Download presentation
Presentation is loading. Please wait.
1
Page 1© Crown copyright 2005 Stratocumulus Adrian Lock
2
Page 2© Crown copyright 2005 Overview Stratocumulus climatology and recent changes in its representation in the Met Office climate model Summary of ‘important’ aspects of the boundary layer parametrization Other issues Cloud scheme complexity Resolution GCM dynamics and grid staggering
3
Page 3© Crown copyright 2005 Inversion Cloud base Tropopause Sub-tropics tropics Inversion Subsidence F q,T Entrainment Increasing SST Borrowed from Pier Siebesma
4
Page 4© Crown copyright 2005 Stratocumulus climatology Extensive cloud cover in the subtropics Significant SW radiative impact (cooling the planet) Errors lead to SST errors in coupled climate models surface temperature errors over land Latest IPCC round includes an almost completely new Met Office model, HadGEM1
5
Page 5© Crown copyright 2005 Climate model JJA total cloud amount HadGAM1(Released 2004) HadGAM1 - HadAM3(Released 1996) HadGAM1 - ISCCPISCCP
6
Page 6© Crown copyright 2005 Climate model improvement in low level cloud amount ISSCP HADAM3 (~1996) HadGAM1(~2004) Thick (scale 0-20%) Intermediate (scale 0-40%) HadAM3 had excessive thick low cloud and too little thin HadGAM1 very much better
7
Page 7© Crown copyright 2005 GCM cross-section intercomparison (Siebesma et al 2004 and now GCSS) GCMs, data from JJA 1998 Cross-section from California to central Pacific General underestimate of stratocumulus? HadGAM1 not too bad HadAM3 wouldn’t look too bad either HadGAM1 x x x x x x x x California Central Pacific HadGAM1 total cloud cover California SSMI (all clouds) HadGAM1 (low clouds)
8
Page 8© Crown copyright 2005 Met Office GCM cloud fractions EUROCS cross-section – 1998 JJA mean Layer cloud fraction Convective cloud fraction California ITCZ HadGAM1 (Released 2004) HadAM3 (Released 1996) ITCZ California HadGAM1 has more realistic boundary layer structure
9
Page 9© Crown copyright 2005 GPCI: SW radiation More spread between models in radiation fluxes than in the clouds themselves HadGAM1 has too much cloud in the Sc-Cu transition region HadGAM1
10
Page 10© Crown copyright 2005 LWP Diurnal cycle validation against satellite Wood et al (2002) analysed mean LWP diurnal cycle from satellite microwave measurements ( ) Lock (2004) compared with HadGAM1 for the data from the EUROCS intercomparison (JJA in NE Pacific) HadGAM1: JJA mean LWP (kgm -2 ) Wood et al area x x x x x X EUROCS points LWP kgm -2
11
Page 11© Crown copyright 2005 Validation against EPIC profiles Comparison of HadGAM1 October mean profiles with mean profiles from EPIC for 16-21 October 2001 at (85W, 20S) But HadGAM1 monthly mean LWP ~ 70gm -2 compared to the observed 150gm -2 HadGAM EPIC
12
Page 12© Crown copyright 2005 Why the improvement from HadCM3 to HadGEM1? Virtually everything is changed! Doubled resolution (vertical and horizontal) New dynamics New Physics: Microphysics Cloud scheme (Smith + parametrized RHcrit) Convection scheme (revised closure, triggering, CMT) Radiation ~ the same?! Plus new boundary layer scheme…
13
Page 13© Crown copyright 2005 The new boundary layer parametrization in HadGEM1 Old scheme (HadAM3) Local Ri-dependent scheme Kept for stable boundary layers in HadGEM1 Non-local parametrization for unstable boundary layers Specified (robust) profiles of turbulent diffusivities, for turbulence driven from the surface and/or cloud-top Explicit BL top entrainment parametrization Mass-flux convection scheme trigger based on boundary layer structure rather than local instability BL mixes in subcloud layer, mass-flux in cloud layer PBL scheme ~ a mixed layer model on a finite-difference grid
14
Page 14© Crown copyright 2005 Mixed layer framework - motivation Marine stratocumulus is often an equilibrium state arising from complicated interactions between many processes that are parametrized in GCMs Need to replicate this balance within GCM parametrizations, often using schemes that currently operate independently Simple conceptual framework: mixed layer model turbulent mixing generally ensures that variables conserved under moist adiabatic ascent are close to uniform in the vertical. For example:
