Questions? Call 800-624-6242
|
|
|
|
|
The following HTML text is provided to enhance online
readability. Many aspects of typography translate only awkwardly to HTML.
Please use the page image
as the authoritative form to ensure accuracy.
Page 7
Capacity of U.S. Climate
Modeling
Background
In October of 1995 four U.S. climate researchers raised concerns
in a letter (see Appendix A) to the USGCRP agency managers that the
U.S. position of leadership in the development, improvement, and
application of climate models had eroded. They offered various
options for progress in these areas, emphasizing a
well-coordinated, distributed national climate modeling program.
While the specific recommendations of that letter are somewhat
different than those of this report, the letter was one of the
primary impetuses for the Climate Research Committee (CRC) of the
National Research Council (NRC) and the USGCRP Program Office to
hold a jointly sponsored forum on the “Quality and
Infrastructure of Climate Modeling in the United States” on
11–12 June 1996 (see Appendix C, which contains the
invitation letter to the forum and the agenda). About 70 scientists
and federal program managers participated. In preparation for the
forum, the USGCRP Program Office distributed a questionnaire to
over 100 members of the climate-modeling community for their
written comments on “the strengths and weaknesses of the
present research and applications, related activities, and
infrastructure for decadal-centennial scale climate modeling in the
United States,” and received about 25 responses. Some of the
findings in this report are based in part on the discussions at the
forum and in answers to the questionnaire.
|
|
|
|
|
Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 7
Page 7
Capacity of U.S. Climate
Modeling
Background
In October of 1995 four U.S. climate researchers raised concerns
in a letter (see Appendix A) to the USGCRP agency managers that the
U.S. position of leadership in the development, improvement, and
application of climate models had eroded. They offered various
options for progress in these areas, emphasizing a
well-coordinated, distributed national climate modeling program.
While the specific recommendations of that letter are somewhat
different than those of this report, the letter was one of the
primary impetuses for the Climate Research Committee (CRC) of the
National Research Council (NRC) and the USGCRP Program Office to
hold a jointly sponsored forum on the “Quality and
Infrastructure of Climate Modeling in the United States” on
11–12 June 1996 (see Appendix C, which contains the
invitation letter to the forum and the agenda). About 70 scientists
and federal program managers participated. In preparation for the
forum, the USGCRP Program Office distributed a questionnaire to
over 100 members of the climate-modeling community for their
written comments on “the strengths and weaknesses of the
present research and applications, related activities, and
infrastructure for decadal-centennial scale climate modeling in the
United States,” and received about 25 responses. Some of the
findings in this report are based in part on the discussions at the
forum and in answers to the questionnaire.
OCR for page 8
Page 8
Policy Context.
The U.S. government has pending before it the ratification of
the Kyoto Protocol, an agreement to limit the emissions of
greenhouse gases (GHGs), which is largely based on the threat GHGs
pose to the global climate. Such an agreement would have
significant economic and national security implications, and
therefore any national policy decisions regarding this issue should
rely in part on the best possible suite of scenarios from climate
models.
This Protocol relating to GHGs is only one of a series of policy
issues under consideration that involve climate and climate change.
The Intergovernmental Panel on Climate Change (IPCC, 1998) lists
several major areas of concern where climate changes would have a
critical impact on policy decisions: ecosystems, hydrology and
water resources, food and fiber production, coastal systems, human
settlements, and human health. Governments, corporations, and the
public are faced with a multitude of decisions in each one of these
fundamental areas of concern. In terms of the daily lives of
individuals, these decisions impact on jobs, food, economic
well-being, livable environments, and general prosperity.
Issues bearing on the formulation of local, national, or global
climate change policies are complex, and not always well defined.
It could be just as disastrous to impose unnecessary restraints on
society out of ignorance as it is to fail to act in time from
indecision. Human, political, economic, and scientific
considerations all come into play. These considerations can be at
odds with each other in fundamental ways. Compromise rooted in
knowledge is essential if progress is to be made.
Policy makers must have solid, credible information to define
the issues, to generate realistic compromises, and to move the
policy debates and decision processes forward. It is essential to
provide the capability to produce well founded forecasts of the
magnitudes and trends in climate change, as well as to identify the
causative factors—natural and anthropogenic. At the core of
creating that capability is the
OCR for page 9
Page 9
use of climate system models. These models, which may
incorporate components including atmospheric, oceanic, and
terrestrial dynamics, radiative characteristics, and chemistry,
provide the only quantitative mechanism by which climate
projections can be put in a context suitable for policy assessment
and decisions.
