Eos, Vol. 90, No.

14, 7 April 2009

northern Gulf of Mexico, Geochim. Cosmochim. Quiñones-­Rivera, Z. J., B. Wissell, D. Justic, and Wiseman, W. J., Jr., N. N. Rabalais, R. E. Turner, and
Acta, 67, 2359–2375. B. Fry (2007), Partitioning oxygen source and D. Justić (2004), Hypoxia and the physics of the
Green, R. E., and R. W. Gould Jr. (2008), A predic- sinks in a stratified coastal ecosystem using stable Louisiana Coastal Current, in Dying and Dead Seas:
tive model for satellite-­derived phytoplankton oxygen isotopes, Mar. Ecol. Prog. Ser., 342, 69–83. Climatic Versus Anthropic Causes, edited by J. C. L.
absorption over the Louisiana shelf hypoxic Rabalais, N. N., R. E. Turner, B. K. Sen Gupta, D. F. Nihoul et al., pp. 359–372, Springer, New York.
zone: Effects of nutrients and physical ­forcing, Boesch, P. Chapman, and M. C. Murrell (2007),
J. Geophys. Res., 113, C06005, doi:10.1029/​ Hypoxia in the northern Gulf of Mexico: Does the
Author Information
2007JC004594. science support the plan to reduce, mitigate and
Hetland, R. D., and S. F. DiMarco (2007), How does control hypoxia?, Estuaries Coasts, 30, 753–772.
Donald F. Boesch and Walter R. Boynton, Univer-
the character of oxygen demand control the Scavia, D., and K. A. Donnelly (2007), Reassessing
sity of Maryland Center for Environmental Science,
structure of hypoxia on the Texas-­L ouisiana con- hypoxia forecasts for the Gulf of Mexico, Environ.
Cambridge; ­E-mail: boesch@​­umces​.­edu; Larry B.
tinental shelf?, J. Mar. Syst., 70, 49–62. Sci. Technol., 41, 8111–8117.
Crowder, Duke University, Beaufort, N. C.; Robert J.
Justić, D., V. J. Bierman Jr., D. Scavia, and R. Het- Science Advisory Board (2008), Hypoxia in the
Diaz, Virginia Institute of Marine Science, Glouc-
land (2007), Forecasting Gulf’s hypoxia: The next northern Gulf of Mexico: An update by the EPA
ester Point; Robert W. Howarth, Cornell University,
50 years?, Estuaries Coasts, 30, 791–801. Science Advisory Board, EPA-­SAB-­08-­003, Envi-
Ithaca, N. Y.; Laurence D. Mee, Scottish Associa-
Lohrenz, S. E., D. G. Redalje, W. J. Cai, J. Acker, ron. Prot. Agency, Washington, D. C.
tion of Marine Science, Oban, UK; Scott W. Nixon,
and M. Dagg (2008), A retrospective analysis Turner, R. E., N. N. Rabalais, R. B. Alexander,
University of Rhode Island, Narragansett; Nancy N.
of nutrients and phytoplankton productivity in G. McIsaac, and R. W. Howarth (2007), Charac-
Rabalais, Louisiana Universities Marine Consortium,
the Mississippi River plume, Cont. Shelf Res., 28, terization of nutrient and organic carbon and
Chauvin; Rutger Rosenberg, Gothenberg University,
1455–1475. sediment loads and concentrations from the Mis-
Fiskebäckskil, Sweden; James G. Sanders, Skidaway
Mississippi River/­Gulf of Mexico Watershed Nutri- sissippi River into the northern Gulf of Mexico,
Institute of Oceanography, Savannah, Ga.; Donald
ent Task Force (2001), Action plan for reducing, Estuaries Coasts, 30, 773–790.
Scavia, University of Michigan, Ann Arbor; and
mitigating, and controlling hypoxia in the north- Walker, N. D., and N. N. Rabalais (2006), Relation-
R. Eugene Turner, Louisiana State University, Baton
ern Gulf of Mexico, 36 pp., Environ. Prot. Agency, ships among satellite chlorophyll a, river inputs,
Rouge
Washington, D. C. (Available at http://​­www​.­epa​ and hypoxia on the Louisiana continental shelf,
.­gov/​­msbasin/​­pdf/​­actionplan2001​.­pdf) Gulf of Mexico, Estuaries Coasts, 29, 1081–1093.

