A National Strategy for Advancing

Climate Modeling Chris Bretherton Committee Chair

University of Washington

Edward Dunlea Study Director

National Academy of Sciences

1

Addressing the Challenges Ahead • Impacts of climate change will continue to be felt and may accelerate in

coming decades – More urgent need for climate change information at smaller spatial scales – Need for better two-way communication with users of climate modeling

output • Advancing climate models to address user demands involves progress on

many fronts – Hardware to allow finer model grids and more model complexity – Software infrastructure enabling efficient use of hardware and model outputs – Scientific understanding encapsulated in computer algorithms – Observations and careful model testing – Communication of results and their uncertainties – Human resources to make all this possible

– Group of Federal agencies asked National Academies to develop a national strategy for the next 10-20 years for how to advance U.S. climate modeling endeavor – NOAA, NASA, NSF, DOE, Intelligence Community

2

Committee on “A National Strategy for Advancing Climate Modeling”

Chris Bretherton (Chair), University of Washington

V. Balaji, Princeton University Thomas Delworth, Geophysical Fluid Dynamics

Laboratory Robert E. Dickinson, University of Texas James A. Edmonds, Pacific Northwest National

Laboratory James S. Famiglietti, University of California,

Irvine Inez Fung, University of California, Berkeley James J. Hack, Oak Ridge National Laboratory James W. Hurrell, National Center for

Atmospheric Research

3

Daniel J. Jacob, Harvard University James L. Kinter III, Center for Ocean-

Land-Atmosphere Studies Lai-Yung Ruby Leung, Pacific Northwest

National Laboratory Shawn Marshall, University of Calgary Wieslaw Maslowski, Naval

Postgraduate School Linda Mearns, National Center for

Atmospheric Research Richard B. Rood, University of Michigan Larry L. Smarr, University of California,

San Diego

NAS report process – 18 months, 5 committee meetings including large community workshop with key stakeholders in climate modeling and user communities, interviews with key officials and scientists, report reviewed by separate group of 13 experts

Lessons Learned from Past Reports • Several reports in late 1990s/early 2000s examined how to improve

effectiveness of U.S. climate modeling • Discussion in several reports since then

• Reports generally most useful if they include practical recommendations – Influence strategic thinking at program level

• Previous investments and efforts in common software infrastructure have paid substantial dividends – Have helped support social integration of the diverse climate modeling

community by supporting bottom-up community co-operation • Previous reports highlight need for routine and reliable climate

information, products, and services – View outside the modeling community is also that more products are needed

• Bottom-up community governance offers new strategies for working-level decision making – Previous reports have consistently called for more coordination and

consolidation of climate modeling agencies and institutions, but this has met with limited success

4

Institutional Arrangements • Committee examined arrangement of climate modeling in U.S. agencies • Strengths

– Stability of support has allowed cadre of scientists to develop state-of-the-art models on long-term basis

– Collaborative activities (e.g., Climate Process Teams) can effectively leverage talents in both universities and national labs

– Healthy diversity of activity and benefits of competing approaches, especially to difficult scientific problems

• Weaknesses – Subcritical in key areas, especially in human resources – Potential duplication of efforts – multiple institutions solving similar software

problems • Why not a single U.S. climate modeling center?

– Various agencies involved in climate modeling have differing missions – Can reduce much of redundancy in other ways (discussion of common software

infrastructure to follow) – Potential for large disruptions in U.S. modeling capability, at least in near-term

• Risks of a move to a single U.S. climate modeling center significantly outweigh benefits, but there is a need to unify/consolidate current efforts

5

A National Strategy for Advancing Climate Modeling

6

Software Infrastructure Current Situation and Next 10-20 Years

• Climate models run on large computers at individual modeling centers – “in-house” shared software infrastructures

• Intercomparisons (CMIP) have improved understanding of model differences and need for intercomparisons will increase

• Evidence that using finer grids for climate models substantially improves simulations, and this requires more computer power

• Transition to massively parallel computing (no processor speedup, but billions of processors)

Recommendation • Climate modeling groups evolve to a common national software

infrastructure

• Increased ability to share and compare model components • More efficient adaptation to new hardware • How to get there?

