Earth's Energy Imbalance: Natural Variability and SST patterns - PowerPoint PPT Presentation

Earth's Energy Imbalance: Natural Variability and SST patterns Cristian Proistosescu JISAO, University of Washington The Earth’s Energy Imbalance and its implications Nov 15, 2018 Toulouse, France Collaborators: Yue Dong, Kyle Armour, Robb Wills, David Battisti - University of Washington Malte Stuecker - IBS, Pusan, South Korea

Energy budget and climate sensitivity Net feedback: radiative radiative R ( T ) = λ T F λ = F − Q forcing response T Inferred Climate Sensitivity heat uptake F 2 × CO 2 T Q ICS = λ Energy Budget at TOA Q = F + λ T Gregory et al 2002

Central estimate based on historical anomalies since pre-industrial F μ + Q μ λ = T μ F = 2.33 W/m 2 Q = 0.71 W/m 2 •2005-2015 average WCRP assessment; in prep

Uncertainty in forcing leads to large uncertainty in inferred sensitivity ( F μ + F σ ) − Q μ 3.05 K (Forcing Only) λ = T μ F = 2.33±0.5 W/m 2 Q = 0.71 W/m 2 •2005-2015 average

Additional uncertainty due to observation error is negligible (1/25) ( F μ + F σ ) − ( Q μ + Q σ ) 3.05 K (Forcing Only) λ = 3.25 K (Forcing, EEI, Temp) T μ + T σ F = 2.33±0.5 W/m 2 Q = 0.71±0.1 W/m 2 •2005-2015 average •Uncertainties add in quadrature

Account for natural variability using decadal averages across CMIP5 piControl runs ( F μ + F σ ) + ( Q μ − Q σ + Q var ) 3.05 K (Forcing Only) λ = 3.25 K (Forcing, EEI, Temp) T μ + T σ + T var 3.28 K (Forcing, EEI, Temp +variability) F = 2.33±0.5 W/m 2 Q = 0.71±0.1±0.08 W/m 2 •2005-2015 average •Uncertainties add in quadrature

•Uncertainty in forcing strongly dominates observational uncertainty and natural variability (uncertainties add in quadrature) •Natural variability comes from models: ‣ What is its structure? ‣ Can we trust it?

CMIP5 piControl variability TAS •Variability in TOA associated with ENSO & PDO •On long time scales: very little additional variability in TOA despite significant low TOA frequency variability in TAS

CAM5 AMIP run with prescribed SST and Sea Ice (SIC) •Atmospheric concentrations (GHG, aerosols) fixed at 2000 level F = 0 (anomalous) •Changes in EEI at TOA driven by changes in SST & SIC Q = R = λ T

CAM5 AMIP run Decompose TOA into mean response to global temperature changes + variability T ( t ) SST [K] λ = < T > < Q > Q = λ ⋅ T ( t ) + Q var

CAM5 AMIP run exhibits significant variability More outgoing radiation recently - more negative feedbacks T ( t ) More Negative Feedback λ ⋅ T ( t ) SST [K] LOWER ICS Q var

Historical SSTs drive much higher interdecadal variability than the SSTs produced by coupled models More Negative Feedback λ ⋅ T ( t ) LOWER ICS Q var Q var

•Uncertainty in forcing strongly dominates observational uncertainty and natural variability (uncertainties add in quadrature) •Natural variability comes from models: ‣ What is its structure? ‣ Can we trust it? - NO! •What causes decadal variability in EEI? (in AMIP runs at least?)

⃗ ⃗ Green’s function approach: TOA radiative response to localized SST anomalies x 1 ) ∂ R ( J = x 2 ) ∂ SST ( 137 fixedSST runs on CAM4 (Dong, Proistosescu, Armour, Battisti, in prep) (Zhou, Zelinka, Klein 2017)

Response to representative patches S � � Dong et al in prep �

Controlled by West Pacific Surface Temp Controlled by local warming

⃗ Global radiative response to localized warming ( downward ) ∂ R x ) ∂ T (

Future warming pattern preferentially excites positive feedbacks (downward) •Decreased outgoing radiation • More positive feedbacks •higher climate sensitivity

Future warming pattern preferentially excites positive feedbacks (downward) Abrupt4xCO2 Feedback W/m2/K year

Pattern effect explains why historical estimates of ECS are low with respect to long term GCM estimates Simulations Historical Probability Proistosescu & Huybers 2017 Armour 2017 Andrews and Webb 2018 T 2 × CO 2 Abrupt4xCO2 Feedback W/m2/K year

Feedbacks increase with time in models, but decrease with time in AMIP runs AMIP- Historical SSTs Feedback W/m2/K year (with respect to pre-industrial level) Abrupt4xCO2 Feedback W/m2/K year

⃗ Global radiative response to localized warming ( downward ) ∂ R x ) ∂ T (

⃗ ⃗ Weight SST field by Green’s function, and take EOFs ∂ R Y = T ( x , t ) ⋅ x ) ∂ T (

Two dominant EOFs: ENSO + extended warm pool warming

Warm pool warming dominates decadal changes in TOA EEI

Warm pool warming dominates decadal changes in TOA EEI EOF #1 Mean warming pattern Constant feedback term CAM5 output Green’s function x EOF #1 Green’s function Reconstruction

Coupled models do not generate sufficient decadal variability in warm pool temperatures piControl AMIP 1 σ 1 σ

Conclusions •Observational Uncertainty not a major contributor to ECS uncertainty •Do we trust uncertainty estimates? • Pattern effect: slowly evolving patterns of SST (multidecadal) are the major source of uncertainty in linking EEI to ECS. •Feedbacks increase with time (higher ECS) in coupled models •Decrease with time in AMIP runs (historical ECS) due to relative Warm Pool warming • Coupled models severely underestimate decadal scale variability in EEI •Associated with underestimation of West Pacific variability •Forced (aerosols?) /Unforced?

SUPPLEMENTAL SLIDES

Sanity Check: CAM5 (prescribed SST) reproduces CERES EBAF radiation over PDO switch: 2014:2017 - 2000:2014 CERES EBAF SW CERES EBAF LW CERES EBAF NET CESM1 AMIP LW CESM1 AMIP NET CESM1 AMIP SW W/m 2

Background variability in atmospheric TAS and TOA T •CAM5 AGCM control: •Simulations with climatological SSTs •White noise variability in TAS and TOA Q

Background variability in atmospheric TAS and TOA + variability driven by changes in SSTs TAS •CAM5 AGCM control: •Simulations with climatological SSTs •White noise variability in TAS and TOA •CESM1.2 control: •Additional variability in TOA associated with ENSO & PDO type signals •On long time scales: very little additional variability in TOA despite significant variability TOA in TAS

Background variability in atmospheric TAS and TOA + variability driven by changes in SSTs TAS •CAM5 AGCM control: •Simulations with climatological SSTs •White noise variability in TAS and TOA •CESM1.2 control: •Additional variability in TOA associated with ENSO & PDO type signals •On long time scales: very little additional variability in TOA despite significant variability TOA in TAS •CMIP5 piControl •Consistent picture