margin-bottom:10px;\n  position: relative;\n}\n.ffx-submenu.on .triangle {\n   display: block;\n}\n.ffx-submenu.on a{\n  color: #4ED06A;\n}\n.ffx-submenu a{\n  color: #666666;\n  font-family: \u5fae\u8f6f\u96c5\u9ed1;\n  font-weight: 400;\n  font-size: 14px;\n  line-height: 18px;\n  padding: 5px 0;\n  text-decoration: none;\n}\n.ffx-submenu a:hover {\n    color: #4ED06A;\n    text-decoration: none;\n}\n.triangle {\n    position: absolute;\n    display: none;\n    right: 10px;\n    top: 0px;\n    width: 0;\n    height: 0;\n    border-left: 4px solid #22e000;\n    border-bottom: 4px solid transparent;\n    border-top: 4px solid transparent;\n}\n.wp-block-columns .wp-block-column.menus {\n   position: relative;\n}\n#left-sidebar {\n  display: none;\n  position: absolute;\n  top: 0;\n  left: 0;\n  width: 100%; \/* \u4e0e\u539f\u6709flex-basis\u4fdd\u6301\u4e00\u81f4 *\/\n  max-height: calc(100vh - 150px);\n  overflow-y: auto;\n  z-index: 100;\n  background: #fff;\n  padding: 15px;\n  border-radius: 8px;\n  box-shadow: 0 4px 12px rgba(0,0,0,0.1);\n}\n.wp-block-columns .wp-block-column.artCss {\n   position: relative;\n   \n}\n#top-sidebar {\n  position: absolute;\n  top: 0;\n  left: 0;\n  width: 100%;  \n  padding-bottom: 10px;\n  margin-bottom: 10px;\n  background: #fff;\n  z-index: 9;\n}\n#top-sidebar .type-sidebars {\n  padding: 10px 0;  \n}\n#top-sidebar .title {\n font-size: 16px;\n font-weight: 600;\n color: #333333;\n margin-right:10px;\n}\n#top-sidebar .sidebars a {\n   font-size: 14px;\n   color: #333;\n   border: 1px solid #F1F1F1;\n   padding: 5px 10px;\n   font-weight: 400;\n   margin: 0px 5px;\n   text-decoration: none;\n   background: #F1F1F1;\n   border-radius: 4px;\n}\n#top-sidebar .sidebars a:hover {\n    background: #3EB457;\n    border-color: #3EB457;\n    color: #FFFFFF;\n}\n#top-sidebar .sidebars a.active {\n    background: #3EB457;\n    border-color: #3EB457;\n    color: #FFFFFF;\n    cursor: default;\n}\n#top-sidebar .sidebars a:first-child {\n  \n}\n#top-sidebar .sidebars a:last-child {\n \n}\n@media (max-width: 1140px) {\n    .wp-block-columns .backtop {\n        right: 10px;\n    }\n}\n\n.backtop {\n    display: none;\n    position: fixed;\n    right: calc((100% - 1140px) \/ 2 + 10px);\n    bottom: 20%;\n    width: 60px;\n    height: 60px;\n    border-radius: 10px;\n    font-size: 14px;\n    color: #333;  \n    line-height: 28px;\n    text-align: center;\n    padding: 10px;\n    gap: 6px;\n    background: #F1F1F1;\n    box-sizing: border-box;\n    border: 1px solid #CCCCCC;\n    cursor: pointer;\n}\n.backtop .arrow {\n   display: block;\n   width:100%;\n   height:16px;\n   background-image: url('data:image\/png;base64,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');\nbackground-repeat: no-repeat;\nbackground-position: center center;\nbackground-size: contain;\n}\n.backtop .text {\n}\n.backtop:hover {\n   background: #FFFFFF;\n}\n<\/style>\n<script>\nfunction changeMenu(event) {\n\nif(document.querySelectorAll('#top-sidebar .bitem') && document.querySelectorAll('#top-sidebar .bitem').length>0) {\n\ndocument.querySelectorAll('#top-sidebar .bitem').forEach(function(val) {\n  val.className = 'bitem'\n})\nevent.currentTarget.className = 'bitem active'\n}\nvar posid = event.currentTarget.getAttribute('pos');\nvar ftop = document.querySelector('.wp-block-columns .wp-block-column.artCss').offsetTop;\nif(posid && document.getElementById(posid) && document.getElementById(posid).offsetTop) {\n   var goPos = document.getElementById(posid).offsetTop -120;  \n   window.scrollTo(0, goPos+ftop);\n}\n\n}\nfunction