– Community-based decision process to determine best path forward

7

Recommendation #1

National Climate Modeling Forum Current Situation and Next 10-20 Years

• Some communication between users and climate model developers (conferences, publications, etc.)

• Continued need for coordinated model development and evaluation • Increasing complexity of user needs

Recommendation • Annual climate modeling “forum”

• Organized by multi-agency organization; USGCRP as logical choice • Bring together climate modelers and users • Provide mechanism for different climate modeling communities (e.g., global and

regional climate modelers) to work together • Opportunities for continuing education of users and strategic discussion

How is this meeting different? • Not envisioned as a meeting of just presentations working groups • Not every modeler required to be there probably representatives from all

modeling centers though • Other meetings do not currently bring U.S. modeling groups and user groups

together in a focused way

8

Recommendation #2

Unified Weather-Climate Modeling Effort Current Situation and Next 10-20 Years

• Weather (< weeks) and climate (seasons to centuries) model time scales are typically separate

• Many of same physical processes are important in both • Unified weather-climate prediction (aka “seamless”) models increasingly used

• UK Met Office as prime example • Validation of climate models often limited by available observations

• Performance on weather and seasonal timescales are useful tests of climate models

Recommendation • Accelerated national modeling effort spanning weather to climate time scales

• Would be one effort within U.S. modeling endeavor • Would be facilitated by collaboration among operational weather forecast centers, data

assimilation centers, climate modeling centers, and the external research community • Testing of ‘fast physics (e.g., clouds, turbulence)’ in climate model at weather

time scales • Reduced weather forecast errors due to ‘climate drift’ • Powerful tool for better synthesizing observations into climate reanalyses

9

Recommendation #3

Climate Interpreters Current Situation and Next 10-20 Years

• Climate model output often difficult for users to access easily and interpret properly

• Some entities (public and private) providing translation of climate model information for users

• But no quality assurance /certification of information provided • Growing demand for climate information

Recommendation • Develop training program for climate model “interpreters”

• Could involve degree or certificate program offered by universities • Possible accreditation through national organization (possibly AMS or AGU)

• Trained interpreters can facilitate two-way communication between climate modelers and users

• Interpreters not envisioned as solution to all users needs/climate services • Rather, training program is crucial step that benefits any system that bridges climate

modeling and user communities

10

Recommendation #4

A National Strategy for Advancing Climate Modeling

11

Science and Uncertainty Research Recommendations

• Climate modeling priorities (large uncertainties, global importance, limit overall progress, societal impact)

• Climate Sensitivity: How much will the planet warm this century and beyond? • Includes Cloud Feedbacks, Carbon Cycle and Ecosystem Feedbacks, Cryosphere, Sea-level rise

• How will climate change on regional scales? How will this affect the water cycle, water availability, and food security?

• How will climate extremes change? • Unites States should vigorously support research on uncertainty

• Understanding and quantifying predictability and uncertainty • Automating approaches to optimization of uncertain parameters within models • Communicating uncertainty to both users of climate model output and decision makers • Developing deeper understanding on relationship between uncertainty and decision

making

Goals • Significant progress for number of scientific questions • Unwise to promise that newer models will invariably result in reduced

uncertainty • Models likely to exhibit an increasingly rich range of behavior, full of surprises and

unexpected results

12

Other Recommendations

Vision for U.S. Climate Modeling in 10-20 Years • Different modeling groups around the country will specialize in different aspects of the

hierarchy of models to address wide range of climate related questions • Regional climate modeling will allow interactive simulation of additional processes not included in the global

model

• Higher-resolution models able to provide information at smaller spatial scales • Global climate models routinely run at 5-10 km resolution in ten years and 1-5 km resolution within 20 years

• Research and operational weather, regional climate and global climate modeling, will be done within a national common software infrastructure

• Climate Modeling Forum will provide organized process for climate model users and stakeholders to help move models in new directions

