Testing Climate Models with CLARREO: Feedbacks and Equilibrium Sensitivity Stephen Leroy, John Dykema, Jon Gero, Jim Anderson Harvard University, Cambridge, Massachusetts 21 October 2008 Talk Outline • Feedbacks and Equilibrium Sensitivity • Climate OSSE – Optimal Methods/Multi-pattern regression – Response: GPS Radio Occultation (RO) – Feedbacks: Clear-sky Thermal IR Spectra • Discussion 21 October 2008 CLARREO: Feedbacks and Sensitivity 2 Climate Feedback 21 October 2008 CLARREO: Feedbacks and Sensitivity 3 Climate Feedback (2) Radiative forcing ΔFrad Longwave cooling ΓΔT Amplification or suppression of greenhouse effect, γΔT SW LW SW Δ +Frad Δγ T +γ TT Δ = Γ Δ ∂F dx ∑i ∑i γ SW = i i i i ∂x dT −1 i LW ⎡ SW LW ⎤ ∂F dx Δ =TF Δ Γγ − γ − γ LW = i rad ⎢ ∑i ∑i ⎥ i ⎣ i i ⎦ ∂xi dT 21 October 2008 CLARREO: Feedbacs and Sensitivityk 4 Feedback Uncertainty Bony, S., et al., 2006: How well do we understand and evaluate climate change feedback processes? J. Climate, 19, 3445-3482. 21 October 2008 CLARREO: Feedbacks and Sensitivity 5 Feedbacks and Climate Prediction Hansen, J. et al., 1985: Climate response times: Dependence on climate sensitivity ean mixing. and ocScience, 229, 857-859. ( 2T CO× ) −T ( 1 × CO ) s ≡ 2 2 ( 2Fradiative CO×2 ) −Fradiative ( 1 × 2 CO ) −1 ⎛ longwave shortwave ⎞ =⎜ Γ∑ −γ i − ∑γ i ⎟ ⎝ i i ⎠ −1 β=s() × C ρ ocean d dT =( Δsβ F = ()βs Δ F T − ) Δ dt imbalance radiative t T t()= T + sβ F Δ ()′ t−β()t e − t′ ′ d t 0 ∫0 rad 21 October 2008 CLARREO: Feedbacs and Sensitivityk 6 Climate OSSE: The Science of a Benchmark Benchmark Climate OSSE Climate Measurement Uncertainty • Traceable to • Simulate trends in • Shortwave forcing observable as international standards • Longwave forcing produced by different • Minimize sampling models • Climate feedbacks & error sensitivity • Explore information content with various contravariant fingerprints Climate OSSE Results • Detection time and accuracy requirements • How measurement constrains climate predictability • Relative redundancy with other benchmark data types 21 October 2008 CLARREO: Feedbacks and Sensitivity 7 Calibration: Double Differencing Hardy, K.R., G.A. Hajj, and E.R. Kursinski, 1994: Accuracies of atmospheric profiles obtained from GPS occultations. Int. J. Sat. Comm., 12, 463-473. 21 October 2008 CLARREO: Feedbacks and Sensitivity 8 Optimal Fingerprinting/Multi-pattern Regression We are limited by the naturally occurring inter-annual variability of the climate system…so optimize. Find signal amplitudes (αm) and uncertainty (Σα) in a data set (d) according to the signals’ patterns (si) against a background of natural variability, the eigenvectors and eigenvalues of which are eμ and λμ. 21 October 2008 CLARREO: Feedbacks and Sensitivity 9 GPS Radio Occultation • Refractivity p p ( 1 ) 10= (N 77 − n . 