GISS ModelE Progress and Plans

WGCM, Princeton, Nov 2016 Gavin Schmidt and team GISS Post-CMIP5 Progress

Goddard Institute for Space Studies

MJO variability and prediction skill Self-generated QBO Enhancements to forcings (irrigation, volcanic, solar) Better use of single forcing runs Greatly improved ocean/sea ice simulations Planned GISS CMIP6 Configurations

Goddard Institute for Space Studies

Multiple configs w/variations for DECK runs: 1. GISS-E2.1 (ready) Variations: OMA vs MATRIX; R vs H ocean; L40 vs L96/102 2. GISS-E3 (mid-2017) C90+L96/102, same oceans; self-generated QBO, MATRIX aerosols, M&G cloud microphysics, cold pool convection 3. GISS-E4 (2018?) C180+L96/102, GO2 (GISS Ocean 2) (cubed sphere/ALE vertical) 1JULY 2012 K I M E T A L . 4645

Newly resolved modes I: MJO

1JULY 2012 K I M E T A L . 4645 1JULY 2012GISS$E K I M E T A L .GISS$E2 GISS$E2.1))))4651

GISS$E2.1)))) FIG. 3. Space–time spectrum of the 158N–158S symmetric component of precipitation divided by the background spectrum for (a) GPCP, (b) AR4a, (c) AR5a, (d) C_AR5a, (e) AR5c, (f) AR5a_Ent1, (g) AR5a_Ent1_Re, and (h) AR5a_Ent2_Re. Superimposed are the dispersion curves of the odd-numbered meridional mode equatorial waves for the equivalent depths of 12, 25, and 50 m.

data are from the International Best Track Archive the Kelvin, equatorial Rossby wave, and MJO bands for Climate Stewardship (IBTrACS) dataset (Knapp (the last being defined as wavenumbers 1–3, periods 30– et al. 2010). 60 days) that can be found in the observations (Fig. 3a). The most significant improvement that AR5a has com- 3. Simulations of the MJO using Model E2 pared to AR4a is its simulation of the Kelvin mode. The Kelvin mode in AR5a is similar to that in observations in a. Simulations of the MJO in AR4 and AR5 versions (Kim et al, 2012) both its amplitude and phase speed; the implied equiv- of Model E2 22 FIG. 9. Phase–longitude diagram of OLR [contour plotted every 3 W m , positive (green) and negativealent (purple)] depth and is surface about evap- 25 m. Compared to AR5a, AR4a oration (W m22)/ 925-hPa moisture convergenceFollowing (kg Wheeler kg21 s2 and1) for Kiladis (a), (c) observations,(1999), wavenumber– and (b), (d) AR5a_Ent1.has a much Phases weaker are from and the faster Kelvin mode, which is also MJO life cycle composite; values arefrequency averaged between diagrams 108S are and constructed108N. to determine the mostly confined to high frequencies (i.e., periods less FIG. 3. Space–time spectrum ofcapability the 158N–15 of8S the symmetric models component to simulate of convectivelyprecipitation divided cou- bythan the background 7 days). Despite spectrum these for improvements, AR5a still (a) GPCP, (b) AR4a, (c) AR5a, (d)pled C_AR5a, equatorial (e) AR5c, waves (f) AR5a_Ent1, and the MJO. (g) AR5a_Ent1_Re, Figure 3 shows and the (h) AR5a_Ent2_Re.lacks the MJO Superimposed mode. are the dispersion curves of the odd-numbered meridional mode equatorial waves for the equivalent depths of 12, 25, and 50 m. a2DphasespaceofthetwoleadingPCsfromtheCEOFsymmetric wavenumber–frequencytropospheric power temperature spectra [nor- increasesFigure (Fig. 3 also 12b). contains The rel- the symmetric components of analysis. In this 2D phase space,malized distance by estimated from the background or- ative humidity power, Wheeler (Fig. 12c) and andthe precipitable wavenumber–frequency water (Fig. spectra of equatorial pre- igin