15
Page 15© Crown copyright 2005 Observed profiles from stratocumulus Price (QJ, 1999) Stevens et al (QJ,2003)
16
Page 16© Crown copyright 2005 Mixed layer model Integrating over the boundary layer and assuming a discontinuous inversion gives the mixed layer model equations: where is the jump across the inversion and is the entrainment rate. Mixed layer model is simple but effective highlights the importance of entrainment
17
Page 17© Crown copyright 2005 Entrainment efficiency (Stevens, 2002) Cloud-top radiative cooling both cools the ML but also drives entrainment and so warms (and dries) it Net effect of radiative cooling on ML evolution is therefore a strong function of the entrainment efficiency Eg. for ‘minimal’ model: Uncertain – parametrizations in the literature give widely different entrainment efficiencies (Stevens, 2002):
18
Page 18© Crown copyright 2005 Entrainment efficiencies (Stevens, 2002) Differences in variability of entrainment efficiency as well as its magnitude AL is noticeably weaker 0.4 0.7 0.9 0.7 0.8 0.5
19
Page 19© Crown copyright 2005 LES Cloud free: = surface heated Smoke clouds: = radiatively cooled = surf heat + rad cool Water clouds: X = rad cool (no b.r.) + = rad cool + buoy rev = buoyancy reversal only Observations = Nicholls & Leighton,1986; Price, 1999; Stevens et al 2003 Verification of entrainment parametrization against LES and observations Parametrized entrainment rate (m/s) Actual entrainment rate (m/s)
20
Page 20© Crown copyright 2005 Entrainment parametrization in GCMs Hard to parametrize on (coarse) GCM grids using traditional down-gradient diffusion because gradients at inversions are large and K(Ri): local stability dependence is not relevant K(z/z i ): very sensitive to the definition of z i K(TKE): accurate TKE evolution hard to resolve Explicit parametrization requires a formulation for w e Uncertain But you know what you are (or should be) getting
21
Page 21© Crown copyright 2005 Numerical handling of inversions All model processes (turbulence, radiation, LS advection) should be coupled to preserve mixed layer budgets i.e., no spurious numerical transport across inversion (Stevens et al 1999, Lenderink and Holtslag 2000, Lock 2001, Grenier and Bretherton 2001, Chlond et al 2004) Explicit BL top entrainment parametrization with flux-coupling A subgrid inversion diagnosis takes tendencies from subsidence, radiation and turbulence and realistically distributes them between the mixed layer and the inversion grid-level – ‘inversion flux-coupling’ x x x x x Entrainment Subsidence Mixed layer GCM inversion grid-level (cools/warms) Real inversion (rises/falls) l profile (idealised mixed layer) Free atmosphere
22
Page 22© Crown copyright 2005 Diagnosis of subgrid mixed-layer model profile X X X X X Inversion grid-level GCM cell-averages Subgrid profile Parcel ascent vl = l ( 1+r m q t ) Flux grid-levels Identify ‘inversion grid-level’ Extrapolate GCM profiles into the inversion grid-level Assume a discontinous subgrid inversion Calculate subgrid inversion height and strength
23
Page 23© Crown copyright 2005 Example calculation of grid-level entrainment fluxes (ignoring subsidence) Radiative flux X X = GCM fluxes Heat Fluxes X X X X X X Inversion grid-level Flux grid-levels = mixed layer model fluxes Subsidence fluxes across the inversion must be coupled similarly X X X X
24
Page 24© Crown copyright 2005 Impact of inversion treatment SCM: Time-height contour plots of liquid water from simulations with subsidence > entrainment but ML budget (so cloud depth) stationary With inversion flux-coupling Without inversion flux-coupling GCM: JJA mean liquid water path With inversion flux-coupling: Impact of removing inversion flux-coupling: Inversion height Errors from not coupling fluxes amount to spurious additional entrainment
