Development of an Accurate 3D Monte Carlo Broadband Atmospheric Radiative Transfer Model

168 Pages·2016·5.17 MB·English

by  Alexandra L. Jones

©2016byAlexandraL.Jones. Allrightsreserved.
DEVELOPMENTOFANACCURATE3DMONTECARLOBROADBAND
ATMOSPHERICRADIATIVETRANSFERMODEL
BY
ALEXANDRAL.JONES
DISSERTATION
Submittedinpartialfulfillmentoftherequirements
forthedegreeofDoctorofPhilosophyinAtmosphericSciences
intheGraduateCollegeofthe
UniversityofIllinoisatUrbana-Champaign,2016
Urbana,Illinois
DoctoralCommittee:
ProfessorLarryDiGirolamo,Chair
ResearchScientistBrianJewett
ProfessorRobertRauber
AssociateProfessorNicoleRiemer
Abstract
Radiation is the ultimate source of energy that drives our weather and climate. It is also
the fundamental quantity detected by satellite sensors from which earth’s properties are
inferred. Radiative energy from the sun and emitted from the earth and atmosphere is
redistributed by clouds in one of their most important roles in the atmosphere. Without
accurately representing these interactions we greatly decrease our ability to successfully
predict climate change, weather patterns, and to observe our environment from space.
The remote sensing algorithms and dynamic models used to study and observe earth’s
atmosphere all parameterize radiative transfer with approximations that reduce or ne-
glect horizontal variation of the radiation field, even in the presence of clouds. Despite
having complete knowledge of the underlying physics at work, these approximations
persist due to perceived computational expense. In the current context of high resolu-
tionmodelingandremotesensingobservationsofclouds,fromshallowcumulustodeep
convectiveclouds,andgivenoureveradvancingtechnologicalcapabilities,theseapprox-
imationshavebeenexposedasinappropriateinmanysituations. Thispresentsaneedfor
accurate 3D spectral and broadband radiative transfer models to provide bounds on the
interactionsbetweencloudsandradiationtojudgetheaccuracyofsimilarbutlessexpen-
sive models and to aid in new parameterizations that take into account 3D effects when
coupledtodynamicmodelsoftheatmosphere.
Developing such a state of the art model based on the open source, object-oriented
framework of the I3RC Monte Carlo Community Radiative Transfer (“IMC-original”)
Model is the task at hand. It has involved incorporating (1) thermal emission sources
ii
of radiation (“IMC+emission model”), allowing it to address remote sensing problems
involving scattering of light emitted at earthly temperatures as well as spectral cooling
rates, (2) spectral integration across an arbitrary range of the electromagnetic spectrum
(“MCBRaT-3D” model) to produce heating rates relevant to atmospheric dynamics, and
(3) developing tools to interface between the model and databases of single scattering
properties of the real atmosphere. Special attention has been paid to practical aspects of
implementationforhighperformancecomputingonBlueWaters. Incrementaltestsofthe
accuracyofeachnewcomponenthavebeenperformedbasedoncarefullydesignedana-
lytical solutions, culminating in the “MCBRaT-3D” model’s ability to reproduce a profile
of broadband atmospheric heating rate from a published intercomparison study and its
initial use to provide benchmarking results for the evaluation of other 3D models. The
models, tools, and databases developed herein provide a significant contribution to the
fieldofatmosphericscience,especiallybymakingthempubliclyavailableforfurtherde-
velopment and use. They will enable quantitative evaluations of radiative transfer mod-
els and parameterizations by providing a highly accurate solution to many 3D radiative
transferproblemsencounteredintheatmosphericsciences.
iii
Tothosewhosufferfromimpostorsyndrome,rememberit’swhattodoandhowtodoitthatare
therealquestions,neverifyoucandoit.
iv
Acknowledgments
This research is part of the Blue Waters sustained-petascale computing project, which is
supported by the National Science Foundation (awards OCI-0725070, ACI-1238993 and
ACI-14-44747) and the state of Illinois. Blue Waters is a joint effort of the University of
IllinoisatUrbana-ChampaignanditsNationalCenterforSupercomputingApplications.
ThisworkwasalsosupportedbyNASAHeadquartersundertheNASAEarthandSpace
ScienceFellowship(NESSF)program-GrantNNX11AL39H.
