Вы находитесь на странице: 1из 24

Science, Technology,

Agrawala et al. / Climate


& Human
Forecasts
Values

Integrating Climate Forecasts and Societal


Decision Making: Challenges to an
Emergent Boundary Organization

Shardul Agrawala
Columbia University
Kenneth Broad
University of Miami
David H. Guston
Rutgers, The State University of New Jersey

The International Research Institute for Climate Prediction (IRI) was created in 1996
with an “end-to-end” mission to engage in climate research and modeling on a seasonal-
to-interannual time scale and to provide the results of this research in a useful way to
farmers, fishermen, public health officials, and others capable of making the best of the
predicted climate conditions. As a boundary organization, IRI straddles the divides
between the production and use of research and between the developed world and the
developing world. This article describes the institutional history of IRI, examining how
the end-to-end mission evolved over time, how it is becoming institutionalized in IRI as a
boundary organization, and the ongoing challenges it presents to managing the bound-
ary between climate variability research and societal applications.

In 1982 and 1983, an El Niño—one of the strongest of the century—


altered climate conditions worldwide and caused flooding, famine, and dis-
ease that cost thousands of lives and an estimated $13 billion in damages
(Broad 1983). Characterized by anomalously warm waters in the tropical
Pacific, El Niño, and its cold-water counterpart, La Niña, are part of a natural,
irregularly recurring climate cycle called the El Niño Southern Oscillation
(ENSO). ENSO “events” alter ocean currents, winds, temperature, and rain-
fall patterns, and they are second only to the seasons themselves in their

AUTHORS’ NOTE: We gratefully acknowledge comments from Phil Arkin, Reid Basher, Mark
Cane, Anne Depaigne, Antonio Moura, Ed Sarachik, Steve Zebiak, and two anonymous
reviewers.

Science, Technology, & Human Values, Vol. 26 No. 4, Autumn 2001 454-477
© 2001 Sage Publications
454
Agrawala et al. / Climate Forecasts 455

influence on global climate patterns (Glantz, Katz, and Nicholls 1991;


Ropelewski and Halpert 1987).1
Early in 1986, Mark Cane and Steve Zebiak, climate modelers at Colum-
bia University’s Lamont-Doherty Earth Observatory (LDEO), discerned in
their mathematical model the signature of another El Niño event for 1986-87.
Because their model had “hindcasted” the 1982-83 and other previous ENSO
events well, they felt confident enough to go public with their prediction for
1987. First published in Nature (Cane, Zebiak, and Dolan 1986) and publi-
cized via a press conference and an article in the New York Times (Eckholm
1986), Cane and Zebiak’s forecast drew heavy criticism from some of their
peers who felt that the reliability of such predictive models was insufficient to
justify the dissemination of results outside of scientific circles. However, the
1987 El Niño did occur. This correct prediction, and the subsequent success-
ful prediction of a weaker ENSO event for 1991-92, gave credibility to the
prospect of seasonal climate forecasting.
Because such events are often associated with characteristic effects on
temperature and rainfall patterns in many parts of the world,2 there were some
initial modest uses of forecasts of the onset of an ENSO. Some members of
the agricultural sector in Peru began to use ENSO forecasts when choosing
whether to plant rice or cotton for the upcoming season (Lagos and Buizer
1992), although the extent of the use of the forecasts is currently under debate
(Lemos et al. 1998).3 Members of the climate forecast community concluded
that climate predictions would be most useful if they were issued on a regular
basis, as regularity would help cultivate a clientele and would also help scien-
tists better assess, and potentially improve, their forecasting skills. An influ-
ential paper correlating ENSO with maize yields in Zimbabwe further
concretized notions that forecasting an El Niño or La Niña could, in fact, be
linked to decision making in agriculture and food security (Cane, Eshel, and
Buckland 1994).4 Such thinking spurred further initiatives by scientists and
government agencies engaged in seasonal forecasting to produce seasonal
climate forecasts and make them available to improve societal responses
worldwide.
This article describes the subsequent efforts to institutionalize such a
capacity in what became known as the International Research Institute for
Climate Prediction (IRI). It focuses on IRI as an example of a boundary orga-
nization (Guston 1999; see also the introduction to this special issue)—an
organization that straddles the shifting divide between politics and science,
drawing its incentives from and producing outputs for principals in both
domains, and internalizing the provisional and ambiguous character of the
distinctions between these domains. It analyzes in detail how the various pro-
ponents of such an institute developed a plan for the prediction and
456 Science, Technology, & Human Values

application of forecasts for societal benefit—the “end-to-end” mission, as it


became known. This article assesses how IRI attempts to negotiate the end-
to-end mission by responding to principals in both producer and user com-
munities. Finally, it identifies and discusses the challenges that IRI, as a still-
emerging boundary organization, faces in attempting to implement its end-
to-end objectives. The article draws on extensive documentary review, inter-
views with IRI staff and key actors in IRI’s development, and the experience
of two of the coauthors at the institute.

Origins of an International Research Institute

Most new organizational initiatives have multiple origins, and the IRI is
no exception. Not only is there a set of multiple actors, but there is also a set of
multiple motives that bear on the forging of links between the producers and
users of climate research. Table 1 provides a summary time line of major
activities in the development of IRI.
The extraordinary 1982-83 El Niño catalyzed scientific and political
interest in developing a forecast capability, highlighting the relevance and
accelerating the planning of the multinational Tropical Ocean and Global
Atmosphere (TOGA) research program. TOGA, in operation from 1985 to
1994, and the monitoring network put in place in the tropical Pacific, greatly
improved the understanding of short-term climatic fluctuations.5 Recog-
nizing that technical advances in prediction had created new opportunities,
the international TOGA Scientific Steering Group recommended in 1989
“the creation of a Center or centers for the study of climate predictability and
the development of quasi-operational prediction schemes” (quoted in
National Oceanic and Atmospheric Administration [NOAA] 1995b, 198).
Over subsequent years, the Intergovernmental TOGA Board (ITB)—jointly
established by the World Meteorological Organization and the Intergovern-
mental Oceanographic Commission—specified that such an organization
“could be linked to a distributed network of regional centers whose role it
would be to interpret and validate model predictions for regional applica-
tions” (quoted in Moura 1995, 3). In 1989, the TOGA Program Advisory
Panel of the U.S. National Research Council supported such an idea and, in
1990, recommended the establishment of such a center.
Meanwhile, in the U.S. government, as early as 1987, the Office of Global
Programs (OGP) of NOAA began to contemplate addressing global climate
change by building an infrastructure of laboratories that would be a hybrid
between government and academia. The director of OGP, J. Michael Hall,
decided to refrain from pushing such a program until his budget reached a
Agrawala et al. / Climate Forecasts 457

