Strategic Guidance for the National Science Foundation's Support of the Atmospheric Sciences (2007)

The committee was charged by the National Science Foundation’s (NSF’s) Atmospheric Science Division (ATM) to assess the balance among the types of activities and modes of support and to make recommendations as to how the balance might need adjustments to ensure the health of the atmospheric sciences into the future. In its interim report the committee recommended that ATM should continue to utilize the current mix of modes of support for a diverse portfolio of activities (i.e., research, observations and facilities, technology development, education, outreach, and applications) (NRC, 2005e). Thus, the committee concluded that the types of activities and the modes of support were appropriate and now addresses the further question of whether the balance among activities or modes should be adjusted.

The committee devoted considerable thought to the appropriate methodology for dealing with this “balance question” in the context of strategic guidance. It decided that the most useful approach would be to create a list of major research accomplishments in the atmospheric sciences, supported at least in part by NSF, and then analyze the role of ATM’s modes and activities. The balance would be judged to be in need of adjustment if various modes or activities had in fact not been crucial in achieving any of these major research accomplishments. Conversely, finding the various modes of support and activities to be well represented among the major achievements of the past decades is good evidence that having a diversity of modes and activities has been a successful strategy. That is, it would show that the mix

of modes and activities has contributed to major advances and it would provide evidence that the balance has been adjusted to accommodate new opportunities and needs. Given that the NSF has multiple mechanisms for assuring that the processes for granting awards are functioning properly, the committee believes that the present task can be addressed by focusing solely on the major scientific results of ATM’s programs.

Statistical criteria are often used when judging certain aspects of scientific quality. For example, the number of highly cited papers by field would be a possible approach to identifying the relative effectiveness of fields or modes. This form of measurement is often applied to the contributions of individual Principal Investigators (PIs). However, the committee concluded that such statistical measures are both too imprecise and too beset with complications and biases to be useful for our purposes. They are, moreover, not the type of measurements that are appropriate for other modes of support for the atmospheric sciences.

The highly significant accomplishments selected by the committee are shown in Table 2-1 (in no particular order). It is important to note that this list of major achievements and the selection of case studies was made without prior examination or consideration of the roles the modes played in each of the achievements. While this list is not exhaustive, the committee believes that enlarging the set of major achievements would not change our conclusions regarding the adjustment to the balance between modes and activities. The committee selected case studies from all disciplines within NSF’s ATM division. While advances in understanding of climate variability and change are certainly among the most significant accomplishments of the past few decades and a few of the case studies cover aspects of climate

science, the broad scope of climate science did not lend itself to a case study. The committee refers the reader to the comprehensive assessments of the Intergovernmental Panel on Climate Change (e.g., IPCC, 2001). Some of the case studies focus on advances in tools, while others emphasize breakthroughs in knowledge and understanding.

The committee notes also that certain significant achievements in atmospheric sciences lend themselves to quantitative assessment; that is, objective measures of progress over time are available. Quantitative improvement measures are immediately apparent for items 1, 2, 4, and 5 in Table 2-1. Such metrics have been treated extensively in the context of mission-oriented programs, such as the U.S. Climate Change Science Program, in the recent National Research Council (NRC) report Thinking Strategically: The Appropriate Use of Metrics for the Climate Change Science Program (NRC, 2005d). However, the committee does not believe it is appropriate or possible to expect all of ATM’s major research accomplishments to fit that model of quantitative assessment. In part, quantitative assessment of research is problematic because some of these accomplishments were not planned and therefore do not fit the goal-driven model outlined in this earlier NRC report. Moreover, the type of quantitative measures appropriate for some of the modes is certainly not the appropriate measure for others (e.g., progress in severe weather forecasting vs. productivity and effectiveness of individual PI grants). This raises the issue of how to objectively compare incommensurate measures, a kind of “apples and oranges” problem. This dilemma is another strong reason the committee chose to focus on major accomplishments in addressing the balance issue.

In what follows, summaries are provided of what the committee believes are among the most important research results of the past several decades. These summaries are then analyzed for the ways in which NSF ATM’s modes and activities contributed to these achievements, and how, in doing so, they occasionally adjusted the balance between modes and activities. It is clear from our analysis of these case studies that NSF ATM has made effective use of its varied modes of support and that the balance between the modes has evolved over time in response to the needs and opportunities of the field. This chapter also includes many testimonials written by some key participants describing in more personal terms how the achievements were made possible by federal agency or private-sector support. It is important to note that the tenacity and dedication of the investigators, whatever the role of NSF support, was an integral factor in many of the research achievements described. Note that this list is not intended to be exhaustive, but the committee believes it is appropriate to the purpose of this chapter.

It is difficult to trace the one seed that began research conducted by the National Oceanic and Atmospheric Administration’s (NOAA’s) National Severe Storms Laboratory (NSSL) and the nearby School of Meteorology at the University of Oklahoma (OU), other universities, and the National Center for Atmospheric Research (NCAR), that has led to dramatic improvements in severe-weather forecasting during the last decade or two (Doswell et al., 1993). Using conventional radar and aircraft, NSSL conducted studies of severe convective storms in the 1960s (Bluestein, 1999a). These studies built upon the Thunderstorm Project conducted in the 1940s (Byers and Braham, 1949); further contributions by Chester Newton, Ted Fujita, and Keith Browning in the 1950s and 1960s at the University of Chicago and Air Force Cambridge Research Laboratories (AFCRL) (e.g., Fujita, 1963; Newton, 1963); the Alberta Hail Studies Project; the National Hail Research Experiment; and radar development by Roger Lhermitte, Dave Atlas, Rod Rogers, Alan Bemis, Pauline Austin, and J. Stewart Marshall in France, at AFCRL, Cornell, MIT, and McGill, among others, at the aforementioned institutions and elsewhere. The advent of meteorological Doppler radar in the late 1960s and the development and use of dual-Doppler analysis techniques in the 1970s at NSSL and NCAR provided the most significant leap in the ability to observe the behavior and internal structure of supercells and other convective storms (Davies-Jones et al., 2001).

Equally important as the developments in Doppler radar, the concurrent development of three-dimensional numerical cloud models in the mid 1970s at several universities provided the potential to study the dynamics of severe convective storms by performing controlled numerical experiments. As an example, collaboration between the University of Illinois at Urbana-Champaign and NCAR led to the development of the workhorse “Klemp-Wilhelmson” nonhydrostatic cloud model, which was used for two decades (Wilhelmson and Wicker, 2001). Advances in the capabilities of computers, particularly supercomputers at NCAR, permitted the model to be used for severe-storm research. Pioneering work at NCAR in the early to mid 1980s identified quantitatively the basic environmental parameters supportive of supercells, the most prolific producers of severe weather. At NCAR, expertise was at hand also to physically interpret the mechanisms responsible for supercell formation and behavior (Klemp, 1987). The roles of environmental vertical wind shear and potential thermal buoyancy in producing storm rotation and propagation were elucidated. In the late 1980s and 1990s scientists at NCAR and at universities also investigated

the behavior of groups of convective storms, mesoscale convective systems, such as squall lines, again using controlled numerical experiments (e.g., Box 2-1; Rotunno et al., 1988; Weisman and Davis, 1998). The roles of low-level vertical shear and an evaporatively produced cold pool of air in controlling storm structure and evolution were described. The results of these experiments led to an increased awareness and understanding of the conditions leading to damaging, straight-line surface winds.

Many field experiments have been conducted (Table 2-2), in large part with support from NSF, both for field operations and for development of new instrumentation (Figure 2-1). In the early 1970s, storm-intercept field programs began at NSSL and OU, with funding initially from NOAA (Bluestein, 1999b). Early collaborative annual spring field programs led to a conceptual model of supercells used by spotters and nowcasters, and in situ verification of severe weather events that eventually instigated the development of a national network of Doppler radars (NEXRAD) and its implementation in the 1990s. After the radars became operational, the accuracy and lead time of short-term (< 1 h) severe-weather warnings improved greatly.

The object of some of these experiments was to study the details of tornado development and other severe-storm features; the object of others was to further understanding of convective storms in general. In many instances, NOAA provided partial or seed support. Quantitative studies in tornadoes began with the Totable Tornado Observatory (TOTO), built by NOAA, in the early 1980s. Pressure falls associated with mesocyclones and thermal aspects of the rear-flank downdraft were documented. The first portable Doppler radar, developed at the Los Alamos National Laboratory, was used in the late 1980s to estimate the maximum wind speed in tornadoes from Doppler spectra. NSF funded part of these efforts well before instrumented storm-intercept projects were recognized by the community to be scientifically valuable. From these efforts, it was determined that the “thermodynamic speed limit” was usually exceeded, thus pointing to the important role of dynamic pressure gradients near the ground in tornadoes.

A scanning, airborne Doppler radar (ELDORA—ELectra DOppler RAdar) was developed in large part at NCAR and used by university and NCAR scientists to probe supercells during VORTEX (Verification of the ORigin of Tornadoes EXperiment) in the mid 1990s (Bluestein

and Wakimoto, 2003). Hitherto unseen details of storm evolution during tornadogenesis were examined for the first time and it was found that surface mesocyclogenesis is not a sufficient condition for tornadogenesis.

At about the same time as VORTEX, several ground-based, mobile Doppler radars were developed for analyzing the structure of the tornado itself. One effort, supported at the University of Massachusetts at Amherst and OU by NSF, led to the development and use of a mobile, high-frequency, ultra-high-resolution W-band radar; the other led to the development of the Doppler-On-Wheels, an X-band radar. The latter was initially supported by NCAR, OU, and NSSL, with some NSF funding, and has been very widely used ever since, not only for severe-storms research, but also in hurricanes and in mid-latitude storms.

Tornadogenesis, which was found to take place on time scales of 10 s or less, in one case appeared to occur when a small-scale bulge in the rear-flank gust front developed, and a small-scale vortex appeared just ahead of it and interacted with a larger-scale low-level mesocyclone. Small-scale shear-induced vortices along the gust front were resolved and hypothesized to potentially play a role in tornado formation. The radial variation of wind speed has been clearly resolved; multiple vortices have been documented, as has the fine structure of weak-echo holes. Since then, other mobile radars have been developed, in part with NSF funding; they promise to add even more significantly to knowledge of tornado structure and formation. It is anticipated that field experiments with these radars, especially during VORTEX-II, which is currently in the planning stage, will further unlock the mysteries of tornado formation and ultimately lead to further improvements in tornado prediction.