Current Small and Intermediate
Modeling Capabilities
The U.S. climate modeling community excels in conceiving and
carrying out process and diagnostic studies that form the basis for
climate model improvement. Likewise, the intermediate climate
modeling research efforts, which a few years ago may have been
referred to as high-end (see footnote 1), have been appropriately
encouraged and supported. Evidence of this work are the coupled
modeling contributions by the United States to the IPCC process, as
well as the development of some of the leading mesoscale models
(e.g., MM5, ETA, and COAMPS).3
U.S. leadership in this area has involved, for example, sensitivity
studies, exploration of new hypotheses, and studies aimed at
quantifying and understanding model uncertainties.
The relatively successful forecasts of the regional climate
anomalies associated with the 1997–1998 ENSO event (COLA,
1998) and adaptations that were made in response to these
forecasts, highlight the societal value of climate
forecasting.4 The benefits that
are being experienced as a result of this capability, underscore
the potential utility of the development of long-term climate
change scenarios, in particular, because future, long-term,
anthropogenic
3 In
highlighting the U.S. intermediate-level modeling expertise, it
should not be overlooked that several foreign, intermediate
modeling efforts, such as those of the ECMWF in weather
forecasting, are at the cutting edge, sometimes leading those of
the United States.
4 The
computational requirements of these models, especially for
operational production of ensembles of simulations, are enormously
large, and the United States is at a competitive disadvantage, in
this regard, compared to the major European modeling centers
because of the greater access to appropriate computational
resources outside the United States.
OCR for page 10
Page 10
climate changes are likely to be larger than those that
currently occur on the seasonal-to-interannual time frame of the
ENSO phenomenon.
Current High-End Modeling Capabilities
and Needs
Some of the earliest and defining climate change simulations and
sensitivity experiments were carried out in the United States, and
the contributions from the U.S. modeling community were essential
to the overall understanding of various climate issues. That
initial productivity has been difficult to sustain because of a
lack of coordination and availability of the requisite
computational and human resources.5 This may explain in part why, in
contrast to some of the foreign modeling centers, U.S. modeling
centers have found it difficult to perform coupled atmosphere-ocean
climate change scenario simulations at the spatial resolutions
(e.g., finer than 500 km × 800 km) of direct relevance to
national policy actions presently being considered to mitigate
future global change. According to discussions at the forum, at
least some in the scientific community expressed the concern that
the United States should have been able to contribute more in terms
of high-end, coupled-model GHG and aerosol simulations to the
recent IPCC assessment.
The computational capabilities that are required to incorporate
various spatial resolutions in climate models are outlined in Table
1. It is apparent from Table 1 that several foreign modeling
centers currently possess greater computing power than that of the
U.S. centers. However, no modeling center currently has the
computational ability to realistically depict
small-scale/high-impact atmospheric processes in multi-century,
transient simulations — the type of simulation required to
reduce uncertainties associated with assessments of the societal
implications of climate change. Simulation of certain atmospheric
features such as mesoscale convective complexes and hurricanes,
even in a rudimentary fashion, requires a model spatial resolution
of 10 km or less, which, in turn, requires
5 The view,
repeatedly expressed at the forum, was that U.S. climate modeling
research is at the forefront in most respects, but not all.
OCR for page 11
Page 11
Table 1. Computational capabilities at some
of the major climate modeling centers
Date
Comput. SpeedA (GFLOPS)
SystemB /
No. of Processors
Approx. Run TimeC
Horiz. Resolution (Km2)D /
Vertical Levels Atmosphere Ocean
Modeling Group
1998
1
Cray J90 / 16
3 days
500×830 / 18
_E
e.g., PSU
1998
2.6
Origin 2000 / 64 (four)
3 days
310×310 / 18
65×65 / 20
DOE
1998
5
Cray C90 / 16
2.5 days
310×310 / 18
>130×270 / 45
NCAR
1998
15
Cray T932 / 26
5 hours
250×420 / 14
190×80 / 18
GFDL
1998
35
Cray T3E / 696
**F
280×420 / 19
140×140 / 20
HC
1998
20–25
NEC SX4 / 32
1.5 days
310×310 / 17
220 × 55 / 25
ABOM
1998
20–25
NEC SX4 / 32
3 days
280×280 / 32
200×200 / 29
CCCMA
1998
75
Fujitsu VPP / 116
14 days
200×200 / 31
56×56 / 20
ECMWFG
2000
1000
Currently unspecified
8 hours
140×140 / 18
30×30 / 20
ACPI
2001
10000
Currently unspecified
8 hours
70×70 / 18
15×15 / 20
ACPI
2003
40000
Currently unspecified
8 hours
30×30 / 18
9×9 / 20
ACPI
PSU = Penn State Earth System Science Center; NCAR
= National Center for Atmospheric Research; GFDL = Geophysical
Fluid Dynamics Laboratory; HC = Hadley Center, U.K.; ABOM =
Australian Bureau of Meteorology Research Centre; CCCMA = Canadian
Center for Climate Modeling and Analysis; ECMWF = European Center
for Medium-Range Weather Forecasting; DOE = Department of Energy,
Los Alamos; ACPI = Accelerated Climate Prediction Initiative
Individuals who supplied information for this
table include David Anderson, David Bader, Bill Buzbee, Robert
Malone, William Peterson, Ronald Stouffer, Vince Wayland, Francis
Zwiers, and members of the CRC.