Table 1. Time Line of Relevant International Research Institute


for Climate Prediction (IRI) Events

1985 Ten-year Tropical Ocean and Global Atmosphere (TOGA) Program


initiated.
1986 Mark Cane and Steve Zebiak predict El Niño event for 1987, which
materializes.
1989, National Research Council (NRC) advisory panel for TOGA program
April promotes concept of international center for climate prediction.
1989, International TOGA Scientific Steering Group (TOGA/SSG)
September recommends creation of climate prediction center(s).
1990, Intergovernmental TOGA Board (ITB) notes TOGA/SSG
January recommendation and suggests a model with a central node and
distributed centers.
1990, NRC panel recommends actions to pursue establishment of such a
Spring center.
• ITB recommends establishment of a multinational center.
• U.S. National Oceanic and Atmospheric Administration (NOAA),
Office of Global Programs (OGP), asks ITB chairman Antonio
Moura to chair a task force to prepare a proposal for such a
center.
• United States expresses interest, at the U.N. Conference on
Environment and Development, in collaborating to create an
international research institute for climate prediction.
1992, United States declares implementation of “a pilot project . . . and
June invites government officials and scientists from all interested nations
to join in developing” IRI.
1992, July Moura’s task group issues its detailed proposal for an international
research institute for climate prediction.
1994, NOAA develops the Proposal to Launch a Seasonal-to-Interannual
September Climate Prediction Program.
1995, NOAA issues a request for proposals for seasonal-to-interannual
March climate prediction program research centers.
1995, NOAA sponsors the International Forum on Forecasting El Niño:
November Launching an International Research Institute in Washington, DC.
1996, June Award of grant to create IRI to Columbia University and Scripps
Institution of Oceanography and the University of California,
San Diego.
1996, Antonio Moura becomes director of IRI.
November
1998, July Groundbreaking on the Monell Building for IRI at Columbia’s
Lamont-Doherty Earth Observatory (LDEO) campus.
1999, Fall IRI Applications Division staffed and running; consolidation of
LDEO and Scripps groups at LDEO.
2000, All IRI divisions move to the Monell Building.
January

NOTE: Time line in part from NOAA (1995b).


458 Science, Technology, & Human Values

sufficient level, but in 1989, he began his advocacy. Hall’s original concern
had been greenhouse warming, but he soon concluded that global warming,
“as posed by scientists, might be a cul-de-sac programmatically.” Believing
that “scientists had to cooperate at the interannual timescale in order to make
people’s lives better,” Hall decided to focus on ENSO rather than global
warming.6
A motive thus shared by the international and domestic actors was advanc-
ing the state of the science through the synergy fostered by a multi-
disciplinary approach to applications. Even at this early stage, “the notion of
‘end-to-end’ was already in the air,” and some actors realized that “this pro-
cess is not linear” and that “research, prediction, and application need to be
done in parallel” for each to be done best.7 The director general of the Brazil-
ian Meteorological Service, Antonio Moura, who chaired ITB (and is now
director of IRI), had noted the “gap between the knowledge generated and the
service to society.” He feared that the science would be incomplete without
applications, and he thus sought to interest colleagues in placing the
researchers involved with prediction and people interested in applications
“under the same roof.”8
A second motive shared by the international and U.S. actors was making
the best of their own bureaucracies. TOGA sought a way to deliver human
benefits from its research program.9 OGP was operating in a relatively con-
servative domestic political context, as the Bush administration tried to avoid
cuts in greenhouse emissions by de-emphasizing anthropogenic global
warming and instead emphasizing the establishment of international mecha-
nisms to foster collaborative research to reduce scientific uncertainty. The
idea for an international research institute on seasonal climate variability
gained currency within this broader political context. Focusing on ENSO was
more politically feasible because neither the origins of the phenomenon nor
the human impacts were in dispute. Moreover, creating the technical and
societal capacity to deal with ENSO impacts meant dealing with climate
variability, which would also facilitate adaptation to the impacts of global
warming.10
A third motive underlying these two was the notion of equity. Over the
previous decade, large international research programs (such as TOGA) had
not sufficiently engaged or benefited developing countries struggling to build
their own research capabilities. The moral debt of the developed to the devel-
oping world had arisen as a bargaining point in negotiations leading up to the
1992 United Nations Conference on Environment and Development
(UNCED), held in Rio de Janeiro.11 The creation of a “scientific commons”
to which the developed world might differentially contribute and from which
Agrawala et al. / Climate Forecasts 459

the developing world might differentially benefit offered an attractive oppor-


tunity (Hall 1999, 48).
These actors and motives converged in 1991 when—at ITB’s behest—
Hall asked Moura to head a task group to prepare a proposal to establish an
institution that would combine scientific advances in seasonal climate fore-
casting with societal decision making. The United States furthered its com-
mitment at UNCED by announcing its intention to lead a collaboration creat-
ing an international research institute for climate prediction.12 In July 1992,
Moura’s group issued its detailed proposal for an international research insti-
tute for climate prediction (NOAA 1992). NOAA (1994) transformed it into a
more formal Proposal to Launch a Seasonal-to-Interannual Climate Predic-
tion Program and launched a two-pronged pilot project: a climate-forecast-
ing program at the Scripps Institution for Oceanography at the University of
California, San Diego, and an international applications training program at
Columbia’s LDEO (Berri 1997).13 NOAA (1995c) then issued the request for
proposals (RFP) to establish an international research institute for climate
prediction and awarded the grant to Columbia and Scripps over two other
major competitors.14

Evolution of IRI’s End-to-End System

Shortly after IRI’s creation, an editorial in Nature identified it as a proto-


type for addressing complex scientific issues of societal concern and, more
pragmatically, using the potential for social relevance as leverage for attract-
ing funds in the (post–Cold War) era in which scientific merit alone was an
insufficient justification for expensive programs in stagnant budgets. “Its
progress should be worth watching,” the Nature article concluded (Anony-
mous 1997, 1).
IRI’s core conceptual innovation—which links research to social con-
cerns—is the notion of an end-to-end system running from climate research-
ers to consumers of climate information, and back again. This collaborative
mode between the producers and consumers of research helps identify IRI as
a boundary organization. Indeed, NOAA’s director Hall describes IRI as an
organization at “two boundaries.” One boundary is in the midst of the end-to-
end mission, between fundamental research and societal applications. The
second is the boundary between the developed and the developing world, as
the research and its technologies are primarily from the developed and the
applications are increasingly targeted to developing countries that are partic-
ularly vulnerable to the impacts of seasonal climate variability (Agrawala
and Broad, forthcoming).
460 Science, Technology, & Human Values

At these boundaries, IRI participates in both the new and old versions of
technology transfer, respectively. In the new version, IRI is actively con-
cerned with moving research beyond the laboratory and facilitating societal
use of information, rather than just assuming that the research results will
flow into applications.15 In the old version, IRI participates in the transna-
tional flow of technical knowledge and skills, usually along a gradient from
North to South. Although implicit in the early planning discussions, the end-
to-end concept evolved over time. As it has become explicit in managing the
two boundaries and the two aspects of technology transfer, however, it has
also raised concomitant challenges.

