LES and Climate Modeling with COMBLE

GISS ModelE3-devel

GISS ModelE2.1

source: Andrew Ackerman / GISS

Ann Fridlind • NASA GISS

with contributions from many ...

subtropical �Sc decks

Background: Clouds in Earth system models (ESMs)

• across ESMs, differences in cloud-related physics contribute substantially to large differences in simulated climate change
• during ESM development, cloud physics parameters are commonly slightly tuned to bring simulations into line with targets such as preindustrial TOA radiative balance (e.g., Schmidt et al. GMD 2017)
• as Earth warms in a 2X CO₂ experiment, ESMs predict differing cloud changes, which feed back on warming (cloud-climate feedbacks)
• e.g., CMIP6 ESM-predicted range of 1.8–5.6 K surface warming was found to be inconsistent with other estimates of 1.5–4.5 K, "tied to the physical representation of clouds" (Zelinka et al. GRL 2020)
• clouds participate in a complex coupled Earth system environment

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov

ModelE3 development approach

Field campaigns 🡪 LES 🡪 SCM

CALIPSO

Global data 🡪 ESM tuning

GMAO/cubed-sphere

ACTIVATE Flight RF13

1 March 2020

mixed-phase cold-air outbreak

Elsaesser et al., in prep.

Tornow et al. (ACP 2021)

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov

NASA GISS large-eddy simulation (LES) model

• Distributed Hydrodynamic Aerosol-Radiation Model for Atmospheres (DHARMA)
• second-order forward-in-time process-splitting with advection in semi-Lagrangian spirit (Stevens and Bretherton, 1997)
• dynamic Smagorinsky subgrid-scale mixing (Kirkpatrick et al., 2006)
• two-stream radiative transfer (Toon et al., 1989)
• microphysics options
• condensational adjustment (liquid only)
• two-moment mixed-phase cloud microphysics after Morrison et al., 2009 (up to five species) with prognostic supersaturation (after Morrison and Grabowski, 2008) and optional substitution of alternative schemes for individual processes
• fixed droplet number or multi-modal one-moment prognostic aerosol
• size-resolved microphysics (Ackerman et al., 1995; Fridlind et al., 2004) with flexible treatment of multiple non-spherical ice classes (Fridlind et al., 2017)
• can be run as a column model (Fridlind et al. 2015) or a parcel model (Ackerman et al., 2015)
• used in scientific studies and as an engineering tool for ESM development

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov

Field campaigns —> LES —> Single-column model (SCM)

5

*Lagrangian (cf. Neggers JAMES 2015, Pithan et al. NatGeo 2019)

^ensemble (cf. Neggers et al. JAMES 2019)

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov

Moving into the Lagrangian reference frame

• Lagrangian and aerosol-aware go hand-in-hand

• SHEBA [Morrison et al. 2011]

Vertically pointing mm-wavelength radar

Fridlind and Ackerman [Ch. 7 in Mixed-Phase Clouds, Ed. C. Andronache, 2018]

• ISDAC [Ovchinnikov et al. 2014]

• M-PACE [Klein et al. 2009]

M-PACE [Klein et al. 2009]

Eulerian �(stationary)

Lagrangian �(moving with cloudy air)

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov

Background: Ice formation in supercooled clouds

• esp. large uncertainties remain in supercooled cloud physics
• ice spontaneously melts at 0°C but does not spontaneously form until near –38°C owing to energetic barriers to forming a more organized state
• within that temperature range, supercooled liquid appears to be a gateway to persistent weak ice formation, attributable to ice-nucleating particle activation (Silber et al. 2020 based on NSA and AWARE data)
• however, ice crystals are commonly orders of magnitude more abundant than can be explained by that weak pathway, especially in clouds with riming and/or large drops (Rangno and Hobbs, 2002; Korolev et al., 2021)
• exactly how much more abundant is often unknown owing to lack of reliable measurements (Korolev et al. 2021; Morrison et al. 2020)

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov

Ice formation approach in ModelE3

• Only physically-based mechanisms and parameterizations
• Avoid unnecessary complexity
• Each mechanism should be demonstrably active in observed case studies
• Heterogeneous freezing mechanisms should be linkable to aerosol properties
• But start with diagnostic INP and tune with machine learning
• DeMott et al. 2010 * f_{scale_iifn}
• f_{scale_iifn} < 1 can crudely account for efficient precipitation scavenging (Fridlind et al. JAS 2012)

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov

* additive to Gettelman and Morrison (2015)

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov

Highly supercooled drizzle over Antarctica

AWARE campaign case study (Silber et al. JGR 2019)

CTT ≈ –25°C

Lubin et al.�(BAMS 2019)

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov

AWARE case study

• SCM performs quite well
• stable conditions common�(Silber et al. ACP 2020; GRL 2021)

see Silber et al. [GMD 2022]

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov

ACTIVATE LES case study selection

• choose 2020-2022 flights with greatest fetch offshore

Tornow et al. �(in prep.)