The results of the numerical-simulation efforts and the storm-intercept field programs have been applied to severe-storm forecasting through the efforts of COMET, a University Corporation for Atmospheric Research (UCAR) program. A number of forecasters who were supported by university NSF grants as students subsequently became employees at National Weather Service Forecast Offices and/or the Storm Prediction Center.

NSF has not only supported observational and basic theoretical work, which have indirectly led to the advances mentioned above, but it also has funded efforts to improve severe-storm forecasting more directly, through small centers at universities. One of the first of 11 NSF Science and Technology Centers (STCs), the Center for the Analysis and Prediction of Storms (CAPS) at OU, pioneered storm-scale numerical weather prediction in which fine-scale observations, principally from Doppler radar, along with unobserved quantities retrieved from the Doppler-radar observations, are used to initialize cloud-resolving models. CAPS also developed the world’s first storm-scale prediction system for massively parallel computers, laying the intellectual and technological foundation for what has become a major

area of inquiry including the next-generation Weather Research and Forecast (WRF) model. Other efforts, funded in large part by NSF, continue at NCAR and elsewhere toward perfecting the WRF. An NSF Engineering Research Center, the Center for Collaborative Adaptive Sensing of the Atmosphere, is based at the University of Massachusetts at Amherst, but also has other academic partners including OU, Colorado State University, and the University of Puerto Rico at Mayaguez. It aims to create a distributed, adaptive network of low-power phased-array Doppler radars on existing infrastructure (e.g., cellular towers) to improve severe-weather forecasting and warnings by sensing the region from the ground to 3 km altitude. This effort is jointly funded by the Engineering and Geosciences Directorates at the NSF. Support from and collaboration with industry has also become an important part of these centers. A systems-level testbed of four radars was installed in Oklahoma during January 2006 and will be expanded in the coming years.

This case study illustrates that ATM’s diverse portfolio of activities and modes of support were instrumental in the improvements in severe-weather forecasting during the last few decades. In addition to individual PI grants and the support of the large national center, the support of small centers was particularly fruitful for the development of radar technology and numerical modeling tools. In supporting these activities and theoretical work at universities, NSF has also provided essential support for graduate education to many students. Many of them have since become employed not only by NOAA as mentioned earlier, but also as researchers and educators at universities, government laboratories, and at NCAR, thus ensuring the existence of future generations who will further improve severe-weather forecasting.

The remarkable accuracy of hurricane landfall forecasts during the 2005 hurricane season was largely thanks to the use of dropwindsondes (Figure 2-2). Starting with their use in hurricane reconnaissance in the 1960s by the U.S. Navy and Air Force, dropsondes have become an important part of both research and operations, involving NCAR, NSF-supported university research, NOAA, the Air Force, and the private sector. Hurricane reconnaissance using dropsondes dates from the 1960s, when the U.S. Navy and Air Force used Bendix-made dropsondes to sample tropical cyclones in the Atlantic and the Pacific. In 1966, University of Arizona researcher Walter Evans modified a Bendix sonde to sample the electric field in thunderstorms and dropped them from the NCAR Queen Air, introducing dropsondes to the university research community. Then Robert Bushnell and colleagues at NCAR designed a sonde with a downward-pointing pitot tube

FIGURE 2-2 RD-93 aircraft dropsonde.

to measure vertical winds in thunderstorms for the National Hail Research Experiment (NHRE).

The NHRE-inspired design started a decades-long effort of dropsonde development by a group of NCAR scientists and engineers (in addition to Bushnell, Harold Cole, Stig Rossby, P.K. Govind, Justin Smalley, Dean Lauritsen, Terry Hock, Walt Dabberdt, and Vin Lally), which is also described in Box 2-2. Advances were spurred by the needs of NSF-sponsored field campaigns, international field campaigns, or requests by the Air Force, NOAA, or the Deutsche Luft-und Raumfahrt (DLR); and by improvements in technology. Wind-measuring capability utilizing the

Omega navigation system was introduced in time for the Global Atmospheric Research Program (GARP) Atlantic Tropical Experiment (GATE); these sondes were also used in the Global Weather Experiment (GWE) and the MONsoon EXperiment (MONEX) in 1978–1979. This capability was improved through use of the Loran navigation for the Genesis of Atlantic Lows Experiment (GALE 1986) and the Experiment on Rapidly-Intensifying Cyclones over the Atlantic (ERICA) in 1989. Starting in 1994, NCAR partnered with NOAA and DLR to develop the Global Position System (GPS) sonde for deployment by the new NOAA Gulfstream-IV aircraft and DLR’s proposed stratospheric research aircraft. While the stratospheric research aircraft was cancelled for cost reasons; DLR has continued to deploy the GPS dropsonde from their Falcon research aircraft. During this same period of time, the response and accuracy of the thermodynamic measurements (temperature, mixing ratio, and pressure) were improved, along with the design of the sonde housing, parachute, and antenna, the conversion to digital mode, and with the major improvements to the onboard data systems. Although the sondes were designed primarily for deployment from aircraft, they were also launched briefly from 80-foot-diameter super-pressure balloons for the GWE (1978–1979).

Manufacturing of the sondes passed between NCAR and the private sector in a stepwise fashion as new versions were designed. The NHRE sondes were manufactured by A.R.F. Products, in Boulder. The Dorsett Electronics Division of LaBarge, Inc. (Tulsa, Oklahoma) manufactured the Omega sondes used in GATE; while Traco, Inc. (Austin, Texas) partnered with NCAR to build the aircraft data system. VIZ (Philadelphia, Pennsylvania) manufactured the thousands of sondes used during the First GARP Global Experiment (FGGE) and MONEX after NCAR tested post-GATE improvements using prototype sondes manufactured by A.R.F. The Loran navigation sondes developed for the GALE and ERICA were manufactured at NCAR. The next-generation sondes, developed for the Air Force and used in the Tropical Oceans Global Atmosphere (TOGA) Coupled Ocean-Atmosphere Research Experiment (COARE) in 1992–1993 and CEntral Pacific EXperiment (CEPEX) in 1993, were manufactured by Radian, Inc. When TOGA COARE PIs from Texas A&M, Colorado State, NCAR, and elsewhere became suspicious of the humidity data from the radiosondes, the dropsondes were useful in verifying the biases. NCAR and Vaisala isolated the cause of the humidity biases. This led to improvements that made the Vaisala radiosondes more robust in tropical environments, providing an enormous benefit to the weather and climate communities. Vaisala also manufactures the GPS sondes used today in meteorological research and hurricane reconnaissance, but NCAR continues to build the data systems.

Improvements in hurricane-track forecasting are largely thanks to the dropsonde (Figure 2-3). In 1982 the National Hurricane Center began to

FIGURE 2-3 The average relative errors (in percent) of CON3 with and without the Omega dropwindsondes (ODWs). CON3 is the average of forecasts from three models: HRD’s barotropic VICBAR model, NCEP’s global spectral model, and the GFDL hurricane model. The numbers just above the zero-skill line are the percentages of improvement of the forecast tracks with ODWs, relative to those without ODWs, where the single- and double-asterisk superscripts indicate significance of this improvement at the 95 and 99 percent significance level, respectively. Numbers just below the zero-skill line are the average track improvements in kilometers, and those in parentheses at the bottom are the numbers of cases for each forecast interval (from Burpee et al., 1996).

assimilate dropsonde data into their operational models on an experimental basis. The results were striking: the reduction in track-error forecasts ranged from 16 to 30 percent—at least as large as the improvement over the previous 20–25 years (Burpee et al., 1996). In 1997, use of dropsondes became operational, with GPS sondes deployed from NOAA’s new G-IV aircraft and the Air Force Hurricane Hunter C-130s. The new sonde afforded unprecedented detail in the wind profiles within and around a hurricane.

Such details provide—for the first time—a detailed picture of the near-surface winds in hurricanes, allowing a better estimate of expected damage (e.g., Franklin et al., 2003). James Franklin of the Natural Hurricane Center considers dropsondes the “most important breakthrough” in cutting the uncertainty in hurricane-track forecasts.

Starting with its application to hurricane forecasting and research, the dropsonde has enabled a new mode of observations to support numerical weather prediction—obtaining data where they are most needed. Collecting data for hurricane forecasting has a long history, but in the last decade, the location of more rigorously determined “adaptive observations” are based on ensemble-modeling and adjoint (“backward in time”) techniques to identify regions of the atmosphere that could produce the most forecast errors. The Air Force C-130s and the NOAA G-IV fly dropsonde missions over the Pacific to improve forecasts of specific events, using flight plans based on objective, ensemble-based targeting techniques. Such adaptive–observation techniques have been developed and tested in field programs starting with the Fronts and Atlantic Storm Track Experiment in 1997, which involved NCAR, NOAA, some universities, and scientists from France, the United Kingdom, and Ireland.

Improvements in the dropsonde and its use continue. NCAR is developing a new, much lighter GPS dropsonde for deployment from a carrier balloon, to be used for THe Observing system Research and Predictability EXperiment (THORPEX) and the African Monsoon Multiscale Analysis. A modification of this sonde will eventually replace the GPS sonde currently used in research and forecasting. The development of the dropsonde is a clear example of effective partnerships with the private sector. ATM, through the resources provided by a large national center, initiated multiple improvements in the design and effectiveness of the technology, and the private sector leveraged these improvements to manufacture a higher-quality instrument. Coordinated use of the improved technology through many international field campaigns has led to great improvements in understanding and forecast capabilities.

In the mid 1980s, a remarkable change in understanding of stratospheric ozone occurred when scientific work by the British Antarctic Survey led by Joseph Farman documented an unprecedented and unexpected depletion of Antarctic ozone (Farman et al., 1985). Ozone appeared to be depleted not by a few percent as models were predicting at that time, but by about a third, and far sooner than any existing theory had anticipated. It was also a surprise that such enhanced depletion was clearly occurring only in the Antarctic, above the world’s coldest continent.