• Climate interpreters will access and analyze large, comprehensive model datasets to generate needed local-scale information and digest it for end users

• Climate observation system to observe climate system and provide validation to models • Infrastructure for data supported on ongoing basis (including improved data archiving and

data synthesis) easy access to climate model output and climate observation data • Comprehensive representation of Earth system in models to accurately depict coming

changes, including human-climate interaction • Climate projection uncertainty will remain a big issue

• Improvements in climate models will likely be gradual, not revolutionary, but they can have huge economic value to the Nation

13

A National Strategy for Advancing Climate Modeling

14

Report: dels.nas.edu/Report/National-Strategy-Climate-Modeling/13430 Questions? Contact Edward Dunlea (edunlea@nas.edu)

Acknowledgments • Sponsors • Committee • Speakers /

workshop attendees

• Reviewers • NRC staff • Numerous

colleagues spoken to throughout study

• BASC

Extra Slides

National Research Council (2012)

15

Need for Climate Information

16

Insert 2 figures

Projected changes in annual runoff 2041-2060

2011 Outlook

• Weather = meteorological conditions at a given time in a given location – Temperature, precipitation, etc.

• Climate = average weather conditions over longer time scales – From seasonal to centennial

• Climate models = computer models that simulate past and future climate

• Climate models are crucial for providing climate information – Only models can project into the future

• Short term climate predictions – E.g., seasonal flood risk predictions

• Longer term climate projections – E.g., runoff projections for mid-century

• Wide variety of users of climate information – Farmers

• Decisions on irrigation needs (weeks/months) • Decisions on what crops to plant (years/decades)

– Hydropower system managers • Millions of people rely on hydropower for electricity • Managers use climate info for water storage decisions

and infrastructure planning – Insurance companies

• Accurately reflecting risks from climate and weather related natural disasters is major issue for insurance industry

• Floods, high winds, extreme precipitation, droughts, etc. – National security planners

• E.g., Navy is planning for how climate change will impact their operating environments (like a seasonally ice-free Arctic), missions, and facilities

– Other examples = mayors of large cities and infrastructure planners

• Economic impact is large – Billions of dollars at stake

17

Need for Climate Information (cont.)

Current State of Climate Modeling • Climate models have made great progress in past several decades

– Climate models are tested against historical climate observations – Model simulations compare well to observations for global-to-continental

scale features • Still long way to go to provide information desired by users

– Local scale information, shorter timescales • Multiple models in U.S. and internationally

– U.S. is leader in international climate modeling community – Climate Model Intercomparison Projects (CMIP) are important part of climate

model development – Multiple U.S. modeling centers: NSF/DOE (NCAR), NOAA (GFDL and NCEP),

NASA (GISS and GSFC), etc. • Diverse approaches can help solve hard problems and address different missions • Lack of unification can be impediment to progress

• Size of investment – USGCRP estimate ~$239M spent by federal agencies on “improving our

capability to model and predict future conditions and impacts” (11% of total climate research spending)

– Several hundred people working in climate model development difficult to get exact numbers

18

Computing Needs Currently

• Demand for supercomputing time from climate modeling community outstrips current capacity

Next 10-20 years • Increasing resolution and complexity of

models will require even more computing capacity

19

Other Recommendations

Computing Needs Recommendation

• Two-pronged approach • Continued use and upgrading of dedicated computing resources at

existing modeling centers • Research into more efficient exploitation of high-end massively

parallel supercomputers • Idea for a dedicated national climate computing facility was

examined • Facility beneficial only if it were created in addition to current

computing capabilities (most attractive in environment of sustained growth for climate research budgets)

Goals • Significant computing capacity at disposal of climate

modeling community • Access to high-end machines for cutting-edge research in

climate modeling

20

Other Recommendations

Observations Currently

• Observations critical for monitoring and understanding variability of and changes to climate system

• Evaluation and improvement of climate models fundamentally tied to quality of climate observing system

• Maintaining climate observing system is international enterprise

• But requires strong U.S. support Next 10-20 years

• U.S. support for climate observations under serious threat

• Number of in-orbit and planned NASA and NOAA Earth observing missions will decline by more than a factor of three by 2020

