Moving from Global Climate Simulations to Regional Model Predictions Moving from Global Climate Simulations to Regional Model Predictions Lawrence Buja National Center for Atmospheric Research Boulder, Colorado Lawrence Buja National Center for Atmospheric Research Boulder, Colorado CAM T341- Jim Hack

The Earth Climate System

Climate of the last Millennium Caspar Ammann NCAR/CGD

Ammann et al.

Figures based on Tebaldi et al. 2006: Climatic Change, Going to the extremes; An intercomparison of model-simulated historical and future changes in extreme events,

Figures based on Tebaldi et al. 2006: Climatic Change, Going to the extremes; An intercomparison of model- simulated historical and future changes in extreme events, /publications/tebaldi- extremes.html /publications/tebaldi- extremes.html

Working Group 1 (The physical basis of climate change) –Warming is unequivocal –“Very likely” that most of the late 20th century warming is due to human emissions. Working Group 2 (Climate change impacts, adaptation and vulnerability) –Large-scale changes in food and water availability –Dramatic changes in ecosystems –Increases in flood hazards and extreme weather Working Group 3 (Mitigation of climate change ) –Some devastating effects of climate change can be avoided through quick action –Existing technologies can balance climate risks with economic competitiveness. The Breakthrough 2007 IPCC Fourth Assessment Report Attribute sources of historical warming Project range of possible non-mitigated future warming from SRES scenarios Quantify Climate Change Commitment Project adaptation needs under various mitigation scenarios Time-evolving regional climate change on short and long-term timeframes Quantify carbon cycle feedbacks IPCC AR4 Before After Conclusion: With the wide public acceptance of the IPCC AR4 findings, the climate science community is now facing the new challenge of quantifying time evolving regional climate change that human societies will have to adapt to under several possible mitigation scenarios, as well as addressing the size of carbon cycle feedbacks with more comprehensive Earth System Models The IPCC AR4 findings are stronger and clearer than any previous report. Conclusion: The strength & clarity of the AR4 conclusions came from better observations, models, and HPC. The strong NSF/DOE partnership was critical to our success.

Geoengineering strategies Space mirrors, (Wood, Angel) High Altitude Sulfur injections Seeding stratocumulus clouds to brighten clouds Sequestration of CO2 Iron Fertilization,... Phil Rasch NCAR We are not proposing that geo-engineering be carried out! We are proposing that the implications should be carefully explored.

Krakatau Santa Maria Pinatubo El Chichòn Agung Global average surface temperature (relative to mean) Major volcanic eruptions °C A1B °C change relative to baseline

“Long Term Measure” for DOE Climate Change Research Deliver improved scientific data and models about the potential response of the Earth’s climate and terrestrial biosphere to increased greenhouse gas levels for policy makers to determine safe levels of greenhouse gases in the atmosphere. Imperative post IPCC: Improved climate/earth system models for regional prediction. E.g. What does a 2 o C rise imply in terms of regional change and impacts? DOE-UCAR Cooperative Agreement DOE/SC/BER Climate Change Prediction Program

Two classes of models to address two time frames and two sets of science questions: 1.Near-Term ( ) higher resolution (perhaps 0.2°), no carbon cycle, some chemistry and aerosols, single scenario, science question: e.g. regional extremes 2. Longer term (to 2100 and beyond) lower resolution (roughly 1.5°), carbon cycle, specified or simple chemistry and aerosols, benchmark stabilization concentration scenarios Science question: e.g. feedbacks

HPC dimensions of Climate Prediction Data Assimilation New Science Spatial Resolution Ensemble size Timescale Better Science (parameterization → explicit model) (new processes/interactions not previously included) (simulate finer details, regions & transients) (quantify statistical properties of simulation) (decadal prediction/ initial value forecasts) (Length of simulations * time step) Lawrence Buja (NCAR) / Tim Palmer (ECMWF)