Research conducted by scientists worldwide rapidly established industrially produced chlorofluorocarbons as the dominant cause of the remarkable phenomenon that came to be known as the “ozone hole.” The enhanced ozone losses in Antarctica compared to other latitudes are linked to the fact that chemical processes that had not been expected can occur on polar stratospheric clouds in that “coldest place on earth.”

Policymakers agreed to an international Montreal Protocol to phase out these chemicals, and by the end of the 1990s, global production of these gases had decreased by more than 90 percent. The evolution of scientific understanding of ozone depletion and related policy decisions has since been heralded as one of the most remarkable environmental success stories of the 20th century.

The NSF, including Jarvis Moyers, played many key and unique roles in the scientific support and management that allowed the history of the ozone hole research to progress from observation, to understanding, and to policy in the short space of a few years. NSF was among those responsible for the development of state-of-the-art instrumentation to measure ozone and many key chemicals and dynamical tracers in Antarctica, sponsored by its grants program and its national center. The National Aeronautics and Space Administration (NASA) and NOAA also played important roles in the development of critical measurement capabilities. But those instruments had not been used in the remote Antarctic and the limited available knowledge of the composition, chemistry, and dynamics of the Antarctic atmosphere posed major challenges to establishing the cause of the ozone hole at the time of its discovery.

Within a year after the discovery of the ozone hole, NSF had sent an expedition of four research teams in a National Ozone Expedition to the Antarctic. The 1986 expedition to the Antarctic was strongly led by NSF, with important interagency contributions from NASA, NOAA, and the private sector.

There was a high risk that the work would bear limited if any fruit, but there was also a potential for high payoff. Important strengths included historical approaches to monitoring (e.g., long-term observations of ozone) as well as linkages to instrument development work by NASA, NOAA, and within the NSF astronomy program. Thus, core capabilities in instrumentation, monitoring, and Antarctic logistics were essential to the success of the expedition, which yielded the first measurements showing greatly enhanced chlorine monoxide (de Zafra et al., 1987) and chlorine dioxide (Solomon et al., 1987) in the Antarctic ozone hole. The vertical profile of the ozone depletion was also measured for the first time (Hofmann et al., 1987), providing key evidence for the role of polar stratospheric cloud chemistry. At the time of its discovery, several competing explanations were suggested, including purely dynamical processes, enhanced reactive nitrogen linked

to solar activity, and anthropogenic halocarbons (see, e.g., WMO, 1989). The expedition’s findings showed that the first two were not consistent with observations, and stand today as among the key initial cornerstones that first established the links between chlorofluorocarbons and the ozone hole.

NSF’s Office of Polar Programs (OPP) and ATM worked jointly to make the expedition occur on an unprecedented rapid time scale. Susan Solomon, head project scientist for the National Ozone Expedition, says “NSF’s contributions to understanding the ozone hole can only be described as extraordinary. The success can be traced to the dedication of the staff, and their agility in evaluating what needed to be done and why, as well as addressing the enormous operational demands of getting research teams to the Antarctic as quickly as possible.”

The following year, NSF-funded investigators also played major roles in another joint interagency campaign, this time using airborne approaches (e.g., Anderson et al., 1989) to further document and demonstrate the key role of anthropogenic chlorine and bromine chemistry on polar stratospheric clouds as the primary cause of the ozone hole. The type of research instruments used in the 1986 expedition are now deployed for monitoring ozone and other chemicals not only in Antarctica but at many sites worldwide, and have contributed to the understanding of Arctic and global ozone depletion as well.

Understanding the Antarctic ozone hole is a case in which the NSF, with significant interagency cooperation, spearheaded an extensive and high-risk research effort to understand and address an issue of vital international importance. Further, the effective partnering of ATM and OPP made it possible to bring the resources and expertise of both divisions to quickly move the research forward.

Development of numerical models for research purposes became widespread starting in the 1960s and 1970s, with individuals from universities, NCAR, NOAA, and elsewhere using numerical codes to simulate the solar interior, synoptic weather, mesoscale weather, severe storms, clouds, and the atmospheric boundary layer. The early models were typically developed by individuals or small groups, with the larger ones run at large computer centers. For the NSF research community, these models were typically run at the NCAR Scientific Computing Facility.

The investment in time and resources to develop a modeling system was so great that community models emerged. The earliest of these, and the most widely used (e.g., Mass and Kuo, 1998), was the mesoscale model first developed at Pennsylvania State University by Richard Anthes and his

students, which has evolved today into Mesoscale Model version 5 (MM5; Box 2-3). In the 1980s, Penn State and NCAR jointly developed MM4, and by the late 1980s, NCAR/MMM started supporting MM4 as a true community model, with well-attended community user classes and workshops. The final version, MM5, was released in 1992, improvements continued until 2004, and the last NCAR MM5 tutorial was held in January 2005 (Kuo, 2004). A look at the MM5 parameterization schemes reveals contributions from a broad community—the Blackadar (Penn State University) and Betts-Miller (Betts: CSU and then independent; Miller: ECMWF) boundary layer parameterization schemes, the Grell (University of Miami, University of Washington) and Kain-Fritsch (Penn State University) convection schemes, and the Noah (NCEP, Oregon State, Air Force Weather Agency, NOAA Office of Hydrology) land-surface scheme being some examples. Between 1995 and 2004 the number of users increased from 100 to over 1,100, and the number of institutions using the model increased from 40 to over 560 (NCAR Annual Scientific Reports 1995 and 2004). Other mesoscale models, particularly the Colorado State RAMS model, are widely used, but there are no formal community training workshops.

MM5 is gradually being replaced by the WRF model. Starting in the late 1990s, development of the WRF model began to provide a common modeling system for research and operations and hence to speed technology transfer. The principal large partners are NCAR, NOAA’s NCEP and Global Systems Lab, the Air Force Weather Agency, the Naval Research Laboratory, University of Oklahoma, and FAA (http://www.wrf-model.org/index.php), and there has been active participation of the university community (Kuo, 2004). The Beta version of WRF was released in 2000. As in the case of MM5, university PIs have played a part in its development, and workshops and tutorials are held each year (NCAR ASR 2004–2005). The WRF effort now includes two overlapping numerical modeling systems, the Non-hydrostatic Mesoscale Model, operated by NCEP, and the Advanced Research WRF, which is used by the academic community for research and the Air Force for research and operations. While there are significant differences between the two, they still share the same physics packages and software framework.

For mesoscale meteorology researchers, the availability of community models, local access to single- and multiprocessor workstations and gridded analysis, and forecast data enables the university investigator to run mesoscale models at universities for research and education (Mass and Kuo, 1998). The increase in computing power now commonly available is one major factor that made it possible to run such models in a wide array of settings.

Moving from mesoscale models to global models, the Community Climate System Model (CCSM) couples the atmosphere, surface, and

oceans on a global scale. Researchers, including many NSF-supported PIs, from 22 universities were participating in the development of the CCSM in 2005, working on land parameterization, atmospheric boundary layer, convection, and radiation schemes, the representation of sea ice, ocean modeling, and some biogeochemistry. Hundreds of scientists meet to discuss their work and plans at the annual CCSM workshop. All model components and the results from major experiments are available on the Web. As of October 2006, there were 297 CCSM publications, authored by individuals at NCAR, universities, and other research entities, frequently in collaboration across these institutions (http://www.ccsm.ucar.edu/publications/bibliography.html).

The CCSM is housed at NCAR and has been supported by DOE, NASA, and NOAA as well as NSF. The larger modeling community plays a significant role in its governance (NCAR, 2001; Kiehl, 2004). The need for stable funding, an in-house team of software and hardware engineers, and capability at or near the limits of current computer technology both in terms of speed and storage dictates a centralized operation for these high-end models (NRC, 2001b). Furthermore, dealing with assessments of anthropogenic climate change, ozone, and regional impacts of climate change are best done at a centralized location (NRC, 2001a). Indeed, more than 2PB of data were stored at NCAR from the third IPCC assessment (Kellie, 2004). Projected demands for computing power for coupled climate models outstrip the projected gains from Moore’s law (Kellie, 2004). Thus it is likely that CCSM and similar models will be centrally located for the foreseeable future.

Other significant community models are being developed. The Whole Atmosphere Community Climate Model involves NCAR (Atmospheric Chemistry Division, High Altitude Observatory [HAO], CGD) and multiple collaborators from the university community, the private sector, and international partners (Hagan, 2004). An NSF STC, the Center for Integrated Space-Weather Modeling (CISM) is developing a set of coupled codes to characterize the environment extending from the upper atmosphere of the Earth to the surface of the Sun. CISM is based at Boston University, and involves seven other colleges and universities, the private sector, and NCAR/HAO (UCAR Quarterly, 2003).

While there is economy in developing a community model that is improved by contributions from users at multiple universities and government laboratories, it can still become a “black box,” that is, used by investigators who do not fully understand the model strengths and limitations. However, for mesoscale models that can be run on university workstations, this can be circumvented by running in a quasi-operational mode and following the successes and failures over a sustained period in a class or for research purposes. Then interesting local effects can be used to

understand and then eliminate model shortcomings (Mass and Kuo, 1998). Furthermore, parameterization schemes for such a community model can be tested at a university department, and then shared with the community through inclusion in the new “official” version. Likewise, while it is not realistic to run a coupled climate model at most university departments, it is possible to work on a physical parameterization scheme, an emissions model, a canopy transfer scheme, or another submodel of the community model. Also possible at a university are analysis of model output, comparisons of output to satellite records, and utilization of satellite data as model inputs. Even with these capabilities by individual PIs and their students, frequent workshops and training sessions are needed, and more substantive collaborations involving more substantial PI residence time at NCAR are needed to ensure necessary exchange of information, ideas, and the communication required to foster ongoing collaborations. Such efforts can be used to avoid duplication of effort and ensure more uniform verification procedures (Mass and Kuo, 1998).

The development of community models is another example that illustrates how the balance among the modes of ATM support has fostered a productive relationship among individual university PIs, a large national center that provides capabilities beyond the reach of a university department’s resources, and interagency partnerships. In the case of limited domain mesoscale models, technological developments eventually allowed university PIs to conduct research independent of the large national center, yet the center still serves as a maintainer of the “official” version of model code. In the case of coupled climate models, NCAR serves an important role as a provider of computing resources and coordinator of research activities.