21

Other Recommendations

Observations Recommendation

• Committee reiterates statements of previous reports that call on U.S. to:

• Continue and augment support for Earth observations • Address potential for serious gaps in the space based

observation system Goals

• Maintain existing long-term datasets of essential climate variables

• Develop innovative new measurements for poorly observed processes

• Support intensive observational studies to enhance process understanding

22

Other Recommendations

Workforce Currently

• Climate model development requires challenging mix of skills

• Demands synthetic knowledge of climate physics, biogeochemistry, numerical analysis, and computing environments

• Requires ability to work effectively in a large group • Model development requires long development times

• Often longer than typical research award or graduate school career

Next 10-20 years • Climate model development facing competition with

other career paths, both for scientists and software engineers

23

Other Recommendations

Workforce Recommendation

• Multi-pronged approach to maintain and/or enhance pipeline of climate model developers

• Graduate fellowships in modeling centers • Extended postdoctoral traineeships of 3-5 years • Rewards for model advancement through clear well-

paid career tracks, institutional recognition, quick advancement, and adequate funding opportunities

Goals • Enticing high caliber computer and climate

scientists to careers in climate model development

24

Other Recommendations

Data Archives Currently

• Growth rate of climate model (and observational) data archives is exponential

• Maintaining access to this data is a growing challenge

Next 10-20 years • Climate research community, decision makers, and

other user communities desire to analyze and use both model output and observational data in increasingly sophisticated ways

• Growing demands imply climate data cannot be managed in an ad-hoc way

25

Other Recommendations

Data Archives Recommendation

• Data-sharing infrastructure for supporting model intercomparisons and simulations of broad interest should be systematically supported

• Operational backbone for climate research and serving the user community

• Including archiving and distributing model outputs to the research and user communities

• Develop a national information technology (IT) infrastructure for Earth System climate observations and model data

• Build from existing efforts • Facilitate and accelerate data display, visualization, and analysis both for

experts and the broader user community Goals

• Substantial research effort into new methods of storage, data dissemination, data semantics, and visualization

• Help bring analysis and computation to the data • Rather than trying to download the data and perform analysis locally

• Enhance access/interface to data for users – climate research community and more broadly

26

Other Recommendations

Science and Uncertainty Research Currently

• U.S. climate models will have to address an expanding breadth of scientific problems

• Uncertainty is a significant aspect of climate modeling

Next 10-20 years • United States can make significant progress

through a combination of increasing model resolution and complexity, as well as improved parameterizations

• Predictability and uncertainty needs to be properly addressed by the climate modeling community

27

Other Recommendations

Dissemination Activities Pre-Release Briefings • NSF Briefing – 9/6/2012

– GEO leadership team + others (OPP, etc.)

– ~ 12 participants • House Science Committee /

GAO Briefing – 9/6/2012 – Originated from GAO study idea

on related topic – ~ 10 participants

• Multi-agency Briefing – 9/6/2012

– Including representatives from NOAA, NASA, DOE, Navy

– ~ 25 participants • OMB/OSTP Briefing – 9/7/2012

– Arranged by Phil Duffy (OSTP at that point)

– ~ 15 participants

Fall Briefings • Sackler Meeting – 9/19/2012

– ~ 50 participants – Joint US NAS – Royal Society meeting

• NCAR Director’s Committee phone briefing – 9/20/2012

– ~10 participants • Intelligence Community phone briefing –

9/21/2012 – Sponsor briefing

• Navy phone briefing – 9/26/2012 – 4 participants – Office underneath the Oceanographer of the Navy

• Public Webinar – 9/28/2012 – 92 participants

• USGCRP IGIM workshop on CMIP5 – 10/4/2012

– ~ 40 participants • Community Leaders telecon discussion –

12/17/2012 – Discussion with leaders in the climate modeling

community – 21 participants – Included representatives from all major global

climate modeling centers

28

> 2500 Visitors to Climate Modeling 101 Website