25
Page 25© Crown copyright 2005 Equilibrium mixed layers – Stevens 2002 Minimal model with entrainment efficiency of 0.5 For this fixed entrainment efficiency, heading towards California (increasing divergence, reducing SST): Strong response: z i decreases, wT surf increases (contours) Weaker response: LWP decreases, wq surf decreases (grey-scales) Minimal model not too far from reality Towards California
26
Page 26© Crown copyright 2005 Mixed layer model sensitivity to entrainment efficiency Minimal model with entrainment efficiency of 0.5 Efficiency of 1: lower LWP, much reduced wT surf Towards California
27
Page 27© Crown copyright 2005 GCM cloud sensitivity to entrainment Std Met Office 2 x w e No flux coupling Test GCM sensitivity to cloud-top entrainment More active entrainment (either explicit or numerical) gives an equilibrium state with smaller surface heat fluxes and less stratocumulus (as in Stevens, 2002) Removing inversion flux-coupling is more severe than doubling w e
28
Page 28© Crown copyright 2005 Surface heat flux sensitivity to entrainment Standard Met Office 2 x w e No flux coupling Expect spurious numerical entrainment to be so increases as you approach the coast Implies wT surf would reduce towards the coast As it does if flux-coupling removed in Met Office GCM As it does in other GCMs from EUROCS Spurious or deliberate increase in parametrized entrainment efficiency?
29
Page 29© Crown copyright 2005 Entrainment efficiencies (Stevens, 2002) None of these proposed parametrizations show entrainment efficiency increasing ‘towards California’ 0.4 0.7 0.9 0.7 0.8 0.5 0.81.0 0.8
30
Page 30© Crown copyright 2005 What about precipitation? But AL is still a ‘weak’ entrainment parametrization Is this being compensated for in HadGAM1 by a spuriously high drizzle rate? Requires a drizzle climatology – satellite borne radar? HadGAM1 No obs!
31
Page 31© Crown copyright 2005 Other issues Cloud scheme Prognostic vs diagnostic – does it matter? Vertical resolution Decoupling – structure within ‘mixed’ layers New Dynamics
32
Page 32© Crown copyright 2005 Other issues – cloud scheme complexity Cloud-top height dependence on cloud scheme: HadGAM1 (Smith) PC2 (prognostic q l and C F )
33
Page 33© Crown copyright 2005 Other issues – partial cloud cover Near Hawaii C F ~0.1 Convection scheme Somewhere in between! C F ~0.5 ? Near LA C F ~1 BL scheme
34
Page 34© Crown copyright 2005 Other issues – cloud scheme number of events (%) all clouds low clouds (>700hPa) all clouds low clouds (>700hPa) AM2p12b ARPEGE CAM 3.0 GSM 0412 HadGAM RACMO2 all clouds low clouds (>700hPa) Histograms of Cloud Cover JJA 1998 JJA 2003 Cloud cover PDFs from GPCI Medium cloud cover Relatively low occurrence in HadGAM1 Mixed layer model would break down Possible cause of excessive cloud problems in Sc-Cu transition? HadGAM1: Latitude 0 Low cloud cover 1
35
Page 35© Crown copyright 2005 Other issues Cloud scheme Vertical resolution How much is sufficient? Currently 38 levels ~280m at 1km ~ cloud thickness Decoupling – structure within ‘mixed’ layers New Dynamics
36
Page 36© Crown copyright 2005 Namibian Stratocumulus region 31 st March 2004 Model Level Lock (2001) Striping in cloud fields not observed in albedo from GERB instrument (not shown)
37
Page 37© Crown copyright 2005 Other issues Cloud scheme Vertical resolution Decoupling – structure within ‘mixed’ layers New Dynamics
38
Page 38© Crown copyright 2005 Change in total cloud cover Increasing the susceptibility to decoupling Control Revised scheme Bias RMS speed error EUROCS diurnal cycle of FIRE1 stratocumulus motivated changes to decoupling diagnosis (include subgrid cloudbase) Operational testing (10 global forecast case studies) showed: Reduced cloudiness, as expected Significant improvement to tropical winds at 850hPa
39
