Introducing ToE MIP Time of Emergence Model Intercomparison Project for Ocean Biogeochemistry

Transcription

1 Introducing ToE MIP Time of Emergence Model Intercomparison Project for Ocean Biogeochemistry Sarah Schlunegger, Keith Rodgers, Jorge Sarmiento Thomas Froelicher, John Dunne, Jim Christian, Matt Long 4th International Earth System Model Conference Hamburg, Germany August 30th, 2017 contact: 1

2 Introduction Changes in the ocean carbon cycle and sink are imminent. But when and where will these changes first be emergent? Emergence calculations done with the CMIP5 ensemble obscure physical structures and linkages between variables. Emergence calculations done with 1 Large Ensemble cannot consider model dependency. Emergence calculations done with Pre-Industrial noise cannot consider variance changes, or retrieve an anthropogenic signal at local scales without smoothing techniques. Here, we use 3 Large Ensembles, evaluate emergence of anthropogenic changes in the components of the ocean carbon pump. Begin calculations at 1990, the approximate beginning of the ocean observing era. Use Large Ensembles as an Observing System Simulation Experiment (OSSE). 2

3 The Ocean Carbon Pump s Introduction 1. Solubility (emergence approximated by Air-Sea CO 2 Flux) 2. Biological a. Hard Tissue (CaCO 3 ) b. Soft Tissue (Organic Carbon) Air-Sea CO 2 Flux PgC yr -1 CaCO 3 Pump GFDL-ESM2M (30) CESM1-BGC (32) CanESM2 (50) Soft Tissue Pump RCP8.5 ensemble documentation: Rodgers et. al McKinley et. al & Long et. al Christian et. al

4 Emergence Calculations 1. Start in Emergence occurs when SNR > 2 (~95% Confidence Interval) where Signal is the mean of the ensemble trends between 1990 and a given year and Noise is the standard deviation of the trends 3. Done at grid-cell level (locally), regionally and globally Methods 4

5 Emergence Calculations 1. Start in Emergence occurs when SNR > 2 (~95% Confidence Interval) where Signal is the mean of the ensemble trends between 1990 and a given year and Noise is the standard deviation of the trends 3. Done at grid-cell level (locally), regionally and globally Equatorial Pacific Air-Sea CO 2 Flux Methods 5

6 Emergence Calculations 1. Start in Emergence occurs when SNR > 2 (~95% Confidence Interval) where Signal is the mean of the ensemble trends between 1990 and a given year and Noise is the standard deviation of the trends 3. Done at grid-cell level (locally), regionally and globally Equatorial Pacific Air-Sea CO 2 Flux Methods start:

7 Emergence Calculations 1. Start in Emergence occurs when SNR > 2 (~95% Confidence Interval) where Signal is the mean of the ensemble trends between 1990 and a given year and Noise is the standard deviation of the trends 3. Done at grid-cell level (locally), regionally and globally Equatorial Pacific Air-Sea CO 2 Flux Methods el Nino la Nina start:

8 Emergence Calculations 1. Start in Emergence occurs when SNR > 2 (~95% Confidence Interval) where Signal is the mean of the ensemble trends between 1990 and a given year and Noise is the standard deviation of the trends 3. Done at grid-cell level (locally), regionally and globally Equatorial Pacific Air-Sea CO 2 Flux Methods el Nino Signal Noise la Nina start:

9 Emergence Calculations 1. Start in Emergence occurs when SNR > 2 (~95% Confidence Interval) where Signal is the mean of the ensemble trends for between 1990 and a given year and Noise is the standard deviation of the trends 3. Done at grid-cell level (locally), regionally and globally Equatorial Pacific Air-Sea CO 2 Flux Methods Signal Noise start:

10 Emergence Calculations 1. Start in Emergence occurs when SNR > 2 (~95% Confidence Interval) where Signal is the mean of the ensemble trends for between 1990 and a given year and Noise is the standard deviation of the trends 3. Done at grid-cell level (locally), regionally and globally Equatorial Pacific Air-Sea CO 2 Flux Methods Signal Noise start: 1990 Emergence! 10

11 Results Time of Emergence Maps GFDL-ESM2M air-sea CO2 flux CESM1-BGC CanESM2 CaCO3 flux soft-tissue pump 11

12 Results Time of Emergence Maps GFDL-ESM2M air-sea CO2 flux CESM1-BGC CanESM2 CaCO3 flux soft-tissue pump red hatching non-emergent 12

13 Fraction of Ocean with Emergent Anthropogenic Trend GFDL-ESM2M CESM1-BGC CanESM2 Results CaCO 3 Pump SST Air-Sea CO 2 Flux Soft-Tissue Pump 13

14 Fraction of Ocean with Emergent Anthropogenic Trend GFDL-ESM2M CESM1-BGC Results CanESM2 SST ToE C/century Signal 14

15 Fraction of Ocean with Emergent Anthropogenic Trend GFDL-ESM2M CESM1-BGC Results CanESM2 Air-Sea CO2 Flux Signal gc/m2/yr/century ToE 15

16 Fraction of Ocean with Emergent Anthropogenic Trend GFDL-ESM2M CESM1-BGC CanESM2 Results CaCO 3 Pump ToE Signal gc/m2/yr/century 16

17 Fraction of Ocean with Emergent Anthropogenic Trend GFDL-ESM2M CESM1-BGC Results CanESM2 Soft-Tissue Pump Signal gc/m2/yr/century ToE 17

18 Regional and Global Emergence of Anthropogenic Trends Results Legend: Soft-Tissue Pump SST CaCO 3 Pump Air-Sea CO 2 Flux GFDL-ESM2M CESM1-BGC CanESM2 Southern Ocean Indian Atlantic Pacific Arctic Global Sof44S" 44s,Eq" NofEq" Nof65N" Global"" Arctic" Paci>ic"" Atlantic"" Indian" Southern"" Ocean" Sof44S" 44s,Eq" NofEq" Nof65N" Global"" Arctic" Paci>ic"" Atlantic"" Indian" Southern"" Ocean" Sof44S" 44s,Eq" NofEq" Nof65N"

19 Regional and Global Emergence of Anthropogenic Trends Results Legend: Soft-Tissue Pump SST CaCO 3 Pump Air-Sea CO 2 Flux GFDL-ESM2M CESM1-BGC CanESM2 Southern Ocean Indian Atlantic Pacific Arctic Global Sof44S" 44s,Eq" NofEq" Nof65N" Global"" Arctic" Paci>ic"" Atlantic"" Indian" Southern"" Ocean" Sof44S" 44s,Eq" NofEq" Nof65N" Global"" Arctic" Paci>ic"" Atlantic"" Indian" Southern"" Ocean" Sof44S" 44s,Eq" NofEq" Nof65N"

20 For large-scale, integrated quantities, the 3 large ensembles give a consistent estimate of how long an observing platform needs to be sustained for detection of anthropogenic signals in the ocean carbon sink, pumps and SST (exception CaCO 3 pump). However, if anticipating changes on smaller scales, or specific locations (i.e. the great barrier reef or coral triangle) is desired, then the inter-model uncertainty in emergence times highlights the necessity of having multiple large ensembles. Conclusions Regional Agreement Local Uncertainty GFDL-ESM2M CESM1-BGC CanESM2 Soft-Tissue Pump 20