Choosing a Convection Scheme for ACME: Recent Progress

Shaocheng Xie1, Phil Rasch2, Wuyin Lin3, J. Yoon2, P. Ma2

1Lawrence Livermore National Lab, 2Pacific Northwest National Lab 3Brookhaven National Lab,

ACME Convection Team S. Xie*, W. Lin, J. Bacmeister, S. Mahajan, R. Neale, Q. Tang, H. Wang, J. Yoon, K. Zhang, Y. Zhang

Prepared by LLNL under Contract DE-AC52-07NA27344. LLNL-PRES-667184

ACME & DOE Collaborators S. Ghan, P. Caldwell (SciDAC) S. Klein (RRM) P. Ma (Vert. Res) Y. Qian (Short Simulations) H. Ma (CAPT)

Developer Collaborators V. Larson (CLUBB) A. Gettelman (CLUBB, MG2) P. Bogenschutz (CLUBB) S. Park (UNICON) G. Zhang (ZM-TRIMEM)

Why Is This Effort Needed? The atmospheric component of ACME starts with CAM5.3, but

we will run with SE Dyn-Core with new parameterizations of aerosols and clouds that were developed through the earlier DOE "polar" project.

ACME is a high-resolution Earth System Model with a resolution of 0.250/60-levels for atmos and 1/100 for ocean)

Candidate convection schemes (e.g. CLUBB, UNICON, and ZM variants) have not been extensively tested at such high resolutions

Sensitivity to dynamical core and sub-grid scale orography has not be extensively explored

Adjust to accommodate the other parameterizations

ne30L30 –> ne120L60+

Protocol for Convection Tests

SCM - Single-Column Model CAPT - Cloud-Associated Parameterization Testbed RRM - Regional Refined Model AMIP - Free run with specified SST and Sea Ice Short ensemble UQ - Optimize a set of parameters

A Quick Update • The SCM/CAPT/RRM/UQ capabilities have been added to

ACME • Implementation of candidate schemes started the

beginning of 2015 and most tests were done in the past few weeks

• Implementations to ACME are successfully. One-degree AMIP runs with FV has produced comparable results as shown in the CESM tests for CLUBB and UNICON.

• Issues are found with dynamical cores, model resolutions, and parameter settings

• Tuning and adjustment are necessary to make these schemes work for ACME

Highlights from Initial Tests with one degree AMIP simulations

Mean Precip Improved by CLUBB AMIP 2008-2009, JJA Total Precipitation

CNTL - GPCP

UNICON - CNTL

MG2 - CNTL CLUBB - CNTL

ZMTM - CNTL

Improved Skill in Capturing Diurnal Variability with CLUBB and UNICON

April-July 2009 AMIP: Precip Diurnal Cycle

ZMTM

NEXRAD

CLUBB MG2

CNTL

UNICON

Slide from Wuyin Lin

CLUBB UNICON

Improved MJO with CLUBB and UNICON, but a lack of Kelvin wave in CLUBB OBS CNTL

Issues with Dyn-Cores, Vertical Resolutions, and Parameters

Difference is large between SE and FV with the use of UNICON in ACME

Mean Cloud at the first month

SE - FV

FV

SE

Slide from Jinho Yoon

CESM1 shows similar difference between SE and FV with UNICON – Not an implementation problem!

ACME_UNICON

SE - FV

FV

SE

CESM1_UNICON

Slide from Jinho Yoon

Differences significantly reduced when the orographic forcing term in UNICON turned off

Mean Cloud at the first month

SE - FV

FV

SE

Is there any difference in treating extra tracers in SE vs. in FV?

Slide from Jinho Yoon

CLUBB shows larger sensitivity to vertical resolutions compared to the default schemes

Annual Mean Precip from 1-year AMIP run

Slide from Po-Lun Ma

NE30L30

L64b-L30

L64a-L30

ACME ACME-CLUBB 3.0 mm/day

0.09 mm/day -0.34 mm/day

-0.35 mm/day

2.47 mm/day

0.01 mm/day

Parameter setting matters Precipitation significantly reduced with ACME v0.1 tuning parameters

With NCAR default settings xcldfrc_rhminl = 0.8975 cldfrc_dp1= 0.1 dcs=195D-6 dust_emis_fact= 0.55 zmconv_c0_lnd = 0.0059 zmconv_c0_ocn = 0.045

2.89 mm/day

2.47 mm/day With ACME V0.1 Tuning xcldfrc_rhminl = 0.91 cldfrc_dp1= 0.025 dcs=600D-6 dust_emis_fact= 1.05 zmconv_c0_lnd = 0.0035 zmconv_c0_ocn = 0.0035

Near Team Plan • Have the initial tests with computational inexpensive

modeling testbeds (SCM, CAPT, RRM, AMIP) done in the mid of July, 2015

• Some tunings and adjustments will be done based on the modeling testbeds and vertical resolution will be determined before long-term 0.25 degree resolution simulations can be conducted

• A one-year AMIP/CAPT-type runs with NE120L60+ will be done by the end of August, 2015

• Short ensemble runs might be used to optimize the parameter settings.

Summary • Candidate convection schemes (CLUBB+MG2, UNICON,

ZM variants) have been successfully implemented into ACME and yield similar results as those tested in CESM1

• Issues have been found with the use of SE dyn-core, vertical resolutions, ACME parameter settings

• Tunings and adjustments are inevitably needed to make selected convection scheme work for ACME

• More rigorous tests (including high resolution tests) will be conducted to further assess these schemes before a recommendation could be made