represents amplitude of theKiladis MJO. (1999)] The strong-MJO of equatorial12d) precipitation also increase. from These obser- systematiccipitation changes from thecaused different by versions of Model E2. The eventdata occurs are from during the March–April International 2000 Best in the Track simulation. Archive thethe decreasing Kelvin, equatorial entrainment Rossby rate wave, (from and AR5a_Ent1 MJO bands to for Climate Stewardship (IBTrACS)vations and dataset several (Knapp versions of(the Model last being E2. Our defined focus as is wavenumbersC_AR5a (Fig. 1–3, 3d) periods represents 30– a version of Model E2 that Hovmo¨ ller diagrams of total,on anomalous the signals (deviationsdistinct from theAR5a) background can be characterized spectrum in asuses enhanced higher stability horizontal in the resolution than that in AR5a by fromet al. the 2010). seasonal cycle) and 20–100-day filtered equa- tropics.60 days) This that result can be is consistent found in the with observations those of Kim (Fig. et 3a).al. torial (158S–158N) precipitation show the eastward pro- (2011b),The most who significant showed improvementthat models with that stronger AR5a has MJOs com- pagation3. Simulations of organized of the precipitation MJO using anomalies Model E2 with phase alsopared had to a AR4a cold bias is its in simulation the upper troposphereof the Kelvin relative mode. The to 21 Kelvin mode in AR5a is similar to that in observations in speeda. Simulations;5ms during of the this MJO period in AR4 (Fig. and 11). AR5 Daily versions restart both its amplitude and phase speed; the implied equiv- filesof are Model saved during E2 the period of this event and used to initialize the 30-day integrations of AR5a. Note that we alent depth is about 25 m. Compared to AR5a, AR4a useFollowing restart files Wheeler during February–May and Kiladis (1999), 2000 to wavenumber– encompass has a much weaker and faster Kelvin mode, which is also thefrequency whole strong-MJO diagrams are period. constructed to determine the mostly confined to high frequencies (i.e., periods less capabilityDuring the of the course models of the to simulate 30-day integration,convectively the cou- than 7 days). Despite these improvements, AR5a still AR5apled equatorialversion systematically waves and deviates the MJO. from Figure the AR5a_Ent1 3 shows the lacks the MJO mode. version.symmetric Figure wavenumber–frequency 12a shows the composite power deviations spectra of [nor- the Figure 3 also contains the symmetric components of troposphericmalized by temperature estimated background from the first power, day of simulation. Wheeler and It the wavenumber–frequency spectra of equatorial pre- indicatesKiladis that (1999)] the tropical of equatorial atmosphere precipitation becomes from stabilized obser- cipitation from the different versions of Model E2. The (warmervations and upper/colder several versions lower of troposphere) Model E2. Our gradually focus is C_AR5a (Fig. 3d) represents a version of Model E2 that untilon the day signals 30. The distinct warming from the aloft background is greater spectrum than the in uses higher horizontal resolution than that in AR5a by cooling below so that the mass-weighted average of FIG. 10. Schematic diagram of the reinitialization experiment. Cold pool parameterization: Formed'from'downdrafts,'used'to'restrict'occurrence Goddard Institute for Space Studies of'weakly'entraining'plumes

