본문

서브메뉴

Understanding Drivers of Ice Mass Loss in Greenland Through Sea-Level and Climate Modeling, Remote Sensing, and Machine Learning
Understanding Drivers of Ice Mass Loss in Greenland Through Sea-Level and Climate Modeling...
Understanding Drivers of Ice Mass Loss in Greenland Through Sea-Level and Climate Modeling, Remote Sensing, and Machine Learning

상세정보

자료유형  
 학위논문 서양
최종처리일시  
20250211152135
ISBN  
9798383564288
DDC  
363
저자명  
Antwerpen, Rafael Marco.
서명/저자  
Understanding Drivers of Ice Mass Loss in Greenland Through Sea-Level and Climate Modeling, Remote Sensing, and Machine Learning
발행사항  
[Sl] : Columbia University, 2024
발행사항  
Ann Arbor : ProQuest Dissertations & Theses, 2024
형태사항  
138 p
주기사항  
Source: Dissertations Abstracts International, Volume: 86-01, Section: B.
주기사항  
Advisor: Tedesco, Marco.
학위논문주기  
Thesis (Ph.D.)--Columbia University, 2024.
초록/해제  
요약Changes in global climate conditions significantly impact ice sheet and glacier mass change leading to global mean sea level (GMSL) change. One of the largest present-day contributors to GMSL is the Greenland ice sheet (GrIS) and it will likely continue to be so in the future. To accurately predict future ice mass changes, it is crucial to understand the response of GrIS to a changing climate and to correctly represent this behavior in climate models. The GrIS' contribution to GMSL can in large part be attributed to the loss of ice and snow mass from the ice sheet surface. The surface mass loss has accelerated in the past decades due to increased surface melting and runoff in response to atmospheric warming. Surface melting is strongly controlled by ice albedo, a complex and dynamic property of ice that regulates the amount of solar radiation that is absorbed or reflected by the surface. Absorbed solar radiation leads to heating and melting of the ice surface. However, we lack a comprehensive understanding of the physical processes controlling ice mass loss, including ice albedo. These processes are, therefore, often simplified or crudely parameterized in climate models and subsequently add to large uncertainties in sea level rise predictions. This uncertainty prevents effective mitigation of and adaptation to the effects of climate change and sea level rise. It is, therefore, essential to advance our understanding of these processes and their representation in climate models. In this dissertation, I describe improvements to our understanding of the behavior of the GrIS and pose improvements to climate modeling capabilities that can lead to a reduced uncertainty of sea level rise projections.In the first chapter, I put constraints on the past response of the GrIS to a changing climate. Understanding the response of the GrIS to times in the past when temperatures were as warm or warmer than today offers insights into its current and future response to climate change. The southwestern GrIS retreated inland beyond its current margin during the (at least regionally) warmer-than-present mid-Holocene, before it readvanced. To investigate the timing and magnitude of southwest GrIS retreat and readvance in response to Holocene warmth, we model the response of the solid Earth and local relative sea level (RSL) to past ice sheet change. I compare model predictions to observations of paleo and present-day RSL and present-day vertical land motion around Nuuk, Greenland. I find that the southwest GrIS minimum extent likely occurred