general_exam_revised
advertisement
The Response of Marine Boundary Layer Clouds to Climate Change in a Hierarchy of Models Chris Jones Department of Applied Math Advisor: Chris Bretherton Departments of Applied Math and Atmospheric Sciences VOCALS RF05, 72W, 20S Overview • Introduction: Marine boundary layer (MBL) clouds and climate sensitivity • Idealized local case studies in a hierarchy of models • The well-mixed MBL from observations • Comparison of model responses to changes in CO2 and temperature • Summary of proposed future work Earth’s Radiation Budget: R = Absorbed Solar Radiation – Outgoing Longwave Radiation Marine boundary layer clouds especially important because… 1. They’re reflective at visible wavelengths MBL clouds (NASA) Earth’s Radiation Budget: R = Absorbed Solar Radiation – Outgoing Longwave Radiation Cloud Fraction Marine boundary layer clouds especially important because… 1. They’re reflective at visible wavelengths 2. They cover a lot of area Global net cloud radiative forcing ~ -20 W m-2 (Loeb et al, 2009) Cloud forcing = R(clear sky) – R(all sky) (images courtesy of Chris Bretherton) Compared to CO2 ~ 2 W m-2 Earth’s Radiation Budget: R = Absorbed Solar Radiation – Outgoing Longwave Radiation Marine boundary layer clouds especially important because 1. They’re shiny (reflect incoming solar radiation) 2. They cover a lot of area 3. They’re hard to realistically represent in global climate models • Interplay between dynamics and physics • Nonlinear • Turbulent • Physics must be parameterized Climate Change: Response to radiative forcing R = Absorbed Solar Radiation – Outgoing Longwave Radiation If radiation budget is perturbed by a radiative forcing π«πΈ (e.g., doubling CO2), the Earth’s mean surface temperature adjusts until balance is restored: Δπ = Δπ + πΔππ Feedback parameter π = π0 + ππΏπ + πππ + ππΌ + ππΆ π0 ≈ −3.2 W m-2 K-1 (Planck) Example: If ΔQ results in more low cloud, that means more reflected solar radiation, less warming (Δππ is smaller for a given ΔQ) and thus a negative cloud feedback Δππ : Global mean equilibrium surface temperature change (“sensitivity to Δπ”) Cloud contribution most uncertain (next slide) Cloud feedbacks dominate climate sensitivity uncertainty in GCMs Clouds dominate overall climate feedback uncertainty Bony et al. (2006) Clouds: - Positive feedback, - Large spread between models Cloud feedbacks dominate climate sensitivity uncertainty in GCMs Clouds dominate overall climate feedback uncertainty Bony et al. (2006) Clouds: - Positive feedback, - Large spread between models Cloud feedbacks dominate climate sensitivity uncertainty in GCMs Clouds dominate overall climate feedback uncertainty Bony et al. (2006) Clouds: - Positive feedback, - Large spread between models Low clouds dominate cloud feedback uncertainty Soden and Vecchi (2011) Parameterizations of Physical Processes Make Profound Impact Equilibrium response to 2xCO2 3.2K climate sensitivity UW turbulence and shallow convection parameterizations largely responsible for increase in climate sensitivity from CAM4 to CAM5 – can our analysis help explain this? 4.0 K climate sensitivity (Gettelman et al., 2011) Objectives of This Research • Use a localized, idealized column-oriented analysis of prototypical MBL cloud regimes to identify and evaluate MBL cloud-climate radiative response mechanisms • Hierarchy of models: – Large eddy simulation (LES): high