Grid Computing for Earth Science eGY

Electronic Geophysical Year
PAGES 117, 119 for scaling up computing power and stor-
age capacity in a way that is impossible for a
single institution to do. Also, grid computing heterogeneous data storage and comput-
The fundamental challenges facing
offers a transparent collaborative platform ing resources connected via the Internet
human­kind at the beginning of the 21st cen-
for users, allowing them to have access to and controlled for management and access
tury require an effective response to the
more resources at a given time. This access by software (middleware) such as gLite,
massive changes that are putting increasing
is especially important for exploiting large ­UNICORE, Globus Toolkit, and GRIA. A grid
pressure on the environment and society.
data sets scattered in several locations, for system is based on long-­term and dynamic
The worldwide Earth science community,
running large statistical jobs, and for sharing collaboration among grid partners (resource
with its mosaic of disciplines and players
data and algorithms among many partners providers and user communities) with a trust
(academia, industry, national surveys, inter-
without the need for conversions. agreement to guarantee security and con-
national organizations, and so forth), pro-
vides a scientific basis for addressing issues Earth science computing and data man- fidentiality. A user must be authorized by a
such as the development of new energy agement needs traditionally have been certification authority and must belong to a
resources; a secure water supply; safe stor- provided for by local and national institu- recognized virtual organization: a user com-
age of nuclear waste; the analysis, modeling, tions. The limitations of cost and the num- munity providing the rights to access to grid
and mitigation of climate changes; and the ber of computer central processing units resources (computing, storage, data, soft-
assessment of natural and industrial risks. In (CPUs) available at any one site can be ware, services). The user can then execute
addition, the Earth science community pro- overcome by geographically distributed simple tasks (jobs) or complex computation
vides short- and medium-­term prediction of systems for accessing data, computing workflow operations by specifying only the
weather and natural hazards in real time, resources, and Web services. Geographi- characteristics of the computing resources
and model simulations of a host of phenom- cally distributed computing began by needed and a logical name for data to be
ena relating to the Earth and its space envi- catering to specialized purposes or partic- accessed via the grid storage.
ular user groups (e.g., the Distributed Euro- Because of its architecture, a grid can
ronment. These capabilities require that the
pean Infrastructure for Supercomputing efficiently tackle a large ensemble of com-
Earth science community utilize, both in
Applications). The Berkeley Open Infra- putations running independently. A grid is
real and remote time, massive amounts of
structure for Network Computing (BOINC), also ideally suited for analyzing and pro-
data, which are usually distributed among
introduced in 1999, pioneered the use, on ducing large data sets and for sharing data
many different organizations and data
a volunteer basis, of the enormous pro- within large teams. Several grid infrastruc-
centers.
cessing power of personal computers tures have already been deployed around
The Earth science community can bene-
(PCs) around the world. Earth scientists the world, for example, in North and South
fit greatly from technology that can provide
adopted BOINC only for specific climate America, Asia, Australia, North Africa, and
ready access to computing resources and
and hydrology applications because it is in 2008 in Senegal.
services, easily managed data and metadata
not general enough and does not handle The largest grid deployment to date,
storage in distributed systems or in data cen-
the problems of PC heterogeneity and con- Enabling Grids for E-­Science (EGEE; http://​
ters, clearly defined data policy, authentica-
fidentiality requirements. ­w ww​.­eu​-­­egee​.­org/), is designed for analyz-
tion, confidentiality, and electronic collabo-
ing petabytes of data that will be produced
ration. Grid infrastructure and systems meet
Grid Computing by the European Organization for Nuclear
these requirements as a distributed resource
Research’s (CERN) Large Hadron Collider
system. Grid computing permits the sharing
Grid computing emerged more than experiment in Geneva, Switzerland. Access
of resources between institutions and allows
10 years ago [Foster and Kesselman, 1998] to EGEE is not restricted to high-­energy phys-
as one type of distributed resource sys- ics and is currently used by other scientific
By P. Renard, V. Badoux, M. Petitdidier, tem. Grid computing consists of a network communities including bioinformatics, Earth
and R. Cossu infrastructure comprising loosely coupled sciences, and astronomy. As of March 2009,
 Eos, Vol. 90, No. 14, 7 April 2009
EGEE is deployed at more than 300 sites.
EGEE provides more than 80,000 CPUs and
more than 20 petabytes of storage, and it
is capable of running up to 100,000 jobs
concurrently.
Grid computing has become a basic
tool for particle physics and biotechnology
researchers, but it still is not used routinely
by Earth scientists.