One of the major successes of funding from the NSF (as well as NOAA) has been the development of the wind profiler. Using radar backscatter of electromagnetic waves in the UHF and VHF from nonthermal fluctuations in the atmospheric refractive index, three-dimensional wind profiles can be obtained nearly continuously and with very high temporal resolution in the troposphere, lower stratosphere, and mesosphere. (Originally these radars were designated as MST radars, referring to the mesosphere, stratosphere, and troposphere). In wind profiling radars, the fluctuations in refractive index arise from clear-air turbulence. Backscattering radars are sensitive to those fluctuations having a scale size of one-half the transmitted wavelength.

The story of the wind profiler begins in the 1940s and 1950s with trying to understand echoes from the clear atmosphere or “angels” observed by

radio scientists engaged in radar studies of the lower atmosphere as well as in long-distance over-the-horizon troposphere radio propagation. Many of these original engineers and scientists, through their own curiosity, explored explanations for the observed clear-air echoes. In the late 1950s and 1960s the work of A. W. Friend, David Atlas, Kenneth Hardy, and many others showed that at least some of the echoes were caused by scattering from turbulent irregularities. It was also recognized that specular reflections could also contribute to clear-air echoes at lower frequencies, especially those echoes observed at vertical incidence. The work of Browning (1971) with the Defford radar in the United Kingdom demonstrated the ability to detect lee waves and Kelvin-Helmholtz waves.

In the early 1970s, the focus of research shifted to longer wavelength radars. The pioneering work at VHF was done at the Jicamarca Radio Observatory located in Peru under the direction of Ronald Woodman (Box 2-4). Originally funded by NSF via a special congressional grant, Jicamarca today remains part of a network of NSF-supported high-powered radars that explore the physics of the atmosphere and ionosphere using state-of-the-art radio techniques. The work at Jicamarca culminated in the classic paper of Woodman and Guillen (1974) which theoretically showed and experimentally confirmed the potential of VHF radars to observe the electrically neutral atmosphere. The next step in the development of the wind profiler consisted of radars that were explicitly designed for the purpose of observing the neutral atmosphere.

A flurry of activity ensued in the 1970s and 1980s with funding from NSF and NOAA. Some of the research is summarized in Gage and Balsley (1978), Balsley and Gage (1980), and Atlas (1990). This activity included the design and construction of wind profilers (MST radars) at Sunset (John Green) and Platteville (Ben Balsley), Colorado, and Chatanika, Alaska, as well as further studies at the Jicamarca Radio Observatory and the Arecibo Radio Observatory. The Platteville, Colorado system was the first continuously running, unmanned wind profiler and served as a prototype for the Poker Flat, Alaska wind profiler, which ran continuously from 1979 to 1986. Design and construction of this wind profiler was funded by NSF and served as the major prototype for wind profilers that came thereafter.

Funding for Poker Flat radar was crucial and was a result of the vision of NSF’s Ron Taylor, who visited Ben Balsley in Boulder where they discussed preliminary results from the Platteville system. Taylor encouraged Balsley to submit a proposal to NSF, which eventually led to the entire NSF-sponsored wind-profiler program, with the advice “… if you don’t ask for something big … you won’t get it.” While it was difficult for NSF to justify this project and expense, the ensuing development provided great advancements in operational weather forecasting and the understanding of atmospheric dynamics.

The Japanese quickly became engaged in the profiler development. Following an extended fact-finding visit to Jicamarca they eventually constructed the large VHF radar at Shigaraki, Japan. This powerful and flexible radar with its phased antenna array continues to be adapted to different experimental configurations to study gravity waves, storm development, precipitation, vertical energy coupling, and atmospheric stability. Soon other countries followed with their own wind profiler development, notably throughout Europe, Australia, Taiwan, and India. Major international programs (Middle Atmosphere Program) and workshops (MST workshop series and international radar schools; tropospheric wind profiling conference series) were established and provided opportunities to discuss new scientific understandings gained from the wind profiler measurements.

These scientific developments led eventually to the building of extensive profiler networks for operational use and additional research in the 1980s and 1990s. For example, NSF supported a trans-equatorial Pacific VHF profiler network primarily for troposphere studies associated with El Niño/Southern Oscillation (ENSO). Major information on ENSO, equatorial precipitation, and equatorial dynamics was obtained with this system of five radars (Piura, Peru; Christmas Island, Kirabati; Ponape, ECI; Biak, Indonesia; and Darwin, Australia). The technology began to be transferred from the research community to the operational side of NOAA and the private sector. The NOAA midwest profiler demonstration network was constructed during this period and data are provided routinely to the National Weather Service for use in forecast models as well as for nowcasting. The private sector has provided special-purpose wind profilers for use to monitor low-level winds near airports, and for air quality monitoring systems.

New experimental techniques continue to be developed. The traditional narrow-beam antenna was augmented with interferometry antenna techniques and a radio acoustic sounding system was added to many wind profilers in order to simultaneously measure both winds and temperature. Upper-atmosphere meteor systems were developed as inexpensive add-ons to the wind profiler to obtain winds in the mesosphere and lower thermosphere. Today a wind profiler can not only measure three-dimensional winds but also provide information about wind variability and vertical structure, temperature structure, storm development, divergence and vorticity, momentum and heat flux, turbulence, atmospheric stability, and precipitation.

Of the numerous science problems that have been addressed using wind profiler technology, two highlights are worth mentioning. The first, the Oklahoma–Kansas tornado outbreak on May 3, 1999, is discussed in numerous papers. During this outbreak, the wind profilers were critical in identifying the evolving atmospheric wind patterns, leading to a quick upgrade in the forecast for severe weather. It is estimated that the death toll,

which was 46, may have been as high as 700 within this region of approximately one million people if warnings had not been issued (NRC, 2002b). The second highlight is the original research by Vincent and Reid (1983), who designed a novel experimental setup to observe mesosphere gravity wave momentum flux by using the Buckland Park wind profiler outside of Adelaide, Australia. This research provided the first measurements of mesosphere gravity wave momentum flux and spurred the development of better gravity wave parameterization techniques that are critical for global and specialized models.

The wind profiler story is one of initial radio technique serendipity combined with persistent engineering and scientific pursuit. Initially supported by NSF funding from a visionary program manager, the beginnings of atmosphere radar observations led to a greatly improved understanding of the dynamic atmosphere. Eventually through technology transfer to the private sector, the technique became an important tool in operational forecasting and other applications.

Terrestrial meteorology has achieved huge advances over the past 50 years as predictions have become more and more reliable. This predictive ability is certainly one of the great success stories of contemporary science, and it has made a tremendous impact on the lives of everyone. Less well known, and still just in its beginnings, is the emergence of space weather as a predictive science. This success story would not have been possible without the support, encouragement, and vision of ATM, including Rich Behnke, as well as Tom Tascione of the Air Force.

Space weather refers to changes in the space environment that can have an impact on humans and their technology. Storms in space can produce radiation levels that are hazardous to spacecraft and astronauts. Storms in space also produce ionospheric disturbances that can degrade GPS accuracy and interfere with radio communications. Additionally, large conductors (pipelines, power grids) are vulnerable to geomagnetically induced currents that are produced by such storms. While many of the vulnerabilities are new, space weather effects have been around since the mid-19th century when it was noticed that magnetic storms were associated with degraded telegraph operations (e.g., Carlowicz and Lopez, 2002). Today, as dependency on space-based systems increased, power grids become more intertwined, and human exploration of the solar system is considered, the ability to predict space weather has never been more important.

The historical experience with the rise of terrestrial meteorology provides a roadmap for the emergence of space weather prediction (Siscoe, 2006). The Upper Atmosphere section of ATM, being aware of this his-

tory, has shepherded this process in the solar and space physics community. Scientists in ATM, working with community leaders, recognized in the early 1990s that space weather prediction was a logical, and needed, product of the basic research that had been funded for decades by NSF, NASA, and other agencies. ATM provided critical leadership for the creation of an interagency plan, the National Space Weather Program (NSWP, 1995), which involved NSF, NASA, NOAA, and the Department of Defense as Co-Chairs, with participation by the DOE, the Department of the Interior, and the Department of Transportation.

The effect of the National Space Weather Program has been dramatic. Since its inception, major documents guiding the research community, such as Space Weather: A Research Perspective (NRC, 1997) and The Sun to the Earth—and Beyond: A Decadal Research Strategy in Solar and Space Physics (NRC, 2002c), have highlighted the need for improved space weather predictive capabilities. Within NOAA, the Space Environment Center (SEC) was transferred into the NCEP, making SEC part of the National Weather Service. And the space physics community has made space weather a major part of its effort. This development is due in large part to targeted funding within ATM and the success of ATM in participating in crosscutting NSF activities and collaborative programs with other agencies, such as NASA, and with the private sector (Box 2-5).

Within ATM, special solicitations for space weather applications have opened up opportunities for researchers to work on more applied science. Such support is crucial for producing products that can cross the “valley of death” in moving from research to application (NRC, 2000). ATM has also been successful in participating in NSF-wide programs to fund multi-investigator space weather efforts. Two small centers have been established through such means. One is an STC, the CISM, headquartered at Boston University. The other is the Center for Space Environmental Modeling (CSEM) at the University of Michigan, which received support from a Knowledge and Distributed Intelligence grant from the NSF and additional support from the Department of Defense Multidisciplinary Research Program of the University Research Initiative. Both of these centers are creating end-to-end models of the space environment from the Sun to the Earth (e.g., Hughes and Hudson, 2004). While many features of the codes differ, both centers are developing things such as magnetosphere models to be used for near-Earth space weather prediction (see Figure 2-4).

Outside of NSF, ATM has also partnered with NASA’s Living with a Star program to solicit proposals to create specific products for space weather prediction, and provided support for the Community Coordinated Modeling Center that makes space weather simulations available to the community at large and which also provides support for validation and metric-based evaluation of the codes. Research by a team led by NCAR

scientists (Dikpati et al., 2004) that led to a major advance in understanding and predicting the solar cycle was supported by joint NSF/NASA funding. There is also a close working relationship with the SEC, and participation in SEC’s Space Weather Week, which brings together researchers, forecasters, and customers of space weather predictions. ATM has also had an innovative partnership with the American Geophysical Union, providing seed money for the creation of a new journal, Space Weather. This journal provides a publication venue where research, applications, and policy can

FIGURE 2-4 Visualizations of magnetosphere simulations from CISM (left) and CSEM (right).

mingle in a way that provides a unique community forum (Lanzerotti, 2003).