goTop(event) {\n  window.scrollTo(0, 0);\n}\n<\/script>\n<script>\n\/\/ \u6dfb\u52a0\u6eda\u52a8\u76d1\u542c\u6765\u5b9e\u73b0\u56fa\u5b9a\u6548\u679c\ndocument.addEventListener('DOMContentLoaded', function() {\n \/\/ const sidebar = document.getElementById('top-sidebar');\n\/\/  const columnsContainer = document.querySelector('.wp-block-columns .wp-block-column.artCss');\n\/\/  const originalOffsetTop = columnsContainer.offsetTop;\n\/\/  const originalOffsetLeft = columnsContainer.offsetLeft;\n  \n  window.addEventListener('scroll', function() {\n    const scrollY = window.scrollY;\n    if(scrollY>340) {\n       document.querySelector('.backtop').style.display='block'\n    }else{\n       document.querySelector('.backtop').style.display='none'\n    }\n \/\/   const containerRect = columnsContainer.getBoundingClientRect();\n\/\/    let top1 = document.querySelector('#pm-10km').offsetTop;\n\/\/    let top2 = document.querySelector('#pm-10km-sub').offsetTop;\n\/\/    let top3 = document.querySelector('#pm25-1km').offsetTop;\n\/\/    let top4 = document.querySelector('#o3-mode').offsetTop;\n \/\/   let top5 = document.querySelector('#global-pm-10').offsetTop;\n \/\/   let top6 = document.querySelector('#fire-pm25').offsetTop;\n \/\/   if (scrollY > originalOffsetTop) {     \n \/\/     sidebar.style.top= (scrollY - originalOffsetTop) + 'px';\n  \/\/    let posName = '';\n \/\/     if(scrollY <= top1 + (top2-top1)\/2+originalOffsetTop) {\n   \/\/       posName = 'pm-10km';\n  \/\/    }else if(scrollY>top1 + (top2-top1)\/2+originalOffsetTop && scrollY <= top2 + (top3-top2)\/2+originalOffsetTop){\n\/\/ posName = 'pm-10km-sub';\n\/\/}else if(scrollY>top2 + (top3-top2)\/2+originalOffsetTop && scrollY <= top3 + (top4-top3)\/2+originalOffsetTop){\n\/\/ posName = 'pm25-1km';\n\/\/}else if(scrollY>top3 + (top4-top3)\/2+originalOffsetTop && scrollY <= top4 + (top5-top4)\/2+originalOffsetTop){\n\/\/ posName = 'o3-mode';\n\/\/}else if(scrollY>top4 + (top5-top4)\/2+originalOffsetTop && scrollY <= top5 + (top6-top5)\/2+originalOffsetTop){\n\/\/ posName = 'global-pm-10';\n\/\/}else if(scrollY>top5 + (top6-top5)\/2+originalOffsetTop){\n\/\/ posName = 'fire-pm25';\n\/\/}\n\/\/if(posName){\n\/\/document.querySelectorAll('#top-sidebar .bitem').forEach(function(val) {         \n\/\/          if(val.getAttribute('pos') == posName) { \n\/\/            val.className = 'bitem active'\n\/\/          }else{\n\/\/            val.className = 'bitem'\n \/\/         }     \n \/\/      })\n\/\/}\n\n\/\/    } else {\n  \/\/     sidebar.style.top = '0px';\n\/\/       document.querySelectorAll('#top-sidebar .bitem').forEach(function(val) {\n \/\/         if(val.getAttribute('pos') == 'pm-10km') { \n  \/\/          val.className = 'bitem active'\n  \/\/        }else{\n   \/\/         val.className = 'bitem'\n   \/\/       }         \n   \/\/    })\n   \/\/ }\n });\n  \n\n});\n<\/script>\n\n\n\n<div class=\"wp-block-cover\" style=\"background-image:url(\/wp-content\/uploads\/2025\/09\/finddata1-300x50-1.png);background-position:50% 15%;min-height:180px;aspect-ratio:unset;\"><div class=\"wp-block-cover__inner-container is-layout-flow wp-block-cover-is-layout-flow\">\n<p class=\"has-text-align-left has-large-font-size wp-block-paragraph\">Methods<\/p>\n<\/div><\/div>\n\n\n\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-is-layout-8f761849 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column artCss is-layout-flow wp-block-column-is-layout-flow\" style=\"flex-basis:100%\">\n<div id=\"top-sidebar\">\n <div class=\"china-M type-sidebars\">\n   <span class=\"title\">China<\/span>\n   <span class=\"sidebars\">\n     <a class=\"bitem active\" pos=\"pm-10km\" onclick=\"changeMenu(event)\">PM<sub>2.5<\/sub> - 10km<\/a>\n     <a class=\"bitem\" pos=\"pm-10km-sub\" onclick=\"changeMenu(event)\">PM<sub>2.5<\/sub> Species - 10km<\/a>\n     <a class=\"bitem\" pos=\"pm25-1km\" onclick=\"changeMenu(event)\">PM<sub>2.5<\/sub> - 1km<\/a>\n     <a class=\"bitem\" pos=\"o3-mode\" onclick=\"changeMenu(event)\">O<sub>3<\/sub> - 10km<\/a>\n   <\/span>\n <\/div>\n<div class=\"gobal-M type-sidebars\">\n   <span class=\"title\">Global<\/span>\n   <span class=\"sidebars\">\n     <a class=\"bitem\" pos=\"global-pm-10\" onclick=\"changeMenu(event)\">PM<sub>2.5<\/sub> &amp; Fire PM<sub>2.5<\/sub> - 10km<\/a>\n     <a class=\"bitem\" pos=\"fire-pm25\" style=\"display:none;\" onclick=\"changeMenu(event)\">Fire PM<sub>2.5<\/sub> - 10km<\/a>\n<a class=\"bitem\" pos=\"ai-forecast\" onclick=\"changeMenu(event)\">AI Aerosol Forecasting - 50km<\/a>\n   <\/span>\n<\/div>\n\n<\/div>\n<span class=\"backtop\" onclick=\"goTop(event)\"><div class=\"arrow\"><\/div><div class=\"text\">TOP<\/div><\/span>\n\n\n\n<div style=\"width:100%;height:120px;\">\n<\/div>\n\n\n\n<h3 id=\"pm-10km\">China PM<sub>2.5<\/sub> 10km prediction model &gt;<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">TAP developed a two-stage machine learning model to predict daily PM<sub>2.5<\/sub>&nbsp;concentrations with complete spatial coverage. The structure of our model is shown below. The data sources include PM<sub>2.5<\/sub>&nbsp;measurements, satellite AOD (aerosol optical depth) retrievals, online CMAQ simulations, meteorological reanalysis data, land use information, and population distribution. The first-stage model predicts the high pollution events and employed the SMOTE algorithm to resample the model training data, thus balanced the proportion of high pollution events and normal events in the model training dataset. The first-stage random forest model trained with the resampled data predicts high pollution events. The second-stage model predicts the residual between CMAQ PM<sub>2.5<\/sub>&nbsp;simulations and PM<sub>2.5<\/sub>&nbsp;measurements with random forest. The prediction of residual rather than the prediction of PM<sub>2.5<\/sub>&nbsp;enlarged the response of predictions to variations in predictors, thus improved the prediction accuracy. The missingness in satellite retrievals was filled by decision-tree based modeling algorithm.<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter size-large is-resized\"><img loading=\"lazy\" decoding=\"async\" src=\"\/wp-content\/uploads\/2021\/03\/Fig1-1-1024x417.png\" alt=\"\" class=\"wp-image-709\" width=\"768\" height=\"313\" srcset=\"http:\/\/tap-dev.makenv.com\/wp-content\/uploads\/2021\/03\/Fig1-1-1024x417.png 1024w, http:\/\/tap-dev.makenv.com\/wp-content\/uploads\/2021\/03\/Fig1-1-300x122.png 300w, http:\/\/tap-dev.makenv.com\/wp-content\/uploads\/2021\/03\/Fig1-1-768x313.png 768w, http:\/\/tap-dev.makenv.com\/wp-content\/uploads\/2021\/03\/Fig1-1-1536x625.png 1536w, http:\/\/tap-dev.makenv.com\/wp-content\/uploads\/2021\/03\/Fig1-1.png 1963w\" sizes=\"auto, (max-width: 768px) 100vw, 768px\" \/><\/figure><\/div>\n\n\n\n<p class=\"wp-block-paragraph\">References:<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>Geng, G., Xiao, Q., Liu, S., Liu, X., Cheng, J., Zheng, Y., Xue, T., Tong, D., Zheng, B., Peng, Y., Huang, X., He, K., &amp; Zhang, Q. (2021). Tracking Air Pollution in China: Near Real-Time PM<sub>2.5<\/sub>&nbsp;Retrievals from Multisource Data Fusion.