Ocean carbon cycle feedbacks in the tropics from CMIP5 models

Transcription

1 Ocean carbon cycle feedbacks in the tropics from CMIP5 models Jerry Tjiputra 1, K. Lindsay 2, J. Orr 3, J. Segschneider 4, I. Totterdell 5, and C. Heinze 1 1 Bjerknes Centre for Climate Research, UiB, Norway 2 National Center for Atmospheric Research, USA 3 Laboratories des Sciences du Climat et de l Environment, IPSL, France 4 Max-Planck-Institute for Meteorology, Germany 5 Met Office, Hadley Centre, United Kingdom 17th Annual CESM Workshop Breckenridge, June, 212

2 Temperature change Ocean uptake Idealized 1% CO2 increase / yr Ocean uptake (acc.) Land uptake (acc.) Land uptake Atmospheric CO2 Feedbacks simulated by CMIP5 models (Arora et al.) Figure 1: Atmospheric CO2 concentration (ppm) used in the 1% increasing CO2 simulations (panel a). Model mean values and the range across the eight participating 2/2 models for simulated temperature change, compared to the control simulation (panel b), The positive feedback is predominantly due to change in surface temperature. Ocean: Thermodynamic change lead to change in solubility, buffer capacity, and change in overturning. Land: Change in terrestrial respiration largely responsible for reduced net uptake. Fully coupled Radiatively coupled (+) Biogeochemically coupled (-) List of models (8): BCC-CSM1-1, CanESM2, IPSL-CM5A-LR, MIROC-ESM, HadGEM2-ES, MPI-ESM-LR, Uvic ESM2.9, and NorESM-ME

3 Significance of future feedback in the equatorial Pacific CO 2 Accumulated (21-299) regional oceanic CO 2 uptake due to change in atmospheric CO 2 (solid-colors) and change in climate (shadedcolors). Climate Roy et al. (211) Simulated change in Revelle Factor due to anthropogenic CO 2 uptake by end of 21 st century. The Revelle Factor describes how the partial pressure of CO 2 in seawater changes to a given change in DIC (Sabine et al., 24). 3/2

4 Climate-carbon cycle fluctuation in the North Atlantic Dave Keeling Ocean uptake Sea-to-air CO 2 flux Ocean pco 2 uptake outgassing Julian day in 25 Julian day in 25 Observational-based study indicates substantial variability (in space and time) in the oceanic CO2 uptake in the North Atlantic Watson et al. (29) 4/2

5 Climate-carbon cycle fluctuation in the North Atlantic Annual pco2(sea) EOF1 Annual pco2(sea) EOF2 Modeling study using a coupled physical-biogeochemical ocean model forced by observed atmospheric variability in the North Atlantic. The study indicates that while over a long term, the atmospheric CO2 control the oceanic uptake, over shorter interannual period, the NAO variability regulates the oceanic uptake. Tjiputra et al. (212) 5/2

6 Questions to address How well do the CMIP5 models simulate the observed mean climate state and dominant variability in equatorial Pacific? El-Nino Southern Oscillation (ENSO) is the strongest natural interannual climate signal. Latif and Keenlyside (29) Does the simulated global carbon cycle respond accordingly to climate variability as observed? How the climate carbon cycle interaction change in the future? 6/2

7 Observed ocean CO2 flux variability in the Equatorial Pacific Cruise tracks for the TAO buoys Longitude year Longitude El-Nino Less outgassing La-Nina more outgassing Regulated by surface DIC convergence (vertical upwelling and meridional advection) Weak influences from biological production and temperature solubility effect Feely et al. (26) 7/2

8 CMIP5 Earth System Models NorESM-ME (Bjerknes Centre, UiB, UiO) ATM (CAM4-Oslo, ~1.9ºx2.5º, L26), OCN (MICOM, ~1º, L53) Land CC: CLM4, Ocean CC: HAMOCC5 HadGEM2-ES (Met Office, Hadley Centre) Atmospheric (~1.6º, L38), Ocean (~1º, L4) Land CC: Jules, Ocean CC: Diat-HadOCC IPSL-CM5-LR (Institut Pierre Simon Laplace) Atmospheric (~3.75ºx1.9º, L39), Ocean(~2º, L31) Land CC: ORCHIDEE, Ocean CC: PISCES MPI-ESM-LR (Max Planck Institute) ATM (ECHAM5, ~1.9º, L47), OCN (MPI-OM, ~1.5º, L47) Land-CC: JSBACH, Ocean-CC: HAMOCC5 CESM1 (National Center for Atmospheric Research) ATM (CAM4, ~1.9ºx2.5º, L26), OCN (POP, ~1º) Land CC: CLM4, Ocean CC: BEC 8/2

9 CMIP5 simulated global mean surface CO 2 fluxes, SST, and DIC MPI-ESM IPSL-CM5A HadGEM2 CESM1 NorESM Takahashi WOA GLODAP fgco 2 [mol C m -2 yr -1 ] SST [degree C] DIC [mol C m -3 ] Historical ( ) 9/2

