PARALLEL AND DISTRIBUTED COMPUTING AND ITS ... - DIM

desirespraytownSoftware and s/w Development

Dec 1, 2013 (3 years and 4 months ago)

347 views

Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


1

/
43

P
ARALLEL AND DISTRIBU
TED COMPUTING AND IT
S APPLICATION TO
CHEMICAL WEATHER FOR
ECAST AND CLIMATE IN

C
HILE

W
ORKING
P
APER

Laura Gallardo, Gonzálo Hernández, Axel Osses, Jaime Ortega, Francisca Muñoz, Juan
Juan Carlos Maureira

(1)

Jorge Carrasco, Enrique Garri
do

(2)

Nancy Hitschfeld, José Piquer, Carlos Hurtado (3)

Mauricio Osses (4)

José Rutllant, Maisa Rojas (5)

Rodrigo Torres (6)



(1) Centro de Modelamiento Matemático, Universidad de Chile

(2) Dirección Meteorológica de Chile

(3)

Departamento de Ciencias de

la Computación, Universidad de Chile


(4)
Departamento de Ingeniería Mecánica, Universidad de Chile

(5)
Departamento de Geofísica

(
6
) COPAS??? Free
-
lance??

S
UMMARY


We propose a technology transfer and capacity building effort that connects front line
res
earch and scientific networks with the development and implementation of badly
needed services and data within the Chilean State apparatus, specifically, the Chilean
Weather Service. Technology aspects refer to increase computer power for both the
Chilean
Weather Service and researchers in academia by means of the implementation
of parallel and distributed computing performed on clusters for weather and climate
applications, and the interconnection by high speed networks of available resources
(clusters), b
oth in Chile and between Chile and foreign excellence centers. Capacity
building refers to training activities, ranging from short term stays of professional staff
to doctoral theses in atmospheric science, applied mathematics and computer science,
all of
them linked to one of the two applications to be developed. One application is the
production of high
-
resolution (spatial and temporal) climate scenarios, i.e., data which
is crucial for risk assessment, vulnerability and adaptation studies, etc. The other

application is the implementation of a new service, namely operational numerical
chemical weather
(air quality)
forecast at the Chilean Weather Service. Such tools and
data are required for environmental management and risk assessments of human,
ecologica
l and agricultural impacts.

The project will be developed along three years (36
months). The results and products will be provisionally delivered after the first half, and
the final products should be operative an tested by the end of the project.


Key
-
wo
rds
:

Networking, grids, clusters, climate, chemical weather forecast, capacity

building

Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


2

/
43

1

B
ACKGROUND AND RATION
ALE


“Science is not longer (if it ever was) an individual enterprise. Computer resources can
neither
be isolated nor funded by individual sourc
es”. These statements are particularly
evident when dealing with weather and climate modeling, where

an intensive and
always amount of computer resources are used.

Therefore, world
-
wide large efforts are
oriented towards networking, communication, and shar
ing technologies, especially grid
and parallel computing (See Table 1). These efforts are large in terms of resources
(human and monetary), institutions, countries, and regions involved. This obeys not
only to cost/benefit considerations but also to the e
ssence of the climate system
that is
an extraordinarily complex system in which the biophysical sub
-
systems: atmosphere,
biosphere, cryosphere, hydrosphere, lithosphere, and non the least anthroposphere
interact
. Altogether, these considerations and charac
teristics have lead to new
technological and scientific paradigms.


Table 1.

Examples of contemporary science and technology projects oriented towards grid/parallel
computing. Notice that in all of them, Earth System Modeling or Weather Forecast is a majo
r application.

Project Name

Region of the world

Link

EGEE: Enabling Grids for E
-
Science in Europe

Europe

http://egee
-
intranet.web.cern.ch/egee
-
intranet/gateway.html


E
-
Infracst
ructures

Europe

http://www.einfrastructures.org/


DEISA: Distributed European
Infrastructure for
Supercomputing Applications

Europe

http://www.deisa.org/


PRAGMA: P
acific Rim
Application and Grid
Middleware Assembly

North America and
Asia Pacific

http://pragma.ucsd.edu/


IPG:
Information Power Grid

North America

http://www.ipg.nasa.go
v/

ES: The Earth Simulator

Japan

http://www.es.jamstec.go.jp/


The new approaches and scientific paradigms are multiple and diverse but the ones that
have been adopted so far are parallel/grid computing and

Earth System Modeling.


Weather forecast
and climate
systems consist of complex mathematical models. The
simulation and numerical study such systems require
s of an

exceptional amount of
computational resources available on supercomputers or massive paral
lel architectures.
Supercomputer systems have evolved from isolated, dedicated, specialized architecture
devices (Cray T3D, Connection Machine CM5) to ensembles of connected,
multipurpose architecture devices: clusters and grid computing.