Origins and Evolution of the End-to-end Concept


Questions regarding the extent to which scientists should reach out to
potential users of their seasonal climate predictions, and whether the needs of
users should shape climate research, began to emerge shortly after the poten-
tial for making El Niño forecasts had been demonstrated. A 1991 report
pointed out,

It would be valuable if the scientists working on interannual [climate] variabil-


ity had a clearer idea of the concerns of those who could use climate predictions
as input to policy decisions. What would they like to see predicted? What form
should forecasts take to be most useful? (Cane and Sarachik 1991, 33)

The very first proposal for an international research institute for climate
prediction, produced a year later, viewed the institute as an energetic core
with ties, on one hand, to a network of research groups engaged in climate
forecasting and, on the other, to regionally distributed “application centers”
where forecast information would be refined and translated into forms rele-
vant for decision making (Moura 1992). Although this proposal identified
both ends, the phrase “end-to-end” itself first appeared two years later when
the United States proposed an international research institute for climate pre-
diction within the context of a highly decentralized international network of
institutions constituting a seasonal-to-interannual climate prediction pro-
gram (SCPP). The SCPP approach, however, was more end-to-middle than
end-to-end, as it went only from “the origins [of climate variability] through
its physical manifestations to its socioeconomic impacts” (NOAA 1994, vii).
The actual use of forecast information in mitigating or adapting to adverse
impacts, the information needs of potential users, and how such needs should
factor back into climate research priorities were not explicitly included in the
SCPP. Nevertheless, the phrase itself took hold.
Agrawala et al. / Climate Forecasts 461

The RFP that NOAA issued in March 1995 adopted the end-to-end lan-
guage of the SCPP. The RFP acknowledged that the goal of the desired “mul-
tinational infrastructure” would be “to generate and transfer useful climate
information and forecasts” and specifically “to prepare and disseminate reg-
ularly an experimental forecast to all interested countries.” It specified that
“Regional Applications Centers, located around the world, . . . will refine the
forecast based on analyses of local and regional conditions and distribute
products of social and economic benefit to the users” (NOAA 1995c, 13407).
Of the six specific areas NOAA suggested “should be addressed in the pro-
posal,” three involved the production of forecast models and three involved
their dissemination and use. Only one of these, “support for the preparation
for new forecast guidance products based on Application Center needs”
(NOAA 1995c, 13408), contained any hint of the reflexive or two-way nature
of the applications chore.
The International Forum on Forecasting El Niño, convened in November
1995 (after the proposals were due to NOAA) to launch IRI in NOAA’s
vision, systematically addressed the question of what end-to-end might mean
operationally. This meeting, the first truly international and multidisciplinary
one, resulted in a more sophisticated notion of an end-to-end process than had
been envisioned by the small community of primarily U.S. physical scientists
who were IRI’s leading advocates. Working Group 2, one of the three such
groups convened at this forum, identified the components of an end-to-end
system as three concentric circles: (1) a core focusing on fundamental
research on climate, human, ecosystem, and decision processes; (2) an appli-
cations ring focusing on climate forecasts, decision analysis, and communi-
cation; and (3) an outer outreach shell connected with decision makers in var-
ious sectors, such as farmers, water resource managers, disaster relief
agencies, and insurers. Among the most significant recommendations from
Working Group 2 was that “IRI requires a multidisciplinary research core
that includes all aspects of generating and applying forecasts, and a range of
demonstration projects on application of climate forecasts” (NOAA 1995a,
183).
The end-to-end research agenda, meanwhile, was spelled out in signifi-
cantly greater detail in an article by Sarachik and Shea (1997). The authors
identified specific research priorities within three component elements:
physical climate prediction, consequences of short-term variability, and
applications of predictions. Despite the new terminology, the components of
this research agenda closely resembled the “integrated assessment” agenda
of global climate change research. For example, the U.S. Climate Impact
Assessment Program of the early 1970s and, more recently, the Intergovern-
mental Panel on Climate Change (IPCC) were both similarly organized
462 Science, Technology, & Human Values

around climate prediction, impacts, and responses (Agrawala 1998a, 1998b).


However, unlike IPCC—which only assesses research from end-to-end—
IRI’s advocates intended it to help coordinate, conduct, implement, and eval-
uate such research. Moreover, the compressed time scale of climate variabil-
ity posed a set of challenges very different from those of global warming. For
research on seasonal climate variations and their impacts to be usable, it must
reach a large, diverse, and dispersed group of potential users in time to influ-
ence their seasonal decisions about planting, land and water use, and so forth.
In contrast, global warming research and assessments have sought to influ-
ence international negotiators operating under the aegis of the Climate Con-
vention, where progress is almost glacial and lacks the temporal urgency of a
subsistence farmer’s decision to plant certain crops based on expectations of
rainfall over the next few months. Thus, the novelty of end-to-end research
was in the different context in which its implementation was sought, as well
as in the concept of performing research (rather than assessment) across the
spectrum.

Operationalizing End-to-End: The Experience Thus Far


In its first three years of operation, IRI has recruited research staff for its
four divisions (modeling, forecasting, monitoring and dissemination, and
applications) and consolidated its bicoastal operations at Columbia Univer-
sity’s LDEO. It simultaneously labored to provide timely forecast informa-
tion about the 1997-98 El Niño and the subsequent 1998-2000 La Niña
events, both of which had significant consequences worldwide. Without
being conscious of the broader theoretical implications of its activity, IRI has
during this period created two products that facilitate the connection between
climate forecasts and their social applications.
IRI produces “net assessment” seasonal forecasts for all regions of the
world based on an evaluation of observational data and the often-divergent
results from a variety of climate models. These net assessments are available
on the World Wide Web and are also presented at regional meetings, called
Climate Outlook Fora, that are organized at the start of key seasons in many
regions in Africa, South America, Southeast Asia, and the Caribbean.
Selected portions of the forecast are also communicated through the monthly
Climate Information Digest, published by IRI, which summarizes key devel-
opments in the climate system and their impacts worldwide. Net assessment
maps follow a standardized graphical format representing the probabilities
that the total rainfall or the average temperature over a three-month season
will be above, near, or below normal. They are a novel product in that they
communicate the best available forecast information, synthesized from
Agrawala et al. / Climate Forecasts 463

leading climate models, with associated scientific uncertainties. IRI has


assumed a key role in empowering particular regions to produce their own
consensus forecasts.
During its pilot phase of activity, IRI also developed an interactive soft-
ware package called Climlab. Distributed free of charge, Climlab allows
users to perform statistical manipulations of local and regional climate data
and to explore the relationship between climatic variables and data specific to
key sectors such as health, water resources, agriculture, and fisheries.16 IRI
also has the comprehensive, Web-accessible Climate Data Library, where
users can manipulate, view, and download a variety of climate-related data
including ocean temperatures and station rainfall and temperature data from
several thousand stations worldwide. These tools play an important role in
IRI’s training and capacity-building efforts, in which it collaborates with
research groups, governments, and international and regional organizations
to host training programs for regional climate forecasting and applications
experts. Typically lasting one to two weeks, the training programs seek to
improve regional capacity to make and apply climate forecasts.
The net assessments, Climlab, and the Climate Data Library are clearly
like boundary objects (Star and Griesemer 1989) or standardized packages
(Fujimura 1992). The critical question for both scholarship and IRI, however,
is the extent to which they have begun to change behaviors on either or both
ends of IRI’s continuum. In part because the applications research group,
comprising the social scientists and sector specialists whose work is expected
to link climate research with social decision making, reached critical mass
only in late 1999, there has been no systematic evaluation of these activities
or documented understanding of interactions between producers and con-
sumers.17 At the time of writing, IRI is therefore still in the early stages of
implementing the end-to-end vision of its creators.
The participation of some IRI scientists in a large multidisciplinary effort
begun in 1996 provided an early prototype of this reflexive activity. The pro-
ject explored the sensitivity of the fisheries sector in Peru and how it uses
forecast information to cope with seasonal variations. In addition to helping
identify societal constraints in the use of forecast information, this effort also
provided the opportunity for an early dialogue in which social scientists
transmitted the information needs of user communities back to the climate
modelers at IRI (Broad and Pfaff 1997).
Even in deciding where to conduct training, IRI must identify regions
where predictive skill exists and then identify key decision makers and sec-
tors to target for education. Only after achieving an understanding of the
region’s basic social structure can IRI design an appropriate training curricu-
lum. Through experience from this and other projects—and as the framework
464 Science, Technology, & Human Values