between 5 and 3 ka and that the historical maximum extent was likely approached between 2 and 1 ka. Comparing this timing to local and regional records of temperature and ice-sheet change suggest that the evolution of the southwestern GrIS presented here was in-phase with the likely evolution of southwestern GrIS mass balance through the Holocene.In the second chapter, I assess the performance of a regional climate model in simulating the spatiotemporal variability of GrIS ice extent and ice albedo in the period 2000-2021. A large portion of runoff from the GrIS originates from exposure of the darker ice in the ablation zone when the overlying snow melts, where surface albedo plays a critical role in modulating the energy available for melting. Ice albedo is spatially and temporally variable and contingent on non-linear feedbacks and the presence of light-absorbing constituents. An assessment of models aiming at simulating albedo variability and associated impacts on meltwater production is crucial for improving our understanding of the processes governing these feedbacks and, in turn, surface mass loss from Greenland. Our findings suggest that the regional climate model Modele Atmospherique Regional (MAR) overestimates ice albedo on average by 22.8 % compared to the ice albedo observations derived from the Moderate Resolution Imaging Spectroradiometer (MODIS). We also find that this ice albedo bias can lead to an underestimation of total meltwater production from the GrIS ice zone of 42.8 %.In the third chapter, I build upon the second chapter and present PIXAL, a physics-informed explainable machine learning architecture for Greenland ice albedo modeling. PIXAL is an Extreme Gradient Boosting (XGBoost) model and is trained on a suite of modeled topographic, atmospheric, radiative, and glaciologic variables from MAR to capture the complex and non-linear relationships with ice albedo observations from MODIS in the period 2000-2021. PIXAL outperforms MAR in modeling ice albedo on the southwestern GrIS. The performance metrics show that PIXAL achieves an R2 of 0.563, an SSIM of 0.620, an MSE of 0.005, and a MAPE of 14.699%, compared to MAR's R2 of 0.062, SSIM of 0.112, MSE of 0.032, and MAPE of 46.202%. Explainable artificial intelligence (XAI) analysis reveals that topographic features, specifically ice sheet surface height and slope, are primary drivers of ice albedo. Near-surface air temperature and runoff further impact ice albedo. These findings highlight that understanding the complex physical processes underlying ice albedo variability is essential for accurate climate modeling and sea level rise predictions. PIXAL represents a crucial advancement in ice albedo modeling and paves the way for improved climate models that can more accurately estimate GrIS ice surface melting and its contribution to sea level rise.Overall, my results have implications for future ice sheet modeling studies targeting Greenland and provide a deeper understanding of the interactions between the climate and the cryosphere and thus of future ice sheet change.
일반주제명  
Climate change
일반주제명  
Environmental science
일반주제명  
Atmospheric sciences
일반주제명  
Computer science
키워드  
Albedo
키워드  
Cryosphere
키워드  
Greenland
키워드  
Ice melting
키워드  
Machine learning
키워드  
Remote sensing
기타저자  
Columbia University Earth and Environmental Sciences
기본자료저록  
Dissertations Abstracts International. 86-01B.
전자적 위치 및 접속  
로그인 후 원문을 볼 수 있습니다.