GISS$E2

YOTC'20=day'MJO'rain hindcast'Hovmöller diagrams

(0.63'20=day correlation'with'TRMM TMI'with'cold'pool'vs. 0.70'TMI=Radar correlation)

GISS$E3 Self-generated stratospheric QBO

Tropical zonal mean winds ERA40

GISS$E2.1(L102)

Rind et al (2014) Ocean model improvements

Reformulation of GM eddy parameterisation GISS-Vertical Mixing Scheme Inclusion of GM vertical dependence G. Danabasoglu et al. / Ocean Modelling 73 (2014) 76–107 83 Evaluation with stand-alone ocean CORE-I/II protocol

ClimatologyGISS V2 GISS V2 Model

Obs

Fig. 6. Time-mean meridional heat transports for the Atlantic Ocean. The black lines denoted by L&Y09 represent implied time-mean transport calculated by Large and Yeager, 2009 with shading showing the implied transport range in individual years for the 1984–2006 period. Direct estimates with their uncertainty ranges from the RAPID Zonaldata (square; Johns mean et al., 2011) and SST from Bryden andgradients Imawaki, 2001 (triangle; B&I01) are also shown. AMOC Heat flux 26ºN

in stark contrast with the other models and observationally-based ticipating models, Fig. 6 implies many model differences in details data. The latitudinalSouthern variations in MHT Ocean for MRI-A reflect its AMOC of surfaceCORE heat fluxes, resulting II:" primarilyDanabasoglu from differences in sim- et al (2014) structure. Such variations seem to be common in the MHT distribu- ulated SSTs. One example is the much larger heat gain in BERGEN tions obtained with some other data assimilation products as well between 10°N and 30°N in contrast with most of the other models (see Munoz et al., 2011). We believe that, as discussed in Msadek where much smaller heat gains or even losses are suggested. The et al., 2013, errors in representations of the NADW cell and, partic- oceanic heat gain evident in most models between 45°N and ularly, in the vertical structure of h (see Fig. 11), are largely respon- 55°N – as indicated by the positive MHT slopes – is associated with sible for the substantially lower MHTs in all model simulations the surface heat fluxes acting to damp the cold SST biases present compared to observational estimates even in simulations with real- in these models (see Fig. 8) due to the incorrect path of the North istic overturning strengths. Although much smaller in its contribu- Atlantic Current (NAC) (e.g., Danabasoglu et al., 2012). tion to MHT, errors in the gyre components can also explain some of As hinted at above, AMOC is the dominant contributor to the the differences (Msadek et al., 2013). We note that non-eddy- Atlantic Ocean MHT (Böning et al., 2001; Msadek et al., 2013). resolving horizontal resolutions of the present models can contrib- The relationship between AMOC and MHT is presented in Fig. 7, ute to low MHTs due to changes in the mean rather than the eddy considering the scatter plot of the maximum AMOC transport heat transport (Kirtman et al., 2012). against MHT at 26.5°N. Here and in subsequent scatter plots show- At equilibrium, there is negligible storage so the positive and ing AMOC strength at 26.5°N, we also include the RAPID data for negative MHT slopes with respect to latitude in Fig. 6 indicate reference purposes only, as the model data represents the 20-year the corresponding latitude bands of zonally-integrated warming time-mean. Thus, these AMOC transports do differ from those of and cooling of the ocean, respectively, by the surface heat fluxes. Fig. 5. Fig. 7 confirms the general tendency of larger MHTs with Assuming such an equilibrium state has been achieved by the par- stronger AMOC transports with a correlation coefficient of 0.89. GISS Ocean 2 (GO2) Model

Goddard Institute for Space Studies

Orthogonal Cubed- Sphere grid C720 goal (1/8º)

Arbitrary Lagrangian Eulerian (ALE) vertical coordinate MATRIX Aerosol model

Goddard Institute for Space Studies

9 Efficacy of forcings in transient runs

Use historicalMisc runs + forcing calculations to assess predictability of TCR+ECS from

NATURE CLIMATE CHANGE DOI: 10.1038/NCLIMATE2888 LETTERS historical transients abTransient climate response Equilibrium climate sensitivity 1.5 1.5 AA (ERF): 1.5 °C AA (ERF): 2.2 °C GHG (ERF): 1.4 °C GHG (ERF): 2.0 °C LU (ERF): 3.3 °C LU (ERF): 3.0 °C Oz (ERF): 0.8 °C Oz (ERF): 1.2 °C Sl (ERF): 0.5 °C Sl (ERF): 0.6 °C 1.0 Vl (ERF): 0.5 °C 1.0 Vl (ERF): 1.0 °C Historical (ERF): 1.2 °C Historical (ERF): 1.8 °C (K) 0.5 (K) 0.5 T T ∆ ∆

AA (iRF): 2.2 °C AA (iRF): 3.8 °C GHG (iRF): 1.6 °C GHG (iRF): 2.4 °C 0.0 LU (iRF): 5.3 °C 0.0 LU (iRF): 4.0 °C Oz (iRF): 0.9 °C Oz (iRF): 1.2 °C Sl (iRF): 2.3 °C Sl (iRF): 2.3 °C Vl (iRF): 0.9 °C Vl (iRF): 1.6 °C Historical runs Historical (iRF): 1.4 °C Historical (iRF): 1.9 °C −0.5 −0.5 −1 012 3 −1 012 3 −2 ∆F (W m ) ∆F−∆Q (W m−2)

cdContributions to TCR Contributions to ECS underpredict GHG GHG 1.4 Oz 1.4 Oz Vl Vl Sl Sl 1.2 LU 1.2 LU AA AA Historical mean Historical mean Vector sum Vector sum sensitivity 1.0 1.0

