The Performance of CAM5 Physics Modules at High Spatial Resolution

P.J. Rasch, J.D. Fast, W.I. Gustafson Jr., B. Singh, R.C. Easter, P.-L. Ma, and M. Shrivastava - PNNL G. Pfister, S. Walters, and J.-F. Lamarque - National Center for Atmospheric Research

Motivation

Acknowledgements: This research was supported by the U.S. DOE’s ESM Program under contract DE-AC06-76RCO 1830 at PNNL and uses software tools developed under the PNNL LDRD program.

Concept for Earth System Modeling Project on Arctic Processes

Objectives

Next Steps Comparing Aerosol Predictions

xx

Weather Research & Forecasting (WRF)

Aerosol Modeling Testbed

high resolution

low resolution medium resolution

typical CAM grid cell 2.5 x 1.9 degrees

Multi-Scale Simulations  There has been relatively little interaction between the cloud-resolving / mesoscale and global modeling communities

 CAM will likely be run at Δx = 10 – 20 km in the future (5 – 10 years from now), but the performance of the current suite of physics modules at those scales are not known

 Rapid development and evaluation of the next generation suite for CAM requires the ability to isolate processes as well the ability to test parameterizations across a range of scales

 Current computing capabilities do not allow global models to be run routinely at mesoscale resolutions, anticipated for use 5 – 10 years from now

  Incorporate the CAM5 parameterization suite into WRF  Use the Aerosol Modeling Testbed to evaluate the CAM5

parameterization suite   Evaluate CAM5 physics suite at higher spatial resolution

more compatible with data   Compare CAM5 physics modules against more complex

and expensive representations using systematic and consistent methodology

  Use performance metrics to identify more desirable parameterization choices for both models

  Increase communication between cloud-resolving / mesoscale and global scale modeling communities

Downscaling for MILAGRO Testbed

Community Atmosphere Model (CAM5)

module

CLM

Park & Bretherton

MOZART

Morrison & Gettleman

RRTMG

MAM

ZM deep & UW shallow

trace gas chemistry

convection

microphysics

aerosols

boundary layer

land surface

radiation

trace gas chemistry

convection

microphysics

aerosols

boundary layer

land surface

radiation

Philosophy: Several parameterizations for each atmospheric process for short-term simulations using range of grid spacings

Philosophy: Single parameterization for each atmospheric process for long-term climate simulations using a coarse grid

CAM5 “package”

downscaling with consistent physics

boundary conditions  Use field campaign data to evaluate how performance varies as a function of resolution

 Simulations with CAM5 package and individual CAM5 modules coupled with more complex treatments

Utilize Data from MILAGRO Field Campaign and Tools from Aerosol Modeling Testbed to Evaluate CAM5 Downscaling

Simulated PM2.5, excluding dust

 Run WRF at high resolution, downscaling from CAM5, for the ISDAC / ARCTAS testbed case

 Compare performance of CAM5 simulations with high-resolution simulations from WRF

Clouds from ARSCL

Clouds from Morrison Microphysics Scheme

C-130 Flight on March 10, 2006 T0 Site in Mexico City

Average PBL Depth at T1 Site observed – radar wind profilers YSU scheme from WRF UW scheme from CAM5

conc

entr

atio

n (µ

g m

-3)

Organic Matter (OM)

Primary + Biomass Burning OM

Secondary OM

Sulfate

Nitrate

mean mean downwind city

CAM5 aerosol model neglects NO3

under-prediction of downwind aerosols

from CAM5 related to its PBL scheme

SOA scheme coupled to MOSAIC aerosols

produced too much SOA

 Emissions identical for red and blue lines, so differences above due to differences in aerosol (MOSAIC vs MAM) and boundary layer (YSU vs UW) treatments

  IPCC POM surface emissions ~10% lower over Mexico, but 3.7 times lower over Mexico City and do not vary diurnally

WRF simulations at Δx ~ 5 km over Barrow

‘clean’ ‘dirty’

aerosols entrained

into clouds

conc

entr

atio

n (µ

g m

-3)

Organic Matter (OM)

Primary + Biomass Burning OM

Secondary OM

Sulfate

Nitrate CAM5 aerosol model

neglects NO3

SOA scheme in CAM5 produced too much SOA

downwind

city

transport and mixing simulated reasonably well

mean

 AMS organic aerosol data from MILAGRO very useful in evaluating SOA treatment in CAM5 compared to more sophisticated ones in WRF-Chem

CAM5 treatment too low during afternoon

CAM5 (2.5 x 1.9o) + IPCC AR5 emissions

SW ambient winds µg m-3

WRF (Δx = 12 km) + CAM5 Physics + CAM5 (IPCC AR5) emissions

CAM CAM

CAM CAM

mostly dust

no dust emissions

observed AMS data (provided by Jose Jimenez, Allison

Aiken, and Pete de Carlo)

‘default’ WRF-Chem, local emissions CAM5 Physics, local emissions CAM5 Physics, IPCC emissions

20

15

10

5

0 6a 20a 10 29 20b 22a 19a 18a 15a 7a

OM Percentiles for All Flights

G-1 C-130

observed (5/25/50/75/95 percentiles) ‘default’ WRF CAM5 Physics

Δx = 3 km domain

Mexico City

PM2.5 at 700 hPa Aerosol Optical Depth

WRF (Δx = 12 km) + CAM5 Physics + local emissions (hi-res)

extract results to compare with data

extract results to compare with data

C-130 Flight Path

 How well will MAM aerosols (MAM in CAM5, MAM in WRF-Chem, MOSAIC in WRF-Chem) compare with aircraft measurements in the Arctic

  In contrast to aerosols, the complexity of the Morrison & Gettleman microphysics scheme from CAM5 is comparable to those in WRF. But the performance as scales change needs to be quantified in addition to comparing the performance with other schemes.

red = 4 µg m-3

18 UTC March 19

2006

(city) (downwind)