The emergence of space weather prediction as a field of applied science is an accomplishment in which ATM played the leading role. ATM was in the lead in identifying the fact that progress in basic research had reached the point that one could think about predictive models, some of which are now being transitioned to operations by the SEC. ATM was in the lead in developing an interagency strategic plan, the National Space Weather Program, which has had a significant impact on the agencies and the community. ATM provided the funding space, either on its own or in collaboration with NSF-wide initiatives or other agencies, within which PIs and small centers could advance the state of the art in space weather modeling. And through innovative grants such as the support for Space Weather, ATM has provided a safe harbor for an emerging community. The emergence of space weather as a predictive science is a clear success of farsighted ATM leadership in the public interest.

The Earth’s lower atmosphere has an amazing capacity to oxidize a wide range of chemical compounds emitted by both human-induced and natural processes. Oxidation of most species in the lower atmosphere is driven by reactions of the hydroxyl radical, OH (IGBP, 2003; Prinn, 2003), although some oxidation is accomplished by other radicals, including Cl, ClO, BrO, IO, NO₃, and HO₂, as well as nonradical oxidants such as O₃ and H₂O₂. Oxidation tends to make most airborne chemical compounds more soluble in water, greatly increasing the efficiency of their removal from the lower

atmosphere both by precipitation (wet deposition) and contact with surface water, soil, and vegetation (dry deposition). Without this vigorous “oxidative capacity,” the lower atmosphere would quickly become choked with anthropogenic and natural pollutants, greatly reducing visibility, degrading the photosynthetic efficiency of plants, and impacting respiration processes in animals and humans. The capacity of the atmosphere to oxidize and efficiently remove the chemical pollution emitted into it is now recognized as one of the planet’s major ecological services. Answering the question “Is the ‘Cleansing Efficiency’ of the atmosphere changing?” has been identified by the International Global Atmospheric Chemistry Project (IGAC) of the International Geosphere-Biosphere Programme (IGBP) as a major challenge for the world’s atmospheric chemistry community (IGBP, 2003).

Prior to the early 1970s, the robust oxidation chemistry of the lower atmosphere was not recognized. The fact that the daytime lower atmosphere can be viewed as a photolysis-driven low-temperature flame was simply overlooked. We now have a much better understanding of the complexities of oxidation processes. Oxidizing radicals are now known to be active in gas phase, heterogeneous (gas/surface), and condensed phase reactions, the latter two involving atmospheric aerosol particles and cloud/fog droplets. Further, we now know that, in polluted urban and industrial areas characterized by abundant volatile organic hydrocarbon and NO_x emissions, this chemistry can run rampant, producing the unhealthy levels of ozone and other oxidants, as well as abundant secondary aerosol particles that together characterize photochemical smog (NRC, 1991, 1998b; NARSTO, 2000; IGBP, 2003). The understanding of oxidation chemistry in the lower atmosphere and how this chemistry changes under varying environmental conditions forms the foundation for photochemical models of the lower atmosphere that are critical to air quality assessment.

The photochemical model predictions of oxidative cleansing must be verified experimentally. Since the OH radical is the major cleansing agent, considerable effort has been expended to measure its atmospheric abundance in order to calibrate and test photochemical models. Over the past ~25 years several successful methods to directly measure local concentrations of atmospheric OH as well as the closely coupled radical HO₂ have been developed. Tropospheric measurements of ambient OH have been performed using open-path differential optical absorption spectroscopy (Mount, 1992), chemical ionization mass spectrometry (CIMS; Eisele and Tanner, 1991), and laser-induced fluorescence (LIF; Davis et al., 1979; Wang et al., 1981; Hard et al., 1984; Stevens et al., 1994). LIF measurements that employ atmospheric sampling through a supersonic expansion can also measure ambient HO₂ by using NO to titrate that radical to form OH (Hard et al., 1984; Stevens et al., 1994). These direct OH measurement techniques are now routinely deployed on aircraft and ships as well

as at ground sites for major photochemistry-oriented field measurements (Box 2-6). The high time resolution data they provide are utilized, along with data on nitrogen oxide and volatile organic compound concentrations, solar radiation, and other chemical and environmental parameters, to directly test photochemical models. In addition, OH reactivity (the inverse of the local OH chemical lifetime) can now be measured using LIF techniques to trace the decay of an induced OH spike in an ambient air sample (Kovacs and Brune, 2001). Such measurements directly characterize the local atmospheric pollutant loading susceptible to oxidation by OH (Figure 2-5).

The NSF has played a vigorous role in promoting and funding U.S. research on tropospheric oxidative capacity. For instance, in the mid 1980s, the agency took a lead role in assembling a steering committee and selecting task groups to define the elements of the nation’s Global Tropospheric Chemistry Program. A key component of that program identified and defined long-term gas phase photochemistry research goals and strategies needed to quantify tropospheric oxidative processes (UCAR, 1986). Since then the agency has consistently sponsored the instrument development, process-oriented field measurements, diagnostic model development and utilization, and basic laboratory experimental and theoretical chemistry and spectroscopy research initiatives outlined in that document. Starting even earlier, in the late 1970s, the agency took a strong lead in supporting direct ambient OH and HO₂ radical measurement techniques, eventually funding or co-funding pioneering LIF instrument development at Georgia Tech, Ford Motor Co., Portland State, and Penn State, as well as supporting the seminal work on CIMS detection at Georgia Tech and NCAR. Currently, field studies and related modeling efforts funded or co-funded by NSF are characterizing the lower atmosphere’s oxidative capacity for a full range of ambient conditions from the remote arctic to the world’s megacities.

The brownish haze associated with many industrial regions and with rural areas subjected to heavy biomass burning is a well-recognized result of human activities. This haze can be transported over long distances to form a regional-scale aerosol layer, such as has been observed in the Arctic, across India and southern Asia, extending from east Asia across the Pacific, and in biomass burning and dust plumes from North Africa that spread over most of the subtropical Atlantic (Ramanathan et al., 2001b). In the late 1990s significant advances were made in understanding such atmospheric aerosol layers and how they affect climate (Box 2-7).

Understanding the sources and fate of tropospheric aerosols has been

a major challenge due to the complexity of their sources, composition, chemical interactions, and physical processing in the atmosphere. Fossil fuel combustion and biomass burning emit particles (e.g., fly ash, dust, and black carbon) and aerosol precursor gases (e.g., SO₂, NO_x, and volatile organic compounds), which form secondary aerosols through gas-to-particle conversion (Ramanathan et al., 2001a). Atmospheric aerosols exist

I have always felt that, although we have stumbled on our own, they have been partners in our successes. They have given us the opportunities to advance the atmospheric sciences, which is what NSF’s ATM should be all about. William Brune atop the ~30 m tower at the Program for Research on Oxidants: Photo-chemistry, Emissions and Transport site at the University of Michigan Biological Station near Pellston, Michigan, in summer 1998.

in a variety of hybrid structures: liquid droplets, externally mixed (a mixture of particles that each have single chemical compositions), internally mixed (each particle includes multiple chemical components), coated particles, or a combination of all of the above. They are subject to heterogeneous chemical reactions, phase changes, atmospheric transport, and removal from the atmosphere through precipitation or dry deposition to the surface.

FIGURE 2-5 Diurnal measurements of OH reactivity illustrating the very high levels of atmospheric pollutants, particularly during the morning rush hour, at ground level in Mexico City, compared to U.S. cities and a rural site (data from W.H. Brune).

Aerosols affect climate both directly, by absorbing, reflecting, and scattering solar radiation, and indirectly, by influencing cloud optical properties, cloud water content, and cloud lifetime. The aerosol direct and indirect forcing may have offset as much as 50 to 75 percent of the greenhouse gas forcing since the Industrial Revolution times (NRC, 2005c). The climate influence of aerosols is one of the largest uncertainties in models of present and future climate. Furthermore, aerosols are a major component of air pollution with well-documented effects on human health, ecosystems, and visibility.

A major catalyst for advancing understanding of atmospheric aerosols was the INDian Ocean EXperiment (INDOEX), which culminated in a 1999 field campaign. The INDOEX campaign brought together researchers from the United States, Europe, India, and the Maldives for an intensive investigation into the factors controlling aerosols over the tropical Indian Ocean and the associated climate impacts. By integrating observations

from satellites, aircraft, ships, surface stations, and balloons with one- and four-dimensional models, the participants made several striking discoveries. They observed remarkably high levels of aerosols extending over most of the South Asian region and the North Indian Ocean, and up to 3 km altitude (Ramanathan et al., 2001a). These aerosols enhance scattering and absorption of solar radiation, while also producing brighter clouds that are less efficient at releasing precipitation. Thus, solar irradiance that would otherwise reach Earth’s surface is either reflected back to space or contributes to warming of the atmosphere directly, thereby changing the atmospheric temperature structure. Further, the aerosols were found to suppress rainfall, inhibit removal of pollutants from the atmosphere, and lead to a weaker hydrological cycle.

As the agency leading the mission, NSF was instrumental in planning and executing INDOEX. NSF program managers helped coordinate the contributions of DOE and NOAA, and worked with the U.S. State Department to coordinate the participation of other nations, especially India and the Maldives. Further, NSF supported and facilitated coordination among the many PIs involved to ensure proper utilization of resources.

In addition to NSF support for PIs and coordination efforts, INDOEX utilized two other modes of NSF support: the Center for Clouds, Chemistry, and Climate (C4) STC and NCAR/UCAR system. The C4 center was important for fostering INDOEX because it provided (1) a multi-institutional, multinational organization; (2) funding flexibility; (3) an established infrastructure, including a center manager; and (4) support for a testbed experiment preceding INDOEX to demonstrate capabilities in the field. Moreover, STC funds helped support the analysis part of INDOEX, whereas many field programs suffer in this regard.