Atmospheric and
climate models in general have evolved from isolated, decoupled,
over
-
simplified codes to modular, coupled, complex codes. These developments have
been further enhanced by the extraordinary growth in observational capabilities of the
climate system and hen
ce by the extraordinary growth in data amounts and data
analyses requirements. Figure 1 illustrates the developments at discussion, i.e., the
evolution in climate modeling and computer resources.


Over the last 10 to 15 years, Chile has faced an unpreceden
ted economical growth and
development. On the one hand, this has lead to fast technological advances in
telecommunications and, at large, in information technology. On the other hand, the
answers and actions required to make this development environmentall
y sustainable
Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


3

/
43

have become more complex and demanding. In particular, sophisticated environmental
systems for weather and climate prediction, including biogeochemical aspects, are
needed and requested from policy makers, industry, citizens, etc.. These syst
ems are
required to address key
-
issues such as future availability of water resources for human
consumption and energy production, air pollution potential and health risks in growing
urban centers, etc.


The primary functions of the Chilean Weather Servic
e (CWS) have been providing
weather and seasonal climate forecasts, performing quality control and archiving
meteorological data., as well as the provision of basic processed data for publication and
for satisfying requirements of different users. To accom
plish these functions, the CWS
has several permanent programs which are financially supported by the State.
The
development of a new and sophisticated computing system as the one proposed here
requires the synergy of the scientific capacity, mainly coming
from the academia, and
the institutional infrastructure of a state organization that ensures the functioning of the
system over time.
In this sense, the CWS appears to be the primary State Institution
expected to provide, use and maintain such systems. How
ever, available human
resources are at the moment insufficient to live up to the expectations associated with
development, implementation and functioning of the new system
s

and products.
Moreover, informatics resource
s
, although increasingly available are
typically utilized
in a sub
-
efficient manner and are difficult to maintain up
-
dated. Also the Chilean
scientific community working on climate and weather, in order to be competitive and
participative in nowadays globalize
d

science requires of a strengtheni
ng of its human
capabilities, and of a far more efficient use of computer and communication resources.
Also, in the areas of grid and parallel computing and communication technology there is
a need of increased know
-
how in Chile. All in all, in our opinion

there is a clear need of
strengthening human capabilities, particularly but not solely at the CWS, and to ensure a
more efficient and flexible use of nowadays informatics and communication
technologies, particularly within the framework of international s
cientific collaboration,
and globalization.



Figure 1
. Evolution in Earth System Modeling and computer power. Source: Presentation by Dr. Guy
Brasseur at
Iberian NCs Seminar, Evora, Spain, April 2004 (
http://www.igbp.kva.se/cgi
-
bin/php/article.show.php?
section_id=65)

Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


4

/
43

Through this project we will
transfer and establish front line p
arallel and distributed
computing technology for chemical weather forecast and climate in Chile both for
scientific and operational use. This will be achieved by connecting var
ious scientific
and technological networks in which the scientists and institutions involved in this
project already participate, with local initiatives and capabilities, particularly at the
Chilean Weather Service. Also, regional
-
scale climate and chemica
l weather forecast
models will be implemented and used to assess climate change scenarios and to provide
a new service through the Chilean Weather Service, namely operational chemical
weather forecast for urban centers. Al these actions will be associated
with capacity
building activities that will
increase the critical mass of professionals and scientists in
the areas of parallel computing, climate, and chemical weather forecast.


In section 2 we discuss several scientific aspects regarding parallel and gr
id computing,
earth system modeling and weather forecast. In Section 3, we present issues, questions
and opportunities to be addressed through this project in Chile. Objectives and
methodological aspects, including expected results, are addressed in Sectio
n 4. An
economical evaluation of the proposed project is presented in Section 5.

2

S
CIENCE AND TECHNOLOG
Y ASPECTS


Here we review the state of the art regarding scientific and technological aspects
relevant for atmospheric forecast and climate modeling.

2.1

C
OM
PUTER
S
CIENCE AND TECHNOLOG
Y

2.1.1

P
arallel and distributed computing, networks and communication (Jo,
Gonzalo)

Computer modeling and simulations have proved to be useful techniques to determine
accurate models of real systems. Several complex models have been d
etermined by
iterative corrections, refinements and adjustments of simpler models by medium or
large scale simulations. These large scale numerical studies require
a
great amount of
computational resources. For this reason, during the last 25 years they ha
ve
incorporated parallel and distributed processing techniques and have been implemented
on massive parallel supercomputers, like Cray T3D and Connection Machine
CM5. In
the last 12 years a new, u
nexpensive technology has emerged as an alternative way to
p
erform medium and large scale parallel and distributed simulations: Clusters or
Networks of high performance personal computers linked by fast or gigabit Ethernet
protocols or another network connectivity, like Myrinet, SCI Dolphin or Quadrics. In
this tec
hnology, a message passing library: MPI (Message Passage Interface) allows a
cluster or network of workstations to work as distributed or shared memory MIMD
supercomputer.


