Table of Contents

Subcommittee on Global Change Research, Participating Agencies and Executive Offices

Ad hoc Working Group on Climate Modeling

Foreword

Executive Summary

1. Background

2. Summary of Findings

3. Summary of recommen- dations

4. Final Comments

Charge to the Working Group 

Main Report

1. Purpose

2. Current Situation

3. Scope of Document / Underlying Definitions and Assumptions

4. Elements of Climate Science

5. Issues of Computational Systems

6. Human resources

7. Management / Business Practices / Institutional models

8. Recommen- dations

9. Reference Documents

10. Endnotes

Full Report (PDF)

Formation of a Climate Service: A Climate Service with a well-defined mission should be chartered to deliver simulation and related data products. This should be a national resource that allows investigation in a controlled scientific environment with well-defined interfaces to both research communities and non-scientific customers.

Build from existing expertise: Given the human resource limits and need to address climate problems in a timely way, the Climate Service must be built upon existing expertise, with clear separation of Climate Service functions from current Agency obligations. The quality of discovery-driven climate-science research in the U.S. is excellent and broad, and a new Climate Service should strive to strengthen and focus the research activities on important national priorities.

Location: The Climate Service should not be located or assigned to any Agency or Center within the current multi-agency framework. We propose that an independent service, which is a concerted federation of the appropriate current Agency capabilities, should be formed. The existing agencies need to act like member states, drawing from a concept successfully used in the European Union.[24]

Business and Management: An integrating management structure with an executive decision making process and a supporting incentive structure must be developed. Supporting business practices must be developed. The external review and oversight process needs to be developed to allow stability and insulation from short-term programmatic volatility. Without a new business model, incremental funding of existing organizations will not provide needed capabilities.

Computational systems: The Climate Service requires dedicated computational resources with the highest level of capability. The computational resources must be aligned with the generation of the Climate Service products and under the management of the Climate Service. If additional computer science or computational science research is needed, then this, too, needs to be managed consistently with the goals of the Climate Service. This research must be application driven with the development of integrated software and hardware platforms to deliver the Climate Service products.

Software: Of the major Elements of a climate-science organization, Modeling, Data, and Computational Systems, the financial investment in the software component of the Computational System Element is most deficient. Many of the software issues are straightforward and require the integration of software engineering personnel and principles into climate-science activities. A software infrastructure needs to be developed to allow multiple groups at multiple institutions to participate in concurrent development in a controlled environment. Similarly software investment is needed to allow better interface with the computing environment -- facilitating runability, maintainability, portability and performance. The community must also develop a strategy that supports development of systems software in the absence of provision by vendors.

Hardware: Near term investment in hardware should be targeted to provide incremental increase in capacity, allowing the exploitation of task parallelism -- running multiple copies of application software.

Long-term investment in hardware is dependent upon many uncertainties in the development of applications software and technology development. Significant systems engineering and design is required to reduce the risk in technology investment. It is incumbent to maintain maximum flexibility, including the development of software that buffers the volatility that arises from changes in technology development.

The U.S. policy on high performance computing adversely affects the climate-science community and places U.S. centers at a competitive disadvantage with centers in other countries. At the least, the usability and performance of Japanese vector computers already delivered to non-U.S. centers assure that by many metrics the U.S. will lag non-U.S. centers for five years. There is substantial risk that this deficit will last longer with pervasive negative impact on U.S. climate-science.

The importation or offshore use of Japanese vector computers would have a profound near-term impact on U.S. capabilities -- decreasing the gap between U.S. and non-U.S. centers.

Software-hardware dependencies: The long-term investment in hardware and software should be approached in a systematic way that focuses on the software suites of the Climate Service. A hardware-centric approach that focuses on the placement of computers simply defined by high theoretical processor performance will not be effective.

