Residual Diffusion Modeling for Km-scale Atmospheric Downscaling

Residual Diffusion Modeling for Km-scale Atmospheric Downscaling | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Residual Diffusion Modeling for Km-scale Atmospheric Downscaling Morteza Mardani, Noah Brenowitz, Yair Cohen, Jaideep Pathak, Chieh-Yu Chen, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-3673869/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 24 Feb, 2025 Read the published version in Communications Earth & Environment → Version 1 posted You are reading this latest preprint version Abstract Predictions of weather hazard require expensive km-scale simulations driven by coarser global inputs. Here, a cost-effective stochastic downscaling model is trained from a high-resolution 2-km weather model over Taiwan conditioned on 25-km ERA5 reanalysis. To address the multi-scale machine learning challenges of weather data, we employ a two-step approach Corrector Diffusion ( CorrDiff ), where a UNet prediction of the mean is corrected by a diffusion step. Akin to Reynolds decomposition in fluid dynamics, this isolates generative learning to the stochastic scales. CorrDiff exhibits skillful RMSE and CRPS and faithfully recovers spectra and distributions even for extremes. Case studies of coherent weather phenomena reveal appropriate multivariate relationships reminiscent of learnt physics: the collocation of intense rainfall and sharp gradients in fronts and extreme winds and rainfall bands near the eyewall of typhoons. Downscaling global forecasts successfully retains many of these benefits, foreshadowing the potential of end-to-end, global-to- km-scales machine learning weather predictions. Earth and environmental sciences/Climate sciences/Atmospheric science/Atmospheric dynamics Earth and environmental sciences/Climate sciences Earth and environmental sciences/Planetary science/Atmospheric dynamics Earth and environmental sciences/Natural hazards Full Text Additional Declarations There is NO Competing Interest. Supplementary Files SupplementaryInformation.pdf Cite Share Download PDF Status: Published Journal Publication published 24 Feb, 2025 Read the published version in Communications Earth & Environment → Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-3673869","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":259299815,"identity":"2f259691-ba86-409c-bee3-0ade8922950c","order_by":0,"name":"Morteza Mardani","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAwklEQVRIiWNgGAWjYLACHiDiB9IHeAxI0SLZQKoWBoMDUAZBoDvt8LMPb2ruyBjf7jE88KbAjoG/vTsBrxaz22nGM+cce8ZjdueMwcE5BskMEmfObiCgJcGYmYftMI/ZjbSEwzwGzAwGErmEtKR/Zub5d5jHeAZYSz0xWnKMmXnbgIolkg8AycNEaSlmnNt3mEfiRvIBoF+O8xDhl/TNDG++Hbbnn5HY/OHNn2o5/vZe/FowAFFRMwpGwSgYBaOAAAAAuRlH/3B1dacAAAAASUVORK5CYII=","orcid":"","institution":"Nvidia","correspondingAuthor":true,"prefix":"","firstName":"Morteza","middleName":"","lastName":"Mardani","suffix":""},{"id":259299816,"identity":"eaee02de-263e-4449-8c96-22f7d4b8af4e","order_by":1,"name":"Noah Brenowitz","email":"","orcid":"","institution":"Nvidia","correspondingAuthor":false,"prefix":"","firstName":"Noah","middleName":"","lastName":"Brenowitz","suffix":""},{"id":259299817,"identity":"2de1b6c5-6fda-48dd-ae08-855a05375be9","order_by":2,"name":"Yair Cohen","email":"","orcid":"","institution":"Nvidia Research","correspondingAuthor":false,"prefix":"","firstName":"Yair","middleName":"","lastName":"Cohen","suffix":""},{"id":259299818,"identity":"d3a27e24-e8d9-4a28-a9b1-c3af10b7de45","order_by":3,"name":"Jaideep Pathak","email":"","orcid":"","institution":"Nvidia","correspondingAuthor":false,"prefix":"","firstName":"Jaideep","middleName":"","lastName":"Pathak","suffix":""},{"id":259299819,"identity":"e0f8e584-793b-436b-8501-8b5fe201e613","order_by":4,"name":"Chieh-Yu Chen","email":"","orcid":"","institution":"Nvidia","correspondingAuthor":false,"prefix":"","firstName":"Chieh-Yu","middleName":"","lastName":"Chen","suffix":""},{"id":259299820,"identity":"b185ead1-48e8-4c53-b30a-5d0c24a3d1b3","order_by":5,"name":"Cheng-Chin Liu","email":"","orcid":"","institution":"Central Weather Agency","correspondingAuthor":false,"prefix":"","firstName":"Cheng-Chin","middleName":"","lastName":"Liu","suffix":""},{"id":259299821,"identity":"88650d1b-ff59-408c-a57f-5ef440b18220","order_by":6,"name":"Arash Vahdat","email":"","orcid":"","institution":"Nvidia Corporation","correspondingAuthor":false,"prefix":"","firstName":"Arash","middleName":"","lastName":"Vahdat","suffix":""},{"id":259299822,"identity":"fa74351e-3c09-4c83-a1e5-5227e2c3aa36","order_by":7,"name":"Karthik Kashinath","email":"","orcid":"","institution":"Nvidia","correspondingAuthor":false,"prefix":"","firstName":"Karthik","middleName":"","lastName":"Kashinath","suffix":""},{"id":259299823,"identity":"50e90aa1-6863-4cdd-b1de-5b41b98d6798","order_by":8,"name":"Jan