resolution cloud resolving model – closest we have to “observations” in local climate change simulations – Single-column model (SCM): ties results to GCM – Mixed-layer model (MLM): simplified model for interpretive purposes • Seek to relate SCM back to parent GCM • Scientific Relevance: Understanding mechanisms of change in GCMs is pre-requisite for constraining through observation and/or improving parameterizations. • Mathematical Relevance: Investigate impacts of various parts of model formulation (e.g., subgrid parameterizations, model resolution, applied large-scale forcings); to what extent can models be used to interpret the behavior of other models? Case studies drawn from CGILS Intercomparison Zhang et al (2010) • S12: Shallow Stratocumulus (Sc) • Well-mixed BL • S11: Transition between Sc and shallow cumulus (Cu) • Onset of BL decoupling • Cu rising into Sc • S6: Shallow Cu Hierarchy of models SCM (SCAM5) GCM (CAM5) Image courtesy of NOAA SCAM5 Vertical Resolution MLM LES (SAM) (S6, courtesy of Peter Blossey) Primitive equations for liquid static energy (π β = ππ π + ππ§ − πΏπ£ πβ ) and total water mixing ratio (ππ = ππ£ + πβ ) in this study Large-scale advection Dynamics Subsidence Tendencies due to physical processes, e.g., • Precipitation • Radiation and clouds • Microphysics • Turbulence Mixed-layer model equations Δππ‘ (Stevens, 2007) • πβ ππ‘ • πππ‘ π + π’ ⋅ π»ππ‘ = − ππ‘ ππ§ ππ§π + π’ ⋅ π»π§π = π€π + ππ‘ • + π’ ⋅ π»β = − π ππ§ π€ ′ β′ + πΉπ π§ π0 π€ ′ ππ‘′ + πΉπ π§ π€π (π§π ) Moist static energy β = π β + πΏππ‘ Water mixing ratio ππ‘ = ππ£ + πβ Inversion (cloud top) Mixed-layer model equations Δππ‘ (Stevens, 2007) Advective cooling/drying • πβ ππ‘ • πππ‘ ππ‘ + π’ ⋅ π»ππ‘ = − • ππ§π ππ‘ + π’ ⋅ π»π§π = π€π + π€π (π§π ) + π’ ⋅ π»β = π − ππ§ surface fluxes π ππ§ π€ ′ β′ + πΉπ π§ π0 π€ ′ ππ‘′ + πΉπ π§ = 1 π§π = π€π Δβ + πΆπ π 1 π§π β0∗ Radiation −β − ΔπΉπ π΅πΏ π0 π€π Δππ‘ + πΆπ π π0∗ − ππ‘ + πΉπ 0 Entrainment Precipitation How reasonable is the well-mixed assumption? Previous project studied the extent of well-mixed vs. decoupled boundary layers using aircraft data from VOCALS field experiment • Classified flight legs as wellmixed or decoupled based on gradient of moisture and temperature quantities October 2008November 2008 (http://www.atmos.washington.edu/~robwood/VOCALS/vocals_uw.html) Well-mixed Decoupling metric(s) πΏπ πΏπβ Decoupled Cloud layer Subcloud layer Profile-based decoupling classification: Well-mixed if πΏπ < 0.5 g kg-1 and πΏπβ < 0.5 K Approximately 30% of region was well-mixed. Well-mixed regions correspond to shallower boundary layers. These conditions are met at S12 location. Jones et al. (2011) Case setup and proposed sensitivity studies Simulation setup • Diurnally averaged summertime insolation • Models run to steady-state • Large-scale forcings specified from observations: – – – – Horizontal divergence Subsidence Sea surface temperature Wind profile CGILS sensitivity studies • Control (CTL) – Mimics current climate • 4xCO2 concentration (4xCO2): – Captures “fast” adjustment • Uniform +2K temp. increase: – Captures temperaturemediated response – Reduced subsidence (P2K) – Subsidence as in CTL (P2K OM0) S12 Results: Cloud Fraction LES Results from CGILS intercomparison MLM Results Preliminary S12 Results: Profiles Liquid static energy SAM LES: MLM: Moisture Cloud liquid Liquid static energy SAM LES: MLM: SCAM5: (L80) Moisture Cloud liquid