A GFLOPS =
109 floating point operations per
second (typical sustained)
B The highest
performance machine at each institution.
C Run time is a
rough approximation of the wallclock time needed to simulate 15
model years on a dedicated computer, assuming optimal
multi-tasking (e.g., running as many separate simulations as
can be accommodated by all of the machine's processors and dividing
the final wallclock time by the number of simulations).
E—= This
climate system component is not included in this particular type of
model run.
F **= Not
available
G The ECMWF model
is designed for operational forecasting, not multi-decadal climate
scenario analysis, like the other models in this table. It is
included to illustrate the large computational capability that has
been devoted to this activity abroad.
OCR for page 12
Page 12
computational throughput more than three orders of magnitude
greater than is presently available to U.S. climate modelers. This
is at the upper range of the 40 teraflop (1012 floating point operations per second)
capability proposed by the Advanced Climate Prediction Initiative
(ACPI, 1998) for the year 2003.6
Current model deficiencies are not only in the realm of spatial,
but also temporal resolution. For example, the current GFDL coupled
model is generally run without a diurnal cycle, thereby precluding
the ability to explicitly resolve critical variables such as daily
maximum and minimum temperatures. The United States possesses the
intellectual ability to put together models capable of better
resolving many of these important climatic features. However, the
current lack of coordinated computational capability limits the
ability of U.S. scientists to develop such policy-relevant
scenarios; it also limits the ability of U.S. scientists to
diagnose and understand the physics of climate and climate
change.
Currently, there are relatively few modeling centers anywhere in
the world capable of producing relatively high-resolution (e.g.,
250–300 km grid spacing), transient climate simulations. The
differences in simulated climate produced by the various structures
and compositions of these few models help to bound the range of
outcomes that the climate system might produce given a certain
forcing scenario. Thus, the state of climate modeling throughout
the world is such that the addition or removal of even a single
model could affect the confidence levels assigned to certain
scenarios of future climate change. In other words, not only would
the United States benefit from enhancements in its modeling
capabilities but the international community would benefit as
well.
In addition to the need for simulations produced by different
climate models, estimates also need to be produced of the
stochastic nature of climate change within a given model, i.e.,
ensembles of simulations, using slightly different initial
conditions. Ideally, a new ensemble should be produced whenever
significant improvements in a
6 The ACPI is
an unfunded Department of Energy plan for increasing U.S. computing
power for climate applications. The system and number of processors
are currently unspecified.
OCR for page 13
Page 13
model's code become available. The production of such ensembles
to perform model diagnostics and climate change assessments in a
timely fashion requires high computational throughput. Further
increasing computational requirements is the use of regional,
high-resolution models nested within medium-to-coarse resolution,
global coupled models. U.S. climate modeling centers currently do
not possess the computational resources required for these types of
simulations.
While trying to catch up with foreign, high-end modeling
efforts, the necessity of adequate model testing should not be
overlooked. Testing, diagnosis, and documentation of model
characteristics must be an intrinsic part of the procedure for
developing climate change scenarios for assessment purposes. Again,
the computational and human resources for this facet are
substantial and are not sufficiently available to U.S. climate
modelers.
Access to Foreign Model Output
A further hindrance to the sub-optimal high-end U.S. modeling
capabilities is that the United States is not assured full, open,
and timely access to output from foreign models. Ready access to
foreign model output could alleviate some of the need for high-end
domestic capabilities. It is acknowledged that in many instances
the output from foreign models is readily forthcoming, such as in
the case of the Atmospheric and Coupled Model Intercomparison
Projects (AMIP/CMIP), the LINK project at the U.K. Climatic
Research Unit, as well as countless other formal and informal data
transmittals. However, the committee is concerned that access to
foreign model output is not guaranteed. In several recent
instances, access to foreign atmospheric data has either been
denied or only occurred at considerable expense to U.S. entities
(see Appendix D). As the commercial value of these data becomes
more apparent, the possibility of greater restrictions exists,
particularly with the movement towards privatization of
meteorological agencies.