Spatial Resolution (x*y*z) Ensemble size Timescale (Years*timestep) Today Terascale Climate Model Petascale 1.4° 160km 0.2° 22km AMR Km Regular Earth System Model 100yr* 20min 1000yr* 3min 1000yr * ? Code Rewrite Cost Multiplier Data Assimilation ESM+multiscale GCRM New Science Better Science HPC dimensions of Climate Prediction ? Lawrence Buja (NCAR) Exascale

Global General Circulation Continental large-scale flow Regional local Paleoclimate BGC/Carbon Cycle Spin-ups MJO convergence Resolve Hurricanes IPCC AR TF IPCC AR IPCC AR TF CCSM Grand Challenge PF Lawrence Buja (NCAR)

Atmosphere sea-ice Land-surface Prognostic Aerosols Land Ocean Carbon/Nitroge n cycles Earth System Model Off-line impacts On-line impacts Without downscaling Impact models AOGCM Downscaling and embedded regional models Gas Chemistry Schematic of an AOGCM (oval at upper left) and Earth System model (in orange oval) and various types of impact models (right). F ROM E SMs T O I MPACTS Ice Sheets Dynamic vegetation Ecosystems (Life) Ocean

TransportationTransportation Forecast Lead Time Warnings & Alert Coordination Watches Forecasts Threats Assessments Guidance Outlook Protection of Life & Property Space Operation RecreationRecreation EcosystemEcosystem State/Local Planning EnvironmentEnvironment Flood Mitigation & Navigation AgricultureAgriculture Reservoir Control EnergyEnergy CommerceCommerce Benefits HydropowerHydropower Fire Weather HealthHealth ForecastUncertaintyForecastUncertainty Minutes Hours Days 1 Week 2 Week Months Seasons Years Initial Conditions Boundary Conditions Seamless Suite of Forecasts Weather Prediction Climate Prediction Climate Change Trenberth

Predictability of weather and climate Trenberth

Stage 2. Historical: run using time-evolving, observed, Solar, GHG, Volcanoes, O Historical Stage 3. Future Scenarios: IPCC Scenarios from end of historical run Commit 2100 B A1B 2100 A Future Scenarios Years TS (Globally averaged surface temperature) Stage control run: 1000 years with constant 1870 forcing: Solar, GHG, Volcanic Sulfate, O control a b cd e 1870 Probablistic Climate Simulations

Future Plans The overarching goal is to ensure that CCSM plays a substantial and credible leadership role in climate change science, and makes substantial contributions to national and international coordinated climate change experiments and assessments The current model development timeline anticipates CCSM4 in 2009 in time to participate in the next set of internationally coordinated mitigation scenario experiments in short term climate change: 30-year climate predictions at higher resolution and a single scenario long term climate change: 300-year climate change simulations at medium resolution and carbon cycle for benchmark mitigation scenarios A next-generation Earth System Model will also be under development during this time period.

Earth System Grid has transformed CCSM data services CCSM3.0 Release (2004) Source Code, Input data and Documentation So easy that it was almost an afterthought. IPCC AR4 (2005-present) Distributed data services through both PCMDI and NCAR Delivered the model data for the IPCC AR 4 (Working Group 1) Changed the World Ongoing CCWG Research ESG data services have been a huge win for us… Cheaper Much easier Allows us to reach huge new research communities (GIS) “Lets the Scientists do Science”

Briefing on Results : USGS Science Strategy to Support U.S. Fish & Wildlife Service Polar Bear Listing Decision: a 6 month effort U.S. Department of the Interior U.S. Geological Survey

Habitat Change Projection: to USGS Report: Durner et al. (2007)

Earth Observations in Climate Models Probabilistic Climate Simulations: –Model Verification, ERBE, SHEBA, GRACE, historical temperature reconstructions New: Life and biogeochemistry Globally, regionally, and pointwise. Annual, monthly, daily, instantaneous –Boundary Conditions Solar, GHG, Sulfates, O 3, dust Deterministic Climate Predictions: –Same requirements as Probabilistic plus –Initial Conditions & Assimilation Ocean: 4-D temperature(tropics) and salinity(high-lats) most important –Argo (2KM depth) global float array big improvement over XBTs ( m) Sea-ice: Have extent, need thickness Land: Water (Snow, Soil, River) and Vegetation (LAI/Land cover) Atmospheric initial state not that important (will follow ocean) –Detailed atmospheric composition and annual cycle

Thanks! Any Questions?