Preliminary S12 Results: Summary 4xCO2 P2K π«π³π’ [m] π«πππ [g m-2] π«ππππ [Wm-2] SAM (LES) -111 -13 +28 SCAM5 (SCM) -176 -12 +54 MLM -68 -9 +14 SAM (LES) +109 +2 -2 SCAM5 (SCM) +70 +1 -7 MLM +114 +32 -30 SAM (LES) -38 -9 +20 +5 -5 +18 -4 -4 +8 P2K OM0 SCAM5 (SCM) MLM • All models exhibit similar steady-state mean sensitivities: • 4xCO2 has lower inversion, thinner cloud (positive cloud feedback) • P2K deepens and thickens relative to control (negative cloud feedback) • P2K OM0 thinner than P2K and slightly thinner than CTL (positive cloud feedback) • Subsidence (large scale dynamics) plays dominant role in P2K response MLM 4xCO2 Sensitivity Mechanism: Increased down-welling LW radiation ο decreased cloud top radiative cooling (~10% decrease) ο Less turbulence (i.e., less entrainment) 4xCO2 CTL ο Lower zi ο Cloud thickness decreases 4xCO2 CTL SCAM5 S12 Resolution Sensitivity Higher resolution does Cloud fraction Default CAM5 Resolution doesn’t sustain a cloud Future Work – Apply MLM to interpreting other LESs involved in CGILS case study – Fully investigate SCAM5 S12 behavior • What’s driving the resolution sensitivity? – Expand analysis to other locations (MLM may not apply) – Parameter-space representation with SCAM • Use SST, Free troposphere lapse rate, CO2 and/or subsidence as control parameters – Find a way to relate the local cloud response in SCAM to the sensitivity in its parent GCM (MODIS satellite image) Questions? Additional slides Future Work (plenty to keep me busy) – Apply MLM to interpreting other LESs involved in CGILS case study (hypothesis: by tuning entrainment efficiency, can I reproduce their mean properties / sensitivities?) – Dig into roots of SCAM5 S12 sensitivity (interpret w/MLM when appropriate) • What’s driving the resolution sensitivity? – Expand analysis to other locations (MLM may not apply) – Parameter-space representation with SCAM, following approach of Caldwell and Bretherton (2009) MLM study • Use SST, Free troposphere lapse rate, CO2 and/or subsidence as control parameters – Find a way to relate the local cloud response in SCAM to the sensitivity in its parent GCM Additional Slides • CRF, adjusted CRF, etc. SCAM5 Default Resolution vs. VOCALS radar strip SAM LES Equations • Prognostic TKE SGS model • Diagnostic cloud water, cloud ice, rain, and snow • Periodic horizontal domain, surface fluxes from Monin-Obukhov similarity theory • ISCCP cloud simulator • Parallel (MPI) Khairoutdinov and Randall (2003) The proposal (remember the proposal? This is a presentation about the proposal …) • Use MLM to interpret output from other LESs (can “tune” parameterizations and entrainment closure as needed) • Investigate sensitivities in each model for each location • Map out primitive parameter-space representation using SCM (like CB09) • Ultimately, most concerned with SCAM, b/c it connects directly to GCM – to what extent can we use this analysis to shed light on the low cloudclimate mechanisms in CAM5? Primitive equations for liquid static energy (π β = ππ π + ππ§ − πΏπ£ πβ ) and total water mixing ratio (ππ = ππ£ + πβ ) in this study Large-scale advection Subsidence Tendencies due to physical processes, e.g., • Precipitation • Radiation and clouds • Microphysics • Surface fluxes • Turbulence Primitive equations • LES: • SCAM: Mixed-layer model equations • Prognostic equations: • π·β π·π‘ • π·ππ‘ π =− π·π‘ ππ§ π·π§π = π€π + π·π‘ • =− π ππ§ π€ ′ β′ + π€ ′ ππ‘′ πΉπ π§ π0 π€π (π§π ) Entrainment closure: π΄ π€∗ π π π§π Δπ π£ π π£0 π€π = • π΄ = π1 1 + π2 π ∗ 1 − =π΄π Δππ Δπ • • • • • ⟨π€ ′ π₯ ′ ⟩ (vertical turbulent flux of x) πΉπ (radiation flux) πΉπ (precipitation) Δπ₯ = π₯ π§π+ − π₯(π§π− ) A: entrainment efficiency • + πΉπ π§ • • β = ππ π + ππ§ + πΏππ£ (Moist static energy) ππ‘ = ππ£ + πβ (total water mixing ratio) π§π : Inversion height exp − ππ ππ π€π ππ π€∗ Mixed-layer model equations Prognostic equations: • π·β π·π‘ • π·ππ‘ π =− π·π‘ ππ§ π·π§π = π€π + π·π‘ • =− π ππ§ π€ ′ β′ + πΉπ π§ π0 π€ ′ ππ‘′ + πΉπ π§ π€π (π§π ) Mixed Layer Assumptions: • Vertically uniform profiles below inversion • Surface fluxes from bulk transfer model • Inversion flux given by π€ ′ π₯ ′ = −π€π Δπ₯ • No turbulence above inversion • Precipitation parameterized following Wood et al • Radiation from RRTMG radiative transfer model • Subsidence, large scale divergence, SST, surface pressure, and free troposphere h, q specified at all times Mixed-layer model equations: subsidence • Sensible heat flux Radiative cooling ππ§π = π€π + π€π (π§π ) ππ‘ BL πβ 1 ΔF R = −π’ ⋅ π»β + π€π Δβ + πΆπ π βπ ∗ − β − ππ‘ π§π π0 πππ‘ 1 = −π’ ⋅ π»ππ‘ + π€π Δππ‘ + πΆπ π ππ ∗ − ππ‘ + πΉπ 0 ππ‘ π§π Advection (cooling,drying) Latent heat flux Entrainment warming/drying Precipitation Contributing Mechanisms for MBL Balance Subsidence Advection EPIC 2001 (Bretherton, et al.) Mixed-layer model: • Well mixed q and h moist thermo variables => vertically uniform. – Bulk aerodynamic formulas for surface flux – Inversion fluxes based on thermo jumps ππ§π subsidence Sensible heat flux Radiative cooling = π€π + π€π (π§π ) ππ‘ BL πβ 1 ΔF R = −π’ ⋅ π»β + π€π Δβ + πΆπ π βπ ∗ − β − ππ‘ π§π π0 πππ‘ 1 = −π’ ⋅ π»ππ‘ + π€π Δππ‘ + πΆπ π ππ ∗ − ππ‘ + πΉπ 0 ππ‘ π§π Advection (cooling,drying) Latent heat flux Entrainment warming/drying Precipitation Sc (top) vs. Cu (bottom) MBL structure (Stevens et al 2007; Stevens 2006) MLM time series for S12 Relevant previous column modeling studies • Caldwell and Bretherton • Zhang and Bretherton • … Model run specifics • Grid resolution – CESM 1.0 (CAM5): 1 deg = 0.9 deg x 1.25 deg x 30 levels – (i.e., ~100 km x 137 km x … [variable]) • Time steps (?) • Length of integration • Numerics / miscellaneous Outline • Introduction – – – – Climate sensitivity, feedbacks, and cloud radiative forcing Why are low clouds important (to climate system, climate sensitivity)? What has been done, and where does this study fit in? Feedback flow chart (?) • Proposal for this study: Localized case studies using a hierarchy of models – CGILS cases – Primitive equations – An assortment of models • GCM (global models, under-resolved,…) • SCM (single column of the GCM) • LES (high-resolution column model – resolve largest, most energetic eddies, models subgrid) • MLM (idealized reduced order model that uses – Decoupling work pepper VOCALS throughout • MLM comparison with LES for S12 (and maybe SCAM?) • Proposed dissertation topic Outline • Introduction – – – – What is climate sensitivity and why do we care? Why are low clouds important (to climate system, climate sensitivity)? What has been done, and where does this study fit in? Feedback flow chart (?) • Proposal for this study – CGILS cases – Primitive equations – An assortment of models • GCM (global models, under-resolved,…) • SCM (single column of the GCM) • LES (high-resolution column model – resolve largest, most