of boundary organizations suggests—it is clear that the connection between


IRI’s activities in modeling, assessment, and applications cannot be
unidirectional.
IRI has recently been attempting the systematic design of projects that
integrate the end-to-end vision. Work on three such integrated projects is cur-
rently under way. One builds on the earlier work in Peruvian coastal fisher-
ies. A second is intended to build the capacity to produce and apply climate
forecasts in the Greater Horn of Africa. The third focuses on both short-term
and longer-term strategies to reduce vulnerability to drought in the state of
Ceará in Northeast Brazil. These integrated projects follow a two-pronged
approach: (1) the identification of a particular user group’s informational
needs about seasonal climate and the establishment of a collaborative dia-
logue between the user group and the physical scientists to identify strategies
and forecast products that can better meet these needs, and (2) the identifica-
tion of the constraints facing the use of currently available climate forecast
information and the development of strategies to reduce these constraints
through improving communication and dissemination channels, demonstrat-
ing the utility of such information, and building additional capacity for the
local conduct and application of climate research.18

Challenges of the End-to-End Mission

The successful implementation of IRI’s end-to-end mission faces several


challenges. Some emanate from the complexities of the physical and socio-
economic systems that the institution is intended to bridge. Still others have
emerged from the ideological, institutional, and political context in which
IRI was created and now operates. A third cluster of challenges revolves
around the integration across natural and social science disciplines.

Limitations in Scientific Understanding of the Physical System


A brief review of the steps involved in the seasonal climate forecasting
process demonstrates the complexity of this endeavor. IRI employs what is
known as a two-tiered approach (Mason et al. 1999). First, mathematical
models use observational data on current ocean temperatures and atmo-
spheric conditions to forecast surface temperatures in the tropical Pacific.
Second, this information is input to other models to predict regional climate
patterns worldwide for the upcoming three-month season. This forecasting
scheme employs models that incorporate the physical dynamics of the oceans
and the atmosphere, as well as ones based on statistical regressions. The
Agrawala et al. / Climate Forecasts 465

performance of particular models varies from region to region and season to


season and might also be a function of the climate variable being forecast.
Results are also extremely sensitive to even small changes in some of the
input variables such as winds and ocean temperatures.
Because of chaos in the climate system, limitations of the various climate
models, and uncertainties in the observational data used to drive them, sea-
sonal forecasts are necessarily probabilistic. As described earlier, IRI net
assessments present forecasts in terms of the probabilities that the total rain-
fall (or the average temperature) for three-month periods over fairly large
regions will fall into the three “tercile” classes—that is, the wettest or hottest
third of years, the normal third of years, and the driest or coolest third, relative
to the historical record. A typical seasonal forecast, for example, might
assign a 45 percent chance that the total three-month rainfall in a region span-
ning several hundred square miles would be above normal, a 30 percent
chance that total rainfall will be near normal, and a 25 percent that it will be
below normal.
This attempt to create a standardized package in the standardized commu-
nication of the net assessment, however, has limitations. Not only do many
users desire greater specificity in terms of probabilities, but they also often
operate on considerably smaller spatial scales: a farm plot, a river watershed,
an administrative district, and so on. The temporal distribution is also prob-
lematic: even rainfall that might classify as normal when totaled over three
months could be catastrophic if it were to fall over just a few days. Further-
more, many users are sensitive to climate extremes, that is, rainfall signifi-
cantly above or below normal. Pulled by this demand among potential users,
IRI also started producing forecasts of such rainfall extremes in 1998.19
Finally, while IRI produces forecasts for all regions, for all seasons, and for
all years, much of the skill of its predictive tools resides only in certain
regions, certain seasons, or years when a strong El Niño or La Niña event is
under way. This specificity exists because current models do not adequately
account for variations other than the ENSO cycle that may have a significant
influence on climate at a regional level. However, credibility with user groups
depends on a sustained track record of reasonably good forecasts in the
region and the scale of the user’s interest. There might thus be an upper bound
to the degree of success possible in IRI’s end-to-end efforts, imposed by lim-
ited understanding of the climate system’s complexity.

Complexities of User Environments


IRI’s mission emphasizes reducing the vulnerability of populations to
hardships caused by drought, floods, disease, and economic instability,
466 Science, Technology, & Human Values

which are often influenced by fluctuations in climate. Even with a hypotheti-


cally perfect forecast, IRI still faces a still more daunting array of societal
constraints long recognized in the economic development literature (see,
e.g., Scott 1998; Sen 1984). The recognition of such constraints is now being
brought to bear on the utilization of climate information (Patt, forthcoming;
Broad and Agrawala 2000; Broad, Pfaff, and Glantz, 2001; Roncoli 2000;
National Research Council 1999; Orlove and Tosteson 1999; Finan 1998;
Rayner and Malone 1998).
As IRI began to engage in applications research and specific projects, it
has moved away from climatic determinism as it became clear that advance
knowledge of the climate was just one of multiple interacting factors that
mediate a group’s ability to prepare for and respond to extreme events. Many
of the regions where climate is most predictable are poor countries in the
tropics, with extreme levels of poverty, corruption, civil strife, and political
instability. Consequently, they are often ill equipped to make use of such
information effectively. A weak state infrastructure and competing resource
needs, combined with a lack of fit in scale between the current lead time of
climate forecasts and institutional (governmental) ability to respond, limits
the efficacy of forecasts (see Orlove and Tosteson 1999). Even within a par-
ticular region, the populations most vulnerable to disease and natural disaster
are often those living in marginal lands or precarious urban settings (e.g., in
shantytowns in dry riverbeds or on unstable hillsides). In such cases, signifi-
cantly reducing vulnerability is an issue of long-term planning and gover-
nance, and a forecast with a short lead time will often be insufficient for most
bureaucracies to arm a comprehensive response.
Furthermore, groups within societies have differential access to and abil-
ity to understand information, especially as the relevant climate information
provided by the IRI is probabilistic and available mainly via the Internet.
Response necessitates that intermediary organizations get the information to
the decision makers in a timely manner. In the situation of a weak state, the
mass media most often fill this role. Experience during the 1997-98 ENSO
has also shown that understanding the probabilistic nature of the information
has a high degree of difficulty in part because of journalists’ trouble with
communicating scientific uncertainty.20 Furthermore, forecasts may be
rejected outright if they are at odds with locally generated forecasts based on
natural indicators, lived experience, and traditional practices or if their source
is not a trusted one.
On a household level, the most vulnerable populations usually lack the
resources to act on advance information. For instance, small-scale farmers
lack the capital to change inputs, have limited access to drought-resistant
seed, or are reliant on others for timing their planting (e.g., they require
Agrawala et al. / Climate Forecasts 467

