1 of 16

1

Radiative Effect of Clouds

on Tropospheric Chemistry in a

Global 3-D Chemical Transport Model

Hongyu Liu

(http://research.nianet.org/~hyl)

National Institute of Aerospace (NIA)

@ NASA Langley

2^nd GEOS-CHEM Users’ Meeting

April 4-6, 2005

Acknowledgements -

LaRC: Jim Crawford, Brad Pierce, Gao Chen

GSFC: Peter Norris, Steve Platnick

Harvard: Jennifer Logan, Daniel Jacob, Bob Yantosca

2 of 16

Outline

2

Research Objectives
GEOS-CHEM (Fast-J)
GEOS-3 Cloud Distributions and Evaluation
Radiative Effect of Clouds

Photolysis Rates
Key Oxidants

Sensitivity to Cloud Vertical Distributions

GEOS1-STRAT 🡪 GEOS-3 🡪 GEOS-4

Conclusions

3 of 16

Research Objectives

3

To improve our understanding of the radiative effect

of clouds on global trop chem

To quantify the radiative effect of clouds on

photolysis rates and key oxidants, as well as

associated uncertainties

To assess the impact of cloud overlap assumptions

4 of 16

4

GEOS-CHEM Global Chemical Transport Model v5.5�(http://www-as.harvard.edu/chemistry/trop/geos)

Driven by the NASA Goddard Earth Observing System (GEOS-3) assimilated meteorology from the Global Modeling and Assimilation Office (GMAO)
Horizontal resolution 1^ox1^o to 4^ox5^o, 48 levels in vertical
Ozone-NO_x-CO-VOC coupled to aerosol (sulfate-nitrate-ammonium and carbonaceous) chemistry [Bey et al., 2001; Park et al., 2004]
Extensively evaluated with surf / in situ / remote sens obs
Sensitivity simulation period: Aug 2000 – Dec 2001
Photolysis rate calculation: Fast-J [Wild et al., 2000] with GEOS-3 surface albedo, 3-D cloud optical depth, and cloud fraction

5 of 16

Global Distribution of Cloud Optical Depth �GEOS-3 vs. Satellite Retrievals

5

Mean Cloud Optical Depth (grid-scale) for March 2001

MODIS

(MOD08_M3)

ISCCP

(D2)

GEOS-3

MODIS

ISCCP

GEOS-3

Probability Distribution Functions

Global Average

GEOS3-OD / MODIS-OD = 0.91

GEOS3-OD / ISCCP-OD = 1.31

6 of 16

Latitudinal Distribution of Cloud Optical Depth �GEOS-3 vs. Satellite Retrievals

6

GEOS-3

MODIS

ISCCP

GEOS-3 cloud OD reasonably agree with MODIS and ISCCP cloud retrieval products, but tend to be larger in the tropics and SH marine stratiform clouds region.

7 of 16

Model Representations of the Vertical Coherence of Clouds

7

LIN: Linear Assumption [Wild et al., 2000]

τ_c' = τ_c · f

RAN: Approximate Random Overlap [Briegleb, 1992]

τ_c' = τ_c · f ^3/2

MRAN: Maximum-Random Overlap

clouds in adjacent layers (a cloud block) are maximally

overlapped; cloud blocks are randomly overlapped.

version 1: Stubenrauch et al. [1997];Tie et al. [2003]
version 2: Collins [2001]; Feng et al. [2004]

grid-scale OD

in-cloud OD

cloud fraction

When looking at the radiative effect of clouds in the model, an important issue is how to represent the coherence or overlap of cloud in the vertical.

The simplest assumption made in current CTMs is the uniform cloud distribution method or LINEAR ASSUMPTION. Grid-scale OD = In-cloud OD * CF. This method assumes that the actinic flux is linearly proportional to the COD, which may introduce a bias. The exact overlap scheme assumes that all layers of clouds are randomly overlapped. It requires a lot of computation time. The Approximate Random Overlap was developed by Briegleb [1992] as a good approximation of the exact random overlap method. The other commonly used method is Maximum-Random Overlap. In this method, clouds in adjacent layers (a cloud block) are maximally overlapped, and cloud blocks are randomly overlapped. A vertical profile of partial cloudiness is converted into a series of configurations. Fast-J is applied to each configuration. Photolysis frequencies are then weighted by the probability of cloud presence in each configuration. Tie et al. [2003] used version 1 in their study of rad effect of clouds on trop chemistry. Feng et al. [2004] used the other version to study the effects of cloud overlap in photochemical models. The effect of diff between ver 1 and 2 seems small. Following Feng et al. [2004], we call these methods as LIN, RAN, and MRAN, respectively.