This issue of data accessibility is important for at least two
reasons, and ultimately points towards the need for a high level
of
OCR for page 14
Page 14
domestic Earth system modeling capability. First, if political
decisions are to be based upon the most current and reliable
information, access to those data must be ensured. The NRC Pathways
report (NRC, 1998b) addressed this in its statement that
“the USGCRP must foster the development and application of
models at the scale of time and space needed to understand and
project the specific mechanisms controlling changes in the state of
the Earth system thus providing the information required to support
important policy processes.” Second, because the
commercial value of data is in part a function of its timeliness,
potential commercial opportunities may be lost if prompt access to
climate model data is not guaranteed. These needs can be met by
furthering the development, running, and testing of high-end models
within the United States.
A further point is that the more difficult it is to access a
model and its output, the more opaque are the model's results. If
the United States is to fully capitalize upon the most recent model
products, it must have researchers directly involved in the
modeling process who understand the details of a given model's
underpinnings so that they can be in a position to comprehend and
interpret nuances of that model's simulations. If U.S. scientists
are not directly involved in the high-end modeling itself, they may
miss opportunities to gain valuable insights into the underlying
processes that are critical to subsequent modeling investigations.
In this regard the issue of accessibility is much more than just a
commercial and political issue; in order to most effectively
advance the science in the United States, researchers need to have
access to both model output and the models to iteratively diagnose
the output, advance our knowledge of climate, and improve the
model's predictive capabilities. To some extent this can be
addressed by enhancing collaborations with foreign climate modeling
groups, but ultimately this can go only so far.
The concerns regarding the need for U.S. high-end modeling
capabilities are also based, in part, on the possibility that
decisions that might substantially affect the U.S. economy might be
founded upon considerations of simulations produced by countries
with different priorities than those of the United States. While
the leading climate models are global in scale, their ability to
represent small-scale,
OCR for page 15
Page 15
regionally dependent processes (e.g., hurricanes and extreme
flooding events) can currently only be depicted in them using
high-resolution, nested grids. It is reasonable to assume that
foreign modeling centers will implement such nested grids to most
realistically simulate processes on domains over their respective
countries which may not focus on or even include the United
States.
Priority Setting
The information gathered from active climate researchers and
agency program managers at the CRC/USGCRP modeling forum indicated
that climate modeling priorities are established primarily within
individual agencies, specifically, DOE, NASA, NOAA, and NSF.
Individual agency program managers appear to be aware of modeling
activities in other agencies through informal personal exchanges of
information and through the USGCRP Integrated Modeling and
Prediction Working Group (IMAP) (No analogous coordinating activity
involving the directors of U.S. climate modeling centers exists.).
Although these limited harmonization efforts may provide some
context for setting funding priorities, we conclude that research
funding decisions are mainly driven by the missions of the
individual agencies without strong interagency coordination.
U.S. funding agencies rely heavily on working scientists to
shape the climate modeling program. This system promotes a healthy
competition among modeling groups and has given rise to a rich
diversity of climate modeling efforts that is highly valued by the
scientific community. This system, however, does not necessarily
promote research that addresses the questions of most importance to
policy makers or U.S. society at large, particularly if no agency
considers a given issue to be among its priorities.
The approach that has been used in the United States to set
research priorities for climate modeling can be contrasted with
some European countries, especially the United Kingdom and Germany,
where a stronger top-down management approach is used for setting
research priorities. These countries, which have smaller GDPs
than
OCR for page 16
Page 16
the United States, have leveraged their funding of research to
more directly serve policy needs (at the partial expense of
fundamental climate research). To facilitate progress, the United
States should establish a set of priorities that carefully balances
both policy and science needs and avoids a top-down prioritization
of research activities driven by short-term agency agendas that
might ultimately dissipate scientific resources.
Coordination
The lack of national coordination and funding, and thus
sustained interest, are substantial reasons why the United States
is no longer in the lead in high-end climate modeling.7 Many scientists at the time of the
forum believed that the current major U.S. modeling centers were
not adequately responding to the challenges of integrating
component models of the atmosphere, oceans, land surface, and
atmospheric chemistry, that are needed for climate change scenario
studies. At that time, some members of the academic community
averred that their expertise was not being effectively utilized in
the development of these comprehensive models. Moreover, the
coordination of model development activities seemed to be
fragmented even within the major climate modeling centers.