external sources of labor for plowing, etc.). Furthermore, most of these small-
scale producers are reliant on existing markets and prices set elsewhere, lim-
iting their ability to alter their decision-making schemes even with new infor-
mation. Finally, the variation between crops, inputs, and terrain of small-
scale farmers makes linking climate and crop models extremely challenging,
given the resolution of forecasts and the diversity of concerns among small-
scale farmers. The precarious day-to-day existence in which many of the
world’s poorest populations—including subsistence farmers and fisher-
men—live not only limits their ability to respond to forecasts but also renders
them extremely vulnerable to poor forecasts.

Dilemmas of Research versus Intervention


The scientific, societal, and institutional constraints on the use of forecasts
give rise to ethical concerns that have caused IRI to reassess its policy on fore-
cast dissemination and applications activities. These include concerns over
issues of equity, sovereignty, and the unintended consequences of forecasts.

Equity. Even if scientific information is a public good, it is not a free good


(Callon 1994), and certain groups may take advantage of others within soci-
ety because of their own superior access to and understanding of climate
information. By disseminating information without first identifying the dif-
ferent potential stakeholders and their differential ability to take advantage of
information, IRI risks favoring one group over another (Pfaff, Broad, and
Glantz 1999; Broad 2000). Most commercial enterprises are better poised to
use the currently provided information than are small-scale (household-
level) users, although the latter group is more vulnerable to climate variabil-
ity. If IRI adopts the economic philosophy that by supporting the large com-
mercial enterprises it will ultimately strengthen the state and its ability to care
for the more marginal populations, it may choose to target such firms, accept-
ing that smaller firms are relatively unequipped to adapt to probabilistic cli-
mate information. However, as many countries move into an extreme free-
market economy with minimal state intervention, such a decision could leave
IRI vulnerable to criticism for simply making the rich richer and the poor
poorer.
Another example involves intergenerational equity. The use of climate
information can help industrial fishermen along parts of the South American
coast anticipate where concentrations of fish stock will be. Thus, they may
increase their extractive capacity to the detriment of small-scale fishing
groups who compete for similar species but lack access to this information or
the equipment necessary to pursue the fish. But overexploitation of this
468 Science, Technology, & Human Values

resource, from increased fishing efficiency, can harm future generations, and
resolving the equity issue by assisting the small-scale fishing groups may
even further deplete the resource.21
A more direct ethical dilemma arises as IRI chooses where to engage in
activities. Should IRI provide technical assistance to nations with repressive
regimes just because there is an opportunity to apply forecast information? A
similar dilemma arises in a regional context. IRI could unintentionally insti-
gate conflict between states over common property resources such as water
or fisheries, or influence the outcome by providing advance information to
one party, thus enabling it to take actions—such as damming water upstream
in expectation of a drought—that harm another party.

Sovereignty. The transboundary nature of climate forecasts raises the sen-


sitive issue that IRI faces in respect to impinging on another state’s sover-
eignty. Experience in some countries has shown that national meteorological
services or scientific agencies are often underequipped to generate and dis-
seminate climate information. In fact, they may choose to withhold informa-
tion from the public (e.g., privatization of public services, where data must be
purchased). Even if IRI were in a position to communicate more directly with
end users, it might be politically hazardous to circumvent governmental
channels. For example, in the global climate change debate, early initiatives
to influence policy ended up being marginalized as they attempted to bypass
national governments (Agrawala 1999). Maximizing the effectiveness of its
forecasts to end users while constructively engaging national governments is
particularly challenging for IRI because some countries are dubious of
“altruistic” U.S. interventions in their internal affairs, and IRI still seems like
a U.S. institution.
Another scenario could create conflict with state sovereignty. If applica-
tions research indicates the potential positive use of a climate forecast, yet the
formal and informal policies of a state are identified that will result in the
forecast’s not being used in the manner envisioned, should IRI attempt to
intervene and influence policy, hand over the best information possible and
leave its use up to the country, or simply decide not to collaborate with that
country?

Unintended consequences. Perhaps the most significant challenge IRI


faces is the possibility of taking extensive measures to minimize these prob-
lems but still encountering negative results because of changes in financial
and political systems, the misrepresentation of forecast information, or other
exogenous factors. As mentioned earlier, limits to the understanding of the
climate system will result in poor forecasts, potentially undermining
Agrawala et al. / Climate Forecasts 469

confidence in the forecasts and in their provider. While six out of ten good
forecasts may be a skillful record for a forecast system, even one poor fore-
cast on which important decisions are based could be disastrous, particularly
for users such as subsistence farmers.
Perverse effects, where the recipients of forecasts take action that may
protect their interests but causes more cumulative harm, are also possible.
During the 1997-98 El Niño, banks in several countries cut loans across the
board based on the rumors of anticipated climate impacts (Pielke 2000).
Credit shortages undermined the ability of individuals and firms to take their
own measures, for example, by planting rice instead of cotton in anticipation
of heavy rains, to respond to the forecasts.
Operationalizing climate forecasts in policy making may also shift the
policy focus toward the short-term management of natural disasters based on
advance warning rather than on long-term solutions to infrastructure devel-
opment, land use, water management, public health systems, and other fun-
damentals that underlie many natural disasters (Cuny 1994). The occurrence
of disasters in the face of an imperfect forecast leaves IRI vulnerable as a
scapegoat. Given its current forecasting skill, IRI must think hard about
whether it should encourage reliance on its forecasts regularly or only in
extreme ENSO events in which it might have more confidence in the fore-
cast.22

