Phenomenology, Simulation and Parameterization of Atmospheric Convection Pier Siebesma Yesterday: “Dry” Atmospheric Convection Today: “Moist” Convection and (shallow cumulus) clouds Dry Convection 1 1. Motivation 2. Equations 3. Moist Thermodynamics Concepts 4. LES of shallow cumulus convection 5. Parameterization of cumulus convection 6. Geometry of Cumulus Clouds 7. (PDF cloud schemes) Dry Convection 2 Tropopause 10km Subsidence ~0.5 cm/s inversion 10 m/s Cloud base ~500m Equator 0 o Ev Trade wind Ev region North o 30 N •Deep Convective Clouds •Shallow Convective Clouds •Stratocumulus •Precipitation •Little precipitation •Vertical turbulent transport •Vertical turbulent transport •Interaction with radiation • Net latent heat production •No net latent heat production •Engine Hadley Circulation •Fuel Supply Hadley Circulation EUROCS intercomparison project on cloud representation in GCM’s in the Eastern Pacific Large Scale Models tend to overestimate Tradewind cumulus cloudiness and underestimate Stratocumulus Deep cu Shallow cu Siebesma et al. (2005, QJRMS) scu 1. Motivation 2. Equations 3. Moist Thermodynamics Concepts 4. LES of shallow cumulus convection 5. Parameterization of cumulus convection 6. Geometry of Clouds 7. (PDF cloud schemes) Dry Convection 5 Grid Averaged Equations of thermodynamic variables t qv t ql t v w v qv w v ql w z qv z ql z Large scale Large scale advection subsidence xi ui xi xi uiqv uiql turbulent transport L cp c e Qrad c e c e Net Condensation Rate Pr Introduce moist conserved variables! l L c p ql qt qv ql l t qt t v l w v qt w •Liquid water potential Temperature •Total water specific humidity l z qt z z z w l Qrad wqt Pr What happened with the clouds? Buoyancy is the primary source for the vertical velocity w t With: g 0 v v (1 0.61qv ql ) Typical numbers: = 0.5K v 1 . 61 q v q l 0 .5 K 0 ~ 3K qv= 1~5 g/kg ql= 0~3 g/kg 0 ~ 1K So we need to go back to “down to earth” variables: Cloud Scheme in LES: All or Nothing { a c 0, { a c 1, ql 0, q l ( q t q sl ), if if q t q sl 0 q t q sl 0 In Climate models we have partial cloud cover so we need a parameterization. 1. Motivation 2. Equations 3. Moist Thermodynamics Concepts 4. LES of shallow cumulus convection 5. Parameterization of cumulus convection 6. Geometry of Clouds 7. (PDF cloud schemes) Dry Convection 10 Conditional Instability •Lift a (un)saturated parcel from a sounding at z0 by dz •Check on buoyancy with respect to the sounding: B g 0 v, p profile d m Stable for unsaturated parcels Unstable for saturated parcels 5.4K/km unstable z stable Conditionally Unstable!!! v d v z m v Introductory Concepts 1: CAPE CAPE = Convective Available Potential Energy. CIN = Convection Inhibition CAPE •CAPE and CIN unique properties of moist convection CIN •Primary Reason why moist convection is so intermittant CAPE and CIN: An Analogue with Chemistry 1) Large Scale Forcing: • Horizontal Advection Activation (triggering) • Vertical Advection (subs) LS-forcing Surf Flux • Radiation CIN 2) Large Scale Forcing: CAPE slowly builds up CAPE Free 3) CAPE Energy •Consumed by moist convection RAD LS-forcing Mixed Layer LFC Parcel Height LNB • Transformed in Kinetic Energy •Heating due to latent heat release (as measured by the precipitation) •Fast Process!! Free after Brian Mapes Introductory Concepts 3: Quasi-Equilibrium (Arakawa and Schubert JAS 1974) dCAPE dt JM b F LS 0 LS-Forcing that slowly builds up slowly The convective process that stabilizes environment Quasi-equilibrium: near-balance is maintained even when F is varying with time, i.e. cloud ensemble follows the Forcing. Forfilled if : tadj << tF wu au Used convection closure (explicit or implicit) tadj JMb ~ CAPE/ tadj : hours to a day. Mb=au wu r :Amount of convective vertical motion at cloud base (in an ensemble sense) Introductory Concepts 4: Earthly Analogue Free after Dave Randall: •Think of CAPE as the length of the grass •Forcing as an irrigation system •Convective clouds as sheep •Quasi-equilibrium: Sheep eat grass and no matter how quickly it grows, the grass is allways short. •Precipitation……….. 