The USGCRP could assume increased responsibility for
identifying, from an interagency perspective, any gaps or
imbalances in the research priorities established by the individual
agencies. The USGCRP is, however, limited in this area because some
agencies have excluded from their USGCRP budgets the computational
and human resources allocated to support some major U.S. climate
modeling efforts. Although ideally any imbalances identified should
be rectified, it appears that the USGCRP does not currently have
the means to ensure this. Thus, this may be a fundamental weakness
of the current approach to setting climate modeling priorities.
7 This view is
supported by Finding 6 of NRC (1998b): “Advances in
developing and most importantly in testing and evaluating models
are needed. The United States is no longer in the lead in
this critical field.”
OCR for page 17
Page 17
As in setting priorities, the establishment of a coordinated
modeling strategy in the United States should carefully balance
both policy and science needs. The implementation of these
priorities should not come at the expense of small and intermediate
modeling, which currently form a solid base of expertise in the
United States. Although better coordination of U.S. climate
modeling activities is advocated and is likely to lead to
substantial enhancements in overall capabilities, coordination
alone is not sufficient. U.S. modelers cannot produce the
high-resolution, multi-decade, ensemble simulations necessary to
perform detailed assessments of anthropogenic climate change
without an increase in the computational capability available to
U.S. scientists.
With the development of coupled models, including the
atmosphere, oceans, and biosphere, there are common problems that
must be addressed by the overall model framework that links the
components of the system. It is possible that movement towards
increased modularity among model components and a common component
interface, sometimes referred to as a “flux coupler,”
might speed improvement of these comprehensive climate models. In
principle, independently developed individual atmosphere or ocean
model components could be interchanged in different combinations to
generate various coupled models that could be assembled for
particular applications. Current coupled model development,
however, is proceeding independently at various laboratories and
universities with little coordination.
Many field programs are justified in part by arguments that
their efforts will lead to model improvements. It is not easy,
however, to fashion new parameterizations from even the most
carefully designed field programs and it is also not certain that
even well-conceived new parameterizations will lead to overall
model improvement. Often the difficult task of developing new
parameterizations is carried out by small, academia-based research
groups. The subsequent labor-intensive testing of the
parameterizations in climate models requires a tailoring of code so
that it complies with the unique requirements of each host model.
This work could again be
OCR for page 19
Page 19
Specialists can efficiently analyze the output from several
different models if standard simulations are performed and if
output is archived in a standard format. Agency program managers
may not need to modify their current strategies much because
international projects such as AMIP and CMIP are already fostering
this and attracting considerable participation and cooperation.
These projects, which should be encouraged, serve to coordinate the
efforts of the broader community of climate scientists evaluating
climate models and make it possible for smaller groups to submit
their models for closer and more comprehensive scrutiny than their
own resources would permit. In principle, a single expert in a
particular area can evaluate the performance of all models in the
intercomparison projects with little more effort than it would take
to evaluate a single model. The efficiency of this approach is
becoming more evident as these projects mature. To facilitate the
intercomparison of separate model analyses, the development of
compatible diagnostic algorithms should be encouraged.
For intercomparison efforts to be successful, the models being
analyzed must use the same initial and boundary conditions so that
it is differences in the representation of model physics that are
being assessed, not differences in the forcings. At present, there
is no uniform set of land-surface data for use as boundary
conditions in the climate models of the major U.S. modeling
centers. Consistency in this regard can be maintained through
enhanced coordination of U.S. high-end modeling efforts among these
centers.
One way to establish the credibility of the climate models used
for making climate assessments is to test their performance when
run in a weather- or short-term climate-forecast mode. However, in
the United States there are serious impediments to cooperation
between the operational forecast facility (NCEP) and the climate
modeling community. Operational forecasting facilities in the
United States currently provide little support for activities other
than operational ones, and therefore, inadequate human resources
for collaboration with external climate modeling groups or
individuals. An increase in collaborative opportunities between
these entities could be beneficial to both.
OCR for page 20
Page 20
Some comments were also received from members of the academic
community who are eager to have early access to output from some of
the computationally most demanding climate model simulations. The
potential scientific benefits resulting from broad participation in
the analysis of these simulations must be balanced by a recognition
that the development of the models used to generate these
simulations can take years, and the scientists who have developed
these models deserve to reap the first rewards of their efforts in
terms of publishable research.