Earth Science Community Needs

Since the International Geophysical Year

in 1957–1958, the Earth science community
has been deluged with data from the world-
wide deployment of instruments. All obser-
vations need to be archived, but synoptic
(time series) observations are particularly
important because they cannot be repeated.
For the past 4 decades, satellite observations
have driven developments in computer sci-
ences for handling, storing, and processing
large volumes of data. The total data archive
of the Earth observation program, man-
aged by the European Space Agency, now
accounts for some 5 petabytes, distributed
geographically among several European Fig. 1. Illustrations of the four Earth science grid applications described in the text. (a) Schematic
data centers. Plans for the next 10–15 years view illustrating the three-­dimensional computation of seismic wave propagation in a reservoir
foresee the accumulation of 10–15 times model for seismic inversion by GeoCluster. Image courtesy of CGGVeritas. (b) Interpretation of
more data per year than at present. As near–­real time seismo­grams, with the “beach balls” being the common graphic representation
models and simulations become increas- of the source mechanism. Image courtesy of E. Clévédé, Institut de Physique du Globe de Paris,
ingly sophisticated, they also generate large France. (c) Visualization of atmospheric ozone content. Image courtesy of the European Space
amounts of data. Model simulations place Agency. (d) Forecasting seawater intrusion in a coastal aquifer. Image courtesy of J. Kerrou, Uni-
heavy demands on computing and storage versity of Neuchâtel, Switzerland. Original color image appears at the back of this volume.
resources, and often require supercomput-
ers and/­or distributed CPU capabilities that adoption of grid computing on a large scale 2. The Institut de Physique du Globe de
cannot be met at the institutional level. As a (see Earth science white paper on grids, at Paris has developed and deployed on EGEE
consequence, observational and simulation ­http://​­www​.­eu​-­­degree​.­eu/​DEGREE/​­internal​ an application to interpret, in near real time,
data are underutilized. This is particularly -­­section/​­wp6/​­DEGREE​-­­D6​.­1​.­2 ​_­v2​.­8​.­pdf). The earthquake data from the worldwide Geo-
apparent in the understanding and predic- first step is to build an Earth science grid scope seismo­meter network (­http://​­geoscope​
tion of climate change, where much unex- community by disseminating grid technolo- .­ipgp​.­jussieu​.­fr). Results are delivered as
plored data await exploitation. gies and to support the deployment of new space-­time location of a seismic rupture, the
applications. The 5-­year target is for a dedi- seismic energy released, the source mecha-
The DEGREE Project cated Earth science grid platform on the grid nism, and earthquake duration (Figure 1b).
infrastructure to share computing and stor- Grid computing allows hundreds of simula-
To identify the technical obstacles facing age resources, data, knowledge, algorithms, tions to be treated concurrently and then to
the adoption of grid methods in the Earth and services over a continuum of time and be combined for final results, reducing deliv-
sciences (the complexity of setting up the over a variety of geographical scales. ery times from 1 week to less than 6 hours,
hardware and installing and configuring suitable for seismic early warning systems.
the grid software, the use of grid middle- Grid Applications in Earth Science 3. In April 1995, the Global Ozone Moni-
ware, and the implementation of Earth sci- toring Experiment was launched on board
ence tools), a consortium of Earth science The following four examples illustrate cur- the European Remote Sensing (ERS) satel-
academic and private institutions, space rent applications of grid computing in the lite. A neural network algorithm was used to
agencies, and computer science institutes Earth sciences. retrieve atmospheric ozone profiles, which
launched the Dissemination and Exploita- 1. GeoCluster (http://​­www​.­cggveritas​.­com/​ were then validated using ground-­based
tion of Grids in Earth Science European proj- ­default​.­aspx​­?cid​­=4​-­13​-­1925) is a software plat- light detection and ranging (lidar) obser-
ect (­DEGREE; http://​­www​.­eu​-­­degree​.­eu/) in form developed by CGGVeritas for seismic vations (Figure 1c). Grid computing was
2006. Over a period of 2 years, 2006–2008, data processing, imaging, and underground used to produce and validate 7 years of ERS
the consortium conducted a survey of the reservoir characterization. It is operated in satellite data, allowing scientists from dif-
tools, software, and protocols commonly a grid environment (http://​­www​.­egeode​.­org; ferent institutions to easily manage about
used by the Earth science community and Figure 1a) using EGEE middleware and dif- 70,000 files, handle the metadata for geospa-
of those provided by grid projects. ­DEGREE ferent infrastructures to accommodate aca- tial queries collocating the satellite’s lidar
addressed three main topics: data manage- demic and business needs. This first commer- observations, and share raw and analyzed