MARC

 008250123s2024        us                              c    eng  d
■001000017163099
■00520250211152135
■006m          o    d                
■007cr#unu||||||||
■020    ▼a9798383564288
■035    ▼a(MiAaPQ)AAI31484060
■040    ▼aMiAaPQ▼cMiAaPQ
■0820  ▼a363
■1001  ▼aAntwerpen,  Rafael  Marco.
■24510▼aUnderstanding  Drivers  of  Ice  Mass  Loss  in  Greenland  Through  Sea-Level  and  Climate  Modeling,  Remote  Sensing,  and  Machine  Learning
■260    ▼a[Sl]▼bColumbia  University▼c2024
■260  1▼aAnn  Arbor▼bProQuest  Dissertations  &  Theses▼c2024
■300    ▼a138  p
■500    ▼aSource:  Dissertations  Abstracts  International,  Volume:  86-01,  Section:  B.
■500    ▼aAdvisor:  Tedesco,  Marco.
■5021  ▼aThesis  (Ph.D.)--Columbia  University,  2024.
■520    ▼aChanges  in  global  climate  conditions  significantly  impact  ice  sheet  and  glacier  mass  change  leading  to  global  mean  sea  level  (GMSL)  change.  One  of  the  largest  present-day  contributors  to  GMSL  is  the  Greenland  ice  sheet  (GrIS)  and  it  will  likely  continue  to  be  so  in  the  future.  To  accurately  predict  future  ice  mass  changes,  it  is  crucial  to  understand  the  response  of  GrIS  to  a  changing  climate  and  to  correctly  represent  this  behavior  in  climate  models.  The  GrIS'  contribution  to  GMSL  can  in  large  part  be  attributed  to  the  loss  of  ice  and  snow  mass  from  the  ice  sheet  surface.  The  surface  mass  loss  has  accelerated  in  the  past  decades  due  to  increased  surface  melting  and  runoff  in  response  to  atmospheric  warming.  Surface  melting  is  strongly  controlled  by  ice  albedo,  a  complex  and  dynamic  property  of  ice  that  regulates  the  amount  of  solar  radiation  that  is  absorbed  or  reflected  by  the  surface.  Absorbed  solar  radiation  leads  to  heating  and  melting  of  the  ice  surface.  However,  we  lack  a  comprehensive  understanding  of  the  physical  processes  controlling  ice  mass  loss,  including  ice  albedo.  These  processes  are,  therefore,  often  simplified  or  crudely  parameterized  in  climate  models  and  subsequently  add  to  large  uncertainties  in  sea  level  rise  predictions.  This  uncertainty  prevents  effective  mitigation  of  and  adaptation  to  the  effects  of  climate  change  and  sea  level  rise.  It  is,  therefore,  essential  to  advance  our  understanding  of  these  processes  and  their  representation  in  climate  models.  In  this  dissertation,  I  describe  improvements  to  our  understanding  of  the  behavior  of  the  GrIS  and  pose  improvements  to  climate  modeling  capabilities  that  can  lead  to  a  reduced  uncertainty  of  sea  level  rise  projections.In  the  first  chapter,  I  put  constraints  on  the  past  response  of  the  GrIS  to  a  changing  climate.  Understanding  the  response  of  the  GrIS  to  times  in  the  past  when  temperatures  were  as  warm  or  warmer  than  today  offers  insights  into  its  current  and  future  response  to  climate  change.  The  southwestern  GrIS  retreated  inland  beyond  its  current  margin  during  the  (at  least  regionally)  warmer-than-present  mid-Holocene,  before  it  readvanced.  To  investigate  the  timing  and  magnitude  of  southwest  GrIS  retreat  and  readvance  in  response  to  Holocene  warmth,  we  model  the  response  of  the  solid  Earth  and  local  relative  sea  level  (RSL)  to  past  ice  sheet  change.  I  compare  model  predictions  to  observations  of  paleo  and  present-day  RSL  and  present-day  vertical  land  motion  around  Nuuk,  Greenland.  I  find  that  the  southwest  GrIS  minimum  extent  likely  occurred  between  5  and  3  ka  and  that  the  historical  maximum  extent  was  likely  approached  between  2  and  1  ka.  Comparing  this  timing  to  local  and  regional  records  of  temperature  and  ice-sheet  change  suggest  that  the  evolution  of  the  southwestern  GrIS  presented  here  was  in-phase  with  the  likely  evolution  of  southwestern  GrIS  mass  balance  through  the  Holocene.In  the  second  chapter,  I  assess  the  performance  of  a  regional  climate  model  in  simulating  the  spatiotemporal  variability  of  GrIS  ice  extent  and  ice  albedo  in  the  period  2000-2021.  