Institutional Heritage and Operational Context


IRI’s development and work continue to be influenced in large part by
NOAA, its primary patron, and Columbia University, its well-endowed aca-
demic home.23 Although IRI has signed agreements that facilitate collabora-
tive research and data sharing with numerous countries and research institu-
tions, it remains financially reliant on NOAA (and to a lesser extent on
Columbia), and its ability to foster a reputation as a truly international institu-
tion remains limited. A recent agreement under which Taiwan pledged $7.5
million over five years to IRI has alleviated the situation somewhat, but the
continued perception of IRI as a U.S. institution increases skepticism of its
underlying goals by those sensitive to U.S. foreign-policy initiatives.
The dependent relationship with NOAA also initially shaped the applica-
tions research agenda, as IRI responded to pilot projects that NOAA had pre-
viously established. On one hand, the pilot projects provided IRI with preex-
isting networks to tap, but on the other hand, they encouraged a reactive mode
at IRI and may have committed resources into paths it might not otherwise
have taken. IRI is also subject to the legal constraints that accompany funds
from the U.S. government, as well as Columbia’s regulations. This dual
470 Science, Technology, & Human Values

context can be beneficial in the implementation of applications projects that


necessitate flexibility and timeliness in arranging contracts, transferring
money overseas, and seeking funds from other U.S. and international agen-
cies. But outside sources of funding are subject to the university’s significant
overhead costs. Employees of IRI are university employees and are subject to
academic employment categories, wages, incentives, and evaluation criteria
that may be at odds with IRI’s mission-oriented approach. The university is,
however, taking measures to remove these obstacles, but changes are slow.
Operating in a university setting, meanwhile, offers access to a range of
experts and to students to work on various aspects of the end-to-end problem.
Through initiatives such as flexibility in allotting discretionary funds and the
creation of permanent (e.g., tenure-track) positions through the Earth Insti-
tute, Columbia is increasingly providing incentives for faculty and students
to involve themselves in the multidisciplinary work IRI conducts. Through
its position at a university, IRI is under continuous scrutiny for its intellectual
integrity and the social implications of its actions. The stable funding base,
along with the pressure to maintain high academic standards, has fostered a
strong reflexive component in the IRI development process that is uncom-
mon among other institutions.

Ideological Heritage and the Challenges


to Multidisciplinary Integration
The idea of an end-to-end system to link advances in climate forecast
capabilities with social applications emanated from a small network of physi-
cal scientists engaged in climate modeling and prediction efforts. A combina-
tion of confidence in their own ability to make seasonal forecasts, optimism
about improvement in the quality and resolution of such forecasts, an ideal-
ized vision of benefiting society, writ large, and a degree of naïveté about the
complexities of implementing such a social mandate together led to the bold
institutionalization of an end-to-end vision at IRI. There have been relatively
few opportunities thus far for alternate paradigms to shape its vision. The
early conceptions of IRI did not include a social science research agenda, and
it was not until the 1995 forum that social scientists—a handful out of more
than one hundred participants—were first involved. As IRI has more fully
staffed its Applications Division, it may better address some of the complexi-
ties of forecast utilization.
Although conceived of as applications research, it had been felt early in
IRI’s history that forecast applications would be at regional and local levels
worldwide and would therefore need to be decentralized. Consequently, the
1992 ITB proposal only envisaged a small applications unit within the core
Agrawala et al. / Climate Forecasts 471

IRI facility to coordinate and manage activities of a distributed network of


regional applications centers. It was only the 1995 International Forum,
which launched IRI, that first recommended that the institute have in-house
capacity to do research related to applications. Thus far, applications efforts
have been directed at collaborative research, management, and implementa-
tion of targeted country-specific projects. But there is a more fundamental
question of whether a metalevel social science research agenda, beyond the
facilitation of two-way interaction between the climate forecasters and end
users, is necessary. IRI is just beginning to formulate a longer-term social sci-
ence research agenda that strikes a balance between research and operational
deadlines and that better integrates physical and social science research.
The suite of challenges to IRI’s successful implementation of its end-to-
end mission parallel the challenges to other boundary organizations, such as
offices of technology transfer in the United States (Guston 1999, 2000). Lim-
itations in the science of seasonal-to-interannual climate forecasting mean
that the technology may not be in the best position to transfer. Although it
may be flexible enough to permit local adaptations, it seems more likely that
it is still too imprecise to encourage application, particularly at a local level.
Some research simply does not have a market, and the choice of when to
move an innovation is a critical one. Yet even if forecasts were technically
perfect, or the marketing choices were made correctly, the environment into
which the innovation passes is complex and cannot be controlled. IRI cannot
manage all the social processes relevant for the success of the forecast and its
applications. And even proficient interventions, defined technically and nar-
rowly, may have profoundly inequitable, destabilizing, or otherwise unpalat-
able social consequences. To date, IRI attempts to manage these issues on a
project-by-project basis, but a poor choice or poor luck in any particular pro-
ject puts the enterprise at risk. Moreover, IRI does not exist as a freestanding
organization, but it is embedded in the norms, rules, and procedures of gov-
ernment-university relations in the United States, for better and for worse.

Conclusion

As its end-to-end mission suggests, IRI is an emergent boundary organi-


zation. It is situated between the relatively different social worlds of climate
modeling and forecasting on one end, and agricultural, health, and other
social and political decision making on the other. It is beginning to create
products, such as net assessment forecasts, Climlab software, and the Cli-
mate Data Library, that serve as boundary objects or standardized packages.
As it matures, IRI is constructing a space that encourages professionals to
472 Science, Technology, & Human Values

continue to create and improve such products—improving both technical


practice and societal decisions as a consequence. It is accountable to the mod-
eling and forecasting community through its production of new research, its
need for technical collaborators, and its need for data and other technical
inputs. It is simultaneously accountable to a dispersed set of decision makers,
without whose endorsement and, eventually, financial support, IRI will
founder.
The end-to-end mission has taken time to evolve, from a thin and some-
what unidirectional vision promoted by physical scientists to a more robust
and interactive process critically involving social scientists. The interactive
process, however, brings difficulties of its own. Despite both the scientific
plaudits and social benefits that could flow from skillful forecasts that also
have high geographical and temporal specificity, climate modeling and fore-
casting are extraordinarily complex, and forecasts are irreducibly probabilis-
tic. Despite both the technical innovations that create more usable climate
information and the research, training, and capacity building that pave the
way for applications, integrating new information into social decisions can
run afoul of numerous political and ethical considerations outside IRI’s con-
trol. Despite a relatively clear vision of the end-to-end mission and the cre-
ation of a relatively autonomous organization to implement it, engaging in
the reflexive practice of applications research and projects might deviate
from the academic culture of a university setting.
The four years of IRI’s operation are brief compared to the operational
challenges facing it, yet alone its substantive mission to apply climate infor-
mation for social benefit. But IRI has shown considerable adaptability in
adjusting its organizational structure and priorities to better reflect the scien-
tific and political complexities associated with its end-to-end mission. It has
begun to diversify the suite of products it offers to users and, even more so,
the strategies it uses to target and communicate with them. It has also become
more focused on the feedback to producers required by a boundary organiza-
tion to maintain its equipoise. Other important reforms IRI could attempt
include refining its vision of its clientele in an effort to minimize the likeli-
hood of troubles arising from issues of equity or sovereignty; linking more
actively to local institutions, which can legitimate applications more than IRI
may be able to; and working more explicitly on alleviating societal vulnera-
bility to climate variations, which may be more systemic and potentially less
divisive than seasonal forecast application. With such reforms, IRI may
emerge as a fully functional boundary organization and achieve its goal of
integrating climate research and societal applications.
Agrawala et al. / Climate Forecasts 473