I have been fortunate enough to spend 4 years as a pre-doctoral fellow, 3 years in the
NESSF program, and 1 year as a Blue Waters Graduate Fellow. I gratefully acknowledge
these programs as they have allowed me to pursue research directions I may not have
had the opportunity to otherwise. A few additional semesters of support were provided
by NSF and NASA grants, specifically NASA grant NNX14AJ27G and NSF grant ATM-
08-54954.
There are a number of individuals who have provided guidance to me as I grew my
scientific computing knowledge. Without that guidance my dissertation research would
have taken a very different path and I would not be able to call myself a computational
scientist. This includes friends: Matt Erickson, Phil Miller, Justin McHugh; Department
ofAtmosphericSciencesStaff: BillChapman,BrianJewett,KenPatten,DavidWojtowicz;
andBlueWaterssupportstaff: GalenArnold,CraigSteffen,andMarkStraka.
Imustalsoacknowledgethesupportobtainedthroughnumerousconversationsabout
3DradiativetransferamongtheDiGirolamoresearchgroupinthecourseofmyPhDdis-
sertation research, most notably with Allison Houghton, Daeven Jackson, Megan King,
v
Max Smith, and Guangyu Zhao. I must especially acknowledge my advisor, Larry Di
Girolamo, who instilled in me both a drive to never settle for anything less than my best
workaswellasaconfidenceinmyabilitytoproducetopofthelinescience.
I am grateful to the Department of Atmospheric Sciences at the University of Illinois
foritscommitmenttosupportingitsgraduatestudentsinallaspectsoftheirprofessional
development, from administrative support from staff, to opportunities for leadership,
suchasthatprovidedtomethroughthedevelopmentoftheMidwestCloudandAerosol
Forum,totheopendoorpolicythatencouragescollaborationandmentorshipacrosssub-
disciplines. Ihavehadnumerousconversationswithfaculty,bothinsideandoutsidemy
committee membership, that have shaped the direction of my research and professional
growth.
Finally, this project has been a long and grueling journey that I would not have had
the strength to endure without the emotional support of my friends and family near and
far. The Wesley United Methodist Church and Foundation have become my family and
have made Champaign-Urbana a place that will always feel like a home to me. Since
nothing compares to the support a peer can provide, I must also thank Luke Bard, Jeff
Curtis, Bobby Jackson, Jason Keeler, Catrin Mills, Sara Strey for walking this path with
me.
vi
Table of Contents
Chapter1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 TheImportanceofRadiationinaCloudyAtmosphere . . . . . . . . . . . . . 1
1.2 CommonlyUsedRepresentationsofRadiation: 1D . . . . . . . . . . . . . . 3
1.3 Problemswith1DRepresentations . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 EvaluatingtheNeedsoftheRadiativeTransferCommunity . . . . . . . . . 11
1.5 What’sBeenDone? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.6 TheRoleofthisProject . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Chapter2 RadiativeTransferBackground . . . . . . . . . . . . . . . . . . . . . . . 16
2.1 BasicRadiativeQuantities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2 RadiativePropertiesoftheAtmosphere . . . . . . . . . . . . . . . . . . . . . 18
2.3 1DRadiativeTransferEquation . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4 3DRadiativeTransfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.4.1 MonteCarloMethods . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.5 SpectralIntegration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.5.1 OperationalApproaches . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.5.2 AdvancedResearchApproaches . . . . . . . . . . . . . . . . . . . . . 32
Chapter3 MonochromaticModel . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.1 The“IMC-original”Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.1.1 CodeBaseStructure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.1.2 CalculationofRadiativeQuantities . . . . . . . . . . . . . . . . . . . 38
3.1.3 AlgorithmHighlights . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.2 The“IMC+emission”Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.2.1 PhysicalBasisforInternalEmission . . . . . . . . . . . . . . . . . . . 42
3.2.2 ComputationalImplementationHighlights . . . . . . . . . . . . . . . 49