6 ×6 K = hPa- 1 + ) ( ×3 363 2 - 1 w 10 K hPa ) T T 2 • “Dry” pressure ∞ p()q h dp ( ) ( 4p . h= 402 × 10−4 N- 1 hPa( dh )≅ m ( p ) + 7521 h K ) d ∫ ∫ h 0 T • Geopotential height ⎡ 1 2 2 1 2 2 ⎤ = Φh (()⎢ −r Ωrs)( − Φrs −)/msl⎥ Ω g 0 ⎣ 2 2 ⎦ 21 October 2008 CLARREO: Feedbacs and Sensitivityk 01 GPS RO Dry Pressure Tendency 21 October 2008 CLARREO: Feedbacks and Sensitivity 11 How Does GPS RO Test GCMs? α = global average surface air temperature, d = GPS RO dry pressure [height] Poleward migration of jet 20 2 streams 15 ] -1 1 km 10 -1 Height [km] 0 5 [K ster Increased ITCZ 0 -1 humidity 90 60 30 0 -30 -60 -90 Latitude 2 1 0 Near perfect 2000 2010 2020 2030 2040 2050 Surface Air Temperature [K] Year tracking of global average surface air temperature 21 October 2008 CLARREO: Feedbacks and Sensitivity 12 Thermal Infrared Spectra 21 October 2008 CLARREO: Feedbacks and Sensitivity 13 Thermal Infrared Spectra (2) ] -1 5 decade -1 0 ster -1 ) -1 -5 (cm Tropics, SRES A1B, clear -2 NCAR CCSM3 GFDL CM2.0 W cm -10 MPI/ECHAM5-OPYC -8 UKMO HadCM3 -15 -20 Radiance Trend [10 500 1000 1500 2000 Frequency [cm-1] 21 October 2008 CLARREO: Feedbacks and Sensitivity 14 How Does Spectral IR Test GCMs? Carbon Dioxide Tropospheric Temperature Cumulative Signal 0.0 0.0 -0.2 -0.2 0.0 rad] -6 -0.4 -0.4 500 1000 1500 2000 500 1000 1500 2000 Water Vapor Stratospheric Temperature -0.2 Radiance [10 0.0 0.0 Δ -0.4 -0.2 -0.2 500 1000 1500 2000 Frequency -0.4 -0.4 500 1000 1500 2000 500 1000 1500 2000 21 October 2008 CLARREO: Feedbacks and Sensitivity 15 Applied Scalar Prediction 1.0 0.2 ] Tropospheric Temperature Signal Stratospheric Temperature Signal -1 0.5 0.1 ster -2 0.0 0.0 -0.5 -0.1 Radiance [W m Δ -1.0 -0.2 1.0 0.2 ] Water Vapor Signal Carbon Dioxide Signal -1 0.5 0.1 ster -2 0.0 0.0 -0.5 -0.1 Radiance [W m Δ -1.0 -0.2 0 5 10 15 20 0 5 10 15 20 Time [years] Time [years] 21 October 2008 CLARREO: Feedbacks and Sensitivity 16 Summary • Trends in GPS radio occultation data bear strongly on global average surface air temperature. • Trends in the outgoing longwave spectrum can be used to monitor longwave forcing and constrain all longwave feedbacks observationally. Optimization in space necessary to reduce detection times. • Work in progress includes simulations in cloudy skies and shortwave trends. 21 October 2008 CLARREO: Feedbacks and Sensitivity 17 Backup Slides 21 October 2008 CLARREO: Feedbacks and Sensitivity 18 Applied Scalar Prediction Find signal amplitudes (αm) and uncertainty (Σα) in a data set (d) according to the signals’ patterns (si) against a background of natural variability, the eigenvectors and eigenvalues of which are eμ and λμ. 21 October 2008 CLARREO: Feedbacks and Sensitivity 19 Applied Scalar Prediction Signal #1 Signal #2 Signal #3 Signal #4 Find signal amplitudes (αm) and uncertainty (Σα) in a data set (d) according to the signals’ patterns (si) against a background of natural variability, the eigenvectors and eigenvalues of which are eμ and λμ. 1000 2000 1000 2000 1000 2000 1000 2000 Eigenvector #1 Eigenvector #2 Eigenvector #3 Eigenvector #4 21 October1000 20082000 1000 2000CLARREO:1000 Feedbacks2000 and1000 Sensitivity2000 20 The Climate Benchmark −1 d=cos Re λ[]cos2 cos() λ− 1 φ cos 1 φ+ 2 1 sin φ 2 sin φ • Lat-lon grid is standardized 6816= . 