Allocating Resources
Productive climate modeling efforts require an appropriate
division of resources between support for personnel (including both
climate and computer scientists) and computer facilities. Also
essential is access to the results of process studies that lead to
improved model formulations, and the collection and analysis of
observational data for use in evaluating models, but these needs
are outside of the scope of this report and will not be considered
here.
Although the total computer resources available to the U.S.
climate modeling community are substantial, inadequate access to
the world's most powerful mainframes by U.S. modelers in
universities and the national centers is significantly limiting
progress. This view is supported by Table 1 and its accompanying
text and by evidence put together by Dr. Bill Buzbee, the director
of NCAR's Scientific Computing Division (see Appendix E). Buzbee
has compared NCAR capabilities to those at GFDL and six research
labs around the world. He has shown that “international
colleagues now enjoy a substantial computational advantage over
U.S. modelers.” This view is further buttressed by the USGCRP
National Assessment program's current reliance on climate change
scenarios developed by foreign modeling groups and by recent
special arrangements to use computers at foreign institutions in
order to produce complementary simulations for the National
Assessment with the NCAR climate system model.
OCR for page 21
Page 21
The ability of the climate community to acquire state-of-the-art
mainframes is severely hampered by a Department of Commerce
“antidumping order” prescribing a financial penalty in
excess of 400 percent on the purchase price of the world's most
powerful commercial supercomputers, which are Japanese in origin.
The climate community has not been provided with the financial or
computational resources to overcome this barrier and has,
therefore, been unable to fully capitalize on the scientific
potential within the United States. In effect, if total resource
availability remains fixed, the cost of any increase in access to
fast machines could require a comparable reduction in the
scientific and technical personnel who develop, test, and apply
models.
In allocating resources for climate modeling, agency managers
should understand that there is an inherent pyramid structure in
climate research. The broad base of understanding that is required
in constructing climate models is obtained through a multitude of
observational programs and individual process studies. The small
and intermediate modeling efforts (usually involving a single
component of the climate system) incorporate the relevant portions
of the underlying research both during model development and in
model applications. The most sophisticated high-end models are
essentially built by integrating the various climate system
components. U.S. agencies spend most of their resources on small
and intermediate modeling. The results of this work are published
in journals and are therefore freely available to the climate
modeling community.
The full benefits of investing in the foundations of the pyramid
can not be realized without sufficient support for the high-end
modeling needed for impacts and policy purposes. A non-trivial
element of a comprehensive high-end modeling system is the
dissemination of the output from these models to the wider climate
research and user communities — something that is largely
unfunded in the United States.
In Europe a relatively high priority has been given to funding
research at the top of the pyramid, which relies in part on the
fundamental research carried out in the United States. There
is,
OCR for page 22
Page 22
unfortunately, a certain nonreciprocity in this arrangement
because the results of the U.S.-funded research can feed directly
into the high-end modeling efforts in Europe, but, as mentioned
previously, the United States does not necessarily have full, open,
and timely access to output from foreign models. Solutions to this
problem should not involve the imposition of access restrictions to
U.S. data.
Simply acquiring hardware alone is not sufficient. We also need
to invest in the development of scientific expertise and the
dissemination of that knowledge. Conversely, while the United
States is an intellectual leader in this field, it needs the
hardware to make effective use of this intellectual capability.
Thus, one of the fundamental reasons for considering additional
investment in U.S. high-end modeling infrastructure is that the
incremental returns on investment could be very large, as the added
effort in high-end modeling could be encouraged to interact with
the existing vast U.S. expertise in small/intermediate modeling.
The synergism of interaction could be expected to yield
substantially more than the sum of the individual modeling
efforts.
Recent Developments Relevant to this
Report: Computational Capabilities and Coordination
Since the climate modeling forum two years ago, there have been
indications of several significant developments and changes in the
U.S. climate modeling effort. Among these are the following:
• An NCAR proposal to acquire an NEC SX-4 computer was
effectively denied by a Department of Commerce “antidumping
order” prescribing a 400 percent financial penalty. This has
precluded certain applications of the NCAR Climate System Model
(CSM) and slowed its use in studies requiring multi-century
simulations and may also retard the production of regionally
resolved climate scenarios for the USGCRP National Assessment. This
lack of routine access by climate researchers to the world's most
powerful computers has become a quite serious problem that
increasingly affects the international competitiveness of the
U.S.
OCR for page 23
Page 23
climate modeling community. This precedent-setting decision
seems to have discouraged other institutions within the United
States from considering the purchase of foreign computers, even
though these computers might prove superior in climate model
applications.
