1. GFDL’s Coupled and Earth System Model Developments High-resolution Atmosphere and Oceans  horizontal (capturing regional, other fine-scale details), vertical (capturing the stratosphere) Climate Models graduating onto Earth Systems Model, including interactive Atmospheric Chemistry (capturing Climate – Air Quality nexus) and Biogeochemistry (capturing Carbon- Climate feedbacks) Need for comprehensiveness and realism in representation of processes

2. Earth system/ comprehensiveness pipeline Precip in 25 km modelSurface NOX in AM3 (200 km) Resolution pipeline Merge in future Model development strategy within GFDL: 2 pipelines developed in tandem push atmospheric and oceanic climate models to higher resolution, while developing and utilizing lower resolution Earth System Models

3. Winter mean precipitation in Western U.S. in 50km model 200 km res. 50 km res.Observations Annual mean precipitation in Europe 200km 50km 2 deg. 1/2 deg.

4. High-Resolution GFDL Model rainfall Current generation Developing generation Observed SST and Wind From Vecchi, Xie and Fischer (2004, J. Clim)Model data: S.J. Lin, I. Held, M. Zhao To resolve societally and physically relevant scales: High-res. atmosphere, ocean, coupled models being developed at GFDL

5. High resolution simulation of Southern Ocean (Hallberg and Gnanadesikan, JPO, 2006) Small vortices affect oceanic carbon uptake heat, transport of heat towards Antarctic continent, marine ecology of Southern Ocean

6. Development Pathway: Coupled Model for AR5 CM3 = AM3 + LM3 + CM2.1 ICE/OCEAN ESM2.1 = CM2.1 + LM3V + OBGC CM2G = CM2.1 with GOLD CM2M = CM2.1 with MOM+ESM2M ESM2G 2008 5yr/day on 500PEs 2400 CPUhr/yr 8yr/day on 140PEs 425 CPUhr/yr 10yr/day on 120PEs 300 CPUhr/yr

7. Why two ocean models? z-coordinate better in weakly stratified regions  -coordinate better on sloping bottom Role of ocean in transient climate change?

8. LM2 -> LM3 Land Model Dynamic vegetation Subgrid land-use heterogeneity Distinct treatment of ground, vegetation, & canopy air Soil water dynamics (liquid & frozen)‏ Multilayer snow pack River network with capacity for tracers and temperature

9. Decadal Predictability Research Ongoing studies with CM2 model to develop improved understanding of a)mechanisms of decadal variability b)decadal scale predictability arising from internal variability Development and use of higher resolution coupled models. Want this to be focus of variability and predictability research New coupled assimilation system

10. More intense hurricanes Drought More rain over Sahel and western India Warm North Atlantic linked to … Two important aspects: a.Decadal-multidecadal fluctuations b.Long-term trend Atlantic Meridional Overturning Circulation (AMOC) North Atlantic Temperature What will the next decade or two bring?

11. Simulated North Atlantic AMOC Index Aerosol only forcing All forcings Greenhouse gas only forcing

12. Projected Atlantic SST Change (relative to 1991-2004 mean) Results from GFDL CM2.1 Global Climate Model Can we predict which trajectory the real climate system will follow?

13. Decadal Predictability Decadal prediction/projection is a mixture of boundary forced and initial value problem Changing radiative forcing (esp. aerosols) will be a key ingredient Some basis for decadal predictability of internal variability, probably originating in ocean Some of predictability will arise from unrealized climate change already in the system Substantial challenge for models, observations, assimilation systems, and theoretical understanding

14. Earth System Modeling What proportion of Fossil Fuel CO 2 emissions will stay in the atmosphere, and for how long? What are the ecological impacts of increased CO 2 and ocean acidification? What are the ecological impacts of climate change? What is the role of land use on carbon cycling? How effective can proposed CO 2 sequestration approaches be (e.g. iron fertilization, deep ocean CO 2 injection, forest preservation)?