&nbsp;<em>Environ Sci <em>Technol<\/em><\/em>, <em>55, <\/em>12106-12115.&nbsp;<strong><span class=\"has-inline-color has-vivid-green-cyan-color\">[<\/span><a rel=\"noreferrer noopener\" href=\"https:\/\/pubs.acs.org\/doi\/10.1021\/acs.est.1c01863\" target=\"_blank\"><span class=\"has-inline-color has-vivid-green-cyan-color\">Link<\/span><\/a><span class=\"has-inline-color has-vivid-green-cyan-color\">]<\/span><span class=\"has-inline-color has-vivid-red-color\"> [<\/span><a rel=\"noreferrer noopener\" href=\"https:\/\/pubs.acs.org\/doi\/pdf\/10.1021\/acs.est.1c01863\" target=\"_blank\"><span class=\"has-inline-color has-vivid-red-color\">PDF<\/span><\/a><span class=\"has-inline-color has-vivid-red-color\">]<\/span><\/strong><\/li><li>Xiao, Q., Zheng, Y., Geng, G., Chen, C., Huang, X., Che, H., Zhang, X., He, K., &amp; Zhang, Q. (2021). Separating emission and meteorological contribution to PM<sub>2.5<\/sub> trends over East China during 2000\u20132018.&nbsp;<em>Atmos Chem Phys<\/em>, 21, 9475-9496.&nbsp;<strong><span class=\"has-inline-color has-vivid-green-cyan-color\">[<\/span><a rel=\"noreferrer noopener\" href=\"https:\/\/acp.copernicus.org\/articles\/21\/9475\/2021\/\" target=\"_blank\"><span class=\"has-inline-color has-vivid-green-cyan-color\">Link<\/span><\/a><span class=\"has-inline-color has-vivid-green-cyan-color\">]<\/span><span class=\"has-inline-color has-vivid-red-color\"> [<\/span><a rel=\"noreferrer noopener\" href=\"https:\/\/acp.copernicus.org\/articles\/21\/9475\/2021\/acp-21-9475-2021.pdf\" target=\"_blank\"><span class=\"has-inline-color has-vivid-red-color\">PDF<\/span><\/a><span class=\"has-inline-color has-vivid-red-color\">]<\/span><\/strong><\/li><li>Xiao, Q., Geng, G., Cheng, J., Liang, F., Li, R., Meng, X., Xue, T., Huang, X., Kan, H., Zhang, Q., &amp; He, K. (2021). Evaluation of gap-filling approaches in satellite-based daily PM<sub>2.5<\/sub>&nbsp;prediction models.&nbsp;<em>Atmos Environ, 244<\/em>, 117921.<strong> <span class=\"has-inline-color has-vivid-green-cyan-color\">[<\/span><a rel=\"noreferrer noopener\" href=\"https:\/\/www.sciencedirect.com\/science\/article\/pii\/S1352231020306555\" target=\"_blank\"><span class=\"has-inline-color has-vivid-green-cyan-color\">Link<\/span><\/a><span class=\"has-inline-color has-vivid-green-cyan-color\">]<\/span> <span class=\"has-inline-color has-vivid-red-color\">[<\/span><a rel=\"noreferrer noopener\" href=\"https:\/\/www.sciencedirect.com\/science\/article\/pii\/S1352231020306555\/pdfft?md5=e6a530ead7de636e672bbe03b084794c&amp;pid=1-s2.0-S1352231020306555-main.pdf\" target=\"_blank\"><span class=\"has-inline-color has-vivid-red-color\">PDF<\/span><\/a><span class=\"has-inline-color has-vivid-red-color\">]<\/span><\/strong><\/li><\/ul>\n\n\n\n<h3 id=\"pm-10km-sub\">China PM<sub>2.5<\/sub> species 10km prediction model &gt;<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">TAP developed PM<sub>2.5<\/sub> species concentrations data set based on the published 10km PM<sub>2.5<\/sub> concentrations with complete spatial coverage. The methodology framework is shown below. The model extract PM<sub>2.5<\/sub> species information from operational CMAQ simulations under the constraint of PM<sub>2.5<\/sub> concentrations. To correct the bias of the CMAQ simulations, TAP first improved the windblown dust module, and then trained models to revise simulated PM<sub>2.5<\/sub> species proportions based on ground measurements and extreme gradient boosting (XGBoost) algorithm. Having obtained more accurate PM<sub>2.5<\/sub> species conversion factors, PM<sub>2.5<\/sub> species concentrations were derived on the basis of PM<sub>2.5 <\/sub>data. The TAP-released PM<sub>2.5<\/sub> species data include sulfate (SO<sub>4<\/sub><sup>2-<\/sup>), nitrate (NO<sub>3<\/sub><sup>-<\/sup>), ammonium (NH<sub>4<\/sub><sup>+<\/sup>), organic matter (OM), and black carbon (BC).