MBL �aerosol

FT �aerosol

DHARMA LES

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov

COMBLE LES case study selection

• choose CAO back trajectories passing near Zeppelin

source: Christian Lackner

source: �Paul DeMott

Tornow et al. (in prep.; see AMS 12B.5)

INP at Andenes

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov

COMBLE observational constraint

​

​

​

Silber et al. �(in prep.)

• use EMC² (Silber et al. GMD 2022) to evaluate LES vs ground-based radar + lidar

LWP confined to cell cores

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov

COMBLE observational constraint

​

​

​

Silber et al. �(in prep.)

• use EMC² (Silber et al. GMD 2022) to evaluate LES vs CALIPSO satellite
• LES clouds too deep + dense

DHARMA LES

CALIPSO

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov

Now onto machine learning tuning...

• LES modeler: We don't even understand how most atmospheric ice crystals are formed!
• Climate modeler: We need to deliver a climate model now...

​

GISS ModelE3-devel

GISS ModelE2.1

source: Andrew Ackerman / GISS

subtropical �Sc decks

Tuning Protocol

17

​

• scale_iifn is one of 45 parameters taken to be poorly constrained
• LES/SCM used to estimate parameter ranges
• new satellite datasets used in tuning
• satellite dataset uncertainties are specified

source: Greg Elsaesser

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov

correlation�with mean of�obs metrics

*

*

*

source: Greg Elsaesser

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov

19

ModelE3 emulator based on 450 1-year atmosphere runs

Latin Hypercube sampling in a 45-dimensional parameter state space. Lots of empty state space; emulator (neural network) fills in the gaps.

P1

P2

P3

Example Penalty State Space

Transect for any given model metric

source: Marcus van Lier-Walqui

After The Machine

• photo of white board at GISS

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov

Aerosol indirect effect and ECS from E3 candidates

• AIE from 2000-2010 AMIP runs, PD minus PI offline aerosol for droplet activation only
• ECS from 30-year Q-flux PI runs

​

​

​

​

21

source: Andy Ackerman

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov

ModelE3 supercooled cloud fraction vs CALIPSO

• COSP simulator modified to see ”precipitation”
• note: cloud ice is continuous with precipitating ice (e.g., Fridlind et al. JAS 2012)
• ”precipitation” also affects cloud feedbacks across ModelE3s

Cesana et al. (GRL 2021, Fig. S6)

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov

AWARE case study

• SCM performs quite well
• stable conditions common�(Silber et al. ACP 2020; GRL 2021)

see Silber et al. [GMD 2022]

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov

A new ground-based lidar/radar simulator: EMC²

• Earth Model Column Collaboratory
• Python open source, community code base
• tool to evaluate supercooled cloud fraction, cloud base and surface precipitation, ...

Silber, Jackson, Collis et al. (GMD, 2021)

from surface:

sounding approach

​

​

from surface:

lidar approach

​

​

from space

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov

COSP simulator tested on LES/SCM AWARE case

Cesana et al. (GRL 2021, Fig. S1)

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov

Ground-based with EMC²

Stanford et al. (ACP, 2023)

McFarquahar

et al. 2021

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov

SCM

LES

ESM

PARCEL + 1D

satellite + COSP

long-term ground + EMC²

AEROICESTUDY

KIT laboratory SIP

calibration

aircraft field campaigns

long-term ground

development

development

• ESM development workhorse

• pre-calibration tool

• simulator testbed

• cloud feedback analysis tool

• primary and secondary ice formation

+ rain formation and mesoscale structure

+ gravity waves, surface fluxes, ice properties,

aerosol-cloud interactions, ...

DOE ARM Summer School • 22 May 2024 • ann.fridlind@nasa.gov