Timeline of Climate Model Development

Chemistry Climate Chemistry Climate BioGeo Chemistry BioGeo Chemistry Software Engineering Climate Variability Polar Climate Polar Climate Land Model Land Model PaleoClimate Ocean Model Ocean Model CCSM Working Groups DevelopmentDevelopment Application Atm Model Atm Model Climate Change CCSM is primarily sponsored by the National Science Foundation and the Department of Energy

Lessons Learned 1. Observational data is very similar to model data

2. Obs data very different from model data Time Value Obs data Model data Lessons Learned

3. Don’t let scientists build their data management and distribution systems on their own! Building robust, useful data systems requires close collaboration between the two communities! …but don’t let the CS folks do it alone, either

4. Effective Data Distribution Systems Require Sustained Investment Home Grown Data Systems Community Data Portal Earth System Grid Initially Cheap $$$ in long term Limited Scale Modest Investment Agile and Right-sized for Many Projects Institutional Scale Large Investment Infrastructure for Large Projects Spans Institutions Lessons Learned

Thanks! Any Questions?

Future HPC Directions finally heading toward "massively" parallel capabilities in our models. –That means improving both performance and memory scaling. We hope to run high resolution (global 1/10 degree) on 10s of thousands of processors. Of late, most of our effort has been dealing with memory scaling because machines like IBM bluegene are limited to 256 Mb to 1Gb of memory per processor. At 1/10 degree, we might be able to have a global array declared at any one time. This has forced us to seriously recode various parts of the model that are usually ignored, like initialization and I/O. we are begining to truly require a paralell I/O capability. –In terms of scaling, we are working on improving the scaling capabilities of models like CAM by improving decomposition strategies and reducing communication cost. Finally, I believe a lot of the scaling success is resulting from the fact that hardware is becoming better balanced. 5 years ago, we had fast processors and less capable memory and communication systems on supercomputers. In the last 5 years, the processor speed increases have slowed, but the memory and communication system performance has been catching up, in a relative sense. (You might want to talk to Rich Loft to see if he agrees with my assessment). The petascale machines of the next couple years look like IBM Bluegene and Cray XT4. –This effort is really focused on the technical ability to run higher resolution on 10s of thousands of processors. That capability will then allow the science to have a chance to evolve at these resolution, and it will also benefit our moderate resolution runs by improving our scaling capabilities. Second, we are migrating CCSM to a more flexible coupling strategy, focusing on single executable (instead of multiple executable) and on the ability to run models sequentially, concurrently, or a combination of the two in order to optimize performance for a given configuration. This will give us an important capability to both improve model performance and also use the hardware resources better. I guess that's how I see our performance efforts right now. All of this is underway right now and will probably be a year or two before we see whether it works out. Let me know if you have any questions about any of my thoughts.

DOE/BER Climate Change Research Goals Improved understanding of the Earth's radiant-energy balance Prediction of regional climate change primarily using global models Quantification of sources and sinks of energy-related greenhouse gases, especially carbon dioxide Improved understanding of the scientific basis of the potential consequences of climate change, Improved understanding of the scientific basis for the potential impacts (ecological, social, and economic) of climate change and the benefits and costs of alternative response options.

CO 2,CH 4 and estimated global temperature (Antarctic ΔT/2 in ice core era) 0 = mean. Source: Hansen, Clim. Change, 68, 269, 2005.

Climate Modeling Basics The IPCC AR4 Simulations and Report Geoengineering Next steps

WG1: Scientific aspects of the climate system and climate change. Summary for Policy Makers:Feb 2007 Full Report:May 2007 WG2:Vulnerability of socio-economic and natural systems, consequences of climate change & options for adapting to it. Summary for Policy Makers:Apr 2007 Full Report:Jun 2007 WG3: Options for limiting greenhouse gas emissions and otherwise mitigating climate change. Summary For Policy MakersMay 2007 Full Report 2007 IPCC AR4: Intergovermental Panel on Climate Change Fourth Assessment Report Assess the state of knowledge on climate change based on peer reviewed and published scientific and technical literature on regular intervals.

Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report (AR4) Simulations NCAR Community Climate System Model (CCSM-3). Open Source 8-member ensembles 11,000 model years simulated “T85” - high resolution ~1 quadrillion operations/simulated year Rate of simulation: 3.5 sim. years/day Data volume for IPCC: ~110 TB Development effort: ~1 person-century

Climate Change Epochs Reproduce historical trends Prove Climate Change is occurring SRES Scenarios Investigate Mitigation Approaches Test Adaptation Strategies Look at Regional Details Work with Gov’t/IndustryPublic Before IPCC AR4 After

Abrupt Transitions in the Summer Sea Ice Observations Simulated 5-year running mean Gradual forcing results in abrupt Sept ice decrease Extent decreases from 80 to 20% coverage in 10 years. “Abrupt” transition Simulation of Future Climate

NCAR Projections of Degradation of Near-Surface Permafrost

Since 1970, rise in:Decrease in: Global surface temperatures NH Snow extent Tropospheric temperatures Arctic sea ice Global SSTs, ocean Ts Cold temperatures Global sea level Glaciers Hurricane intensity Drought & Heat waves Extreme high temperatures (Trenberth) Precipitation in extratropics Rainfall intensity Water vapor Warming is “unequivocal” “Very likely” the observed 20 th century warming is due to human emissions. IPCC AR4 WG2: Climate Change 2007: The Physical Science Basis.

Water shortages affecting hundreds of millions of people will worsen with increased temperatures. Declines in food production in the equatorial regions. 30% of plant and animal species facing extinction under a 1.5 to 2.5° rise. Increased risks of coastal flooding the sea level rise Extreme weather events to become more frequent and intense detecting human health and well-being Adaptation strategies to address unavoidable global warming outcomes. Mitigation strategies to avoid/minimize delay future emissions warming IPCC AR4 WG2: Climate Change 2007: Impacts, Adaptation and Vulnerability. Large-scale changes in food and water availability Changing ecosystems. Escalating flood hazards Increases in extreme weather

Alternative energy. Energy efficiency. Improved Industrial and agricultural practices Geoengineering options are risky and unproven Near term stabilization reduction has large long-term benefits Policies and economic incentives and technology transfer Additional research required to address some gaps in knowledge IPCC AR4 WG3: Climate Change 2007: Mitigation of Climate Change. Quick action can avoid some devastating effects Achievable with existing technologies, balancing economic costs with climate risks

Strength and clarity of IPCC AR4 message: Greater trust in our models = f( Science, HPC ) More realistic processes Higher resolution More ensembles -> less uncertainty Long historical simulations More trust in the observations Satellites/obs showing a consistent picture of global warming worldwide Definitive resolution of problems in the observations Recent Temperature trends in the lower atmosphere (*) Historical sfc temperature reconstruction (Mann's hockey stick) {**} Natural variability: No observed trend in recent solar activity Better agreement between the models and observations Glaciers, nature's independent climate integrators, showing increased edge melting and buildup in the centers consistent with Tebaldi & Meehl’s warmer/wetter world conclusion

Changes in Sea Ice Coverage Meehl et al, 2005

Nature vol March 2007 The Government is coming around

The Energy Sector is Reversing Course

The Public is now engaging at a level never seen before

Climate Modeling Basics The IPCC AR4 Simulations and Report Geoengineering Next steps

Geo-engineering The intentional large-scale manipulation of the global environment. The term has usually been applied to proposals to manipulate the climate with the primary intention of reducing undesired climatic change caused by human influences. Geoengineering schemes seek to mitigate the effect of fossil-fuel combustion on the climate without abating fossil fuel use; for example, by placing shields in space to reduce the sunlight incident on the Earth.(D. Keith,1999. Geoengineering. Encyclopaedia of Global Change. New York) Humans are already manipulating the environment deliberately, but generally not with the intention of inducing climate change. Phil Rasch, NCAR