energetic eddies, models subgrid) • MLM (idealized reduced order model that uses – Decoupling work pepper VOCALS throughout • MLM comparison with LES for S12 (and maybe SCAM?) • Proposed dissertation topic Our approach: • Consensus that we need better understanding of the processes underlying low-cloud response to climate change (i.e., GCM intercomparison studies demonstrate clearly the global average low cloud response is a big uncertainty, but individual models differ in parameterizations of cloud processes, and climate-change output diverges widely between models) • Use IDEALIZED LOCAL CASE STUDIES (drawn from CGILS intercomparison) to investigate cloud sensitivity in a hierarchy of models (LES, SCM, and MLM) to climate-change inspired tests, with the goals of: – Understanding mechanisms behind cloud sensitivity (i.e., do LES and SCM agree? Can this behavior be constrained by observations? Is improved parameterization, informed by LES necessary?) – Connecting these back to the GCM behavior of a given model. Proposal: use a hierarchy of models to investigate low cloud response to climate perturbations • Local analysis: – Focus on 3 regions used in CGILS intercomparison study representing 3 low cloud regimes with idealized large scale forcings – Use 3 types of column models to investigate cloud sensitivity to a variety of perturbations: • Ultimate goal: Connect these back to GCM Profiles Well-mixed Decoupling metric(s) πΏπ πΏπβ Decoupled Cloud layer Surface layer drizzle Subcloud legs Decoupling metric: Δzb = π§π − πΏπΆπΏ (actual cloud base – “well-mixed” cloud base) Radar reflectivity (drizzle proxy) C-130 flight path (grey) Cloud base (lidar-derived) LCL (“well-mixed cloud base”) We use vertical profiles and subcloud level legs (courtesy of Rob Wood) Inversion Jumps • Lock (2009) and others have suggested high values of ππ Δπβ π =1+ πΏ Δππ‘ induce strong entrainment and Sc cloud breakup. • Strong entrainment might also favor decoupling. Inversion base Inversion “top” Δπβ Δππ‘ Decoupling not correlated with inversion jump parameter ππ πΏπβ π =1+ πΏ πΏππ‘ • Use REx C-130 profiles to calculate jumps/decoupling, adjacent subcloud legs to calculate cloud fraction. Restrict to flights before 10:00 LT in left panel. Blue = well-mixed Red = decoupled Hollow = POC Dash = Lock (2009) LES results • κ > 0.4 often (but not always) goes with broken cloud. • For κ < 0.5 there is no obvious correlation of κ and decoupling. • POC and non-POC distributions overlap Shiny clouds MODIS Visible Image Marine Boundary Layer (MBL) clouds: CGILS Cases (focus on S12 this talk) • S12: Shallow Stratocumulus (Sc) • Well-mixed BL => mixed-layer model appropriate • Focus of remainder of this talk • S11: Transition between Sc and shallow cumulus (Cu) • Onset of BL decoupling • Cu rising into Sc • S6: Shallow Cu Mixed-layer model equations Δππ‘ horizontal advection • πβ ππ‘ • πππ‘ ππ‘ + π’ ⋅ π»ππ‘ = − • ππ§π ππ‘ + π’ ⋅ π»π§π = π€π + π€π (π§π ) + π’ ⋅ π»β = π − ππ§ surface fluxes π ππ§ π€ ′ β′ + πΉπ π§ π0 π€ ′ ππ‘′ + πΉπ π§ = 1 π§π = π€π Δβ + πΆπ π 1 π§π β0∗ Radiation −β − ΔπΉπ π΅πΏ π0 π€π Δππ‘ + πΆπ π π0∗ − ππ‘ + πΉπ 0 Entrainment Precipitation Marine Boundary Layer (MBL) Clouds (Infrared satellite image, courtesy of Rob Wood) Marine Boundary Layer (MBL) Clouds NASA MODIS Satellite Image Questions? Marine boundary layer clouds: 1. Reflect incoming solar radiation 2. Cover a large fraction of the surface