UCAR played an important role in the field campaign logistics, particularly in the deployment of the aircraft, communication between platforms (i.e., ships, aircraft), and in providing meteorological forecasts in the field. NSF program managers led the design of this interagency program involving the participating PIs and UCAR in determining what instruments would be included and how best to deploy them. NCAR scientists brought to bear several modeling tools during the INDOEX campaign (e.g., chemical forecasts that helped guide flight plans) and to analyze the afterwards (e.g., aerosol assimilation models).

The INDOEX campaign combined with the focused research on aerosols at the C4 STC has had several significant scientific impacts. It has resulted in dozens of scientific papers, establishing a new paradigm for how air pollution and climate change are linked and stimulating a new area of interdisciplinary research. Furthermore, the field observations allowed for initial validation of how climate models treat aerosol forcing. Several follow-up research efforts are under way, including the establishment of

numerous observatories in the Indo-Asia-Pacific region to monitor aerosol pollution and a UNEP-sponsored Project Atmospheric Brown Cloud.

The identification of the importance of tropospheric aerosols is a particularly good example of how small centers, such as the C4 STC, can foster major scientific breakthroughs and have the flexibility to respond to unforeseen avenues of research. Further, it illustrates how ATM successfully facilitated international cooperation through the use of large national centers and domestic interagency coordination. Support for numerous types of activities played important roles in this research, including technology development, field programs, and laboratory research.

In 1958, Charles Keeling began making measurements of atmospheric carbon dioxide (CO₂) at the Mauna Loa Observatory. The long-term monitoring record from this site has become an icon of global climate change. Keeling’s first measurements at the Mauna Loa Observatory began in 1958 and were funded by Dr. Henry Wexler of the U.S. Weather Bureau as part of the bureau’s efforts during the first International Geophysical Year to measure CO₂ at remote locations. The possibility of making continuous measurements of atmospheric CO₂ built on improvements in instrumentation using infrared gas analyzers and Keeling’s carefully developed manometric technique for precisely calibrating analyzers. Keeling’s interest in atmospheric carbon dioxide measurements was fueled by his earlier observations of the variability in CO₂ concentrations near the ground in Pasadena and Big Sur State Park in California, the Olympic Pennisula, in Washington, and the high mountains of Arizona (Keeling, 1958) as well as insights from reading Climate Near the Ground (Geiger, 1957). It is also interesting to note that Keeling worked with Sam Epstein to make ¹³C isotopic measurements of his first samples, foretelling the importance of both atmospheric carbon dioxide and the accompanying carbon isotopes in understanding the global carbon cycle almost 50 years after the first Mauna Loa measurements (Keeling, 1958).

Once Dr. Keeling had secured funding for his atmospheric carbon dioxide measurements, getting the measurements set up required his attention to competing demands (Keeling, 1998). The backing from the U.S. Weather Bureau allowed Keeling to purchase four instruments—one for deployment at Mauna Loa, one for deployment in Antarctica at the Little America Station, one for deployment on the Scripp’s Oceanographic Institute research ship, and a fourth for use in the laboratory to cross calibrate with Keeling’s laborious manometric technique. Dr. Roger Revelle, director of Scripps Oceanographic Institute, succeeded in attracting Dr. Keeling to

Scripps to pursue his CO₂ studies there. Despite the difficulty of making measurements in Antarctica, it became the first monitoring site established in 1957. Dr. Revelle was convinced that shipboard and aircraft measurements were critical to understanding what was then thought to be the substantial spatial variability of CO₂. Revelle’s interests competed with Keeling’s determination to set up the Mauna Loa site. The U.S. Weather Bureau was able to provide a full-time employee to assist Keeling in meeting these competing demands and Keeling began the Mauna Loa measurements in 1958.

While the Mauna Loa data record is remarkably continuous over almost 50 years, the continuity of the funding record required the participation of a number of funding agencies and backers, including the international community. The publication of the first description of the seasonal cycle of CO₂ in 1960 was crucial for building support for his efforts (Keeling, 1960). The Mauna Loa and South Pole measurements were supported through 1962 by funding from the NSF and the U.S. Weather Bureau. In 1963, the Weather Bureau funding was discontinued and the entire program was funded by NSF. In 1961 and 1969, Keeling went to Europe on sabbatical, which was important for stimulating international interest in these measurements and for the establishment of Keeling’s laboratory as the central laboratory for the WMO’s CO₂ calibration effort. In 1971, NOAA installed an additional CO₂ analyzer at Mauna Loa. Also in 1971, NSF reduced funding for Keeling’s laboratory by 50 percent, deeming that atmospheric CO₂ measurements were routine. The head of the WMO, Dr. Christian Junge, helped to restore NSF funding and worked to secure funding from the newly formed UNEP to calibrate CO₂ measurements worldwide based on the scientific arguments laid out in Keeling’s 1970 paper (Keeling, 1970).

The next big scientific and funding breakthroughs came with Keeling’s publication of 14 years worth of CO₂ measurements from both Mauna Loa and the South Pole (Keeling et al., 1976). That same year, the Atomic Energy Commission established the Energy Research and Development Agency (ERDA) to argue for nuclear-powered electricity generation. ERDA pursued studies of the carbon cycle because “the burning of fossil fuels might be more dangerous to mankind than any perceived side effects of nuclear energy” (Keeling, 1998, p. 56). ERDA became the DOE and the precedent was set for DOE to pursue studies of the carbon cycle and fund the Mauna Loa CO₂ measurements in 1978. This occurred just as the NSF was decreasing its support for Keeling’s measurements because they were again considered to be routine. In 1980–1982, NOAA and DOE contributed 80 percent of the funding for the Mauna Loa measurements and NSF funding was slated to be eliminated at the end of 1982.

More than two decades of measurements at Mauna Loa laid the groundwork for what would be the next big breakthroughs in the carbon

cycle: the return to the measurement of carbon isotopes; the measurement of CO₂ in ice cores; the development of models of the carbon cycle with north–south resolution of sources and sinks of atmospheric carbon; and, eventually, the three-dimensional representations of the global carbon cycle (e.g., Box 2-8; Fung, 1986; Heiman and Keeling, 1986; Fung et al., 1987; Keeling et al., 1989). However, continual funding for Mauna Loa would be problematic. In 1981, NOAA’s responsibility for funding of Mauna Loa measurements was transferred to DOE. In 1983, DOE indicated that the agency was withdrawing funding. Once more, Keeling successfully reapplied to NSF for support of the Mauna Loa measurements and the WMO CO₂ program calibration. NOAA funding was secured once more, but only for one year. After considerable argument and discussion in 1984, DOE began to fund the Mauna Loa CO₂ measurements once again and that funding would continue through 1994. In the funding hiatus of 1982, the Electrical Power and Research Institute began to support the Mauna Loa CO₂ measurements as well.

The challenges in maintaining funding for continuous measurements of CO₂ at Mauna Loa continued until Keeling’s death. The history of funding for the Mauna Loa record underscores the resilience rooted in maintaining healthy relationships with multiple funding agencies, the importance of establishing and maintaining international partnerships, and effective interdisciplinary collaboration within the scientific community. Charles Keeling’s persistence and passion for his subject are a testimony to the difference a single individual can make. For its part, ATM was able to ensure that support for individual PIs, facilities, and instruments continued, by adapting to the changing contributions and priorities of other agencies.

Jacob Bjerknes, in a series of papers between the late 1960s and mid 1970s, laid out important groundwork for understanding El Niño, or El Niño/Southern Oscillation (ENSO) as a coupled atmosphere–ocean phenomenon. The work of several others following him (e.g., Wyrtki, 1975; McCreary, 1976; Busalacchi and O’Brien, 1981; Cane, 1984) elaborated the manner in which the tropical Pacific Ocean responds to changing patterns of wind during El Niño events. In parallel, a number of modeling studies reinforced Bjerknes’ conclusions concerning the influence of sea surface temperature anomalies on the tropical and extratropical atmosphere associated with El Niño (e.g., Rowntree, 1972; Lau, 1981; Zebiak, 1982; Shukla and Wallace, 1983). The seminal work by Rasmussen and Carpenter (1982) provided a coherent description of the systematic evolution of oceanic and atmospheric anomalies during El Niño events, as derived from collected historical observations over several decades.

The outsized El Niño event of 1982, with its similarly outsized impacts felt throughout much of the globe, did much to galvanize global attention, and to redouble the research community’s resolve to better understand the phenomenon and its potential predictability. Through this one event, the importance of ENSO to societies worldwide came into much sharper focus.

Following a very few pioneering efforts to model atmosphere–ocean interactions underlying ENSO (e.g., McWilliams and Gent, 1978; Lau, 1981; McCreary, 1983), a number of new studies were undertaken in the mid 1980s to provide a more comprehensive understanding of ENSO dynamics and associated predictability. One line of research, introduced by Barnett (1981, 1984) and later by Graham et al. (1987), applied advanced statistical methods to identify systematic lead-lag relationships between atmospheric and oceanic variables, and to exploit them to develop statistical prediction models for El Niño related sea surface temperature patterns. This work indicated real predictability associated with El Niño onset, at lead times of several months.

The second line of research involved further development of physically based models. Several were developed, with various simplifying assumptions (e.g., Anderson and McCreary, 1985; Cane et al., 1986; Schopf and Suarez, 1987). The approach taken by Cane and Zebiak (Cane et al., 1986; Zebiak and Cane, 1987) proved particularly useful and led to the first dynamical El Niño predictions. The most important simplification in this case was to model only the departures of atmospheric and oceanic states, relative to the observed, seasonally varying mean climatological state. In so doing, a relatively simple dynamical model was capable of simulating realistic features of El Niño, including aperiodic oscillations with a spectral peak near the four-year period, and the characteristic spatial patterns and magnitude of anomalies. These authors introduced an El Niño forecasting system based on this model in 1986, and at that time also produced a one-year lead forecast indicating that a moderate-amplitude El Niño event would develop later that year. The forecast proved substantially correct, though the onset was more than two months later in nature. Routine predictions with this system were initiated the following year, and have continued (with several revisions) to present.