Why do this study? –A worry that we are transforming our energy system far too slowly to avoid the risk of catastrophic climate change. –What might be deployable in a planetary emergency to mitigate some of the effects of greenhouse gas warming? –We are not proposing that geo-engineering be done! We are proposing that the implications be explored (see Crutzen study and Cicerone commentary, Climatic Change, 2006) Phil Rasch NCAR

Global Averaged Annual Averaged Surface Temperature change (flawed representation for evap and sedimentation) Phil Rasch

Cautions There are obviously very important Moral, Ethical, Legal issues to be considered here Arguments have been made that we shouldn’t even be considering these types of “solutions”. –Geoengineering will undercut society’s resolve to deal with emissions –Action would change climate in different ways for different nations. People in threatened regions most likely to use intervention Worst case outcomes –unanticipated impacts (e.g. CFC and ozone hole) –forestall global warming only to find we had triggered a new ice age Phil Rasch - NCAR

DOE ESnet4 Configuration Core networks: Gbps in , Gbps in Cleveland Europe (GEANT) Asia-Pacific New York Chicago Washington DC Atlanta CERN (30 Gbps) Seattle Albuquerque Australia San Diego LA Denver South America (AMPATH) South America (AMPATH) Canada (CANARIE) CERN (30 Gbps) Canada (CANARIE) Europe (GEANT) Asia- Pacific Asia Pacific GLORIAD (Russia and China) Boise Houston Jacksonville Tulsa Boston Science Data Network Core IP Core Kansas City Australia Core network fiber path is ~ 14,000 miles / 24,000 km 1625 miles / 2545 km 2700 miles / 4300 km Sunnyvale Production IP core (10Gbps) ◄ SDN core ( Gbps) ◄ MANs (20-60 Gbps) or backbone loops for site access International connections IP core hubs Primary DOE Labs SDN (switch) hubs High speed cross-connects with Ineternet2/Abilene Possible hubs

Climate Modeling Basics The IPCC AR4 Simulations and Report Geoengineering Next steps

DOE CCRD Directions Less emphasis on climate change detection and attribution More emphasis on decision support for policy makers provide decision-makers with scientific information on "acceptable" target levels for stabilizing atmospheric CO2 possible adaptation and mitigation strategies for the resulting climates before or after stabilization. Areas 1. Advance climate & Earth systems modeling 2. Improve understanding and model representations of climate and Earth system processes that can affect climate 3. Improve understanding of human impacts and consequences of climate change 4. Improve capabilities and infrastructure for conducting climate change research

Likelihood of warming Stabilizing temperature requires stabilizing atmospheric CO2, Limiting warming to 2°C requires stabilization at ppm CO2 (Meinshausen et al., realclimate.org) A short overshoot of atmospheric CO 2 might be compatible with the 2°C target.

Preparing for IPCC AR5

5 HPC dimensions of Climate Prediction (Tim Palmer, ECMWF) Data assimilation/ initial value forecasts Simulation complexity All require much greater computer resource and more efficient modelling infrastructures Resolution Ensemble size Timescale

Data assimilation/ initial value forecasts Simulation complexity Spatial (x*y*z) Resolution Ensemble size Timescale (Years*timestep) Today Terascale 2015 Exascale CM: 5 Components Petascale 1.4° 160km 0.2° 22km 1000 (AMR) Km ESM = CM+ 10 Components (Life, $$$) ESM + Afternoon session (multiscale GCRM) 100yr/ 20min 100??? 1000yr 3min 1000yr ? Code Rewrite

1 st generation Earth System Model Atmosphere Ocean Coupler Sea IceLand C/N Cycle Dyn. Veg. Ecosystem & BGC Gas chem. Prognostic Aerosols Upper Atm. Land Use Ice Sheets

Global General Circulation Continental large-scale flow Regional local Paleoclimate BGC/Carbon Cycle Spin-ups MJO convergence Resolve Hurricanes IPCC AR TF IPCC AR IPCC AR TF CCSM Grand Challenge PF

