Climate model parameterizations Phil Austin Tuesday, July 27 source: Bony et al., 2006 1/30 Summary from Monday: Monday summary 2/30 Summary from Monday: I Climate models use F = ma, conservation of energy (and entropy) and conservation of mass to calculate the evolution of temperature, humidity, wind speed, cloud cover, etc. on a grid. Monday summary 2/30 Summary from Monday: I Climate models use F = ma, conservation of energy (and entropy) and conservation of mass to calculate the evolution of temperature, humidity, wind speed, cloud cover, etc. on a grid. I The energy conservation and water conservation equations are: dhm =d dt dqT = −rate of precip removal + rate of precip evaporation dt where hm is the moist static energy, which depends on the temperature, vapor, and potential energy of the air, qT is the total specific humidity ((mass vapor)/(mass air)), and d is the diabatic heating is due to radiation. Monday summary 2/30 Today’s lecture: parameterizations Coverage: Tuesday intro 3/30 Today’s lecture: parameterizations Coverage: I Tuesday intro Calculating the radiative heating/cooling of the atmosphere 3/30 Today’s lecture: parameterizations Coverage: I Calculating the radiative heating/cooling of the atmosphere I Diagnosing cloud fraction and cloud overlap Tuesday intro 3/30 Today’s lecture: parameterizations Coverage: I Calculating the radiative heating/cooling of the atmosphere I Diagnosing cloud fraction and cloud overlap I Using satellites to validate climate/forecast models Tuesday intro 3/30 What is a parameterization? Parameterizations 4/30 What is a parameterization? I How do we calculate the effect of fluctuations in wind, temperature, specific humidity and radiation that occur at scales smaller than the 200 km grid size? These sub-grid scale fluctuations can transfer energy and water to and from the resolved scale flow Parameterizations 4/30 What is a parameterization? I I How do we calculate the effect of fluctuations in wind, temperature, specific humidity and radiation that occur at scales smaller than the 200 km grid size? These sub-grid scale fluctuations can transfer energy and water to and from the resolved scale flow For example: Parameterizations 4/30 What is a parameterization? I I How do we calculate the effect of fluctuations in wind, temperature, specific humidity and radiation that occur at scales smaller than the 200 km grid size? These sub-grid scale fluctuations can transfer energy and water to and from the resolved scale flow For example: I Parameterizations The reflectivity of a grid cell depends on the spatial distribution of the cloud liquid water 4/30 What is a parameterization? I I How do we calculate the effect of fluctuations in wind, temperature, specific humidity and radiation that occur at scales smaller than the 200 km grid size? These sub-grid scale fluctuations can transfer energy and water to and from the resolved scale flow For example: I I Parameterizations The reflectivity of a grid cell depends on the spatial distribution of the cloud liquid water The kinetic energy of air in the grid scale depends on convective plumes, or the waves developed by air flowing over mountains 4/30 What is a parameterization? I I How do we calculate the effect of fluctuations in wind, temperature, specific humidity and radiation that occur at scales smaller than the 200 km grid size? These sub-grid scale fluctuations can transfer energy and water to and from the resolved scale flow For example: I I I Parameterizations The reflectivity of a grid cell depends on the spatial distribution of the cloud liquid water The kinetic energy of air in the grid scale depends on convective plumes, or the waves developed by air flowing over mountains The mass of liquid water in a grid scale depends on removal by falling precipitation 4/30 Coupling between processes Parameterizations 5/30 Main atmospheric processes and variables Parameterizations 6/30 heating rate d: observed energy fluxes Current estimated net imbalance of ≈ 1 W m−2 . How are these fluxes changing over time? (Kiehl et al., 2009) Radiation 7/30 Start with longwave emitters/absorbers: H2 O, CO2 Radiation 8/30 H2 O transmission: US standard atmosphere Water vapor rotates as well as bends. This makes its transmission spectrum especially complicated. It absorbs at a broad range of wavelengths, and remits radiation depending on the wavelength and temperature. Radiation 9/30 H2 O transmission: US standard atmosphere Water vapor rotates as well as bends. This makes its transmission spectrum especially complicated. It absorbs at a broad range of wavelengths, and remits radiation depending on the wavelength and temperature. I absorptivity= a = 1 - transmissivity Radiation 9/30 H2 O transmission: US standard atmosphere Water vapor rotates as well as bends. This makes its transmission spectrum especially complicated. It absorbs at a broad range of wavelengths, and remits radiation depending on the wavelength and temperature. I absorptivity= a = 1 - transmissivity I emissivity = = absorptivity (“Good absorbers are good Radiation 9/30 Modeled emission: Looking down at the surface (T=300 K) from 12 km, no atmosphere. Longwave flux = σT 4 Radiation 10/30 Now add 270 ppm CO2 and a temperature profile. 