• Responding to encouragement to interface more effectively
with the outside community, NCAR has made notable changes, as
evidenced by their annual Climate System Model (CSM) Workshop and
their CSM working group meetings, with heavy involvement of the
academic community. Furthermore, the publication of a series of
papers describing results from the NCAR CSM in a special journal
issue (J. Climate 11[6], 1998) is a good sign that
U.S. coupled atmosphere-ocean modeling efforts are progressing.
• A new DOE initiative (ACPI, 1998) is under development,
which, if funded, will attempt to increase by a few orders of
magnitude the amount of computing power available for climate
modeling applications. Under this initiative, massively parallel
computing machines, currently under development for other DOE
purposes, would be applied to climate modeling, including the
production of multi-century, high-resolution simulation
ensembles.
• At the request of USGCRP and with support from NSF, NCAR
has agreed to develop some climate change scenario runs with the
CSM for the USGCRP National Assessment. Some of these runs are
being completed in Japan and Australia because of the current
scarcity in the United States of the kind of computing resources
needed for this type of model.
• A sharply upgraded version of the GFDL MOM3 ocean model
has recently been released to its worldwide user community, which
includes most of the major climate modeling centers.
• Two workshops were recently held at the Goddard Institute
for Space Studies (GISS) — one on ocean modeling and the other
on land surface modeling — to encourage involvement of the
academic community in the GISS modeling effort.
• In August 1998, a workshop was held at NCEP, sponsored
jointly by NCEP and NSF, to investigate the question of whether a
common modeling infrastructure could enhance the degree of
collaboration between NCEP and the weather and climate research
OCR for page 24
Page 24
communities in the United States.
• An ad hoc working group, currently chaired by Steven
Zebiak and Robert Dickinson, has been formed in response to one of
the recommendations of the NCEP/NSF meeting that a group is needed
to formulate coding and data standards to facilitate exchanges of
data and promotion of interactions between modeling groups and the
academic community. At a first meeting in Tucson in October 1998,
initial steps were taken to develop standards for model
parameterizations and approaches for facilitating involvement of a
wider community in their development. Steps were also taken to
develop agreements on standard data formats. In addition,
suggestions were presented as to how standard data sets for model
boundary conditions should be developed.
With respect to this report, these developments, among others,
indicate that the climate modeling effort is evolving. In some
cases (e.g., the denied purchase of an SX-4), a problem identified
by the climate modeling community has been exacerbated, but in
another, (e.g., the potential increase in computational resources
through the DOE initiative), there are indications that U.S.
computational capabilities could dramatically increase sometime in
the future. Emerging efforts to encourage more collaboration across
institutional barriers is promising. The U.S. climate modeling
community will likely remain behind the rest of the world in terms
of computational facilities for the next several years.
Nevertheless, the United States can maintain many aspects of
scientific leadership through its major satellite-based climate
observing system and fostering of research and development in
climate processes. Unfortunately, the U.S. community may, to some
extent, have to be content to see these advances implemented in
foreign models.
Conclusions.
Through our analysis of the discussions at the climate modeling
forum, responses to the USGCRP questionnaire, personal contacts
with the climate modeling community, and deliberations within the
CRC, we have reached initial conclusions in our evaluation
OCR for page 25
Page 25
of the organization and infrastructure of climate modeling in
the United States. These conclusions are reported in the context of
the three questions referred to the CRC by the USGCRP program
managers.
1. Do USGCRP agencies have a coordinated approach for
prioritizing from a national perspective their climate modeling
research and assessment efforts?
We find that:
• USGCRP agencies do not have a coordinated
approach.
Climate modeling priorities within the USGCRP are primarily
established by individual agencies with substantial input to each
agency from climate researchers, but with little formal
inter-agency coordination. There is no effective integrating
national strategy and little formal consideration of the needs of
the policy community.
2. Are resources allocated effectively to address such
priorities?
We find that:
• There are few monetary resources dedicated to high-end
climate modeling. Further, there is insufficient access to
computers powerful enough to take advantage of the U.S.
intellectual capability to design and run the climate models needed
to answer critical science and policy questions. In addition, there
is no coordinated mechanism for establishing the priority of these
questions. The lack of resources and the inefficient assignment
of those that are available are hampering progress, both in
theoretical understanding of climate and in performing simulations
of direct relevance to policy decisions concerning natural and
anthropogenic climate variability and change.
3. How can the U.S. climate modeling community make more
efficient use of its available resources?
OCR for page 26
Page 26
We find that:
• First, a national set of goals and objectives that are
agreed to by the USGCRP agencies is essential. These goals and
objectives would be aimed at establishing the major themes for
climate research, and would be set in the context of scientific
priorities and national policy decisions. By formulating these
goals and objectives, the USGCRP agencies should also agree to seek
coordinated funding initiatives directed at achieving them.