74 km • Uncertainty derived fom r t testsdenpenedin 62º26.53" N 114º23.85" W 46º57.00" N 7º27.00" E 21 October 2008 CLARREO: Feedbacs and Sensitivityk 12 How Not to Monitor Climate… • Tape measure calibration is unstandardized • Tape measure subject to failure (by scissors) 21 October 2008 CLARREO: Feedbacks and Sensitivity 22 How Does Spectral IR Test GCMs? 0.0 rad] -6 -0.2 Radiance [10 Δ -0.4 0 500 1000 1500 2000 Frequency [cm-1] 21 October 2008 CLARREO: Feedbacks and Sensitivity 23 How Does Spectral IR Test GCMs? 0.0 rad] -6 -0.2 Radiance [10 Δ -0.4 0 500 1000 1500 2000 Frequency [cm-1] 21 October 2008 CLARREO: Feedbacks and Sensitivity 24 How Does Spectral IR Test GCMs? Anthropogenic Forcing Climate Response 0.0 0.0 0.0 rad] rad] rad] -6 -6 -6 -0.2 -0.2 -0.2 Radiance [10 Radiance [10 Radiance [10 Δ Δ Δ -0.4 -0.4 0 500 1000 1500 2000 0 500 1000 1500 2000 Frequency [cm-1] Frequency [cm-1] -0.4 0 500 1000 1500 2000 Frequency [cm-1] 21 October 2008 CLARREO: Feedbacks and Sensitivity 25 How Does Spectral IR Test GCMs? (3) 0.0 rad] -6 Radiance [10 -0.2 Δ -0.4 500 1000 1500 2000 Frequency 21 October 2008 CLARREO: Feedbacks and Sensitivity 26 How Does Spectral IR Test GCMs? (3) 0.0 rad] -6 Radiance [10 -0.2 Δ -0.4 500 1000 1500 2000 Frequency 21 October 2008 CLARREO: Feedbacks and Sensitivity 27 How Does Spectral IR Test GCMs? (3) 0.0 rad] -6 Radiance [10 -0.2 Δ -0.4 500 1000 1500 2000 Frequency 21 October 2008 CLARREO: Feedbacks and Sensitivity 28 Water Vapor-Longwave Feedback Precision After 20 Years Linear Trend Analysis Correlation Analysis (W m-2 K-1) (W m-2 K-1) Truth Data Truth Data GFDL CM2.0 3.30 ± 1.85 3.20 ± 1.85 2.75 ± 0.20 2.53 ± 0.18 GISS E-H 2.63 ± 0.81 2.95 ± 0.62 2.61 ± 0.10 2.94 ± 0.12 MIROC3.2 2.81 ± 0.85 2.53 ± 0.62 2.68 ± 0.13 2.49 ± 0.10 ECHAM5 3.14 ± 1.60 3.53 ± 1.81 2.98 ± 0.08 3.36 ± 0.10 CCSM3 2.80 ± 0.92 2.81 ± 0.91 2.66 ± 0.17 2.66 ± 0.16 HadCM3 3.10 ± 1.48 2.65 ± 1.15 2.78 ± 0.09 2.74 ± 0.11 …scales as (Δt)-3/2 …scales as (Δt)-1/2 21 October 2008 CLARREO: Feedbacks and Sensitivity 29 Water Vapor-Longwave Feedback Precision After 20 Years Linear Trend Analysis Correlation Analysis (W m-2 K-1) (W m-2 K-1) Truth Data Truth Data GFDL OptimizationCM2.0 3.30 ± 1.85 : Little3.20 ± 1.85 as of2.75 yet, ± 0.20 but 2.53 ± 0.18 GISS E-Hinclusion2.63 of± 0.81 the spatial2.95 ± 0.62 dimension2.61 ± 0.10 2.94 ± 0.12 MIROC3.2should 2.81yield ± 0.85 substantial2.53 ± 0.62 improvement.2.68 ± 0.13 2.49 ± 0.10 ECHAM5 3.14 ± 1.60 3.53 ± 1.81 2.98 ± 0.08 3.36 ± 0.10 CCSM3 2.80 ± 0.92 2.81 ± 0.91 2.66 ± 0.17 2.66 ± 0.16 HadCM3 3.10 ± 1.48 2.65 ± 1.15 2.78 ± 0.09 2.74 ± 0.11 …scales as (Δt)-3/2 …scales as (Δt)-1/2 21 October 2008 CLARREO: Feedbacks and Sensitivity 30 Accuracy Requirements, Detection Times • With observations traceable to international standards, one evaluates the uncertainty (accuracy) of individual measurements in a timeseries.