The software investment depends greatly on the hardware options. With the restriction of U.S. centers to distributed-memory, commodity-based processors a large investment is needed in applications software, and there is substantial risk that the performance of the software is intrinsically limited. This investment is significantly higher than that necessary for non-U.S. centers. If Japanese computers were available to U.S. centers, it would not serve as substitute for significant investment in software infrastructure.

Integration: There should be the formation of a Climate Service that is roughly parallel to the National Weather Service. Earlier reports (see reference list) suggested the formation of, perhaps, 1-3 climate-focused activities as well as fully evolved capabilities in seasonal to interannual prediction. Issues of chemistry and data assimilation were not addressed in these earlier reports.

We recommend two major core simulation activities. The first is focused on weather and should build from the National Weather Service. The second is focused on climate, and while it builds from existing expertise, the exact components and location of the Climate Service is dependent upon addressing a number of organizational and management issues discussed throughout this document. With strong leadership and a clear vision, many of the principal components of Climate Service can be aligned, in the near term, from existing product-oriented activities in the mission agencies. It is critical that initial steps be made to develop a credible and competitive high-end climate capability, and we are concerned that potential Agency and political positioning over the location and running of a potential Climate Service will delay its formation.

The Weather Service and the Climate Service should undertake the development of a unifying infrastructure to allow effective transfer of expertise and algorithms. There is significant potential benefit from more thorough integration of weather and climate activities, as well as numerous other sub-disciplines that have developed to significant maturation as a field. On the time scale of ten years, a useful vision is one of a unified national capability for weather and climate modeling with focused centers for specific applications.

With this call for a more managed product-driven Climate Service, we maintain that a robust and diverse discovery-driven research capability must be sustained. If "operations" were to come to dominate the entire climate-science community there would substantial negative impacts. A balance is needed between product-driven and discovery-driven activities, which each benefiting the other. Due to the complexity of both the scientific and programmatic aspects of these fields, integration and mergers will be difficult to manage.

The issues of integration are amongst the most difficult to reconcile. There are many difficult conflicts that must be considered when deciding what to include in a climate service, and/or when to include it. Of the underlying issues there are two that directly impact our recommendation: 1) the shortage of human resources, and 2) the ability to support an integrated product-driven capability in the current multi-agency culture.

The shortage of human resources motivates tight integration as there are not enough people to populate several comprehensive modeling efforts. The leadership, management, and business challenges motivate keeping the implementation as simple as possible and expansive integration increases the risk of failure. The need for simplicity suggests, therefore, that the nascent Climate Service should focus only on climate simulation and that chemistry, data assimilation, carbon cycle modeling, etc., should not be included. However, these other activities would continue to be advocated and developed to meet both programmatic and scientific needs. This would place them in direct competition for human and monetary resources as well as further entrenching the fragmenting processes that are already in place. Further, efforts in other countries are already including chemistry and data assimilation in their climate activities and if a nascent U.S. service did not, then there would be a competitive disadvantage from the beginning. This conflict between simplicity and inclusion of activities broader than traditional climate simulation is complex and requires more deliberation than possible in this document. Further, some strategic positions need to be taken at Agency and higher levels on what should be included. We make the following observations.

Seasonal-to-Interannual Prediction: The issues of how to fit seasonal-to-interannual activities together with longer term focused, i.e. decadal, centennial, etc., are amongst the most controversial. Seasonal-to-interannual activities are on the brink of operational utility, and could be viewed in several ways, for instance, an entity unto itself, an extension of operational weather forecasting, or the foundation of a climate service. There is clearly a strong link between any climate service and seasonal-to-interannual activities, as one metric of a credible climate model will be its ability to perform seasonal-to-interannual predictions. If operational capabilities are truly on the brink, then alliance with NOAA operational activities is called for. NOAA has the primary operational capability in the nation and, whether or not this operational capability is as robust as necessary, the generation in the near-term of an independent operational capability is not justified. If the Nation is to generate additional operational capabilities, because the current capabilities are deemed inadequate, then we feel this needs to be done in an integrated systematic way that would benefit not only seasonal-to-interannual activities, but also other climate activities and, in fact, should also address those perceived inadequacies in the current weather operations.[25] There would remain the need to provide seasonal-to-interannual expertise within the proposed Climate Service as well.

