Vertical subgrid-scale heterogeneity and cloud overlap parameterizations

for low-level clouds in GCMs and RCMs

Jean JOUHAUD, Jean-Louis DURESNE, Jean-Baptiste MADELEINE

LMD/IPSL, Paris, France

CFMIP workshop, 30 Sept – 4 Oct, Mykonos, Grece

Too Few, too Bright low-cloudsGCMs (RCMs?) simulate cumulus type clouds with too small cloud fraction and too high cloud reflectance.

Satellite Observations

(Konsta et al. 2016)

More frequent

0

0 0.2 0.4 0.6 0.8 1

Cloud cover

Low level, tropical clouds

Too Few, too Bright low-cloudsGCMs (RCMs?) simulate cumulus type clouds with too small cloud fraction and too high cloud reflectance.

Satellite Observations

(Konsta et al. 2016)

LMDZ5B.x

More frequent

0

0 0.2 0.4 0.6 0.8 1

Cloud cover0 0.2 0.4 0.6 0.8 1

Cloud cover

Working hypothesis: the vertical heterogeneity of cloud properties is misrepresented in models

Low level, tropical clouds Low level, tropical clouds

z

y

vertical slice of a cloud simulated by a

LES

𝑪𝑭𝒔𝒓𝒇

𝑪𝑭𝒗𝒐𝒍

Cloud vertical heterogeneity

Two cloud fractions (CF):• by volume 𝑪𝑭𝒇𝒗𝒐𝒍

• by surface 𝑪𝑭𝒔𝒓𝒇 (by area

seen from above)

𝑪𝑭𝒗𝒐𝒍: depends on microphysics𝑪𝑭𝒔𝒓𝒇: depends on 𝑪𝑭𝒗𝒐𝒍 and

on vertical geometry of clouds

𝑪𝑭𝒔𝒓𝒇 > 𝑪𝑭𝒗𝒐𝒍 but in GCM and

RCM layers, 𝑪𝑭𝒔𝒓𝒇 = 𝑪𝑭𝒗𝒐𝒍

How to correct this?

z

y

vertical slice of a cloud simulated by a

LES

𝑪𝑭𝒔𝒓𝒇

𝑪𝑭𝒗𝒐𝒍

Cloud vertical heterogeneity

Two cloud fractions (CF):• by volume 𝑪𝑭𝒇𝒗𝒐𝒍

• by surface 𝑪𝑭𝒔𝒓𝒇 (by area

seen from above)

𝑪𝑭𝒗𝒐𝒍: depends on microphysics𝑪𝑭𝒔𝒓𝒇: depends on 𝑪𝑭𝒗𝒐𝒍 and

on vertical geometry of clouds

𝑪𝑭𝒔𝒓𝒇 > 𝑪𝑭𝒗𝒐𝒍 but in GCM and

RCM layers, 𝑪𝑭𝒔𝒓𝒇 = 𝑪𝑭𝒗𝒐𝒍

How to correct this?

z

y

vertical slice of a cloud simulated by a

LES

𝑪𝑭𝒔𝒓𝒇

𝑪𝑭𝒗𝒐𝒍

GCM layerz

y

Cloud vertical heterogeneity

Need to introduce subgrid heterogeneity of clouds in GCMs

z

y

vertical slice of a cloud simulated by a

LES

𝑪𝑭𝒔𝒓𝒇

𝑪𝑭𝒗𝒐𝒍

GCMz

y

Cloud vertical heterogeneity

z

y

vertical slice of a cloud simulated by a LES

𝑪𝑭𝒔𝒓𝒇

𝑪𝑭𝒗𝒐𝒍

GCMz

ySubgrid heterogeneity and cloud overlap are intended to represent the same geometrical characteristics

Cloud vertical heterogeneity

1) Parameterization of subgrid heterogeneity

For cumulus clouds, based on LES simulations

𝐶𝐹𝑠𝑢𝑟𝑓 = 1 + 0,0044. ∆𝑧 𝐶𝐹𝑣𝑜𝑙

(Neggers et al. 2011, Jouhaud et al. 2018)

1) Parameterization of subgrid heterogeneity

For cumulus clouds, based on LES simulations

𝐶𝐹𝑠𝑢𝑟𝑓 = 1 + 0,0044. ∆𝑧 𝐶𝐹𝑣𝑜𝑙

(Neggers et al. 2011, Jouhaud et al. 2018)

For all clouds, based on satellite observations

𝐶𝐹𝑠𝑟𝑓 =1

1 + 𝑒−𝑓1

𝐶𝐹𝑣𝑜𝑙− 1

f depends on Δz, Δx and wind shear(Brooks et al. 2005) CFvol

CF

srf

For each atmopheric layer i:

Microphysic 𝐶𝐹𝑣𝑜𝑙(i)

SG heterogeneity 𝐶𝐹𝑠𝑟𝑓(i)

𝐶𝐹𝑠𝑟𝑓,𝑡𝑜𝑡?

𝐶𝐹𝑠𝑟𝑓,𝑡𝑜𝑡?

