1. INTRODUCTION

A major acceleration of U.S. climate modeling is required to support the development of effective national energy and environmental policies. To meet this need, the Accelerated Climate Prediction Initiative (ACPI) has been proposed to significantly increase the rate of development, improvement and application of comprehensive coupled climate models through the provision of teraflop-scale computing resources. The ACPI is in turn a part of the federal initiative for Information Technology in the Twenty-first Century (IT2) (OSTP, 1999).

The specific objective of ACPI is to provide the most accurate possible predictions of the changes of regional climate and climate variability likely to occur as a result of future anthropogenic changes of atmospheric composition. The scientific and computational challenges presented by ACPI are discussed in a recent report (Department of Energy, 1998), which also cites the importance of applying the results of climate models to the needs of the user community. This document and a recent report of the National Research Council (1998) discuss the outstanding research questions, and point out the need for a coordinated national strategy if the U.S. climate modeling community is to contribute effectively to a national climate change assessment and to remain competitive in its capability to model and predict climate. As a goal-oriented project focussed on the high end of climate modeling, ACPI?s success will be measured by demonstrated improvements in our ability to make regionally-useful forecasts of long-term climate changes, and by the degree to which it can usefully supplement and interact with on-going climate modeling and research programs.

The charge given to the present Committee by the U.S. Global Change Research Program in October 1998 was to develop a plan for the implementation of ACPI in consultation with the U.S. climate modeling and research community. The purpose of this report (in which brevity was seen as a virtue) is to present our recommendations to the USGCRP, to the DOE as the principal implementing agency, to the NSF, NOAA and NASA as the major cooperating agencies, and to the U.S. climate modeling community at large.
 
 

2. FINDINGS

In the course of its work the Committee found a broad national concensus in support of ACPI, and received many valuable suggestions regarding the content of the program and its implementation. Perhaps the most important finding is that the U.S. climate modeling and research community recognizes the importance of the scientific, engineering and application challenges posed by ACPI, and sees the initiative as a unique opportunity to build a nationally-coordinated program.

Other findings with broad community support that are relevant to ACPI implementation are:

3. RECOMMENDED PROGRAM STRUCTURE AND IMPLEMENTATION

As a nationally-coordinated project, ACPI requires a clear statement of the management strategy to be used, and clear guidelines for review and coordination of its various components. An overview of the Committee?s recommendations for the ACPI program structure is given in Figure 1, and specific recommendations for individual parts of the project are given in the following paragraphs.

3.1 The Joint ACPI Implementation Agreement

A first step in the recommended ACPI implementation is the negotiation of a written agreement among the sponsoring federal agencies to jointly support and implement the project. This agreement should lay out the overall management and coordination strategy of the project, and identify any specific interests and/or responsibilities of each agency. The agreement should permit the inclusion of additional agency sponsors in the future. If ACPI is to successfully unify participants into a coherent national program, we recommend that the joint implementation agreement also define the relationships of the supporting agencies with the components of the ACPI management and project structure described below.

3.2 The ACPI Management Body

The core of our recommendations for ACPI implementation is the formation of a Management Body. Such a body would be charged by the sponsoring agencies via a cooperative agreement or other instrument with responsibility for the organization, management and performance of the overall program. The Management Body would be committed to the provision of effective leadership for the project and to its coordination with the OSTP IT² initiative, the SSI, the USGCRP, and other relevant national and international programs.

The ACPI Management Body needs to operate under an interagency agreement that effectively binds the sponsoring agencies to a common initiative while permitting them to exercise adequate oversight, coordination and review. If an existing interagency group were given this task, it would have to be empowered to exercise effective decision authority, and would have to overcome the inherent programmatic conflicts that have often diluted the effectiveness of such groups. An alternative would be to create a non-governmental project office or corporation under the direction of a respected senior climate modeler with experience in large project management and a willingness to devote full time to the project. In this option, the Director would have the operational management of ACPI as his/her sole function, and could be dismissed upon completion (or mismanagement) of the ACPI project.

For these reasons we recommend that a non-governmental project office or corporation be established as the ACPI Management Body, with its Director empowered by the sponsoring agencies with the authority and responsibility for the scheduling, accountability and leadership of the overall program. The Director would be appointed by (and report to) the sponsoring agencies and would be advised by an Advisory Committee which would also be appointed (and report to) the sponsoring agencies.

3.3 The ACPI Modeling and Research Consortia

We recognize that only a few high-end modeling groups will be able to commit themselves to the sustained development of comprehensive coupled models, but we recommend that there be more than one such group. However, in view of the larger community of climate scientists who could bring their expertise to bear in relevant aspects of climate modeling, we recommend that each high-end modeling group form a consortium with external researchers who would participate in the development, diagnosis and improvement of the high-end model. In addition to the performance of extended climate change simulations with progressively improved coupled models, we recognize the importance of the observation-based evaluation of such models and of the performance of research and experimentation on basic scientific questions related to them as supplements to already successful agency programs. We therefore recommend that ACPI support research within the modeling and research consortia that would lead to the improvement of our understanding of the roles played by critical physical, chemical and biological processes in climate change and that would facilitate their incorporation into high-end models. We also recognize the need for acceleration of the diagnosis of such models, and recommend that the consortia involve expertise in the numerical weather prediction, seasonal-to-interannual forecast, and paleoclimate communities to permit evaluation of the high-end models? performance on these time scales. It is foreseen that the consortia?s evolving models could become the de facto "core" coupled models of the national climate community.