0.8 0.8 (K) (K) T T ∆ ∆ 0.6 0.6

0.4 0.4

0.2 ERF 0.2 ERF iRF iRF 0.0 0.0 0 1 2 345 0 1 2 345 ∆F (W m−2) ∆F−∆Q (W m−2) Figure 1 | Model historical and single-forcing transient and equilibrium sensitivities. a, Non-overlapping ensemble average decadalMarvel'et'al,'2016 mean changes in temperature and instantaneous radiative forcing for GISS-E2-R single-forcing ensembles (filled circles) (defined with respect to 1850–1859). TCR is calculated from the slope of the best-fit line. Also shown are 1996–2005 temperature changes and e￿ective radiative forcing (open circles). In this case, TCR is the quotient of the temperature and ERF estimates. Following ref. 4, straight grey contours show isolines of TCR from 0 to 4 for the iRF case. b, Same, but changes in the rate of ocean heat uptake are subtracted from forcing changes. ECS is calculated from the slope of the best-fit line (for iRF) or from the quotient (for ERF). c, 1996–2005 average 1T and instantaneous (filled arrows) and e￿ective (white arrows) radiative forcing for each single-forcing experiment. The transient climate response for each experiment in each case is the slope of the line. The vector sum of the single-forcing values does not substantially di￿er from the historical values (circles) and the TCR of the sum and historical experiments is less than that of the GHG-only experiment. The published GISS-E2-R TCR (1.4 C) is shown as a dashed black line. d, Same as c, but the x axis shows the di￿erence of 1996–2005 average forcing and estimated ocean heat uptake. The slope of each line is the equilibrium climate sensitivity. The published GISS-E2-R ECS (2.3 C) is shown as a dashed black line.

For each decade, we plot the temperature anomaly versus forcing despite evidence27 that this may result in an underestimate of the (for TCR, Fig. 1a) or the dierence between forcing and ocean heat ‘true’ values. uptake anomalies (for ECS, Fig. 1b). Using The ratios of single-forcing TCR and ECS to CO2-only TCR and ECS define transient and equilibrium ecacies, respectively13. 1F 1T 1F 1T 1Q (1) = TCR ; = ECS + These are measures of the enhancement (or suppression) of the we calculate as the slope of the best-fit line in both cases4. Using climate response to the forcing relative to the climate response to only the first and last decades gives comparable results. The TCR CO2. Supplementary Table 1 lists the transient and equilibrium and ECS are then given by ecacies calculated from the GISS-E2-R single-forcing runs, along with uncertainties derived from the five-member ensembles for F2 CO2 F2 CO2 each forcing. TCR ⇥ ECS ⇥ (2) = TCR ; = ECS The global mean climate responses to dierent forcings may dier because of the character of the forcings themselves (such as where F2 CO is the model forcing for CO2 doubling. These linear their geographical or vertical distribution) and because dierent ⇥ 2 methods assume that both ECS and TCR are constant in time, forcings induce dierent patterns of surface warming or cooling,

NATURE CLIMATE CHANGE | VOL 6 | APRIL 2016 | www.nature.com/natureclimatechange 387 © 2016 Macmillan Publishers Limited. All rights reserved Forcing improvements

Goddard Institute for Space Studies

Irrigation (water added to land surface, either from rivers or groundwater) Greater differentiation in LU (crops, pasture etc.) Volcanic forcing by emission Solar forcing uncertainty Aerosol forcing - uncertain pre-cursor emissions and atm. processing InteractiveZonal Mean simulation Stratospheric Aerosol Optical Depth − Modelof V2

50 Goddard Institute for explosive Space Studies 0 Latitude volcanoes −50 Pinatubo AOD via GISS E2.1 + MATRIX 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Time Zonal Mean Stratospheric Aerosol Optical Depth − Model master

50

0 Latitude

−50

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Time Sato: Zonal Mean Stratospheric Aerosol Optical Depth

50

0 Latitude

−50

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Time

0.000 0.000 0.000 0.000 0.000 0.000 0.002 0.007 0.013 0.025 0.050 0.125 0.250 0.350 0.500 [AOT]

12 MIP foci Goddard Institute for Space Studies

1) DAMIP - single forcing ensembles (also SolarMIP/VolMIP/LUMIP) 2) RFMIP - Essential complement to understanding responses for all relevant expts. 3) AerChemMIP 4) CFMIP 5) PMIP - ‘out-of-sample’ evaluations

13 GISS CMIP concerns

Goddard Institute for Space Studies

Forcing variations and expansion Greater (controlled) structural variations in models Greater interactions Better stratosphere and trop/strat coupling Feedback to model groups from users ?❓ Tracking of data use (DOI or PIDs) ? Complete enough simulations to multiply/constrain ECS ? Derived data connection to original files