ment, job management, and portals to inte- cial grid application, which began operating data [Iapaolo et al., 2007].
grate Earth science and grid tools. Survey in 2007, has demonstrated grid computing in 4. Grid computing can help emerging
results, which are published on the ­DEGREE a field where data security and confidentiality countries access high-­p erformance com-
Web site, were discussed with grid project are extremely important. The application also puting and solve severe environmental
developers, and applications were proposed has demonstrated the enormous benefit to problems such as seawater intrusions in
to test critical functionalities. An outcome users of ready access to computing resources coastal aquifers. The finite element code
is a grid road map for the Earth sciences and the updated software, and the ease of known as Coupled Variable Density and
that proposes a series of steps leading to the collaboration with distant colleagues. Saturation 3-­D (­CODESA-­3D) has been
 Eos, Vol. 90, No. 14, 7 April 2009
used to compute probabilistic maps of sea- transparent collaborative platform for users, W. Som de Cerff (Royal Netherlands Meteo-
water intrusion in Tunisia’s Korba aqui- allowing them to access more resources at rological Institute (KNMI)), L. Fusco and
fer (Figure 1d) by using a Monte Carlo a given time. This is especially important for J. Linford (European Space Agency),
method. The simulations include flow and the exploitation of large data sets scattered L. Hluchy and V. Tran (Institute of Informat-
density-­dependent transport processes in in several locations; for running large statisti- ics, Slovak Academy of Sciences, Slova-
a 3-­D heterogeneous coastal aquifer [Ker- cal jobs; and for sharing data and algorithms, kia), C. Plevier (Dutchspace, Netherlands),
rou et al., 2007]. In this example, grid com- without the need for conversion, among large H. Schwichtenberg (Fraunhofer Institute
puting maintains two key advantages over numbers of partners. for Algorithms and Scientific Computing,
classical distributed computing. First, a Grid computing is currently available Germany), and M. Zhizhin (Institution of
very large number of Monte Carlo simu- and can meet most of the technical require- the Russian Academy of Sciences Geophysi-
lations, numbering at least in the hun- ments for the Earth sciences. Although cal Center (RAS)), without whom this work
dreds, can be run in parallel with substan- some technical gaps still exist, grid devel- could not have been done.
tial gains in time and accuracy. Second, opers are aware of them and are working
the grid analysis can be controlled sim- to meet Earth science needs in the next References
ply from a Web browser (e.g., ­http://​­www​ generation of grid development. Signifi-
.­eumedgrid​.­org) by collaborating scientists cantly more effort will be required before Foster, I., and C. Kesselman (Eds.) (1998), The Grid:
located in Europe and northern Africa. transparent grid usage will be widespread Blueprint for a New Computing Infrastructure,
in the Earth sciences. The vision is that the 677 pp., Morgan Kaufmann, San Francisco, Calif.
Iapaolo, M., et al. (2007), GOME ozone profiles
A Vision for the Future grid infrastructure must provide, within the
retrieved by neural network techniques: A global
next 5 years, a dedicated platform for shar- validation with lidar measurements, J. Quant.
The above examples show that grid com- ing knowledge, algorithms, data, and ser- Spectrosc. Radiat. Transfer, 107, 105–119.
puting can fulfill most of the computing vices over a wide range of time and spatial Kerrou, J., G. Lecca, F. Murgia, and P. Renard
requirements of Earth scientists and offers scales. Such a platform will help provide (2007), Grid-­enabled simulation of the impact of
new ways for efficient collaboration. Grids efficient and timely answers to many funda- exploitation uncertainty on the seawater intru-
such as EGEE—consisting of clusters and mental challenges facing mankind. sion of the Korba aquifer (Tunisia), in IST-­Africa
farms of CPUs—cannot handle massive 2007 Conference Proceedings, edited by P. Cun-
computations requiring parallel computing, Acknowledgments ningham and M. Cunningham, Int. Inf. Manage.
Corp., Dublin.
shared memory, and intense communication
between the processors. Other grids, such as The work presented in this article was
TeraGrid (http://​­www​.­teragrid​.­org), can ful- funded by the European Community (con- Author Information
fill these needs. tract DEGREE-­IST-­2005-­034619). We grate-
Philippe Renard and Vincent Badoux, Univer-
Grid computing permits the sharing of fully acknowledge Charles Barton for pol- sity of Neuchâtel, Neuchâtel, Switzerland; ­E-mail:
resources between institutions and for scal- ishing our text and all the partners of the ­philippe​.­renard@​­unine​.­ch; Monique Petitdidier,
ing up the computing power and storage project: G. Lecca (Center for Advanced Laboratoire Atmosphères, Milieux, Observations
capacity in a way that is impossible for a sin- Studies, Research and Development in Sar- Spatiales, Paris, France; and Roberto Cossu, Euro-
gle institution. Also, grid computing offers a dinia, Italy), G. Vetois (CGGVeritas, France), pean Space Agency, Frascati, Italy