Notes

1. Although the scientific community dubs occurrences of El Niño Southern Oscillation


(ENSO) as “events,” such phenomena typically unfold over a period of several months and occur
approximately every two to ten years.
2. The strength of these relationships varies greatly and is strongest closer to the phenome-
non’s origin in the central Pacific. The effects of La Niña—the cold phase of the ENSO cycle—
on seasonal rainfall patterns are often, but not necessarily, opposite to those associated with El
Niño.
3. For a review of other uses of ENSO forecasts, see Glantz (1996).
4. Subsequent research has focused on linking the ENSO phenomenon with fluctuations in
a number of socially significant variables such as the incidence of certain diseases (such as
malaria and dengue), water resources, and fish stocks.
5. For a thorough overview of recent developments in seasonal forecasting and the interna-
tional perspective on their provision and use, see Carson (1998); for a review of the Tropical
Ocean and Global Atmosphere (TOGA) research program, see Anderson et al. (1998).
6. Telephone interview with J. Michael Hall, Director, Office of Global Programs, National
Oceanic and Atmospheric Administration, 23 November 1999.
7. Hall interview.
8. Interview with Antonio Moura, Director, International Research Institute for Climate
Prediction, 5 November 1999, Palisades, New York.
9. The International Research Institute for Climate Prediction (IRI) was, in part, conceived
to “help bring the scientific achievements of the [World Climate Research Programme] to a more
rapid and complete fruition and lead to a fulfillment of societal needs” (Moura and Sarachik
1997, 345).
10. One possible consequence of global warming is increased climate variability, including
stronger ENSO events and other weather extremes.
11. Prior to the United Nations Conference on Environment and Development, the president
of Uruguay, Luis Alberto LaCalle, acknowledged his interest in greenhouse warming but noted
that his greatest responsibility in climate was “to worry about conditions at next year’s harvest”
(quoted in Hall 1999, 45). Such remarks underscore the importance many developing countries
place on the impacts of seasonal climate variability as opposed to climate change, and they
implicitly communicate the message about equity.
12. The logic of the hegemonic United States leading the creation of a global “scientific com-
mons” is worthy of Olson (1971).
13. The 1994 proposal was somewhat less ambitious than the 1992 proposal, in response in
part to domestic political considerations in the United States and in part in consideration for the
roles of national meteorological services, which felt that IRI might impinge on their functions.
14. There was some question about whether to use the normal grant-making process or to
engage in some more elaborate establishment of a variety of pilot centers and the construction of
a more elaborate international infrastructure prior to site selection, but TOGA worried about the
time this method would take, and everyone worried about the possible collapse of political sup-
port once site selection took place, as had happened with the Superconducting Supercollider.
Hall interview.
15. Actively pursuing applications, IRI participates in a new model of research productivity
that supplants the older, passive model. See Guston (2000, especially chaps. 5 and 6) for new
visions of research productivity.
474 Science, Technology, & Human Values

16. The primary focus of Climlab, however, is to explore statistical correlations, which need
not necessarily imply causal linkages.
17. Akin to the use of licensing and marketing experts by previously identified boundary
organizations in new technology transfer (Guston 1999), IRI requires competencies beyond the
researchers’ generating the innovation to be transferred to social scientists, who collaborate with
the producers and the potential users to create opportunities for both applications and the reflex-
ive improvement of research based on users’ criteria. IRI has started to de-emphasize its divi-
sional boundaries to foster a more interactive dialogue between the producers in prediction and
the facilitators in applications.
18. Many of the activities within these integrated projects will be conducted through partner-
ships with donor agencies such as USAID, World Bank, and the U.N. Development Programme;
national and state governments; nongovernmental organizations; and international and regional
scientific and technical organizations.
19. Instead of forecasting probabilities for terciles, IRI forecasts the probability of rainfall
and temperature above the eighty-fifth and below the fifteenth percentiles.
20. See Friedman, Dunwoody, and Rogers (1999) for a recent survey of the media and scien-
tific uncertainty. This literature also points out that scientists have strategic interests in the use of
uncertainty as well. See, for example, Shackley and Wynne (1996) on the specific issue of uncer-
tainty and climate change.
21. To some extent, IRI initially had an idealized notion of potential users, such as small-
scale fishermen or small farmers, although there is increasing recognition that more commercial
or larger-scale organizations will have a greater capacity to apply its forecasts.
22. Note, however, that this proposition contradicts the early rationale that regularity would
make for a more usable forecasting system.
23. Through the creation of the Columbia Earth Institute, the university is attempting to inte-
grate the physical and social sciences with the explicit goal of creating social outcomes through
fundamental changes in basic research.

References

Agrawala, S. 1998a. Context and early origins of the Intergovernmental Panel on Climate
Change. Climatic Change 39 (4): 605-20.
. 1998b. Structural and process history of the Intergovernmental Panel on Climate
Change. Climatic Change 39 (4): 621-42.
. 1999. Early science-policy interactions in climate change: Lessons from the Advisory
Group on Greenhouse Gases. Global Environmental Change 9 (2): 157-69.
Agrawala, S., and K. Broad. Forthcoming. Technology transfer perspectives on climate forecast
applications. Knowledge and Society.
Anderson, D.L.T., E. S. Sarachik, P. J. Webster, and L. M. Rothstein, eds. 1998. The TOGA
decade: Reviewing the progress of El Niño research and prediction. Washington, DC: Amer-
ican Geophysical Union.
Anonymous. 1997. Cooperation will help get the message across. Nature 388(6337):1.
Berri, G. 1997. Pilot project for an international research institute (IRI). ENSO Signal, October,
2-4.
Broad, K. 2000. El Niño and the anthropological opportunity. Practicing Anthropology 22
(4):20-23.
Agrawala et al. / Climate Forecasts 475