Kautz","email":"","orcid":"","institution":"NVIDIA","correspondingAuthor":false,"prefix":"","firstName":"Jan","middleName":"","lastName":"Kautz","suffix":""},{"id":259299824,"identity":"46bc5513-e827-49b3-a90d-2fc9ee14b9ba","order_by":9,"name":"Mike Pritchard","email":"","orcid":"","institution":"Nvidia, University of California Irvine","correspondingAuthor":false,"prefix":"","firstName":"Mike","middleName":"","lastName":"Pritchard","suffix":""}],"badges":[],"createdAt":"2023-11-27 23:25:09","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-3673869/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-3673869/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s43247-025-02042-5","type":"published","date":"2025-02-24T05:00:00+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":77105346,"identity":"10ea84b6-6430-4827-9cbd-9a32436c9edc","added_by":"auto","created_at":"2025-02-25 08:07:02","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":16297631,"visible":true,"origin":"","legend":"","description":"","filename":"diffusionsdownscalingsubmission.pdf","url":"https://assets-eu.researchsquare.com/files/rs-3673869/v1_covered_67762c92-590f-4d6a-9891-836f2c3411fc.pdf"},{"id":49153283,"identity":"41486a61-f4a1-4f2f-8316-83de8d0aefb8","added_by":"auto","created_at":"2024-01-04 02:20:23","extension":"pdf","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":25572007,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryInformation.pdf","url":"https://assets-eu.researchsquare.com/files/rs-3673869/v1/faeaa5974858a61d03ec47a1.pdf"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Residual Diffusion Modeling for Km-scale Atmospheric Downscaling","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-3673869/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-3673869/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003ePredictions of weather hazard require expensive km-scale simulations driven by coarser global inputs. Here, a cost-effective stochastic downscaling model is trained from a high-resolution 2-km weather model over Taiwan conditioned on 25-km ERA5 reanalysis. To address the multi-scale machine learning challenges of weather data, we employ a two-step approach Corrector Diffusion (\u003cem\u003eCorrDiff\u003c/em\u003e), where a UNet prediction of the mean is corrected by a diffusion step. Akin to Reynolds decomposition in fluid dynamics, this isolates generative learning to the stochastic scales. \u003cem\u003eCorrDiff\u003c/em\u003e exhibits skillful RMSE and CRPS and faithfully recovers spectra and distributions even for extremes. Case studies of coherent weather phenomena reveal appropriate multivariate relationships reminiscent of learnt physics: the collocation of intense rainfall and sharp gradients in fronts and extreme winds and rainfall bands near the eyewall of typhoons. Downscaling global forecasts successfully retains many of these benefits, foreshadowing the potential of end-to-end, global-to- km-scales machine learning weather predictions.\u003c/p\u003e","manuscriptTitle":"Residual Diffusion Modeling for Km-scale Atmospheric Downscaling","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-01-04 02:20:16","doi":"10.21203/rs.3.rs-3673869/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"communications-earth-and-environment","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"commsenv","sideBox":"Learn more about [Communications Earth and Environment](https://www.nature.com/commsenv/)","snPcode":"","submissionUrl":"","title":"Communications Earth \u0026 Environment","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Communications Series","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"08bde373-8577-4509-998c-214ec57627f1","owner":[],"postedDate":"January 4th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":27384208,"name":"Earth and environmental sciences/Climate sciences/Atmospheric science/Atmospheric dynamics"},{"id":27384209,"name":"Earth and environmental sciences/Climate sciences"},{"id":27384210,"name":"Earth and environmental sciences/Planetary science/Atmospheric dynamics"},{"id":27384211,"name":"Earth and environmental sciences/Natural hazards"}],"tags":[],"updatedAt":"2025-02-25T08:06:23+00:00","versionOfRecord":{"articleIdentity":"rs-3673869","link":"https://doi.org/10.1038/s43247-025-02042-5","journal":{"identity":"communications-earth-and-environment","isVorOnly":false,"title":"Communications Earth \u0026 Environment"},"publishedOn":"2025-02-24 05:00:00","publishedOnDateReadable":"February 24th, 2025"},"versionCreatedAt":"2024-01-04 02:20:16","video":"","vorDoi":"10.1038/s43247-025-02042-5","vorDoiUrl":"https://doi.org/10.1038/s43247-025-02042-5","workflowStages":[]},"version":"v1","identity":"rs-3673869","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-3673869","identity":"rs-3673869","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}