The Hills Are Alive: Earth Science approach to experimental design through
cultivating a collaborative group that
in a Controlled Environment includes representation from hydrology, geo-
morphology, soil geochemistry, atmospheric
science, ecology, and genomics. Several
PAGE 120 the unique infrastructure of the Biosphere 2 planning workshops have already occurred
project. Biosphere 2 is a large-­scale Earth sci- (some of which were jointly supported by
The structure of Earth’s critical zone, ence facility near Tucson that encompasses the Hydrological Synthesis Center; http://​
which is the interface between the solid about 3.15 acres of land and houses five natu- ­cwaces​.­geog​.­uiuc​.­edu/​­synthesis/​­index​.­html),
Earth and its fluid envelopes and involves the ral biomes. Sealed off to the outside world, Bio- and others are scheduled for the near future
coevolution of biota, soils, and landforms, is sphere 2 allows scientists to exert precise cli- (http://​­www​.­b2science​.­org/​­earth​-­hillslope​
governed by processes important to hydrol- mate and mass balance control at large scales .­html). A key focus to date has been on
ogy, geology, biology, and atmospheric sci- [Osmond et al., 2004]. The facility’s name stems understanding spatial variability, tempo-
ence [National Research Council, 2001] (Fig- from the Earth’s biosphere (biosphere 1); the ral dynamics, and interactions (including
ure 1). Earth surface scientists have long goal of Biosphere 2 is to be a microcosm of the abiotic-­biotic couplings) within hillslopes
recognized that temperature, chemical, and interaction between life and landscape seen using modeling assessments.
gravitational gradients drive energy and water on Earth, such that critical zone interactions Three 33-­meter × 18-­meter environmen-
fluxes, thus controlling systems evolution, can be studied at large spatial scales. tally controlled bays will be available to sci-
but understanding the critical zone has been To facilitate this study, scientists from the entists who would like to propose projects
tackled primarily from disciplinary perspec- University of Arizona will construct experi- for experimentation. The long-­term goal is to
tives [Brantley et al., 2006]. Interdisciplinary mental landscape units—hillslopes—within improve our understanding of the processes
research is needed, and many such efforts, Biosphere 2. They will also build corre- that lead to surface and subsurface structure
such as the U.S. National Science Founda- sponding system models that couple criti- of the critical zone. Workshops have guided
tion’s recent watershed-­scale Critical Zone cal zone hydrology, geo­chemistry, geo- design parameters, such as hillslope geom-
Observatories and the National Ecological morphology, and biology. This program etry (slope angle, planar or complex shape),
Observatory Network, are in formative stages. and facility provide a new opportunity to soil composition (mineral assemblage and
By and large, these facilities focus on utilizing advance understanding of critical zone texture), vegetation type (herbaceous, woody
land surface complexity to elucidate process processes through controlled large-­scale plants), and key details of climate forcing.
knowledge. Unfortunately, incorporating such experimentation. Focused numerical modeling was also used
complexity occurs at the expense of the con- to inform decision making on design param-
trol that characterizes true experimentation. Experimental Design eters. For example, groups from the Univer-
At the University of Arizona, a science pro- sity of Arizona; Oregon State University; Uni-
gram is being built to bridge the gap between Scientists working on this project are versity of Québec; University of Illinois at
laboratory-­and field-­scale studies by utilizing specifically pursuing an interdisciplinary Urbana-­Champaign; University of California,
 Eos,Vol.
Eos, Vol.90,
90,No.
No.1,14, 7 April 2009
2 January 2009

