an atmospheric “mesoscale”:where convection meets waves

(rotation optional)

Brian MapesUniversity of Miami

for oceanographers• In a moist convecting atmosphere, small scale

vertical motions don’t just carry fluxes, they cause latent heating

• OK, you can view it as a vertical flux of water substance upward past the condensation level.

• Spectral space: energy injection across scales • Physical space: feedback small updrafts

• UV catastrophe of conditional instability (Lilly 1961)• Smallest updrafts, broadest subsidence (Bjerknes 1938)

mesoscale convection• “Mesoscale” convection events (meso =

middle, in between L~H “convective” scale and N/f H “Rossby radius”) are less theoretically tidy than parcel or exp(ikx) UV catastrophe, but profound & real

• convectively coupled internal waves

Convectively coupled gravity waves in 2D CRMNo preferred hor. scale

Stefan Tulich (Mapes et al. 2009 JMSJ)

3 decades

5 decades

}

Scale Interactionscascade... ...pas?

mesoscale convection• Are these things coherent aspects (the spectral

tail) of the large-scale flow, or an emergent metaphenomenon bubbling up from convection?

• Implication: is it better to spend computing DOFs to resolve the mesoscale? Or rather on little hi-res but periodic “sample” patches of convective-scale flow, coupled across an enforced scale separation? – (MMF or “super-parameterization”)

3D global simulations

• GEOS-5 global AGCM at 5km mesh size» by Bill Putman, Max Suarez, others at NASA GSFC

• 20-day run analyzed here• Cubed sphere grid, nonhydrostatic• GCM physics left on – mostly

• subgridscale plumes hobbled by entrainment• disabled subgrid orographic gravity wave drag

comparison to satellite imageryhttp://earthobservatory.nasa.gov/

IOTD/view.php?id=44246&src=eoa-iotd

predicted cloud features for February 6, 20102 weeks into simulation

5km GCM: detailed examinations• 1. Tropical mesoscale rain events: case studies

– One scale selected in to analysis: 250km events• Rebin rainrate to 2.5deg, find 10 largest maxima

– in ~20 day simulation period (Jan-Feb 2010)– in 15N-15S, to minimize cyclone dominated cases

• Extract space-time cubes around these events– (+/-18h, +/- 3 degrees) – 10 wettest cases, plus composite mean case

• 2. Vertical flux [wq], partitioned by scale» through simple coarse-graining (rebinning)

http://www.rsmas.miami.edu/personal/ssong/research/HR_250kmevents.htm

animation

Tropical cyclone: 1 case in top 10 (in 15N-15S belt, 250km scale)

http://www.rsmas.miami.edu/personal/ssong/research/HR_250kmevents.htm

Anim: composite of 10 cases

• m

composite basis

HOURS RELATIVE TO MAX 250km RAIN

99% is from resolved condensation process: good

Tropical radar observations (EPIC 2001)Time scale is hours even for small space scales

Mesoscale is real (if broadband)

cell: <1h

MCS: 10h8km radius

96km radius

Mapes and Lin 2006 MWR

• m

HOURS RELATIVE TO MAX 250km RAIN

T’(t,p): 250km area mean

leading nose

p (hPa)

HOURS RELATIVE TO MAX 250km RAIN

250km water vapor mixing ratio (t,p)

W

Low-level “valve” on convection

RH(t,p)

HOURS RELATIVE TO MAX 250km RAIN

p (hPa)

W

W

trimodal: shallow, medium, deep similar to obs (if a bit off in exact heights)

Mapes et al. 2006 DAO

GCRM detailed examinations• 1. Tropical mesoscale rain events: case studies

– One scale built in to analysis: 250km– Rebin rainrate to 2.5deg, find 10 largest maxima

• in ~20 day simulation period (Jan-Feb 2010)• in 15N-15S, to minimize cyclone dominated cases

– Extract space-time cubes around these • (+/-18h, +/- 3 degrees) • 10 cases, and composite mean case

• 2. Vertical enthalpy flux, partitioned by scale» through simple coarse-graining (rebinning)

2. Enthalpy flux• Enthalpy = sensible heat + latent heat

– CpT + Lqv

• Flux thru 500mb level balances ~23 Wm-2 radiative cooling above that level

– sensible heat flux Cp [wT] ~ 7 Wm-2– latent heat flux L [wq]: ~ 16 Wm-2

• destined to condense up there

Latent flux across 500mbsnapshot

by scales resolved in 80km rebinning

sub-80km = total flux minus the above

Latent fluxsnapshot

by scales resolved in 250km rebinning

sub-250km = total explicit flux minus above

Latent fluxsnapshot

by scales resolved in 500 km rebinning

sub-500km = total explicit flux minus above

sub-80km and super-80km scales conspire to carry flux: convection occurs in mesoscale clusters

Flux partitoned by scales• Vapor flux by convective (5-80km) scales is

colocated with flux in >80km scale mesoscale updrafts.

• Small scales mainly just add a bit (10 - 40%) to the flux by mesoscale mean updrafts

• Might this be true at still-finer scales? • Borrowed slides (with permission, and email

discussion last 2 days) from Chin-Hoh Moeng (NCAR)

• Marat Khairoutdinov (Stony Brook) ran “Giga-LES”

• Moeng et al. 2009, 2010 JAMES

Split the LES flow into: “resolvable” grid-scale (GS) & “unresolved” scale (SGS)

)()(~)( xfxfxf SGS is the difference.

apply “smoothing”

CRM resolvableGiga-LES

Moeng et al. 2010 JAMES

Apply “smoothing” with a width of 4 km

GS: CRM-grid scales

GSSGS(w-var) SGS: CRM-SGS

SGS (wq-cov)

GS

GS

SGS(q-var)most of w-kinetic energy in SGS

~ half of moisture flux in SGS

large scales small scalesMoeng et al. 2010 JAMES

• SGS flux

Moeng et al. 2010- JAMES

Moeng et al. 2009 JAMES

• SGS flux is in clouds

• condensedwater path (vertical integral)

Flux partitoned by scales• Vapor flux by convective (5-80km) scales is

colocated with flux in >80km scale mesoscale cloud system updrafts.

• Small scales mainly just add a bit (up to 40%) to the flux by mesoscale mean updrafts

• Vapor flux by sub-convective (0.1-4km) scales is colocated with >4km scale convective cloud updrafts.

• Small scales mainly just add a bit (~40%) to the flux by convective mean updrafts

Flux part summary

• Mesoscale updrafts are moist, fluxing q up• Convective updrafts are inside, adding to it• Sub-drafts inside the convective drafts: ditto• Q: How might poorly-resolved convection be

distorted by having to carry the flux of missing sub-scales? (and can param’z’n fix it?)

• Q2: Is subgrid param’z’n a flux amplifier? Is that safe numerically?

Summary• Deep convection – gravity wave interactions are

common: a “mesoscale”• Broadband (meso synoptic, in tropics)

– -5/3, but NOT a swirls-advecting-vorticity cascade– has a velocity scale, not a length scale– multicellular: hours, not minutes (not just H/w)

• “Mesoscale convection”, convective cells, and sub-cellular drafts all conspire to carry geophysically (radiatively) demanded vertical energy flux– Do we need to resolve them all? Or might truncation +

parameterization suffice?