• Second, a concerted effort by the relevant agencies is
needed to establish a coordinated national strategy for climate
modeling. That strategy may call for the allocation of
resources to be distributed at a number of locations, and it must
have the capability to deal with the complexity of high-end climate
modeling. A number of formats exist for the establishment of such a
resource. Appendix A identifies several possibilities, and there
likely are others. What must be encouraged in such an endeavor is
the increased coordination and integration of activities between
national laboratories and universities. What must be avoided is a
top-down prioritization of research activities driven by short-term
agency agendas.
The current approach to climate modeling in the United States
produces a rich diversity of research driven by individual
researchers. The purpose has to be to focus that research, not
subject it to the “problem of the month,” which
ultimately will dissipate scientific resources. While difficult to
specify a priori, a carefully considered balance must be struck
between policy- and science-driven research. Several examples of
how the recommended coordination might be manifested are given in
the body of this report.
• Third, in order to optimally use existing scientific
capabilities, adequate resources, including adequate supercomputing
capabilities, need to be provided to the
OCR for page 27
Page 27
climate modeling community.
At present, the U.S. modeling community on the whole is not
supported to produce climate change scenarios for the GHG-driven
climate change assessments, such as IPCC and the USGCRP National
Assessment. This is in part because of the limited funding for
these activities and in part because of the inability of U.S.
climate modeling centers to acquire state-of-the-art
supercomputers. U.S. scientists do participate to the extent
possible in climate change assessment activities by reprogramming
resources within their limited budgets. Participation in these
activities is of necessity on a volunteer, often uncoordinated and
normally aperiodic basis. Unfortunately, standard tenure-track
systems, which emphasize frequent, first-authored publications, do
not always reward such participation. Longer-term research, which
may require years of effort to achieve results, in fact, is
penalized in the race to produce early papers.
The provision of financial resources should be based upon
peer-reviewed proposals that advance the main themes of the agreed
upon science and policy objectives. Those resources need to be
committed for periods commensurate with the time required to
achieve definitive results.
We agree with certain aspects of the discussions on funding in
Appendix A. Merely adding funds to budgets is not effective. We
agree with other statements in Appendix A that “the option of
accelerating progress by simply adding funding will fail without
also making major changes in the management and institutional
cultures of existing centers.”
• Fourth, the reliance of the United States upon other
countries for high-end climate modeling must be redressed. This
issue is rooted in both science and policy, but can only be
resolved through governmental intervention. The European modeling
centers, for example, benefit from pioneering research by the
United States in intermediate climate
OCR for page 28
Page 28
modeling; the CRC strongly supports this element of free and
open exchange of climate data. Unfortunately, the United States is
not guaranteed equivalent access to European model output or
methodologies, although, at present, access to transient climate
change simulations from foreign models is generally available to
the U.S. research community. The concern over data access is due in
large part to European national policies that restrict the free and
open exchange of some information in favor of enhancing national
commercial advantages (see Appendix D).
Until the gap in climate modeling capabilities between the
United States and other countries is closed, decisions that could
substantially affect the U.S. economy might be based upon
interpretations of simulations (e.g., nested-grid runs) produced by
countries with different priorities than those of the United
States.
There is real concern that if U.S. scientists lose involvement
in advanced modeling activities, they will miss opportunities to
gain valuable insights into the underlying processes that are
critical to subsequent modeling investigations Further, the state
of climate modeling throughout the world is such that the addition
or removal of even a single model would affect the confidence
levels assigned to certain scenarios of future climate change. In
other words, not only would the United States benefit from
enhancements in its modeling capabilities, the international
community would benefit from these efforts as well.
In essence, the CRC finds that the United States lags behind
other countries in its ability to model long-term climate
change. What computational and intellectual capability it does
possess is neither well focused nor well financed. Those
deficiencies have a significant and negative impact on the United
States:
1. to predict future climate states and thus:
a) assess the national and international value and impact of
OCR for page 29
Page 29
climate change;
b) formulate policies that will be consistent with national
objectives and be compatible with global commitments;
2. to most effectively advance understanding of the underlying
scientific issues pertaining to climate variability and change.
Thus, to facilitate future climate assessments, climate treaty
negotiations, and our understanding and predictions of climate, it
is appropriate to develop now a national climate modeling
strategy that includes the provision of adequate computational and
human resources and that is integrated across agencies.
Representative terms from entire chapter:
modeling centers
