Towards a McICA representation of cloud-radiation interactions in the ECMWF model

ECMWFRadiation and Clouds: Towards McICA? 20060608 1

Towards a McICA representation of cloud-radiation interactions

in the ECMWF model

Radiation: J.-J. Morcrette

Cloud processes: Adrian Tompkins

ECMWFRadiation and Clouds: Towards McICA? 20060608 2

McICA

In long seasonal runs and high-resolution 10-day forecastsHow do the model survive noise in radiative heating rate?How do the model survive noise in layer cloud fraction?

Tests with 31x10-day FC at TL319L60 from 20010401 to 20010501Tests with 4-month simulations at TL95 L60 for same period

control (control)random perturbation within Gaussian distribution (the relevant

quantity x -> x (1+*ran)=2 CF (1-CF) applied on x = CF (random1)=1.5 CF |HRtot| applied on x = HR (random2)=2 CF sqrt (HRLW

2+HRSW2) applied on x = HR (random3)

ECMWFRadiation and Clouds: Towards McICA? 20060608 3

McICA: How does the model survive radiative noise?

NH SH

Tropics India

Arabia E. Asia

Anomaly correlation Z 1000 hPa

NH SH

Tropics India

Arabia E.Asia

Anomaly correlation Z 500 hPa

TL319 L60 31 x 10-day FCs

ECMWFRadiation and Clouds: Towards McICA? 20060608 4

McICA: How does the model survive radiative noise?

NH SH

Tropics Tropics

NH SH

Arabia E. Asia Arabia E. Asia

IndiaIndia

Mean error T 850 hPa Mean error T 200 hPa

TL319 L60 31 x 10-day FCs

ECMWFRadiation and Clouds: Towards McICA? 20060608 5

McICA: Hoes does the model deal with radiative noise?

Systematic perturbation:

Re +1 mDe +10 m

TL95 L60 starting24-hour apart from20010401 to 2001030

Results averaged over JJA

Ref=Control Systematic perturbation

Difference Perturb-Control Student t-test

ECMWFRadiation and Clouds: Towards McICA? 20060608 6

McICA: Hoes does the model deal with radiative noise?

Systematic perturbation:

Re +0.1 mDe +1 m

Randomperturbation:

random3

Difference Perturb.-Control

Difference Random-Control

t-test

t-test

ECMWFRadiation and Clouds: Towards McICA? 20060608 7

McICA: How does the model survive radiative noise?

For each variable, first column is differenceSecond is area with difference significant at > 95% levelThird is area with difference significant at > 97.5 % level

No particular problem in either forecast or long run modeThe McICA approach can then be used (Pincus et al., 2004, JGR)

OLR ASW STR SSRControl 247.1 0.950 0.975 226.5 0.950 0.975 -54.9 0.950 0.975 140.1 0.950 0.975Re1/De10 1.3 17.1 11.1 3.0 24.0 18.4 -0.4 11.6 7.0 3.2 36.8 28.3Re0.1/De1 0.1 4.7 2.3 0.5 3.3 1.7 -0.1 6.2 3.6 0.5 9.2 4.3Random3 0.0 4.7 2.1 0.1 2.7 1.4 0.0 4.7 2.4 0.1 7.4 3.3

ECMWFRadiation and Clouds: Towards McICA? 20060608 8

What is McICA?

Monte-Carlo Independent Column Approximation

The CKD approach for 1-D PPH columns is

The ICA approach for domain averages is (ICA: Independent Column Approx.)

Combining (1) and (2) gives

Assuming clear- and cloudy-sky columns of gas, and if there are Nc cloudy

columns, (3) can be written as

K

k

knkn Fc1

,F Correlated-k distributed absorption coefficientsas in RRTM

N

n

nN 1

F1

F

(1)

(2)

N

n

K

k

knkFcN 1 1

,1

F (3)

Nc

n

K

k

cldknk

NcN

n

K

k

clrknk FcFc

N 1 1,

1 1,

1F

ECMWFRadiation and Clouds: Towards McICA? 20060608 9

What is McICA?

Which can be simplified to

The hypothesis is that can be given by

In which case, it follows (see Barker’s May 2002 presentation) that

cldcclrc

Nc

n

K

k

cldknk

cc

K

k

clrknkc

AA

FcN

AFcA

FF)1(

1)1(F

1 1,

1,

cldF

K

k

cldkNcrdmkFc

1},,...,1{

K

k

Nc

n

cldknk

c

cldKNcKKNc

cldKKK

cldNcNc

cldTcld

FcN

FcfFcfFcfFcfT

E

1 1,

,,,1,11,11,1,111,1

1

......1

limF

The model is unbiased in the ICA sense, so for T=K * Nc large enough, an unbiased value can be obtained using a different random cloud profile for each k-coefficient

ECMWFRadiation and Clouds: Towards McICA? 20060608 10

McICA: Tests with 1-D radiation code: LW

North Slope of Alaska

OLR SDLW Differences McICA-Ref

South Great Plains

ECMWFRadiation and Clouds: Towards McICA? 20060608 11

McICA: Tests with 1-D radiation code: LW

Trop. West Pacific: Manus

Trop. West Pacific: Nauru

OLR SDLW Differences McICA-Ref

ECMWFRadiation and Clouds: Towards McICA? 20060608 12

It does work in the LW : not yet in SW!

ARM-TWP NauruOLR

SDLW

Box100

McICA

ARM-SGPOLR

SDLW

Box100

McICA

ECMWFRadiation and Clouds: Towards McICA? 20060608 13

ECMWF Plans: Statistical Scheme

These explicitly specify the probability density function (PDF) for the total water qt (and sometimes also temperature)

sq

ttstc dqqPDFqqq )()(

qt

x

q

sq

qt

PD

F(q

t)

qs

Cloud cover is integral under

supersaturated part of PDF

sq

tt dqqPDFC )(Assumes no

supersaturation

LOTS OF ISSUES FOR IMPLEMENTATION: contact Adrian for his thoughts!!!

ECMWFRadiation and Clouds: Towards McICA? 20060608 14

Can use PDF information consistently in other schemes: Radiation, microphysics…

Example of use (with Rob Pincus): Use “cloud generator” to split cloudy column into many subcolumns to investigate effect of subgrid variability on ECMWF microphysics

ECMWFRadiation and Clouds: Towards McICA? 20060608 15

What do we expect?

qliq

Warm rain autoconversion

dqliq

dtSundqvist

Range of values

If subcloud variability is

ignored

Taking variability

into account

Lower autoconversion if subgrid variability neglected hence expect higher mean cloud thickness

ECMWFRadiation and Clouds: Towards McICA? 20060608 16

Instead: Sensitivity opposite to expected effect. Dominated by ice microphysics (q0.16 ice to snow) and accretion terms – i.e. Complex, esp. with multiphase microphysics