<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter size-large is-resized\"><img loading=\"lazy\" decoding=\"async\" src=\"\/wp-content\/uploads\/2022\/02\/MethodologyFramework.v2.png\" alt=\"\" class=\"wp-image-1206\" width=\"768\" height=\"313\"\/><\/figure><\/div>\n\n\n\n<p class=\"wp-block-paragraph\">References:<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>Geng, G., Zhang, Q., Tong, D., Li, M., Zheng, Y., Wang, S., &amp; He, K. (2017). Chemical composition of ambient PM<sub>2.5<\/sub> over China and relationship to precursor emissions during 2005\u20132012.&nbsp;<em><em>Atmos Chem Phys<\/em><\/em>,&nbsp;<em>17<\/em>(14), 9187-9203. <span class=\"has-inline-color has-vivid-green-cyan-color\"><strong>[<\/strong><\/span><a href=\"https:\/\/acp.copernicus.org\/articles\/17\/9187\/2017\/\"><strong><span class=\"has-inline-color has-vivid-green-cyan-color\">Link<\/span><\/strong><\/a><span class=\"has-inline-color has-vivid-green-cyan-color\"><strong>]<\/strong><\/span><span class=\"has-inline-color has-vivid-red-color\"><strong>[<\/strong><\/span><a href=\"https:\/\/acp.copernicus.org\/articles\/17\/9187\/2017\/acp-17-9187-2017.pdf\"><strong><span class=\"has-inline-color has-vivid-red-color\">PDF<\/span><\/strong><\/a><span class=\"has-inline-color has-vivid-red-color\"><strong>]<\/strong><\/span><\/li><\/ul>\n\n\n\n<ul class=\"wp-block-list\"><li>Liu, S., Geng, G., Xiao, Q., Zheng, Y., Liu, X., Cheng, J., &amp; Zhang, Q. (2022). Tracking daily concentrations of PM<sub>2.5<\/sub> chemical composition in China since 2000. <em>Environ Sci Technol, <\/em>56, 16517\u201316527<em>.<\/em> <strong><span class=\"has-inline-color has-vivid-green-cyan-color\">[<\/span><\/strong><a href=\"https:\/\/pubs.acs.org\/doi\/full\/10.1021\/acs.est.2c06510\"><span class=\"has-inline-color has-vivid-green-cyan-color\"><strong>Link<\/strong><\/span><\/a><strong><span class=\"has-inline-color has-vivid-green-cyan-color\">]<\/span><span class=\"has-inline-color has-vivid-red-color\">[<\/span><\/strong><a href=\"https:\/\/pubs.acs.org\/doi\/pdf\/10.1021\/acs.est.2c06510\"><span class=\"has-inline-color has-vivid-red-color\"><strong>PDF<\/strong><\/span><\/a><strong><span class=\"has-inline-color has-vivid-red-color\">]<\/span><\/strong><\/li><\/ul>\n\n\n\n<h3 id=\"pm25-1km\">China PM<sub>2.5<\/sub> 1km prediction model &gt;<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">With the TAP PM<sub>2.5<\/sub> at 10km resolution, the complete-coverage daily PM<sub>2.5<\/sub> predictions at 1km spatial resolution from 2000 to the present were produced by fusing high-resolution satellite remote sensing retrievals and environmental spatial data (e.g. road maps) with a machine learning model. The structure of our model is shown below. To correctly reflect the temporal variations in land cover characteristics that significantly affected the pollution emissions, continuous annual geoinformation datasets, including the annual road distribution maps and ensemble gridded population distribution maps in China were constructed by spatial and statistical models. Then the random forest model was trained by the residual between 10km TAP PM<sub>2.5<\/sub> predictions and PM<sub>2.5<\/sub> measurements in order to enlarge the response of predictions to variations in predictors.