[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-128246-en":3,"doc-seo-128246-105":30,"detail-sidebar-cat-0-en-105":92},{"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":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},128246,2336475104736,"Quinn","https://ap-avatar.wpscdn.com/avatar/22000c4c5e0e5b17e70?x-image-process=image/resize,m_fixed,w_180,h_180&k=1786591360781797222",8,"Research & Report","Scaling Across Multiple Dimensions to Accelerate Subgrid Machine Learning Parameterization Development - Dissertation","Scaling Across Multiple Dimensions to Accelerate Subgrid Machine Learning Parameterization Development advances ML-based subgrid parameterization for Earth system modeling by strengthening evaluation design and leveraging large-scale ensembles. The dissertation develops an end-to-end pipeline with offline validation and online coupling tests, then studies the sample-size requirements for robust detection under noisy training outcomes. It further crowdsources progress using seminal datasets and benchmarks, culminating in hybrid physics–ML climate simulation results from a major Kaggle competition.","UC Irvine  \nUC Irvine Electronic Theses and Dissertations  \nTitle  \nScaling Across Multiple Dimensions to Accelerate Subgrid Machine Learning Parameterization Development  \nPermalink  \n[https://escholarship.org/uc/item/8279t561](https://escholarship.org/uc/item/8279t561)  \nISBN  \n9798291546208  \nAuthor  \nLin, Jerry  \nPublication Date  \n2025-08-22  \nCopyright Information  \nThis work is made available under the terms of a Creative Commons Attribution License, available at [https://creativecommons.org/licenses/by/4.0/](https://creativecommons.org/licenses/by/4.0/)  \nPeer reviewed  \n[eScholarship.org](eScholarship.org) Powered by the California Digital Library  \nUniversity of California  \nUNIVERSITY OF CALIFORNIA,  \nIRVINE  \nScaling Across Multiple Dimensions to Accelerate Subgrid Machine Learning  \nParameterization Development  \nDISSERTATION  \nsubmitted in partial satisfaction of the requirements  \nfor the degree of  \nDOCTOR OF PHILOSOPHY  \nin Earth System Science  \nby  \nJerry Lin  \nDissertation Committee:  \nProfessor Mike Pritchard, Co-Chair Professor Francois Primeau, Co-Chair Professor Gudrun Magnusdottir  \nChapter 2 © 2025 American Geophysical Union All other materials © 2025 Jerry Lin  \nDEDICATION  \nTo my mom and my oldest sister, whose endless encouragement helped me persist during  \nperiods of great personal uncertainty.  \nAnd to my late father, who would have been immensely proud to see me finish a PhD.  \nTABLE OF CONTENTS  \nPage  \nLIST OF FIGURES v  \nLIST OF TABLES vii  \nACKNOWLEDGEMENTS viii  \nVITA x  \nABSTRACT OF THE DISSERTATION xiv  \n1 Introduction 1  \n1.1 Breaking Deadlock and Superparameterization ................. 1  \n1.2 The Promise of Full-Physics Emulation for Subgrid Parameterization .... 2  \n1.2.1 A Host of Competing Approaches .................... 3  \n1.3 Scope and Organization of the Dissertation ................... 6  \n2 Navigating the Noise: Bringing Clarity to ML Parameterization Design with O(100)  \nEnsembles 8  \n2.1 Introduction .................................... 10  \n2.2 Methods ...................................... 12  \n2.2.1 Reference Climate Simulation ...................... 12  \n2.2.2 Training, Validation, and Offline Test Data ............... 13  \n2.2.3 End-to-End Pipeline and Analysis .................... 14  \n2.2.4 NN Configurations ............................ 15  \n2.3 Results (in-distribution) ............................. 21  \n2.3.1 Offline Results .............................. 21  \n2.3.2 Online Results .............................. 23  \n2.3.3 Persistent Online Zonal Mean Biases .................. 28  \n2.4 Estimating necessary sample size for robust detection ............. 30  \n2.5 Conclusion ..................................... 31  \n2.6 Open Research .................................. 35  \n2.7 Acknowledgments ................................. 36  \n3 Crowdsourcing ML Parameterization Innovation with a Seminal Dataset, Benchmark, and Competition 37  \n3.1 Introduction .................................... 38  \n3.2 Methods ...................................... 40  \n3.2.1 ClimSim NeurIPS Datasets and Benchmarks Paper .......... 40  \n3.2.2 LEAP-Atmospheric Physics Using AI (ClimSim) Kaggle competition 44  \n3.3 Impact ....................................... 48  \n3.3.1 Reception ................................. 48  \n3.3.2 Follow-Up Work and The Road Ahead ................. 48  \n3.4 Conclusion ..................................... 49  \n4 Harvesting ideas for hybrid physics-ML climate simulation from a $50,000 Kaggle Competition 50  \n4.1 Introduction .................................... 52  \n4.2 Methods ...................................... 54  \n4.2.1 Architectures ............................... 55  \n4.2.2 Architecture-Agnostic Configurations (AAC) .............. 61  \n4.3 Results ....................................... 