Chemistry, Data Assimilation, etc.: Chemistry and data assimilation are fast becoming important aspects of the climate problem. There is legal mandate for ozone assessments, and the Nation is struggling to provide the infrastructure to provide these assessments. Foreign centers are increasingly incorporating these more comprehensive capabilities in their climate programs. Ideally a climate service capability would include both a chemistry and data assimilation capability from the beginning. However, we feel that these capabilities need to be brought in incrementally. Chemistry capabilities should be brought in sooner rather than later. The programmatic and scientific issues of data assimilation are more complex and inclusion in a Climate Service at the time of formation increases the organizational complexity too much. We express concern that initiatives to develop capabilities in carbon cycle modeling and weather research compete for the same resources as the climate community.

Again, the issues of integration are complex and beyond the scope the current document. If the tenets outlined in this document are accepted, then we propose that the issues of integration be considered in the implementation planning. Success would require statement of strategic desires at the Agency level and above, as well as inclusion of the practitioners in the affected communities.

Budget: On the order of 150 scientists, software engineers, and application-directed computational scientists, programmers and computer scientists need to be dedicated to the Climate Service. This number is dependent on the level of integration that is sought in the Climate Service (see Recommendation 9). This would require approximately $20M.

The money for computing systems is more difficult to define. We propose that on the order of $10-12 M per year needs to be allocated for hardware and related services for high-performance platforms. Similar amounts of should be allocated for other computational capabilities and services within the Climate Service[26]. Thus the total cost for computational resources is $20-24M. This is a lower limit.

1.5 million for facilities

$7.5 million to integrate with research and customer community

Total, approximately $50 M

This capability cannot be built by redirection of research funds that are linked with the USGCRP and will require infusion of new funds. The "mission" agencies, NOAA, NASA, and DOE, already fund some activities that are product oriented. NSF, through the Community Climate System Model Program, has taken on the responsibility to provide a research facility for the climate community. These existing capabilities could contribute to the core of a putative climate service. This requires the Agencies to clearly distinguish the funds that might support this core from those funds that support their discovery-related research and manage them appropriately. With the consideration of these existing product-oriented activities, which are broader than those just associated with USGCRP, the needed capabilities can be built from a combination of existing and new funds. The new funds must support the development of product-driven activities and not simply appear as enhancements to existing research activities. Finally, the current discovery-driven research programs generally support high quality and important investigations; therefore, to build a product-driven research institution at the expense of the discovery-driven programs would undermine the underlying research environment that is broadly cited as the greatest strength of the U.S.

Implementation: The details of implementation will require significant planning and be dependent on a number of interrelated decisions that must be made by the Agencies. Strong leadership, both within the Agencies and at a level higher than the Agencies, will be required.

The implementation can and should be incremental. In fact, we believe that with the definition of a stable vision and leadership there are a number of existing activities that could form the core of a future Climate Service. There are already moves by all of the Agencies to better integrate and unify modeling and computational activities. If these can be orchestrated towards a long-term vision, then substantial steps can be taken while the details of the Climate Service are developed and evolved. Again, without a new business model and management strategies within which to organize the Climate Service, there is a danger of simply rearranging the current activities, which will not be successful.

Finally, we emphasize that is artificial to speak of a climate-science capability, a national climate service, without integration of modeling and data (i.e. observational) activities. As charged, we addressed the data activities, but they were not explored in as much depth as the modeling activities. We state, explicitly, that many of the same underlying problems affect the environmental data undertakings of the U.S. as affect the modeling community, and integrated, systematic solutions are ultimately needed. Additional funding is crucial to both develop foundation climate observing systems and to integrate and maintain existing data sets for climate applications.