Page 117
Fig. 1. High chlorophyll a concentrations along the inner shelf off Louisiana and Texas (orange
and red) on 11 June 2008 (Indian Space Research Organization Oceansat 1 Ocean Color Moni-
tor image provided by the Louisiana State University (LSU) Earth Scan Laboratory; http://www
.esl.lsu.edu/imagery/ocm/) in relation to an overlay of the bottom hypoxia (< 2 milligrams per
liter, gray area outlined by dashes) during 21–27 July 2008. In June, chlorophyll a concentrations
ranged from 22 micrograms per liter (inshore) to 7 micrograms per liter (offshore) (N. N. Raba-
lais, unpublished data, 2008).

Page XXX

Page 119
Fig. 1. Illustrations of the four Earth science grid applications described in the text. (a) Schematic
view illustrating the three-dimensional computation of seismic wave propagation in a reservoir
model for seismic inversion by GeoCluster. Image courtesy of CGGVeritas. (b) Interpretation of
near–real time seismograms, with the “beach balls” being the common graphic representation
of the source mechanism. Image courtesy of E. Clévédé, Institut de Physique du Globe de Paris,
France. (c) Visualization of atmospheric ozone content. Image courtesy of the European Space
Agency. (d) Forecasting seawater intrusion in a coastal aquifer. Image courtesy of J. Kerrou, Uni-
versity of Neuchâtel, Switzerland.