1. Motivation 2. Equations 3. Moist Thermodynamics Concepts 4. LES of shallow cumulus convection 5. Parameterization of cumulus convection 6. Geometry of Clouds 7. (PDF cloud schemes) Dry Convection 16 History of LES of cumulus topped PBL 1. Sommeria, G. (1976) J. Atm Sci. 33, 216-241 2. Sommeria, G and Lemone, M.A (1978) J. Atm Sci. 35, 25-39 3. Beniston, M.G. and Sommeria G (1981) J. Atm Sci. 38, 780-797 4. Bougeault, Ph (1981) J. Atm Sci. 38, 2414-1438 5. Nicholls, L, Lemone, M.A. and Sommeria, G. (1982) QJRMS 108, 167-190 6. Cuijpers J,W,M and Duynkerke, P.G, (1993) J. Atm Sci. 50, 2894-3908 7. Siebesma and Cuijpers J,W,M (1995) J. Atm Sci. 52, 650-666 GCSS; LES intercomparison studies of shallow cumulus: Experiment Case year BOMEX Steady state Trade wind cu 1997 ATEX Trade wind cu topped with Scu 1998 ARM Diurnal Cycle Cumulus 2000 RICO Precipitating trade wind cu 2006 Siebesma et al. JAS 2003 Stevens et al. JAS 2001 Brown et al. QJRMS 2002 Van Zanten et al. in preperation •No observations of turbulent fluxes. •Use Large Eddy Simulation (LES) based on observations BOMEX ship array (1969) 0 t t large t LES scale observed observed To be modeled by LES Nitta and Esbensen 1974 JAS •11 different LES models •Initial profiles •Large scale forcings prescribed •6 hours of simulation Is LES capable of reproducing the steady state? •Large Scale Forcings •Mean profiles after 6 hours •Use the last 4 simulation hours for analysis of ……. •Turbulent Fluxes of the conserved variables qt and l w l w L cp wql wqt wqv wql Cloud layer looks like a enormous entrainment layer!! Convective Mass flux decreasing with height mass flux = cloud core fraction * core velocity LES: “clouds in silico” Siebesma et al JAS 2003 x x = Recently validated for “Clouds in vivo” (Zhang, Klein and Kollias 2009) clouds “in vivo” ARM mm-cloud radar Updraft mass flux = updraft fraction * updraft velocity Conditional Sampling of: •Total water qt •Liquid water potential temperature l Lateral Mixing between clouds and environment 1. Motivation 2. Equations 3. Moist Thermodynamics Concepts 4. LES of shallow cumulus convection 5. Parameterization of cumulus convection 6. Geometry of Clouds 7. (PDF cloud schemes) Dry Convection 26 Mass Flux decomposition a u u u e e e u u e a u (1 a ) e Courtesy : Martin Kohler (ECMWF) M u e w a w u (1 a ) w e a w u ( u ) sub-core flux env. flux Siebesma and Cuijpers JAS (1995) M-flux e In general: bulk approach: Cloud ensemble: approximated by 1 effective cloud: How to estimate updraft fields and mass flux? The old working horse: Betts 1974 JAS Arakawa&Schubert 1974 JAS Tiedtke 1988 MWR Gregory & Rowntree 1990 MWR Kain & Fritsch 1990 JAS And many more…….. Entraining plume model: c z 1 M M z (c ) for l , qt M 2 1 wc 2 z b wc aB, 2 B g 0 v v Plus boundary conditions at cloud base. Different tendency to form cumulus anvils is caused by differences in the vertical structure of model mass flux: , values fixed Mixing; Flexible structure M M Siebesma et al 2007 (JAS) Tiedtke (1989) in IFS EDMF-DualM Neggers et al 2009 (JAS) Standard (schizophrenic) parameterization approach: w K z w M ( u ) t z w S This unwanted situation has led to: •Double counting of processes •Problems with transitions between different regimes: dry pbl shallow cu scu shallow cu shallow cu deep cu Deterministic versus Stochastic Convection (1) •Traditionally convection parameterizations are deterministic: •Instantaneous large scale Forcing and mean state is taken as input and convective response is deterministic •One to one correspondency between sub-grid state and resolved state assumed. •Conceptually assumes that spatial average is a good proxy for the ensemble mean. ~500 km Deterministic versus Stochastic Convection (2) •However: Cloud Resolving Models (CRM’s) indicate that operational resolutions show considerable fluctuations of convective response around the