Summary: We’ve Changed the World. Broad acceptance of the IPCC AR4 is a breakthrough -Impossible without strong interagency collaborations -Moving to studying solutions means increasing –Model complexity/realism/comprehensiveness The next step will require many new partnerships –Science, Engineering, Technology, and Application –Regional, National & International Partnerships –Development Agencies –Economics/Financial Sectors –Energy Industry The NSF/DOE Partnership is more important than ever

NCAR Scientific Support facilities 2. Supercomputer/Network Resources 3. Scientific Models - National Science Foundation FFRDC Staff, 500 Scientists/Engineers - 4 Boulder-area campuses 1. Observational Facilities

NSF/DOE IPCC Project NCAR, ORNL, NERSC, ES 6-Year Timeline 2002: Climate Model/Data-systems development 2003: Climate Model Control Simulations 2004: IPCC Historical and Future Simulations 2005: Data Postprocessing & Analysis 2006: Scientific Synthesis 2007: Publication Observations of the Earths Climate System Simulations Past, Present Future Climate States

CCSM3.5 modifications to the deep convection scheme by Neale and Richter HadiSST Observations Control Neale+Richter HadiSST Obs CCSM3.5 - An Interim Version for Carbon-Nitrogen Work Peter Gent, NCAR Mariana Vertenstein, NCAR William Collins, LBL and UC Berkeley (presenter) CCSM3.5 shows a much improved ENSO Model Component Improvements Relative to CCSM3.0 Ocean Component (POP) Transition to POP2 with CCSM specific modifications and ecosystem and passive tracer capability Increased vertical resolution (60 levels) and topography changes Modified near-surface eddy flux scheme Land Component (CLM) Addition of coupled carbon-nitrogen cycle Improvements to hydrology (CLM3.5 Hydrology Project) Adopt SIMTOP (TOPMODEL-based surface runoff) Adopt SIMGM (groundwater model) New frozen soil scheme (freezing point depression, permeability of icy soil Added soil evaporation resistance term (function of soil moisture) ICE COMPONENT (CICE) Diabatic Layer Transition Layer Interior (quasi- adiabatic) -z Eddy-induced velocity profile (GM 1990) Diffusivities Horizontal Isopycnal Horizontal/Isopycnal This replaces the standard approach in the past of applying near-surface taper functions for the isopycnal and thickness diffusivities. GRACE 1 GRACE 2 CLM3 CONTROL CLM3 LMWG Hyd Water Storage in CLM3.5 Transition to CICE4.0 Included Software engineering enhancements Introduction of sub-blocking flexibility improves scalability and permits the introduction of threading Improved ridging parameterization Improved snow treatment CAM3 -> CAM3.5 Changes Deep Convection Convective momentum transports Dilute instead of undilute plume calculation Cloud Fraction Polar cloud freeze-drying Calculation updated after condensation New chemistry module for prognostic aerosols New chemistry module for greenhouse gases and prescribed aerosols ATMOSPHERE COMPONENT (CAM) Current and Near Term Simulations Have assembled an interim version, CCSM3.5, so that a carbon-nitrogen cycle can be run in an up-to-date version of CCSM. Present day and 1870 control integrations (at 1.9x2.5_gx1v5 resolution) are presently being run on the Cray XT3 (Jaguar) at ORNL. Performance of 32 model years/day for 1870 control Performance of 40 model years/day for present day integration In the near future, control integrations with the full carbon-nitrogen cycle included will also be run, along with 20th and 21st century integrations. Spin up procedure for CCSM3.5/BGC involves greater complexity than previous non-BGC simulations. Conclusions from Control Simulations Significantly reduced some major biases. ENSO frequency Mean tropical Pacific wind stress and precipitation High latitude (Arctic) temperature and low cloud biases Much improved surface hydrology in CLM3.5. Improved ocean and sea ice components. Acknowledgements go to our SciDAC lab partners for their critical and ongoing contributions to the CCSM project.