[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-82972-en":3,"doc-seo-82972-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},82972,687197207639,"Asher","https://ap-avatar.wpscdn.com/davatar_a8503ba1806abce46bf441b54a3ca4cd",8,"Research & Report","Domain Adaptive Climate Downscaling Under Temporal Distribution Shift","Deep-learning climate downscaling learns mappings from historical low-resolution (LR) to high-resolution (HR) climate data, yet deployment to future periods introduces temporal out-of-distribution (OOD) shift. The study analyzes daily temperature downscaling over the Continental United States using paired LR-HR simulations and proposes a temporal domain-adaptive framework. It combines supervised HR reconstruction on historical data with domain alignment between historical and future distributions. Across future validation periods, the method outperforms statistical and deep-learning bias correction, with largest gains during stronger distribution shifts, improved spatial patterns in complex terrain, higher spatiotemporal correlation, and reduced upper-tail bias, supporting robustness under non-stationary climate change.","arXiv :2607 .05645v 1 [ cs .LG] 6 Jul 2026  \nDomain-Adaptive Climate Downscaling Under Temporal Distribution Shift  \nShuochen Wang 1 , Nishant Yadav2 , and Auroop R. Ganguly1, 3  \n1 Sustainability and Data Sciences Laboratory, Northeastern University  \n2Microsoft  \n3AI for Climate and Sustainability, The Institute for Experiential AI, Northeastern University  \nAbstract  \nDeep-learning-based climate downscaling aims to learn relationships from historical low-resolution (LR) and high-resolution (HR) climate data to generate HR climate projections. However, this setting faces a temporal out-of-distribution (OOD) challenge: models trained on historical data are commonly applied to future projections whose distributions may differ substantially from the training period.  \nThis study investigates temporal OOD shift for daily temperature downscaling over the Continental United States using paired LR-HR model simulations. We propose a temporal domain-adaptive downscaling framework that combines supervised HR reconstruction on historical data with domain alignment between historical and future climate distributions. Experiments across future validation periods show that the proposed domain-adaptive model consistently outperforms statistical and deep-learning-based bias-correction methods, with the largest gains occurring when the temporal distribution shift is strongest. Spatial analyses indicate stronger improvements over high-elevation and topographically complex regions, along with higher spatiotemporal correlation with the HR target. The extreme analysis shows that domain adaptation also reduces upper-tail temperature bias relative to thenon-adaptive model. These results demonstrate that temporal domain adaptation can improve the robustness of HR climate projections under non-stationary climate conditions.  \n1 Introduction  \nGlobal Climate Models (GCMs) are essential tools for understanding the Earth system, simulating historical climate variability, and projecting future climate change under different forcing scenarios. Despite advances in Earth’s system representation, GCMs from the current generation Coupled Model Intercomparison Project (CMIP) remain too coarse for many regional and local applications, with horizontal grid spacings often on the order of tens to hundreds of kilometers [1, 2] . At these resolutions, fine-scale effects associated with topography, coastlines, land-sea contrasts, and unresolved sub-grid-scale physical processes cannot be explicitly represented and must instead be approximated through parameterizations, which can introduce biases and smooth local climate variability [2, 3] . Downscaling methods have therefore been developed to bridge the scale gap between low-resolution (LR) GCM outputs and the high-resolution (HR) climate information needed for regional impact assessment [4, 5] . These methods are commonly grouped into dynamical and statistical approaches. Dynamical downscaling uses Regional Climate Models (RCMs) to refine global model outputs through HR physical simulations [6, 7, 8] . A major advantage of RCMs is that they retain a processbased representation of regional climate dynamics, including interactions with terrain, land-surface properties, and coastlines [9] . However, RCMs are computationally expensive and difficult to scale across large ensembles, emission scenarios, long simulation periods, and complex model configura  \nPreprint. Under review.  \ntions. Their projections are also affected by multiple uncertainty sources, including internal climate variability, scenario uncertainty, and systematic errors inherited from the driving GCMs [9, 10] . Statistical downscaling, by contrast, infers empirical relationships between coarse-scale predictorsand fine-scale climate observations or simulations [5] . Classical methods include regression-based models [11], canonical correlation analysis [12], analog methods [13, 14], and bias correction or bias-correction spatial disaggregation approaches","cbCainxAybcYyCo8","https://ap.wps.com/l/cbCainxAybcYyCo8","pdf",6235540,2,1,25,"English","en",105,"# Abstract\n# Introduction\n## Background: GCM coarse resolution and scale gap\n## Downscaling approaches: dynamical vs statistical\n## Data-driven and deep-learning downscaling\n## Challenge: temporal out-of-distribution shift","[{\"question\":\"What problem does the paper address in climate downscaling?\",\"answer\":\"It addresses temporal out-of-distribution (OOD) shift, where models trained on historical climate distributions are applied to future periods whose distributions may differ substantially.\"},{\"question\":\"How does the proposed method improve downscaling performance?\",\"answer\":\"It uses a temporal domain-adaptive framework that combines supervised HR reconstruction on historical data with domain alignment between historical and future climate distributions.\"},{\"question\":\"Where are the largest improvements observed, and what do the results show about extremes?\",\"answer\":\"The largest gains occur when the temporal distribution shift is strongest, with stronger spatial improvements over high-elevation and topographically complex regions. The extreme analysis indicates reduced upper-tail temperature bias compared with a non-adaptive model.\"}]",1784184398,63,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"domain-adaptive-climate-downscaling-under-temporal-distribution-shift","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,47,50],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":20},"https://docshare.wps.com/document/","Document",{"item":48,"name":12,"@type":43,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":43,"position":52},"https://docshare.wps.com/document/domain-adaptive-climate-downscaling-under-temporal-distribution-shift/82972/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-23","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does the paper address in climate downscaling?","Question",{"text":75,"@type":76},"It addresses temporal out-of-distribution (OOD) shift, where models trained on historical climate distributions are applied to future periods whose distributions may differ substantially.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the proposed method improve downscaling performance?",{"text":80,"@type":76},"It uses a temporal domain-adaptive framework that combines supervised HR reconstruction on historical data with domain alignment between historical and future climate distributions.",{"name":82,"@type":73,"acceptedAnswer":83},"Where are the largest improvements observed, and what do the results show about extremes?",{"text":84,"@type":76},"The largest gains occur when the temporal distribution shift is strongest, with stronger spatial improvements over high-elevation and topographically complex regions. The extreme analysis indicates reduced upper-tail temperature bias compared with a non-adaptive model.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]