105 W m−2 is trapped by the atmosphere Radiation 11/30 Now add tropical water vapor and trap another 61 W m−2 Radiation 12/30 Finally, double the CO2 to 540 ppm and trap another 4 W m−2 Radiation 13/30 Greenhouse warming: 1 layer energy balance emissivity=absorptivity is determined by the concentrations of the greenhouse gasses. Total emission is determined by the emissivity and the vertical temperature profile: Longwave Flux = σT 4 . Equilibrium greenhouse 14/30 Now double CO2 and raise absorptivity to 81% Equilibrium greenhouse 15/30 Now double CO2 and raise absorptivity to 81% I Tsurface changes about 1.2 K for 4 W m−2 net forcing. This is the zero feedback response of the planet. Equilibrium greenhouse 15/30 Now double CO2 and raise absorptivity to 81% I Tsurface changes about 1.2 K for 4 W m−2 net forcing. This is the zero feedback response of the planet. I Repeat this with an atmosphere that is warmer than the surface and find that the extra emission produces net cooling, because emission from the warmer air is larger than absorption from surface. Equilibrium greenhouse 15/30 Could CO2 absorption saturate at large concentrations? No. Harvey (2001) Equilibrium greenhouse 16/30 Next step: cloud layering and cloud fraction (Kiehl et al., 2009) Cloud fraction 17/30 Cloud fraction We need to know the cloud fraction as a function of height for situations like this: Cloud fraction 18/30 or this: Cloud fraction 19/30 or this: Cloud fraction 20/30 Parameterizing cloud fraction: statistical cloud schemes Cloud fraction 21/30 Parameterizing cloud fraction: statistical cloud schemes I Cloud fraction Basic idea: use cloud simulations at very high resolution (25-100 m), and small domains (6 km - 300 km), to establish a relationship between the gridcell specific humidity and energy and the cloud fraction (Charboreau and Bechtold, 2003) 21/30 Parameterizing cloud fraction: statistical cloud schemes I Basic idea: use cloud simulations at very high resolution (25-100 m), and small domains (6 km - 300 km), to establish a relationship between the gridcell specific humidity and energy and the cloud fraction (Charboreau and Bechtold, 2003) I Define the saturation deficit, Q ≈ (q vapor − q critical )/σq , where σq is the standard deviation of qvapor , q is the gridcell average, and qcritical is a critical value. Cloud fraction 21/30 Parameterizing cloud fraction: statistical cloud schemes I Basic idea: use cloud simulations at very high resolution (25-100 m), and small domains (6 km - 300 km), to establish a relationship between the gridcell specific humidity and energy and the cloud fraction (Charboreau and Bechtold, 2003) I Define the saturation deficit, Q ≈ (q vapor − q critical )/σq , where σq is the standard deviation of qvapor , q is the gridcell average, and qcritical is a critical value. I Get qcritical , σq from simulations like this animation (cloud fields.avi) Cloud fraction 21/30 Two simulated storms over land and water TOGA simulates large scale convection over the tropical ocean, while ARM is summertime thunderstorm convection over Oklahoma. Simulate the cloud systems for 60 hours, saving cloud fraction, qT , σq . Cloud fraction 22/30 Cloud fraction (N) vs. Q1 The result: the dependence of cloud fraction on excess saturation Q1 is similar for TOGA-tropical (circles) and ARM-Oklahoma (+). CGCM4 uses this to get the fractional cloudiness in each grid cell in the model. Cloud fraction 23/30 Validating cloud statistics: the A-train An orbiting constellation of satellites measure emission, absorption, scattering using radiometers (Aqua), radars (Cloudsat) and lidars (CALIPSO). Cloud fraction validation 24/30 An example scene: Mixed phase ice/water clouds from satellite Cloud fraction validation 25/30 Same scene: A-B transect using radar and lidar Ice crystals on the left side block the lidar, are but barely visible to Cloud fraction validation 26/30 Three climate models vs radar/lidar observations: cloud fraction Chepfer, 2009 Cloud fraction validation 27/30 ECMWF forecast model (top), satellite lidar (bottom) Cloud fraction validation 28/30 CGCM4: Monte Carlo independent column approximation Barker et al., 2004 Cloud layering 29/30 Summary Summary 30/30 Summary I Summary The radiative heating rate depends on the vertical and horizontal distribution of greenhouse gasses (including water vapor) and clouds 30/30 Summary Summary I The radiative heating rate depends on the vertical and horizontal distribution of greenhouse gasses (including water vapor) and clouds I Cloud distributions are specified by a sub-gridscale parameterization that links the grid-averaged specific humidity to cloud fraction I The parameters are determined by modeling the same kind of meteorologcial situation with high resolution cloud models, and global cloud statistics with satellites. 30/30 Summary Summary I The radiative heating rate depends on the vertical and horizontal distribution of greenhouse gasses (including water vapor) and clouds I Cloud distributions are specified by a sub-gridscale parameterization that links the grid-averaged specific humidity to cloud fraction I The parameters are determined by modeling the same kind of meteorologcial situation with high resolution cloud models, and global cloud statistics with satellites. I Cloud layers are overlapped by randomly constructing vertical profiles with the correct vertical correlation and averaging the radiative fluxes from many realizations of the profiles in each grid cell. 30/30