63  \n4.3.1 Offline R2 Evaluation ........................... 63  \n4.3.2 A new state-of-the-art in online zonal mean biases ........... 65  \n4.3.3 Common online pathologies ...","cbCaidsd2k5nrfCB","https://ap.wps.com/l/cbCaidsd2k5nrfCB","pdf",30202275,1,134,"English","en",105,"# List of Figures\n# List of Tables\n# Acknowledgements\n# Vita\n# Abstract of the Dissertation\n# Introduction\n## Breaking Deadlock and Superparameterization\n## The Promise of Full-Physics Emulation for Subgrid Parameterization\n# Navigating the Noise: Bringing Clarity to ML Parameterization Design with O(100) Ensembles\n## Methods\n## Results (in-distribution)\n# Crowdsourcing ML Parameterization Innovation with a Seminal Dataset, Benchmark, and Competition\n## Methods\n## Impact\n# Harvesting ideas for hybrid physics-ML climate simulation from a $50,000 Kaggle Competition\n## Methods\n## Results\n# Conclusion","[{\"question\":\"What problem does the dissertation address in ML subgrid parameterization development?\",\"answer\":\"It focuses on accelerating and improving subgrid machine learning parameterization by tackling design challenges and making evaluation clearer despite noisy outcomes.\"},{\"question\":\"How does the work evaluate ML parameterization models?\",\"answer\":\"It uses an end-to-end pipeline combining offline reference climate simulations, training/validation/testing splits, and online coupling results, including analysis of persistent zonal mean biases.\"},{\"question\":\"What role do datasets, benchmarks, and competitions play?\",\"answer\":\"The dissertation uses crowdsourcing through a seminal dataset, benchmark, and a Kaggle competition to stimulate and consolidate parameterization innovations, producing hybrid physics–ML climate simulation improvements.\"}]","Scaling Across Multiple Dimensions to Accelerate Subgrid Machine Learning Parameterization Development - Dissertation | PDF",1785946097,338,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":87,"head_meta":89,"extra_data":91,"updated_unix":28},"scaling-across-multiple-dimensions-to-accelerate-subgrid-machine-learning-parameterization-development-dissertation","",{"@graph":36,"@context":86},[37,54,69],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/scaling-across-multiple-dimensions-to-accelerate-subgrid-machine-learning-parameterization-development-dissertation/128246/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-23","2026-08-05",true,{"@type":66,"interactionType":67,"userInteractionCount":20},"InteractionCounter",{"@type":68},"ViewAction",{"@type":70,"mainEntity":71},"FAQPage",[72,78,82],{"name":73,"@type":74,"acceptedAnswer":75},"What problem does the dissertation address in ML subgrid parameterization development?","Question",{"text":76,"@type":77},"It focuses on accelerating and improving subgrid machine learning parameterization by tackling design challenges and making evaluation clearer despite noisy outcomes.","Answer",{"name":79,"@type":74,"acceptedAnswer":80},"How does the work evaluate ML parameterization models?",{"text":81,"@type":77},"It uses an end-to-end pipeline combining offline reference climate simulations, training/validation/testing splits, and online coupling results, including analysis of persistent zonal mean biases.",{"name":83,"@type":74,"acceptedAnswer":84},"What role do datasets, benchmarks, and competitions play?",{"text":85,"@type":77},"The dissertation uses crowdsourcing through a seminal dataset, benchmark, and a Kaggle competition to stimulate and consolidate parameterization innovations, producing hybrid physics–ML climate simulation improvements.","https://schema.org",{"og:url":52,"og:type":88,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":90,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":93},[94,98,102,106,111,116,121,124,129,132,136],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":95,"show_sort_weight":96,"slug":97},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":99,"show_sort_weight":100,"slug":101},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":103,"show_sort_weight":104,"slug":105},"Exam",70,"exam",{"id":107,"doc_module":4,"doc_module_name":46,"category_name":108,"show_sort_weight":109,"slug":110},5,"Comic",60,"comic",{"id":112,"doc_module":4,"doc_module_name":46,"category_name":113,"show_sort_weight":114,"slug":115},6,"Technology",50,"technology",{"id":117,"doc_module":4,"doc_module_name":46,"category_name":118,"show_sort_weight":119,"slug":120},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":122,"slug":123},30,"research-report",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":126,"show_sort_weight":127,"slug":128},9,"Religion & Spirituality",20,"religion-spirituality",{"id":127,"doc_module":4,"doc_module_name":46,"category_name":130,"show_sort_weight":127,"slug":131},"World Cup","world-cup",{"id":133,"doc_module":4,"doc_module_name":46,"category_name":134,"show_sort_weight":133,"slug":135},10,"Lifestyle","lifestyle",{"id":137,"doc_module":4,"doc_module_name":46,"category_name":138,"show_sort_weight":107,"slug":139},19,"General","general"]