15. GFDL’s earth system model (ESM) for coupled carbon-climate Land physics and hydrology Ocean circulation Atmospheric circulation and radiation Land physics and hydrology Ocean ecology and Biogeochemistry Atmospheric circulation and radiation Chemistry – CO 2, NO x, SO 4, aerosols, etc Ocean circulation Plant ecology and land use Climate Model Earth System Model Sea Ice Land Ice Sea Ice Land Ice

16. IPCC AR4 WG1 Chapter 10

17. Ocean processes represented in GFDL’s current ESM Coupled C, N, P, Fe, Si, Alkalinity, O 2 and clay cycles Phytoplankton functional groups –Small (cyanobacteria) / Large (diatoms/eukaryotes) –Calcifiers and N 2 fixers Herbivory - microbial loop / mesozooplankton (filter feeders) Variable Chl:C:N:P:Si:Fe stoichiometry Carbon chemistry/ocean acidification Atmospheric gas exchange deposition and river fluxes Water column denitrification Sediment N, Fe, CaCO 3, clay interactions

18. Current land processes represented in GFDL’s current ESM Plant growth –Photosynthesis and respiration – f(CO 2, H 2 O, light, temperature) –Carbon allocation to leaves, soft/hard wood, coarse/fine roots, storage Plant functional diversity –Tropical evergreen/coniferous/deciduous trees, warm/cold grasses Dynamic vegetation distribution –Competition between plant functional types –Natural fire disturbance – f(drought, biomass) Land use –Cropland, pastures, natural and secondary lands –Conversion of natural and secondary lands and abandonment –Agricultural and wood harvesting and resultant fluxes in collaboration with PU, UNH and USGS (Schevliakova et al., subm GBC; Malyshev et al., in prep)

19. Work in progress: repeating these runs with a comprehensive ecosystem model Nutrients NO3 NH4 Si PO4 Fe Phytoplankton Diazotrophs Small plankton Large Plankton DOP,DON Zooplankton (parameterized) DOP, DON Sinking particles (POM, CaCO3, Opal, Fe) Oxygen Dissolved components Oxygen CaCO3 Remineralization/ dissolution Burial (CaCO3, Fe) N fixation Deposition (N, Fe) Runoff (CaCO3,N) Denitrification Dunne et al., in prep.

20. ObservationsModel Log(Chl) NO 3 PO 4

21. Moving towards the future of Earth System Modeling (ESM3 and beyond) Coupling with atmospheric chemistry Coupling with river biogeochemistry Integrated elemental cycles beyond carbon –N, P, Fe, CH 4 Seasonal fire dynamics Coastal and estuarine interactions Ecological prediction of hypoxia, harmful algal blooms and fisheries capacity variability

22. River Routing Hypoxia events Harmful algal blooms Land-use and ecology Nitrogen runoff Atmospheric chemistry Future Model Applications: Ecological Prediction Example

23. Doubles every ≥2 years 10x every ≥7 years

24. FY2007 Project Summary Project% of R&D HPCS (Princeton) CPU-hours per year Climate Scenario Analysis 10 3,800,000 Climate Scenario Generation 20 7,600,000 Software Infrastructure Development 5 1,900,000 Next Generation Ocean Model R&D 15 5,700,000 Long-term Climate Model R&D 40 15,300,000 Seasonal Climate Modeling R&D 10 3,800,000 TOTAL 100 38,100,000

25. DEC-CEN Computing Gaps Computing gaps are associated primarily with requirements for –Increased resolution To meet demands for information on water resources, extreme events, and ecosystems at regional scales. CM2.4, the target model for GFDL’s Decadal Prediction research, is 25x more expensive to run than CM2.1, one of its AR4-class models –Additional comprehensiveness Fully interactive cycles of many chemical species in the atmosphere, land, rivers, and ocean Direct and indirect aerosol effects Ice sheets (eventually) ESM2.1 is ≤2x more expensive to run than CM2.1 CM3 is 5x more expensive to run than ESM2.1 –Ensemble members Using different initial conditions Increasing number of assessments –Technology will provide about 6-8x between AR4 and AR5 State-of-the-art computing is crucial for recruiting and retaining top- notch scientific talent

26. The Challenges  Increasing the realism, capturing the complexities and addressing the key uncertainties  Coupling of the components  Climate, Earth System Models  Performing high-resolution simulations  Increased ensemble member integrations  Meeting timelines (e.g., for major assessments) Global Climate Model Critical Resources  Computational (“Computer-ware”)  Scientific talent (“Brainware”)

27. END