10 GLODAP IPSL-CM5A HadGEM2 NorESM WOA Takahashi CMIP5 simulated global mean surface CO2 fluxes, SST, DIC, and TALK Taylor diagram MPI-ESM CESM1 NorESM-ME IPSL-CM5A-LR $ HadGEM2-ES MPI-ESM-LR CESM1 fgco2 [mol C m-2 yr-1] SST [degree C] 1/2 DIC [mol C m-3] Historical ( )

11 Mean and deviation of equatorial Pacific SST Maps of (contour-lines) annual mean SST computed from historical simulation from January 187 to December 25. The contour lines are in 2 C intervals and the thick contour lines represent the 26 C isotherms. The colors are standard deviation of seasonally-detrended monthly SST [ C]. Observation are taken from HadISST data over the same period. historical 11/2 [ C]

12 Mean of equatorial Pacific CO 2 flux [mol C m 2 yr 1 ] Maps of annual mean sea-to-air CO 2 flux (in [mol C m 2 yr 1 ] units) from the observation [Takahashi et al., 29] and as simulated by the models in the equatorial Pacific. Model values are computed from monthly CO 2 flux from January 1995 to December 25 of historical simulation. historical 12/2

13 Nino3.4 index variability Observation (HadISST) NorESM-ME HadGEM2-ES Five months running mean Nino3.4 index Observation (HadISST) NorESM ME HadGEM2 ES.752 Power [ C 2 (cycles month 1 ) 1 ] Period [years] HadISST NorESM ME HadGEM2 ES IPSL CM5A LR MPI ESM LR CESM1 HadISST (.714) NorESM (.871) HadGEM2 (.743) IPSL-CM5A (.779) MPI-ESM (.77) CESM1 (.99) Frequency [cycles month 1 ] IPSL-CM5A-LR MPI-ESM-LR IPSL CM5 LR MPI ESM LR Consistent 3-6 years intervals. The Hadley model only simulates one spectrum peak at around 7 years. All suggest higher amplitudes than observed. Historical 13/2

14 Vertical temperature and DIC profiles El-Nino La-Nina NorESM-ME HadGEM2-ES IPSL-CM5A-LR MPI-ESM-LR Anomalies Colors: Potential temperature [ C] Contours: Dissolved Inorganic Carbon [µmol C L -1 ] Thetao anomaly [degree C] 25-yrs picontrol 14/2

15 Simulated relationships between ENSO and sea-air CO 2 flux Sea-to-air CO 2 flux (lines) with Nino 3.4 index (color) Mean CO 2 flux [Pg C/yr].62±.5.5±.2.52±.2.7±.3.79±.12 picontrol.43 (Takahashi) 15/2

16 Simulated relationships between ENSO and sea-air CO 2 flux Cross correlation of annual (solid-lines) seato-air CO 2 flux and (dashed-lines) surface pco 2 in the equatorial Pacific Ocean with Nino 3.4 index over the 25-years of preindustrial control simulations. Positive lags Nino 3.4 index is leading. Correlation NorESM-ME Lags [years] Correlation Correlation Lags [years] HadGEM2-ES MPI-ESM-LR Lags [years] Correlation Correlation Lags [years] IPSL-CM5A-LR CESM Lags [years] picontrol 16/2

17 Simulated ENSO change in the 21 st century Standard deviation of Nino3.4 index for 3-yr windows. The majority of models suggest weakening in ENSO variability under most future RCP scenarios, suggesting less ENSO-related carbon cycle variations in future e 4 2.3e 3 2.1e 3 3.5e 3 NorESM ME e 3 1.6e 3 2.5e 3 1.2e e 4 2.3e 4 7.7e 4 1.9e 3 RCP2.6 RCP4.5 RCP6. RCP HadGEM2 ES.4 IPSL CM5A LR e 3 3.e 3 1.9e e 5 1.4e MPI ESM LR Year.4 CESM Year 17/2

18 Simulated mean ENSO change in the 21 st century 4 NorESM ME NorESM ME (RCP8.5) HadGEM2 ES 4 HadGEM2 ES IPSL CM5A LR El Nino 4 IPSL CM5A LR 2 La Nina MPI ESM LR 4 MPI ESM LR /2

19 Long term mean state change in the equatorial Pacific All models simulate persistent future increase in SST, which translates to decrease in CO 2 solubility. Increase in stratification lead to reduced biological production in all models, hence decrease biological pump rate (1-2%). If consistent with AR4 in simulating decrease of AMOC, the export of carbon through westward advection to the Indian Ocean might be reduced, leading to relatively higher surface pco 2 Therefore, over longer term, the positive climate-carbon cycle feedback in the equatorial Pacific is expected. 19/2

20 Summary CMIP5 models are consistent in simulating global positive climate-carbon cycle feedback under high CO 2 world in the future. Five models studied here produce quite realistic Nino3.4 index variability. The climate-carbon cycle interaction in the equatorial Pacific is also well simulated by most of the models. Discrepancies in the interannual climate-carbon cycle fluctuations are mostly due to different representative of the ecosystem modules. Weakening of ENSO variability under future global warming scenarios are suggested by the majority models, suggesting less role of CO 2 flux to ENSO variations in the future. Over long time-scale, feedbacks due to change in SST, ocean circulation, and biological production will dominate. The land is known to have the opposite carbon cycle variability associated with ENSO, thus full analysis from both land and ocean carbon cycle will be necessary. 2/2