ensemble mean. •This suggests that a deterministic (micro-canonical) approach might be too restricitive for most operational resolutions. pdf Mass Flux Plant and Craig 2006 JAS More Sophisticated Parameterization (2) (Plant and Craig 2007) Parameterization: •Select N cloud updrafts stochastically according to the pdf •Calculate impact for each updraft by using a cloud updraft model •Note : N is a function of the resolution •Only tested in 1D setting 1. Motivation 2. Equations 3. Moist Thermodynamics Concepts 4. LES of shallow cumulus convection 5. Parameterization of cumulus convection 6. Geometry of Clouds 7. (PDF cloud schemes) Dry Convection 35 Is this a Cloud?? ….and, how to answer this question? Fractal Geometry “Shapes, which are not fractal, are the exception. I love Euclidean geometry, but it is quite clear that it does not give a reasonable presentation of the world. Mountains are not cones, clouds are not spheres, trees are not cylinders, neither does lightning travel in a straight line. Almost everything around us is nonEuclidean”. Benoit Mandelbrot Instead of Area-Perimeter analyses of cloud patterns (1) Procedure: •Measure the projected cloud area Ap and the perimeter Lp of each cloud •Define a linear size through l Ap • Perimeter dimension define through: For “ordinary” Euclidean objects: log L p Slope: Dp= 1 log l Lp l Dp Area-Perimeter analyses of cloud patterns (2) •Pioneered by Lovejoy (Science 1982) •Area-perimeter analyses of projected cloud patterns using satellite and radar data •Suggest a perimeter dimension Dp=4/3 of projected clouds!!!!! •Confirmed in many other studies since then… Instead of Consequences: •Cloud perimeter is fractal and hence self-similar in a non-trivial way. •Makes it possible to ascribe a (quantitative) number that characterizes the structure •Provides a critical test for the realism of the geometrical shape of the LES simulated clouds!!!! Slope 4/3 Similar analysis with LES clouds •Measure Surface As and linear size l V 1/ 3 of each cloud •Plot in a log-log plot • S (l ) l D s •Assuming isotropy, observations would suggest Ds=Dp+1=7/3 Siebesma and Jonker Phys. Rev Letters (2000) Result of one cloud field Repeat over 6000 clouds Some Direct Consequences Surface area can be written as a function of resolution (measuring stick) l : l S (l ) S L L 2 Ds , l L Ds 7 / 3 With L=outer scale (i.e. diameter of the cloud) and measured with L. SL L 2 the normalizing area if •Euclidian area SL underestimates true cloud surface area S(l=) by a factor L 2 Ds 100 •LES model resolution of l=50m underestimates cloud surface area still by a factor 5!!! •Does this have consequences for the mixing between clouds and the environment??? Resolution dependence l0 for transport over cloud boundary (1) Transport = Contact area x Flux T (l0 ) S (l0 ) F (l0 ) turbulence S (l0 ) K (l0 ) diffusive flux Subgrid diffusion resolved advection S ( l0 ) K (l0 ) c x c x Consequences for transport over cloud boundary (2) T (l0 ) S (l0 ) F (l0 ) l0 S (l0 ) S L L 2 Ds S (l0 ) K (l0 ) c l0 l0 K ( l 0 ) l 0 u ( l 0 ) l 0 u ( L ) L (Richardson Law) l0 T ( l0 ) c u ( L ) S ( L ) L 7 / 3 Ds !!!!! No resolution dependancy for Ds=7/3!! 1/ 3 Is this shear luck ???? Not really: Repeat the previous arguments for l0 T ( ) c u ( L ) S ( L ) L 7 / 3 Ds Boundary flux T only Reynolds independent if Re 3 /4 7 /3 - D s Ds 7 / 3 which completes a heuristic “proof” why clouds are fractal with a surface dimension of 7/3. Gradient Percolation Dp=4/3 A stronger underlying mechanism ? (Peters et al JAS 2009) Dry Convection 47 high Conceptual models Statistical Mechanics Self-Organised Criticality Level of Interface & Mixed layer “understanding” microphysical models Models/Parameterizations models or conceptualisation Direct Large Global Numerical Eddy Climate Simulation Simulations Simulations low Laboratory experiments Atmospheric Field Profiling campaigns stations High Satellite Simulations Observations data Low resolution Scale Hierarchy