<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter size-large is-resized\"><img loading=\"lazy\" decoding=\"async\" src=\"\/wp-content\/uploads\/2025\/09\/QQ20250910-133508.png\" alt=\"\" class=\"wp-image-1254\" width=\"768\" height=\"313\"\/><\/figure><\/div>\n\n\n\n<p class=\"wp-block-paragraph\">References:<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>Xiao, Q., Geng G., Liu, S., Liu, J., Meng, X., &amp; Zhang, Q. (2022). Spatiotemporal continuous estimates of daily 1\u2009km PM<sub>2.5<\/sub>&nbsp;from 2000 to present under the Tracking Air Pollution in China (TAP) framework.&nbsp;<em>Atmos Chem&nbsp;<\/em>Phys,&nbsp;<em>22,&nbsp;<\/em>13229\u201313242. <span class=\"has-inline-color has-vivid-green-cyan-color\"><strong>[<\/strong><\/span><a href=\"https:\/\/acp.copernicus.org\/articles\/22\/13229\/2022\/acp-22-13229-2022.html\"><strong><span class=\"has-inline-color has-vivid-green-cyan-color\">Link<\/span><\/strong><\/a><span class=\"has-inline-color has-vivid-green-cyan-color\"><strong>]<\/strong><\/span><span class=\"has-inline-color has-vivid-red-color\"><strong>[<\/strong><\/span><a href=\"https:\/\/acp.copernicus.org\/articles\/22\/13229\/2022\/acp-22-13229-2022.pdf\"><strong><span class=\"has-inline-color has-vivid-red-color\">PDF<\/span><\/strong><\/a><span class=\"has-inline-color has-vivid-red-color\"><strong>]<\/strong><\/span><\/li><\/ul>\n\n\n\n<h3 id=\"o3-mode\">China O<sub>3<\/sub> prediction model &gt;<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">TAP developed a machine learning model to predict full-coverage daily maximum 8-h average O<sub>3<\/sub>&nbsp;concentrations by fusing data from multiple sources. The model structure is shown below. The predictors include O<sub>3<\/sub>&nbsp;measurements, satellite O<sub>3<\/sub>&nbsp;vertical distribution profile, CMAQ simulations, WRF meteorological simulations, Normalized Difference Vegetation Index (NDVI), night light and population distribution. First, two random forest model were developed to describe the associations between O<sub>3<\/sub>&nbsp;measurements and all the predictors, with and without the satellite O<sub>3<\/sub>&nbsp;vertical distribution profile. Due to the missingness in satellite retrievals, the predictions from the model with satellite retrievals were spatiotemporally discontinuous. Then an elastic-net model was developed to fuse the predictions from the two random forest model in order to improve the prediction accuracy as well as fill the missingness. The last step simulate the residual of the fused predictions with a spatiotemporal Kriging interpolation.<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter size-large is-resized\"><img loading=\"lazy\" decoding=\"async\" src=\"\/wp-content\/uploads\/2025\/09\/QQ20250910-133520.png\" alt=\"\" class=\"wp-image-1255\" width=\"768\" height=\"313\"\/><\/figure><\/div>\n\n\n\n<p class=\"wp-block-paragraph\">Reference<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li>Xue, T., Zheng, Y., Geng, G., Xiao, Q., Meng, X., Wang, M., Li, X., Wu, N., Zhang, Q., &amp; Zhu, T. (2020). Estimating spatiotemporal variation in ambient ozone exposure during 2013\u20132017 using a data-fusion model.