MODIS visible satellite image Reflective GFDL Clouds in climate models - change in low cloud amount for 2ο΄CO2 CCM model number from Stephens (2005) Well-mixed Decoupling metric(s) πΏπ πΏπβ Decoupled Cloud layer Subcloud layer Approximately 30% of profiles in VOCALS-REx were well-mixed (blue) Δπ§π = π§π − πΏπΆπΏ: thickness the cloud would have if it was well-mixed Climate Change: Response to radiative forcing R = Absorbed Solar Radiation – Outgoing Longwave Radiation If radiation budget is perturbed by a radiative forcing Δπ, the Earth’s mean surface temperature adjusts until balance is restored: Δπ = Δπ + πΔππ Δππ : Global mean equilibrium surface temperature change Radiative forcing (e.g., increased CO2) Feedback parameter π = π0 1−πππ −ππΏπ −ππΌ −ππΆ −β― Feedback parameter π = π0 + ππΏπ + πππ + ππΌ + ππΆ Example: If ΔQ results in more low cloud, that means more reflected solar radiation, less warming (Δππ is smaller for a given ΔQ) and thus a negative cloud feedback π0 ≈ −3.2 W m-2 K-1 (Planck) Cloud contribution most uncertain Earth’s Radiation Budget: R = Absorbed Solar Radiation – Outgoing Longwave Radiation Marine boundary layer clouds especially important because… 1. They’re reflective at visible wavelengths 2. They cover a lot of area (Infrared satellite image, courtesy of Rob Wood) Climate Change: Response to radiative forcing R = Absorbed Solar Radiation – Outgoing Longwave Radiation If radiation budget is perturbed by a radiative forcing Δπ, the Earth’s mean surface temperature adjusts until balance is restored: Δπ = Δπ + πΔππ Radiative forcing (e.g., increased CO2) π = feedback parameter Example: If ΔQ results in more low cloud, that means more reflected solar radiation, less warming (Δππ is smaller for a given ΔQ) and thus a negative cloud feedback Likewise, less low cloud => positive feedback (amplifies warming) Climate sensitivity Δππ : Global mean equilibrium surface temperature change due to 2xCO2 Cloud feedbacks dominate climate sensitivity uncertainty in GCMs Clouds dominate overall climate feedback uncertainty Bony et al. (2006) Clouds: - Positive feedback, - Large spread between models Low clouds dominate cloud feedback uncertainty Soden and Vecchi (2011) Earth’s Radiation Budget: R = Absorbed Solar Radiation – Outgoing Longwave Radiation Marine boundary layer clouds especially important because ... MBL clouds (NASA) IPCC (2007) The Models • LES (high resolution): System for Atmospheric Model (SAM) – – – – – • High resolution cloud resolving model Largest, most energetic eddies resolved Subgrid-scale turbulence is modeled The closest we have to “observations” for climate change simulations Parallel effort by Peter Blossey and Chris Bretherton for CGILS LES intercomparision SCM (single column of global model): SCAM5 (CAM5 GCM, operating in single column mode) – Single grid column from the GCM – Approximately 1 degree horizontal resolution, 30 vertical levels – Parameterize subgrid physical processes • MLM (idealized, interpretive model): – Idealized reduced order model applicable in Sc region (S12) when MBL remains “well-mixed” – When applicable, good for diagnosing / interpreting sensitivities in other models Earth’s Radiation Budget: R = Absorbed Solar Radiation – Outgoing Longwave Radiation Marine boundary layer clouds especially important because… 1. They’re reflective at visible wavelengths (NASA MODIS visible satellite image in Eastern Pacific)