3.3 Verification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.3.1 Radiance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.3.2 Flux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.3.3 FluxDivergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
Chapter4 BroadbandIntegration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.1 PhysicalandComputationalBasisfor“MCBRaT-3D”Algorithm . . . . . . . 89
4.1.1 ComputingtotalincidentoremittedFlux . . . . . . . . . . . . . . . . 90
4.1.2 AssigningPhotonFrequency . . . . . . . . . . . . . . . . . . . . . . . 91
vii
4.1.3 ComputationalImplementationHighlights . . . . . . . . . . . . . . . 94
4.2 Verification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.2.1 DistributionofPhotonsSpectrallyandSpatially . . . . . . . . . . . . 95
4.2.2 MagnitudeofIncomingorEmittedBroadbandIrradiance . . . . . . 96
4.2.3 RadiativeQuantities . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
Chapter5 Applicationofthe“MCBRaT-3D”ModeltotheRealAtmosphere . . . 105
5.1 Inputstothe“MCBRaT-3D”Model . . . . . . . . . . . . . . . . . . . . . . . . 106
5.1.1 PhysicalDomainFile . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
5.1.2 OpticalPropertyDatabases . . . . . . . . . . . . . . . . . . . . . . . . 106
5.1.3 SolarSourceFunction . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
5.1.4 Namelist . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
5.2 OnlineConstructionofDomainsfromInputs . . . . . . . . . . . . . . . . . . 109
5.2.1 RayleighScatteringComponents . . . . . . . . . . . . . . . . . . . . . 109
5.2.2 GaseousAbsorptionComponents . . . . . . . . . . . . . . . . . . . . 111
5.2.3 ParticleAbsorptionandScatteringComponents . . . . . . . . . . . . 112
5.3 ComputationalHighlights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.4 GaseousAbsorptionDatabase . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
5.5 WaterDropletAbsorptionandScatteringDatabase . . . . . . . . . . . . . . 122
5.6 Testingthe“MCBRaT-3D”ModelwithaRealisticCase . . . . . . . . . . . . 126
5.6.1 SettingUpCIRCCase7 . . . . . . . . . . . . . . . . . . . . . . . . . . 126
5.6.2 CIRCCase7Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
5.7 “MCBRaT-3D”ModelLimitations . . . . . . . . . . . . . . . . . . . . . . . . 135
Chapter6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
6.1 AdditionalBenchmarkResults . . . . . . . . . . . . . . . . . . . . . . . . . . 141
6.1.1 ImportanceofSpatialResolutiontoBroadbandHeatingRates . . . . 142
6.1.2 PlaneParallel/ICAvs3DBroadbandHeatingRates . . . . . . . . . . 143
6.2 ComputationalChallengesandBestPractices . . . . . . . . . . . . . . . . . . 147
6.2.1 DealingwithDataSize . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
6.2.2 EfficientSimulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
6.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
viii
Chapter 1
Introduction
1.1 The Importance of Radiation in a Cloudy Atmosphere
Radiationistheultimatesourceofenergythatdrivesatmosphericcirculationsatavariety
of spatial and temporal scales. It is the uneven distribution of absorbed radiation on our
spherical planet between the poles and the equator (figure 1.1) that kicks off the Hadley
Circulation, the most basic model for large scale circulation in earth’s atmosphere. The
average solar flux received at the equator is much larger than that received at the poles
becauseofthedifferenceinorientationoftheearth’ssurfaceatthoselocationsrelativeto
thedirectionofincomingsunlight. Thedeficitbetweenabsorbedandemittedradiationat
thepolesandthesurplusneartheequatormeansthatheatmustbetransportedpoleward
by other means, such as atmospheric and oceanic circulations, to avoid ever increasing
coolingatthepolesandwarmingattheequator. Turningtoweatherforanotherexample,
the difference in net radiation absorbed by water and land, due to the differences in ab-
sorptivity and emissivity of the materials, sets off sea and lake breezes that keep coastal
areascoolduringwarmdays.
Theverticaldistributionofabsorbedradiationintheatmosphereandatthesurfaceis
controlled largely by the distribution of clouds, since they have such a strong impact on
the absorbed solar and emitted thermal radiation, are so temporally and spatially vari-
able,andyetaresoplentifulintheatmosphere. Figure1.2showstheglobalannualmean
energy balance of the earth. Almost a quarter of incoming solar radiation is redirected
back to space by atmospheric components, largely clouds. Clouds also act to reduce
1