Broad, K., and S. Agrawala. 2000. The Ethiopian famine: Uses and limits of seasonal climate
forecasts. Science 289 (5485): 1693-94.
Broad, K., and A. Pfaff. 1997. Prediction, profit-motive and policy in the Peruvian fishery.
Columbia Earth Institute: EARTHmatters, May, 11-16.
Broad, K., A. Pfaff, and M. G. Glantz. 2001. Effective and equitable dissemination of seasonal-
to-interannual climate forecasts: Policy implications from the Peruvian fishery during El
Niño 1997-98. Paper submitted for review.
Broad, W. J. 1983. Tracking El Nino: Is more destruction to come? New York Times, 2 August,
C1.
Callon, M. 1994. Is science a public good? Science, Technology, & Human Values 19 (4): 395-
424.
Cane, M. A., G. Eshel, and R. W. Buckland. 1994. Forecasting Zimbabwean maize yield using
eastern equatorial Pacific sea surface temperature. Nature 370:204-5.
Cane, M. A., and E. S. Sarachik, eds. 1991. Prospectus for a tropical-oceans-global-atmosphere
(TOGA) program on seasonal-to-interannual prediction. Special report no. 4, National Oce-
anic and Atmospheric Administration Climate and Global Change Program, January.
Cane, M. A., S. E. Zebiak, and S. C. Dolan. 1986. Experimental forecast of El Niño. Nature
321:827-32.
Carson, D. J. 1998. Seasonal forecasting. Quarterly Journal of the Royal Meteorological Society
124:1-26.
Cuny, F. C. 1994. Disasters and development. Dallas: Intertect Press.
Eckholm, E. 1986. Meteorologists say El Nino will hit in summer and fall. New York Times, 14
March, B5.
Finan, T. J. 1998. Of bird nests, donkey balls, and El Nino: The psychology of drought in north-
east Brazil. Paper presented at the American Anthropological Association annual meeting,
2-6 December, Philadelphia.
Friedman, S. M., S. Dunwoody, and C. L. Rogers. 1999. Communicating uncertainty: Media
coverage of new and controversial science. Mahwah, NJ: Lawrence Erlbaum.
Fujimura, J. 1992. Crafting science: Standardized packages, boundary objects, and “transla-
tion.” In Science as culture and practice, edited by A. Pickering, 168-211. Chicago: Univer-
sity of Chicago Press.
Glantz, M. H. 1996. Currents of change: El Niño’s impact on climate and society. Cambridge,
UK: Cambridge University Press.
Glantz, M. H., R. Katz, and N. Nicholls 1991. Teleconnections linking worldwide climate anom-
alies. New York: Cambridge University Press.
Guston, D. H. 1999. Stabilizing the boundary between US politics and science: The role of the
Office of Technology Transfer as a boundary organization. Social Studies of Science 29 (1):
87-112.
. 2000. Between politics and science: Assuring the integrity and productivity of research.
New York: Cambridge University Press.
Hall, J. M. 1999. International equity and the earth sciences. Earth Matters, fall, 44-49.
Lagos, P., and J. Buizer. 1992. El Niño and Peru: A nation’s response to interannual climate vari-
ability. In Natural and technological disasters: Causes, effects and preventive measures,
edited by S. K. Majumdar, G. S. Forbes, E. W. Miller, and R. F. Schmalz, 223-38. Easton:
Pennsylvania Academy of Science.
Lemos, M. C., D. Liverman, T. Finan, R. Fox, and N. Renn. 1998. The social and policy implica-
tions of seasonal forecasting: A case study of Ceará, Northeast Brazil. Unpublished progress
report to the National Oceanic and Atmospheric Administration, Silver Spring, MD.
476 Science, Technology, & Human Values

Mason, S. J., L. Goddard, N. E. Graham, E. Yulaeva, L. Q. Sun, and P. A. Arkin. 1999. The IRI
seasonal climate prediction system and the 1997/98 El Nino event. Bulletin of the American
Meteorological Society 80 (9): 1853-73.
Moura, A. D., ed. 1992. International Research Institute for Climate Prediction: A proposal.
Paper prepared in response to a request by the Intergovernmental Tropical Ocean and Global
Atmosphere Board, World Meteorological Organization, Geneva, Switzerland.
. 1995. Guest editorial. ENSO Signal, August, 3.
Moura, A. D., and E. S. Sarachik. 1997. Seasonal-to-interannual climate prediction and applica-
tions: New institutions, new possibilities. WMO Bulletin 46 (4): 342-47.
National Oceanic and Atmospheric Administration (NOAA). 1992. International Research
Institute for Climate Prediction: A proposal. Silver Spring, MD: NOAA, Office of Global
Programs.
. 1994. A proposal to launch a seasonal-to-interannual climate prediction program. Sil-
ver Spring, MD: NOAA, Office of Global Programs.
. 1995a. International Forum on Forecasting El Niño: Launching an International
Research Institute, 6-8 November. Executive summary. NOAA, Office of Global Programs.
. 1995b. International Forum on Forecasting El Niño: Launching an International
Research Institute, 6-8 November. Forum proceedings. NOAA, Office of Global Programs.
. 1995c. Notice: Seasonal-to-interannual climate prediction program research centers,
program announcement. Federal Register 60 (48): 13407-10.
National Research Council. 1999. Making climate forecasts matter. Washington, DC: National
Academy Press.
Olson, M., Jr. 1971. The logic of collective action and the theory of groups. Cambridge, MA:
Harvard University Press.
Orlove, B. S., and J. Tosteson, 1999. The application of seasonal to interannual climate forecasts
based on El Niño-southern oscillation (ENSO) events: Lessons from Australia, Brazil, Ethi-
opia, Peru and Zimbabwe. Working paper 99-3, Institute of International Studies, Berkeley,
CA.
Patt, A. Forthcoming. Understanding uncertainty: Forecasting seasonal climate for farmers in
Zimbabwe. Risk: Decision and Policy.
Pielke, R. A. 2000. Guest editorial: A warning about seasonal forecasting. ENSO Signal 13.
Available: http://www.esig.ucar.edu/signal/13/guest.html.
Pfaff, A., K. Broad, and M. G. Glantz. 1999. Who benefits from climate forecasts? Nature 397
(25 February): 645-46.
Rayner, S., and E. L. Malone, eds. 1998. Human choice and climate change, vol. 1. Columbus,
OH: Battelle.
Roncoli, M. C., ed. 2000. Anthropology and climate change: Challenges and contributions (spe-
cial issue). Practicing Anthropology 22.
Ropelewski, C. F., and M. S. Halpert. 1987. Global and regional scale precipitation patterns
associated with the El Niño/southern oscillation. Monthly Weather Review 115:1606-26.
Sarachik, E. S., and E. Shea. 1997. End-to-end seasonal-to-interannual prediction. ENSO Sig-
nal, May, 4-6.
Scott, J. 1998. Seeing like a state: How certain schemes to improve the human condition have
failed. New Haven, CT: Yale University Press.
Sen, A. 1984. Resources, values, and development. Cambridge, MA: Harvard University Press.
Shackley, Simon, and Brian Wynne. 1996. Representing uncertainty in global climate change
science and policy: Boundary-ordering devices and authority. Science, Technology, &
Human Values 21 (3): 275-302.
Agrawala et al. / Climate Forecasts 477

Star, S. L., and J. R. Griesemer. 1989. Institutional ecology, “translation,” and boundary objects:
Amateurs and professionals in Berkeley’s Museum of Vertebrate Zoology, 1907-39. Social
Studies of Science 19:387-420.

Shardul Agrawala is an associate research scientist at the International Research Insti-


tute for Climate Prediction at Columbia University. He has previously worked on scien-
tific assessments of global climate change at Princeton University, Harvard University,
the Intergovernmental Panel on Climate Change, and the International Institute of
Applied Systems Analysis. His recent publications have appeared in Science, Climatic
Change, Global Environmental Change, and Global Governance.

Kenneth Broad, an anthropologist, is an assistant research professor at the University of


Miami, Rosenstiel School of Marine and Atmospheric Science, Division of Marine
Affairs and Policy, and an adjunct associate research scientist at the International
Research Institute for Climate Prediction at Columbia University. His research focuses
on the relationship between humans and the environment, and his articles have appeared
in Science, Nature, and Knowledge and Society.

David H. Guston is an associate professor of public policy at Rutgers, The State Univer-
sity of New Jersey and a faculty associate of the Belfer Center for Science and Interna-
tional Affairs at Harvard University’s Kennedy School of Government, as well as of
Columbia University’s Center for Science, Policy, and Outcomes. He is the author of
Between Politics and Science: Assuring the Integrity and Productivity of Research
(2000).

Вам также может понравиться