&nbsp;<em>Environ Sci <em>Technol<\/em>, 54<\/em>, 14877-14888. <strong><span class=\"has-inline-color has-vivid-green-cyan-color\">[<\/span><a href=\"https:\/\/pubs.acs.org\/doi\/10.1021\/acs.est.0c03098\"><span class=\"has-inline-color has-vivid-green-cyan-color\">Link<\/span><\/a><span class=\"has-inline-color has-vivid-green-cyan-color\">]<\/span> <span class=\"has-inline-color has-vivid-red-color\">[<\/span><a href=\"https:\/\/pubs.acs.org\/doi\/pdf\/10.1021\/acs.est.0c03098\"><span class=\"has-inline-color has-vivid-red-color\">PDF<\/span><\/a><span class=\"has-inline-color has-vivid-red-color\">]<\/span><\/strong><\/li><\/ul>\n\n\n\n<ul class=\"wp-block-list\"><li>Xiao, Q., Geng, G., Xue, T., Liu, S., Cai, C., He, K., &amp; Zhang, Q. (2021). Tracking PM<sub>2.5<\/sub> and O<sub>3<\/sub> Pollution and the Related Health Burden in China 2013\u20132020.&nbsp;<em><em>Environ Sci Technol<\/em><\/em>. <span class=\"has-inline-color has-vivid-green-cyan-color\"><strong>[<\/strong><\/span><a href=\"https:\/\/pubs.acs.org\/doi\/10.1021\/acs.est.1c04548\"><strong><span class=\"has-inline-color has-vivid-green-cyan-color\">Link<\/span><\/strong><\/a><span class=\"has-inline-color has-vivid-green-cyan-color\"><strong>]<\/strong><\/span><span class=\"has-inline-color has-vivid-red-color\"><strong>[<\/strong><\/span><a href=\"https:\/\/pubs.acs.org\/doi\/pdf\/10.1021\/acs.est.1c04548\"><strong><span class=\"has-inline-color has-vivid-red-color\">PDF<\/span><\/strong><\/a><span class=\"has-inline-color has-vivid-red-color\"><strong>]<\/strong><\/span><\/li><\/ul>\n\n\n\n<h3 id=\"global-pm-10\">Global PM<sub>2.5<\/sub> and fire PM<sub>2.5<\/sub> - 10km&gt;<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">First, global high-resolution wildfire emission inventories are used to drive the GEOS-Chem model, to quantify the relative contribution of wildfire emissions to the PM<sub>2.5<\/sub> concentration. Meanwhile, by incorporating multiple sets of ground observation data, reanalysis data, satellite observations, and wildfire emission data, a three-layer machine learning model is developed using random forest, SMOTE, and residual correction techniques. Further, through operational data preprocessing, GEOS-Chem simulation, and rolling training of the machine learning model,  global near-real-time daily PM<sub>2.5<\/sub> and fire PM<sub>2.5<\/sub> at 10km resolution are retrieved synchronously.<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter size-large is-resized\"><img loading=\"lazy\" decoding=\"async\" src=\"http:\/\/tap-dev.makenv.com\/wp-content\/uploads\/2025\/09\/model_framework-1-1024x910.png\" alt=\"\" class=\"wp-image-1952\" width=\"768\" height=\"659\"\/><\/figure><\/div>\n\n\n\n<p class=\"wp-block-paragraph\">Reference<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li> Zhang, Q., Wang Y., Xiao Q., Geng G., S. J. Davis, Liu,X., J. Yang, J. J. Liu, W. Y. Huang, C. P. He, B. H. Luo, R. V. Martin, M. Brauer, J. T. Randerson, and K. B. He (2025), Long-range PM<sub>2.5<\/sub> pollution and health impacts from the 2023 Canadian wildfires.&nbsp;<em><em>Nature<\/em><\/em>. <strong><span class=\"has-inline-color has-vivid-green-cyan-color\">[<\/span><a href=\"https:\/\/www.nature.com\/articles\/s41586-025-09482-1\"><span class=\"has-inline-color has-light-green-cyan-color\">Link<\/span><\/a><span class=\"has-inline-color has-vivid-green-cyan-color\">]<\/span><\/strong><\/li><\/ul>\n\n\n\n<h3 id=\"ai-forecast\"> AI-GAMFS &gt;<\/h3>\n\n\n\n<p class=\"wp-block-paragraph\">A Machine-learning-driven Global Aerosol-Meteorology Forecasting System(AI-GAMFS), developed by the Chinese Academy of Meteorological Sciences, has been operationally deployed on the TAP platform. With 1.3 billion parameters, the model employs an innovative architecture that combines Vision Transformer (ViT) and U-Net networks. By integrating key technologies such