Page 39© Crown copyright 2005 Improved tropical winds hypothesis Decoupling implies moister near surface implies reduced surface moisture fluxes in subtropical stratocumulus areas Implies less moisture transported to tropics implies reduced tropical rainfall implies reduction in current model excessive tropical hydrological cycle implies beneficial reduction in trade wind strength Implies mixed layer structure is important Change in total cloud coverChange in surface moisture flux
40
Page 40© Crown copyright 2005 Other things Cloud scheme Vertical resolution Decoupling – structure within ‘mixed’ layers New dynamical formulation included: Semi-implicit, semi-Lagrangian dynamics Revised timestep (parallel calculation of ‘slow’ physics increments and BL diffusion coefficients) Revised vertical grid staggering (better handling of normal modes and control of computational modes) Showed a general improvement in low cloud But no clean comparison (eg. with the same physics)
41
Page 41© Crown copyright 2005 HadSM3 Cloud Response along GCSS Pacific Cross Section transect Coupled atmosphere/slab-ocean models Sc in HadSM4 (old dynamics) very different from HadGSM1 (new dynamics) HadSM4 includes Lock et al (2000) scheme but not flux-coupling
42
Page 42© Crown copyright 2005 Stratocumulus issues Met Office model appears to have overcome the 1 st order problem (it has stratocumulus-like clouds in some of the right places) but… Entrainment Still large spread in parametrizations (Stevens 2002) Still large spread in LES (Duynkerke et al 2004 or Stevens et al 2005) Is a special treatment of fluxes across inversions necessary in GCMs? Need observed cloud-top height to validate NWP (eg. lidar or radar?) Drizzle Large term in mixed layer budget How well is it represented in GCMs (lack of observed climatology)? Role and parametrization of aerosols (cf land/sea split in NWP)? Stabilising feedback on turbulent dynamics (Ackerman et al 2004)? Mixed layer structure in subtropics is important for downwind deep convection in tropics Beneficial impact from changes to decoupling Testing revisions to non-local part of flux parametrization How well-mixed are the wind profiles?
43
Page 43© Crown copyright 2005 More parametrization issues What level of cloud scheme complexity is necessary? Smith scheme couples liquid water and cloud fraction too strongly Is a prognostic scheme necessary for BL clouds? Inhomogeneity – important for SW and drizzle How to model partial cloud cover? Massflux versus turbulent diffusion Would unification help? Communication would be easier (eg. Cu detraining into Sc) Current massflux schemes seem incompatible with the top-down mixing predominant in stratocumulus Debate is fuelled by a lack of understanding (limited case studies - need more extensive examination of parameter space) If we knew (physically) how to parametrize BL clouds in general, wouldn’t the choice be made on numerical stability and accuracy?
44
Page 44© Crown copyright 2005 Questions?
45
Page 45© Crown copyright 2005 Stratocumulus forecasting Important for forecasting surface temperatures Case study of widespread low cloud over Europe in December 2004 (eg. 12Z on the 10 th ) Low cloud fraction in global UM T+48 x Budapest x Thanks to Martin Kohler
46
Page 46© Crown copyright 2005 Stratocumulus forecasting Boundary layer depth evolution mimics the observed General underestimate of boundary layer depth by ~100m (< grid size) General cold moist bias Analysis tends to ‘decouple’ 12Z 10 th December Day of December 2004 Surface mixed layer depth (m)
47
Page 47© Crown copyright 2005 Spin-up of cloud in forecast models Significant underestimate of cloud cover in the analysis followed by spin-up, at all horizontal resolutions
48
Page 48© Crown copyright 2005 - SW + + LW - CFMIP cloud feedback classes: Low positive class contributes most to uncertainty in sensitivity
49
Page 49© Crown copyright 2005 Lidar versus IR retrieval – Wylie, from Internet!
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.