A  large  portion  of  runoff  from  the  GrIS  originates  from  exposure  of  the  darker  ice  in  the  ablation  zone  when  the  overlying  snow  melts,  where  surface  albedo  plays  a  critical  role  in  modulating  the  energy  available  for  melting.  Ice  albedo  is  spatially  and  temporally  variable  and  contingent  on  non-linear  feedbacks  and  the  presence  of  light-absorbing  constituents.  An  assessment  of  models  aiming  at  simulating  albedo  variability  and  associated  impacts  on  meltwater  production  is  crucial  for  improving  our  understanding  of  the  processes  governing  these  feedbacks  and,  in  turn,  surface  mass  loss  from  Greenland.  Our  findings  suggest  that  the  regional  climate  model  Modele  Atmospherique  Regional  (MAR)  overestimates  ice  albedo  on  average  by  22.8  %  compared  to  the  ice  albedo  observations  derived  from  the  Moderate  Resolution  Imaging  Spectroradiometer  (MODIS).  We  also  find  that  this  ice  albedo  bias  can  lead  to  an  underestimation  of  total  meltwater  production  from  the  GrIS  ice  zone  of  42.8  %.In  the  third  chapter,  I  build  upon  the  second  chapter  and  present  PIXAL,  a  physics-informed  explainable  machine  learning  architecture  for  Greenland  ice  albedo  modeling.  PIXAL  is  an  Extreme  Gradient  Boosting  (XGBoost)  model  and  is  trained  on  a  suite  of  modeled  topographic,  atmospheric,  radiative,  and  glaciologic  variables  from  MAR  to  capture  the  complex  and  non-linear  relationships  with  ice  albedo  observations  from  MODIS  in  the  period  2000-2021.  PIXAL  outperforms  MAR  in  modeling  ice  albedo  on  the  southwestern  GrIS.  The  performance  metrics  show  that  PIXAL  achieves  an  R2  of  0.563,  an  SSIM  of  0.620,  an  MSE  of  0.005,  and  a  MAPE  of  14.699%,  compared  to  MAR's  R2  of  0.062,  SSIM  of  0.112,  MSE  of  0.032,  and  MAPE  of  46.202%.  Explainable  artificial  intelligence  (XAI)  analysis  reveals  that  topographic  features,  specifically  ice  sheet  surface  height  and  slope,  are  primary  drivers  of  ice  albedo.  Near-surface  air  temperature  and  runoff  further  impact  ice  albedo.  These  findings  highlight  that  understanding  the  complex  physical  processes  underlying  ice  albedo  variability  is  essential  for  accurate  climate  modeling  and  sea  level  rise  predictions.  PIXAL  represents  a  crucial  advancement  in  ice  albedo  modeling  and  paves  the  way  for  improved  climate  models  that  can  more  accurately  estimate  GrIS  ice  surface  melting  and  its  contribution  to  sea  level  rise.Overall,  my  results  have  implications  for  future  ice  sheet  modeling  studies  targeting  Greenland  and  provide  a  deeper  understanding  of  the  interactions  between  the  climate  and  the  cryosphere  and  thus  of  future  ice  sheet  change.
■590    ▼aSchool  code:  0054.
■650  4▼aClimate  change
■650  4▼aEnvironmental  science
■650  4▼aAtmospheric  sciences
■650  4▼aComputer  science
■653    ▼aAlbedo
■653    ▼aCryosphere
■653    ▼aGreenland
■653    ▼aIce  melting
■653    ▼aMachine  learning
■653    ▼aRemote  sensing
■690    ▼a0404
■690    ▼a0768
■690    ▼a0725
■690    ▼a0984
■71020▼aColumbia  University▼bEarth  and  Environmental  Sciences.
■7730  ▼tDissertations  Abstracts  International▼g86-01B.
■790    ▼a0054
■791    ▼aPh.D.
■792    ▼a2024
■793    ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17163099▼nKERIS▼z이  자료의  원문은  한국교육학술정보원에서  제공합니다.

미리보기

내보내기

chatGPT토론

Ai 추천 관련 도서


    신착도서 더보기
    최근 3년간 통계입니다.

    소장정보

    • 예약
    • 소재불명신고
    • 나의폴더
    • 우선정리요청
    • 비도서대출신청
    • 야간 도서대출신청
    소장자료
    등록번호 청구기호 소장처 대출가능여부 대출정보
    TF11253 전자도서 대출가능 마이폴더 부재도서신고 비도서대출신청 야간 도서대출신청

    * 대출중인 자료에 한하여 예약이 가능합니다. 예약을 원하시면 예약버튼을 클릭하십시오.

    해당 도서를 다른 이용자가 함께 대출한 도서

    관련 인기도서

    로그인 후 이용 가능합니다.