as global attention mechanisms, spatiotemporal encoding modules, and relay forecasting strategies (as detailed in the following technical roadmap), AI-GAMFS effectively captures complex aerosol-meteorology interactions, significantly enhancing high-dimensional environmental-meteorological feature extraction capabilities. Trained on a 42-year advanced global aerosol reanalysis dataset and initialized with GEOS-FP (GEOS Forward Processing) analysis data, AI-GAMFS delivers rapid and efficient forecasting. It performs 8 daily rolling forecasts, each requiring only 36 seconds on a single GPU. The system provides 5-day global forecasts at 3-hour intervals for key aerosol components (dust, sulfate, black carbon, organic carbon, and sea salt), along with their optical properties, surface concentrations, and related meteorological variables. Covering 54 forecast variables at a 0.5\u00b0\u00d7 0.625\u00b0 resolution, AI-GAMFS substantially enhances operational efficiency and timeliness in global aerosol-meteorology coupled forecasting.\n<\/p>\n\n\n\n<div class=\"wp-block-image\"><figure class=\"aligncenter size-large is-resized\"><img loading=\"lazy\" decoding=\"async\" src=\"http:\/\/tap-dev.makenv.com\/wp-content\/uploads\/2026\/02\/Fig.1.png\" alt=\"\" class=\"wp-image-2085\" width=\"768\" height=\"313\"\/><\/figure><\/div>\n\n\n\n<p class=\"wp-block-paragraph\">Reference\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\"><li> Gui, K., Zhang, X., Che, H., Li, L., Zheng, Y., An, L., Miao, Y., Zhao, H., Dubovik, O., Holben, B., Wang, J., Gupta, P., Lind, E S., Toledano, C., Wang, H., Wang, Z., Wang, Y., Huang, X., Dai, K., Xia, X., Xu, X., &amp; Zhang, X. (2026). Advancing operational global aerosol forecasting with machine learning.&nbsp;<em><em>Nature<\/em><\/em>. <strong><span class=\"has-inline-color has-vivid-green-cyan-color\">[<\/span><a href=\"https:\/\/www.nature.com\/articles\/s41586-026-10234-y\"><span class=\"has-inline-color has-light-green-cyan-color\">Link<\/span><\/a><span class=\"has-inline-color has-vivid-green-cyan-color\">]<\/span><\/strong><\/li><\/ul>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n<\/div>\n\n\n\n<div class=\"wp-block-column is-layout-flow wp-block-column-is-layout-flow\" style=\"flex-basis:0%\"><\/div>\n<\/div>\n","protected":false},"excerpt":{"rendered":"<p>China PM2.5 - 10km PM2.5 Species - 10km PM2.5 - 1km O3 &hellip;<\/p>\n<p class=\"more-link\"><a href=\"http:\/\/tap-dev.makenv.com\/?page_id=601&#038;lang=en\" class=\"themebutton2\">Read More<\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"footnotes":""},"class_list":["post-601","page","type-page","status-publish","hentry"],"_links":{"self":[{"href":"http:\/\/tap-dev.makenv.com\/index.php?rest_route=\/wp\/v2\/pages\/601","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/tap-dev.makenv.com\/index.php?rest_route=\/wp\/v2\/pages"}],"about":[{"href":"http:\/\/tap-dev.makenv.com\/index.php?rest_route=\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"http:\/\/tap-dev.makenv.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/tap-dev.makenv.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=601"}],"version-history":[{"count":199,"href":"http:\/\/tap-dev.makenv.com\/index.php?rest_route=\/wp\/v2\/pages\/601\/revisions"}],"predecessor-version":[{"id":2106,"href":"http:\/\/tap-dev.makenv.com\/index.php?rest_route=\/wp\/v2\/pages\/601\/revisions\/2106"}],"wp:attachment":[{"href":"http:\/\/tap-dev.makenv.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=601"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}