2) Parameterization of cloud overlap

For each atmopheric layer i:

Microphysic 𝐶𝐹𝑣𝑜𝑙(i)

SG heterogeneity 𝐶𝐹𝑠𝑟𝑓(i)

For total cloud layer𝐶𝐹𝑣𝑜𝑙,𝑡𝑜𝑡 = Σ 𝐶𝐹𝑠𝑟𝑓(i)

𝐶𝐹𝑠𝑟𝑓,𝑡𝑜𝑡 SG heterogeneity

1.0

𝐶𝐹𝑠𝑟𝑓,𝑡𝑜𝑡? 𝐶𝐹𝑠𝑟𝑓,𝑡𝑜𝑡

2) Parameterization of cloud overlap

For each atmopheric layer i:

Microphysic 𝐶𝐹𝑣𝑜𝑙(i)

SG heterogeneity 𝐶𝐹𝑠𝑟𝑓(i)

For total cloud layer𝐶𝐹𝑣𝑜𝑙,𝑡𝑜𝑡 = Σ 𝐶𝐹𝑠𝑟𝑓(i)

𝐶𝐹𝑠𝑟𝑓,𝑡𝑜𝑡 SG heterogeneity

1.0

𝐶𝐹𝑠𝑟𝑓,𝑡𝑜𝑡? 𝐶𝐹𝑠𝑟𝑓,𝑡𝑜𝑡

2) Parameterization of cloud overlap

Exponential random overlap approach (Hogan & Illingworth 2000) decorelation length ∆𝑧0 :

𝐶𝐹𝑡𝑜𝑡 𝑖, 𝑗 = 𝛼𝐶𝐹𝑚𝑎𝑥 + 1 − 𝛼 𝐶𝐹𝑟𝑎𝑛𝑑𝑜𝑚 with 𝛼 = 𝑒−

𝛿𝑧 𝑖,𝑗

∆𝒛𝟎

∆𝒛𝟎 ?

1.0

For each atmopheric layer i:

Microphysic 𝐶𝐹𝑣𝑜𝑙(i)

SG heterogeneity 𝐶𝐹𝑠𝑟𝑓(i)

For total cloud layer𝐶𝐹𝑣𝑜𝑙,𝑡𝑜𝑡 = Σ 𝐶𝐹𝑠𝑟𝑓(i)

𝐶𝐹𝑠𝑟𝑓,𝑡𝑜𝑡 SG heterogeneity

∆𝒛𝟎 =−Δ𝑧𝑡𝑜𝑡

ln𝐶𝐹𝑠𝑟𝑓,𝑡𝑜𝑡 − 𝐶𝐹𝑟𝑎𝑛𝑑𝐶𝐹𝑚𝑎𝑥 − 𝐶𝐹𝑟𝑎𝑛𝑑

𝐶𝐹𝑟𝑎𝑛𝑑 , 𝐶𝐹𝑚𝑎𝑥

∆𝒛𝟎 Δ𝑧𝑡𝑜𝑡

𝐶𝐹𝑠𝑟𝑓,𝑡𝑜𝑡𝐶𝐹𝑠𝑟𝑓,𝑡𝑜𝑡

2) Parameterization of cloud overlap

1.0

For each atmopheric layer i:

Microphysic 𝐶𝐹𝑣𝑜𝑙(i)

SG heterogeneity 𝐶𝐹𝑠𝑟𝑓(i)

For total cloud layer𝐶𝐹𝑣𝑜𝑙,𝑡𝑜𝑡 = Σ 𝐶𝐹𝑠𝑟𝑓(i)

𝐶𝐹𝑠𝑟𝑓,𝑡𝑜𝑡 SG heterogeneity

∆𝒛𝟎 =−Δ𝑧𝑡𝑜𝑡

ln𝐶𝐹𝑠𝑟𝑓,𝑡𝑜𝑡 − 𝐶𝐹𝑟𝑎𝑛𝑑𝐶𝐹𝑚𝑎𝑥 − 𝐶𝐹𝑟𝑎𝑛𝑑

𝐶𝐹𝑟𝑎𝑛𝑑 , 𝐶𝐹𝑚𝑎𝑥

∆𝒛𝟎 Δ𝑧𝑡𝑜𝑡

𝐶𝐹𝑠𝑟𝑓,𝑡𝑜𝑡𝐶𝐹𝑠𝑟𝑓,𝑡𝑜𝑡

2) Parameterization of cloud overlap

2 to 5 km

Single column model (SCM)

LESLMDZ5B.x oldLMDZ5B.x new

Cloud liquid water content (g/kg)

LESLMDZ5B.x oldLMDZ5B.x new

Cloud fraction(%)

𝑪𝑭𝒔𝒖𝒓𝒇 𝑪𝑭𝒗𝒐𝒍

Single column model (SCM)

LESLMDZ5B.x oldLMDZ5B.x new

Cloud liquid water content (g/kg)

LESLMDZ5B.x oldLMDZ5B.x new

Cloud fraction(%)

𝑪𝑭𝒔𝒖𝒓𝒇 𝑪𝑭𝒗𝒐𝒍

Cloud fraction increases

Cloud reflectancedecreases

LMDZ5B.x oldLMDZ5B.x new

Single column model (SCM)

LMDZ5B.xSatellites Observations

LMDZ5B.xwith sub-grid

3D model, with COSP

More frequent

0

LMDZ5B.xwith sub-grid

+ exp-ran overlap

o Cloud vertical heterogeneity strongly affects the fraction and brightness of low level clouds

o We consider sub-grid vertical heterogeneity and cloud overlap as two parameterisations of a same phenomena

o These parameterisation reduces the too few too bright bias: reflectance of clouds decreases, clouds with intermediate CF (20-50%) are more frequent

o Cloud macrophysic properties are as important as cloud microphysic

o Impact on climate sensitivity?

o Extension to high level clouds?

Conclusion