8 of 16

Effect of clouds on J[O¹D] calculated by off-line Fast-J�(Test case of Feng et al. [2004])

8

Cloud layers at

3-4km: f = 0.2, 0.9

2-3km: f = 0.1, 0.8

Latitude = 45N

Albedo=0.1

Mean OD = 54

Enhancement above cloud

Reduction below cloud

a). Small cloud fraction:

LIN > RAN > MRAN

b). Large cloud fraction:

Small differences between schemes

SZA=0

CLEAR

RAN

MRAN

LIN

a). small cloud fraction

b). large cloud fraction

9 of 16

Changes (%) in daily mean J[O¹D] due to cloud�June 2001

9

RAN and MRAN differs by ~2%

10 of 16

Changes (%) in daily mean J[NO₂] due to cloud�June 2001

10

RAN and MRAN differs by ~2%

11 of 16

Changes (%) in daily mean OH due to cloud�June 2001

11

12 of 16

O₃ and CO changes (%) due to cloud, June �(GEOS-CHEM vs. MOZART-2)

12

GEOS-CHEM

MRAN

[this work]

MOZART-2

MRAN

[Tie et al., 2003]

O₃

CO

hPa

Why are the sensitivities to cloud so different ?

13 of 16

13

GEOS-CHEM [this work] MOZART-2 [Tie et al.,2003]

LIN RAN MRAN LIN MRAN

OH 0.99 0.13 -0.52 88.09 20.31

O₃ 4.89 3.15 3.65 12.07 8.55

NO_x 5.58 3.46 3.26 -4.17 -3.13

HO₂ -2.27 -1.60 -1.47 16.52 5.89

CH₂O 5.55 3.85 4.77 -14.56 -5.78

CO 0.81 1.33 2.26 -31.40 -9.01

J[O¹D] -3.30 -2.15 -3.44 44.98 13.38

J[NO₂] -4.38 -3.23 -3.76 62.24 13.84

J[CH₂O] -2.30 -1.74 -2.38 54.56 13.75

Global (troposphere) mean changes (%) due to cloud, June

14 of 16

GEOS1-STRAT, GEOS-3 and GEOS-4 Cloud Optical Depths per KM (June)

14

GEOS1-S (1996)

GEOS-3 (2001)

GEOS-4 ( 2001)

ISCCP

MODIS

GEOS-3

GEOS-4

GEOS1-S

Global Average:

GEOS-3 / GEOS1-S = 5.1

GEOS-3 / GEOS-4 = 1.9

GEOS1-S COD: too small

optically too thin

15 of 16

Changes (%) in daily mean OH due to cloud (LIN, June)

15

changes of global mean OH due to cloud

GEOS1-S: -1% GEOS-3: 1% GEOS-4: 14%

GEOS1-S, 1996

GEOS-3, 2001

GEOS-4, 2001

Cloud vertical distribution is more important than the magnitude of COD in terms of the radiative impact on global tropospheric chemistry!

16 of 16

Conclusions

16

The dominant radiative effect of clouds is to influence the vertical redistribution of the intensity of photochemical activity while the global average effect remains modest. This contrasts with previous studies.

For global CTMs, the approximate random overlap scheme

(τ_c' = τ_c · f ^3/2)is a good approximation of the maximum-random overlap scheme; the former is computationally much cheaper.

Using the maximum-random overlap scheme or the random overlap scheme (vs. linear assumption) reduces the impact of clouds on photochemistry, but global average effect remains modest.

Cloud vertical distribution is more important than the magnitude of cloud optical depth in terms of the radiative impact on global tropospheric chemistry.