For these reasons, we recommend that the ACPI computing resources be provided pri-marily to the consortia (although their membership may change from time to time), and that the consortia?s proposals for use of the resources be reviewed to ensure the appropriateness and readiness of the models, certification of the proposing consortium?s intention and ability to undertake sustained model development and evaluation and its intention to participate in the ACPI infrastructure (which should include documentation of the model, incorporation of an agreed level of code modularity as now being developed by the U.S. modeling community, and generation of an agreed set of model simulation statistics). Use of the ACPI Computing Facility would also commit users to permit open access to their models? codes and results, and to submit them to a designated center for quality control, dissemination and storage.

3.4 The ACPI Computing Facility

Although there are many technical decisions to be made in regard to the ACPI Computing Facility, we regard its location and management as important parts of ACPI implementation. We therefore recommend that the Facility be located at an existing high-performance computing facility that has experience in serving the climate modeling and broader user communities, so that full advantage may be taken of existing computing and networking structures. While the Facility need not be at an institution that leads one of the modeling and research consortia, its siting should support resident consulting staff and visitors in a research-oriented environment and facilitate interaction with both the ACPI consortia and regional climate centers.

3.5 The ACPI Regional Climate Centers

The primary role of the ACPI Regional Climate Centers is seen as the provision of data, diagnoses and specialized products from high-end model simulations for use in analyses and assessments of the regional impacts of climate change and variability, in cooperation with the local/regional user commmunity. We recommend, however, that the Centers also undertake research on regional climate modeling in cooperation with the ACPI modeling and research consortia and with other relevant groups, and undertake appropriate educational outreach and public relations activities. We further recommend that there be three to five such Centers, each located at an existing research institution with the capacity and interest to develop effective cooperation with the research, educational and user communities in designated parts of the U.S.

3.6 The ACPI Infrastructure

If ACPI resources are to be used effectively for the reduction of climate models? uncertainties in projecting regional climate changes, it is necessary that a national software and communications infrastructure be developed. This infrastructure should provide a significantly enhanced capability to store, access, transfer, diagnose and visualize the results of high-end climate model simulations, should provide an effective software and communications network linking the project?s components and participants, and should be developed as part of the IT² initiative (OSTP, 1999).

3.7 Coordination, Scheduling and Training

In view of the anticipated progressive increase in the performance of the ACPI computing systems (which are projected to reach 5 teraflops in 2001 and 40 teraflops in 2003), it is important that the modeling groups using these resouces be prepared to implement a computational strategy that is as efficient as possible for their model while retaining appropriate levels of code clarity and modularity. We therefore recommend that the participating groups? models be periodically benchmarked on the ACPI computers, and that provision be made to assist these groups in improving their codes. Since the sustained performance of the ACPI computers will effectively set the pace at which increasingly computer-intensive climate simulations can be made, we recommend that an appropriate set of ACPI milestones or program deliverables be developed and updated on a regular basis. We also recommend that steps be taken to increase the pool of scientists involved in climate modeling and its related computational and applications aspects, perhaps by the provision of ACPI fellowships to train a new generation of climate researchers.

3.8 International Cooperation

While ACPI is designed to accelerate the pace of advanced climate modeling in the U.S. in support of national policy needs, we recommend that it be undertaken in the spirit of international cooperation that has long characterized climate research. The inclusion of foreign research colleagues in the modeling and research consortia and in the diagnosis and validation of the models? results should threfore be encouraged, and appropriate coordination should be maintained with foreign climate modeling groups and other organizations that share the scientific and applications goals of ACPI. We also recommend that ACPI be seen as a significant enhancement of U.S. participation in international research programs such as the WCRP, IGBP and IHDP and in the IPCC assessments.

4. RECOMMENDED PRIORITY ACTIONS

Even though funding for ACPI is not yet assured, we recommend that the following priority actions be taken immediately by the sponsoring agencies:

It is also recognized that the ACPI project?s relative emphasis may change in response to scientific and technological advances, and that the program?s first year will be a learning experience with limited resources. Nevertheless, we recommend that the available funds be allocated in such a way that all elements of the program are at least started in FY00, with priority given to the establishment of the Computing Facility, the Climate Modeling and Research Consortia, and the Regional Climate Centers. Significant program enhancements are envisaged in subsequent years.

Acknowledgements

The Committee would like to thank the many members of the U.S. climate modeling, research and computations community who have contributed their views on ACPI and its implementation, and whose comments have helped to improve an earlier draft of this report. The Chairman of the Committee would also like to thank the directors and staff of COLA, GFDL, GISS, GSFC, IRI, LANL, LBNL, LLNL, NCAR and SIO for their hospitality during his recent visits to their institutions.

References