The late 1980s were a period of great excitement in the climate community. A multitude of El Niño prediction systems (both statistical and dynamical) were developed and routine predictions of El Niño were established with several of these systems. Several of the more successful models were analyzed in some detail to understand better the dynamical basis for the empirically observed predictive skill. A major new, near-real-time observing system was designed and substantially deployed to monitor the upper ocean and surface ocean–atmosphere conditions in the equatorial Pacific. Its purpose was to provide more detailed study of ENSO physics, and to supply

necessary information to the predictive models (McPhaden and Hayes, 1990). And, in order to study coupled processes in the so-called Warm Pool region of the western tropical Pacific—processes poorly captured by existing models and believed to be a limiting factor in prediction—the proposal was put forward to undertake a major observational field campaign: the Coupled Ocean Atmosphere Response Experiment (COARE; Godfrey et al. [1998] document the many accomplishments and findings of this major program). Finally, the international research community developed the concept of an International Research Institute, which would transition El Niño and seasonal climate predictions into operational forecasts, and address the myriad issues at the interface of climate and societal needs, allowing the perceived benefits of this new climate knowledge and information to be realized in practice. All of these activities took place within the context of the international TOGA research program undertaken by the World Climate Research Program during the period 1985–1995. The U.S. participation in this program was organized through an interagency process, and included major contributions from NOAA, NSF (ATM and OCE), and NASA.

The key research that laid the foundation for the first efforts at El Niño prediction was undertaken through grants to individual or small groups of PIs, and was supported by NOAA, NSF (ATM and OCE), and NASA. This initial research was the formative work in understanding the coupled processes and how to forecast El Niño; it opened the door to operational prediction and its multiple societal benefits. The basic investment in research early on catalyzed investments in the observing systems and modeling efforts by other agencies. Most notably, NOAA advanced two extremely important programs: the Tropical Atmosphere Ocean observing system, and the coordinated modeling and prediction program known as the TOGA program on prediction (PIs were still supported through individual grants). UCAR provided access to facilities and field support throughout much of these efforts. Finally, the major field campaign TOGA COARE was international in scope, but very substantially supported by NSF/ATM, NSF/OCE, NOAA, and NASA. The U.S. interagency coordination in all of these efforts was outstanding and highly effective.

Observations of the solar surface reveal oscillations of the order of five minutes (Harvey, 1995). These surface oscillations, a manifestation of resonant oscillations within the Sun, provide a window into the Sun’s interior structure and dynamic that has not only led to a golden age in solar physics and stellar physics, but it has also led to fundamental new understanding in atomic physics. There are about 10⁷ pressure modes oscillating in the Sun

with typical amplitudes of ~1 cm s^–1 (Harvey et al., 1996). Through a complex but straightforward multistep process, sequences of images of the Sun are isolated into the Sun’s normal modes, which can then be used to obtain information about the structure and circulation of the Solar interior.

The basic ideas for helioseismology came from the university community and one of the Astronomy federally funded research and development centers (Harvey, 1995). In 1960, Robert Leighton of Caltech discovered that the surface of the Sun was characterized by small-scale patterns that oscillated radially with a period of about five minutes. In the early 1970s, Roger Ulrich of UCLA and John Leibacher of the National Solar Observatory (NSO) and Robert Stein of Michigan State University demonstrated that the oscillations were caused by acoustic waves generated in the solar interior—the Sun “rings like a bell.” In 1975, Franz Deubner observed the predicted patterns and used it to demonstrate that the then current solar models were incorrect.

Obtaining the surface oscillations is a challenge because of their small amplitudes (Harvey, 1995; Gough et al., 1996; Harvey et al., 1996). Furthermore, longer time series—significantly longer than the consecutive hours of daylight in low latitudes—are needed to determine the normal modes from the portion of the Sun visible from Earth than if the whole surface could be sampled. Thus, the early measurements were made at the South Pole supported by NSF Polar Programs and by a “relay” involving several observatories around the world. These early measurements, however, could only sample those modes with wavelengths of the order of the diameter of the Sun.

The next step, obtaining long-term high-resolution observations of the Sun’s surface, involved two major projects: the Global Oscillation Network Group (GONG), sponsored by NSF/Antarctic Submillimeter Telescope (AST) and run by NSO, and The SOlar and Heliospheric Observatory (SOHO), sponsored by the European Space Agency (ESA) and NASA. A competition was held for the best design for the GONG instruments; the winning design, the Fourier tachometer, was developed by Timothy Brown of HAO/NCAR along with Jack Evans and others at NSO.

The GONG network of six identical solar-imaging telescopes has been collecting continuous data of the Sun’s surface since 1995; the GONG Advisory Panel included scientists from NCAR and a number of universities. Some complementary efforts aimed at getting longer temporal baselines are the U.K. Birmingham Solar Oscillation Network; while efforts aimed at getting structure deeper in the Sun include the Mount Wilson-Crimean-Kazakhstan mini-network, the high-degree helioseismometer operated by the NSO at Kitt Peak, and the low-and intermediate-degree experiment (LOWL), operated at NCAR’s HAO at Mauna Loa Observatory. Scientists from several U.S. universities are involved in helioseismology, as are

scientists in Australia, Denmark, France, Germany, India, Japan, Taiwan, and the United Kingdom. The Solar Terrestrial Research Program in ATM provides an average of roughly $400K–500K per year for helioseismology through its regular grants program, and through the National Space Weather Program (NSWP), Research at Undergraduate Institutions, and NSF faculty early career development program (CAREER) awards.

In its first decade, helioseismology has ushered in a golden age in solar research. Solar structure and motions have been clarified. One of the first results from helioseismology was that the convection zone extends downward to 0.713 of the solar radius, significantly deeper than many earlier solar structure models had predicted (Harvey, 1995). Helioseismology has also demonstrated that the zonal flow at the surface (long documented by following the trajectories of sunspots and other surface features) changes little with radius through the convection zone; but there is considerable radial shear between the convection zone and the radiative interior, through a layer called the “tachocline.” Helioseismic measurements have also shown that small variations in solar rotation—so-called torsional oscillations— occur throughout the convection zone, with periods tied to the solar cycle period. Beyond differential rotation, helioseismic inversions have revealed several other motions, including poleward meridional below the photosphere and other “solar weather” patterns near the surface. Inversions are also beginning to reveal the thermal structure below “active regions,” the sites of sunspots. Now solar activity on the side of the Sun opposite the Earth is routinely detected by so-called “far-side imaging.”

These findings inspired a new generation of individuals and groups around the world to develop idealized models and full-blown general circulation models of the outer layers of the Sun, designed to explain the solar dynamo and the solar cycle.

In 2006, Mausumi Dikpati, Giuliana de Toma, and Peter Gilman, of NCAR’s HAO used a so-called “flux-transport” dynamo model that assimilated data from the previous three solar cycles (borrowing a technique from numerical weather prediction) to predict successfully the relative amplitudes of the new cycle for each of the last nine solar cycles, and to project the amplitude of the next solar cycle in 2012 (Figure 2-6) (e.g., Box 2-9; Dikpati et al., 2006). These predictions and future results coming from helioseismology-inspired research will provide significant contributions to predictions of Space Weather.

In addition, helioseismology has offered a new way to constrain the distribution of elements making up the Sun. The elemental abundance has historically been determined by using what we know—luminosity measurements, solar radius, and solar spectral results giving clues to the chemical composition—to build one-dimensional solar and stellar structure models. More recently, helioseismology has been used to determine the sound

FIGURE 2-6 (a) Observed sunspot area, smoothed by ~1-year Gaussian running average, plotted as a function of time; (b) simulated toroidal (zonal) magnetic flux at the bottom of the solar convection zone, which is the source of sunspots. Solid red area and curve are for steady meridional flow; dashed red curve is for time-varying flow since 1996 incorporated. SOURCE: Dikpati et al. (2006).

speed, providing an independent route to constrain the composition. Until recently, both approaches were consistent with the same solar abundance of elements—until Australian Martin Asplund and colleagues (Asplund et al., 2005) used spectral data interpreted with the help of three-dimensional simulations to revise the solar chemical composition to reduce the percentage of atoms heavier than helium. Sarbani Basu of Yale University, funded by a CAREER grant through NSF/ATM/Solar Terrestrial Research, was

on a team that found a potential way out of this impasse. By raising the abundance of neon (the most uncertain of the heavier element abundances), they have shown that the rest of the revised abundances are potentially consistent with both helioseismic and spectral results (Bahcall et al., 2006). She and H.M. Antia are now attempting to measure the total solar heavy element abundance through helioseismology (Antia and Basu, 2006). This

magazines such as New Scientist and National Geographic in 2004, have given me great encouragement that this research is of interest and value to society. After our prediction for the next solar cycle was published in Geophysical Research Letters on March 6, 2006, the work received great attention in the worldwide press.When reports were translated into Bengali and reported in “Songbad Protidin” (the daily newspaper in Calcutta), it caused my mother to become very excited about the recognition her daughter was getting.

is important because stellar models and models of the universe are tied to solar abundance models.

The reach of helioseismology extends beyond solar physics to particle physics. Secondary students all over the world are taught that the Sun is powered by a reaction that converts hydrogen to helium, releasing neutrinos in the process. In the 1960s, Raymond Davis of Brookhaven

National Laboratory (University of Pennsylvania after 1984) set up an experiment to detect solar neutrinos, 4,800 feet below the surface in the Homestake Gold Mine, in Lead, South Dakota. Neutrinos were detected, but only one-third the amount predicted assuming that the commonly accepted helium-to-hydrogen conversion was true; stimulating new detectors around the world—Kamiokande in Japan, SAGE in the former Soviet Union, GALLEX in Italy, and Super Kamiokande (http://nobelprize.org/physics/laureates/2002/davis-autobio.html). Those explaining the “missing” neutrinos invoked a rapidly rotating solar core, contrary to the thinking of the solar physics community. The LOWL instrument revealed a core rotating at a rate similar to the outer layers of the Sun, indicating something else was needed to explain the missing neutrinos. Also, John Bahcall, Sarbani Basu, and Marc Pinsonneault argued that the fact that the sound speed and density in the core of standard solar models are so close to those inferred from helioseismic measurements, also implies that explaining the missing neutrinos would mean invoking nonstandard neutrino physics rather than nonstandard solar models. Finally, in 2001–2002, scientists at the Sudbury Neutrino Observatory in Ontario, Canada, found evidence that the neutrinos could oscillate among three forms. The 2002 Nobel Prize in Physics was awarded to Davis and Masatoshi Koshiba of the University of Tokyo “for pioneering contributions to astrophysics, in particular for the detection of cosmic neutrinos.”

Responsibility for funding helioseismic studies has been shared at NSF between ATM and AST, and within the United States has been shared between NSF and NASA. The partnership between NSF’s ATM and AST divisions exemplifies a successful intra-agency partnership across directorates within NSF. The helioseismic observations are made both from space and from worldwide ground-based networks, and analysis of the data is carried out at many U.S. and foreign universities who meet regularly to discuss the latest results. The ground-based GONG network has recently been upgraded to provide much higher spatial resolution with NSF/AST money and within two years the SOHO satellite will be superceded by the new Solar Dynamics Observatory satellite (NASA Living with a Star Program, ESA). NSF/ATM and NASA continue to support helioseismic studies with grants to university PIs.

Given the relatively short instrumental climate data record, the ability to test climate models depends strongly on the availability of paleoclimate records. Advances in paleostudies over the past few decades have significantly extended the data record, providing a context for the instrumental climate record and the evolution of atmospheric composition (e.g., CO₂,

CH₄, NO₂). Paleorecords are used to infer the impact of anthropogenic contributions on long-term climate trends and to test climate models, particularly regarding the sensitivity of climate to CO₂.

A major milestone in the field of paleoclimatology was the demonstration that the concentrations of oxygen-18 and deuterium accumulated in deep cores of glacier ice can be used as an indicator of past temperature (Dansgaard, 1964). Radioisotopic dating methods, such as the radiocarbon dating methods developed in the early 1970s for dating sediment or lake cores, were applied to these data sources in order to construct consistent chronologies of past temperature. Since that time, many other radioisotopic dating methods have been tailored and calibrated for specific application to paleoclimate datasets (Cronin, 1999).

The first ice cores were obtained from Vostok, Antarctica, in the 1970s by Russian scientists. In the 1980s, French and American scientists subsequently joined this effort, which was supported primarily by NSF’s OPP (Box 2-10). The most recent Vostok drilling yielded the longest record, dating to 420,000 years B.P. (Petit et al., 1999). Similar drilling expeditions to Greenland have yielded data records on atmospheric composition and temperature trends on millennial time scale. These records have revealed the climate’s variability over millenial cycles, and have provided evidence for the correlation between global temperatures and CO₂ concentrations. These records also demonstrate the importance of Milankovitch cycles in regulating climate (Imbrie and Imbrie, 1979; Berger et al., 1984).

Microfossil records from ocean cores led to the discovery of past systematic changes in sea surface temperatures and changes in the amount of glacial ice stored on the continents. Such records also permit the reconstruction of past global ocean currents, which is important because of the influence of the global thermohaline circulations on climate. Some of these breakthrough discoveries were made during the large CLIMAP (CLimate: Mapping, Analysis, and Prediction) program, a multi-institutional consortium effort funded by the NSF and led by J. Imbrie, J.D. Hays, N. Shackelton, and A. McIntyre. This effort led to a follow-up called COHMAP, which was supported primarily by Climate Dynamics Program of NSF and by DOE (carbon dioxide research division). An important and surprising finding was revealed by the Greenland Ice Core Project, where scientists detected the ability of abrupt climate shifts (5–10°C) during an interglacial period (Dansgaard et al., 1989). For centennial, decadal, or even year-to-year resolution in past climate variability, paleo proxies such as tree-rings, coral records, and lake or bog sediments are used (NRC, 2002a, 2006b).

Many of the field-intensive drilling expeditions, such as ocean sediment or glacial ice core drilling, were supported by international efforts and funded by multiple agencies such as NASA, NOAA, NSF, and DOE. While OPP plays the principal role in funding ice core drilling operations and the

analysis of such ice cores, NSF’s ATM supports most other paleoclimate studies, such as tree-ring, lake sediments, and coral paleoclimate studies. In fact, ATM has consistently led in the support of tree-ring research (e.g., Figure 2-7).

The paleoclimate program is in a unique position because, although about 25 percent of all proposals submitted to ATM are in this area, the program has one of the smaller budgets in this division. However, the program is successful because the cross-division collaborations within GEO and NSF reflect the strongly interdisciplinary nature of this area of scientific research.

in the southwestern Himalayas near the source of the Ganges and Indus Rivers, made possible in part through funding from NSF’s ATM-ESH program. We have had many dedicated program mangers over the years who promoted the paleoclimate community and their faith in this effort has produced a rich record of paleoclimate in regions where it once was unknown.

We turn now to the following questions: What light do these major accomplishments shed on the role of NSF ATM in its support of atmospheric sciences? In particular: What do they imply about the balance between the various modes of support, whether that balance has in fact been adjusted over time, and whether they provide evidence that there is a need to alter the balance?

The first observation is that NSF ATM has played a role in every one of these major accomplishments. In a few cases, ATM played only a minor

FIGURE 2-7 Tree-ring chronologies for several regional composites. The time series have been loosely grouped according to latitude bands and normalized to a common period. The bottom two panels in the right column show grouped replication plots for both North America and Eurasia. NOTE: ALPS = Alps, CNTA = Central Alaska, CNWT = Central Northwest Territory, CSTA = Coastal Alaska, ICE = Icefields, JAEM = Jaemtland, LAB = Labrador, MAN = Manitoba, MON = Mongolia, NWNA = Northwest North Alaska, POL = Polar Urals, QUE = Quebec, SA = Southern Alaska, SEW = Seward, TAY = Taymir, TORN = Tornetraesk, WRA = Wrangells, YAK = Yaktutia, YUK = Yukon. SOURCE: D’Arrigo et al. (2006). Reproduced by permission of American Geophysical Union; copyright 2006.

or supporting role but in the majority of these cases, NSF ATM’s role has been central. Furthermore, the case studies demonstrate that all the modes of support—PI grants, including those for exploratory projects and in response to focused solicitations; small centers; the large national center; cooperative observing facilities; and field programs—have been important to one or more of these major science achievements. Likewise, each major achievement benefited from several modes. For example, much of the early

work in climate modeling was supported by individual PI grants, however, with increasing complexity and the need for ever larger computing power, interagency and intersector support became increasingly important and the available computing facilities at NCAR ever more central to running the models. It is also difficult to envision that the space weather research community would have made the advances without the combination of modes such as small PI grants, support from NCAR for model runs, observing facilities, ATM initiatives, and the pioneering work supported by the small center.

Thus, the range of available modes has been a tremendous and necessary asset for the atmospheric sciences. This is a reflection of the nature of atmospheric science and its development over the past decades as discussed in detail in Chapter 3; but it is also evident that ATM’s portfolio of modes of support and the balance among the modes has evolved with the state of the science. For example, at a very early stage, severe weather research was an interagency effort, mostly between NOAA and NSF’s ATM, and within ATM supported by individual grants to University scientists, who worked in close collaboration with NCAR scientists. As it became necessary to integrate field observations, modeling capability and instrument development to advance severe weather research, the field was ripe to take advantage of new modes such as NSF’s STC leading to the development of the world’s first storm-scale prediction system. Another example is carbon cycle research, which began with a single PI effort originally funded by the U.S. Weather Bureau. Eventually, it became an interagency, cross-disciplinary, and multimode effort to support direct CO₂ measurements, ice core measurements, and the development of carbon cycle models.

Grants to individuals and teams of PIs were instrumental in all of the achievements, while the large national center contributed to nearly all of them. This reflects in part the fact that these have been the two dominant modes of support utilized over the past 40 years. But, more importantly, it reflects that these two modes have been effective at fostering a productive research environment. In addition, in more than half of the major case studies, the science was significantly advanced by field programs, often large efforts requiring significant coordination among researchers, different agencies, and in many cases different nations. The U.S. participation and interagency coordination during TOGA, an effort to further the understanding of ocean–atmosphere processes related to ENSO, exemplifies the success and importance of international and interagency field campaigns in advancing atmospheric research.

Some of the newer modes, such as small centers and cooperative observing facilities, have not been available as long and are a smaller portion of the ATM funding portfolio. Even so, the three small centers that have been established in the atmospheric sciences (CAPS, C4, and CISM) have

each been engaged in research that either led to a significant leap in understanding, as in the case of CAPS and C4, or else are helping us to bring to fruition a major achievement, as in the case of CISM. Another advantage of these small centers is the explicit role for technology transfer. Indeed, atmospheric science is special in that one of the key transfer targets is the federal government.

The value of partnerships with other disciplines, agencies, and nations is also apparent in reviewing these case studies. Every major achievement analyzed required coordination with other agencies, including NOAA, NASA, DOE, EPA, and the Department of Defense. In some cases, broad interagency programs like the U.S. Global Change Research Program or the NSWP have played an important role in focusing research objectives and applying the collective resources of several agencies. Given the range of partnerships employed in these case studies, it is fair to conclude that NSF has been effective in fostering collaboration.

An important lesson to be gleaned from the research activities leading to these major accomplishments is that ATM has adjusted the balance from time to time as opportunities, needs, and scientific progress made necessary and possible. For example, when it became apparent that a concerted, coordinated effort could lead to significant advances in space weather predictions, ATM supported members of the scientific community in their bid for an STC, resulting in the recently formed CISM. The creation of the interagency U.S. Global Change Research Program in the late 1980s is another example of NSF ATM, in coordination with other agencies, identifying the need for greater organization and coordination, and then taking the steps to address this need. In general, ATM has been responsive to evolving needs and has effectively interacted with the community in choosing new directions. It does not in any way detract from this conclusion to note that NSF as a whole has been moving, over the past several decades, to emphasize collaborative and interdisciplinary research.

In summary, it is clear from the analysis of the set of major scientific and applied breakthroughs in atmospheric science considered in this chapter that NSF ATM has made effective use of its varied modes of support and that the balance between the modes has evolved over time in response to the needs and opportunities of the field. The committee expects that ATM will continue to evolve the balance between its modes of support as atmospheric science and its applications evolve.