Large scale simulations of some complex systems, like chemical weather forecast an
d
climate models, require exceptional amount of computational resources not available in
a single computer or even in a cluster or local network of w
orkstations. Such kind of
large
-
scale numerical study can be performed only on
metacomputers: a
networked
v
irtual supercomputer constructed dynamically from geographically distributed clusters
or computer networks linked by high
-
speed communication networks. Actually, most of
the metacomputers have been implemented and administered by the Grid Computing
and Glo
bus technology.

Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


5

/
43


Four major current grid projects are Pragma, DataGrid, DataTag and Information Power
Grid:



PRAGMA

(Pacific Rim Applications and Grid Middleware Assembly)
has been
founded as an open organization in which Pacific Rim institutions will
colla
borate more formally to develop grid
-
enabled applications and will deploy
the needed infrastructure throughout the Pacific Region to allow data, computing,
and other resource sharing. Based on current collaborations, PRAGMA will
enhance these collaboration
s and connections among individual investigators by
promoting visiting scholars' and engineers' programs, building new
collaborations, formalizing resource
-
sharing agreements, and continuing trans
-
Pacific network deployment.



DataGrid

is a project funded by

the European Union that aims to enable access
to geographically distributed computing power and storage facilities belonging
to different institutions. This will provide the necessary resources to process
huge amounts of data coming from scientific experi
ments in three different
disciplines: High Energy Physics, Biology and Earth Observation. Within the
Earth Observation applications
, the European Space Agency
(ESA)
missions
involve the download, from space to ground, of about
100 Gigabytes of raw
images p
er day
. Dedicated ground infrastructures have been set up to handle the
data produced by instruments onboard the satellites. The analysis of
atmospheric
ozone

data has been selected as a specific testbed for the DataGrid. More
generally, the project will d
emonstrate an improved way to access and process
large volumes of data stored in distributed European
-
wide archives.



The
DataTAG
project will create a large
-
scale intercontinental Grid test bed
that will focus upon advanced networking

issues and interopera
bility between
these intercontinental Grid domains, hence extending the capabilities of each
and enhancing the worldwide program of Grid development.
The project will
address the issues which arise in the sector of high performance inter
-
Grid
networking, i
ncluding sustained and reliable high performance data replication,
end
-
to
-
end advanced network services, and novel monitoring techniques. The
project will also directly address the issues which arise in the sector of
interoperability between the Grid middl
eware layers such as information and
security services. The advance made will be disseminated into each of the
associated Grid projects.



Information Power Grid (IPG)

is a high
-
performance computation and data
grid that integrates geographically distributed

computers, databases, and
instruments

lead by the National Aeronautics and Space Administration (NASA)
.
NASA has pioneered the development of grids and inter
-
grid accessibility. Grids
are started when people from separate organizations agree to share sele
cted
computational, data, or instrument resources. Middleware
-

the software that
enables grid computing
-

needs to be installed, and then improved and expanded.
Security and compatibility issues need to be worked out by the grid
administrators. Access to
resources is based on grid certificates issued to
qualified users.

Developping sophisticated applications for the Grid (i.e. widely distributed) represents a
multi
-
faceted challenge. In particular, distributed applications are inherently complex to
develop
, maintain and evolve. A large part of this issue comes from the unability of
Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


6

/
43

current programming paradigms such as object
-
oriented programming to properly
modularize several concerns related to the distributed nature of the system. Industrial
application
servers, like J2EE app servers as JBoss, IBM WebSphere, or ObjectWeb
Jonas, are so succesful precisely because they make it possible, to some extent, for
programmers not to worry about such concerns in their business code. Still, this is
provided in an ad
-
hoc manner. This makes it completely impossible to use their
technologies for a wider area than just application servers. Grid computing in particular
is badly served by these technologies. What is needed is a more uniform, powerful
framework for Grid appl
ications that makes it possible to better modularize distributed
applications. In this setting, application servers are just a particular family of
applications within a brand area.

Please merge

2.1.2

Mixed element meshes over complex topographies and cities.


The development of meshing technologies has become an intense theoretical and
practical research area. The study of mesh generation issues, initially tackled by
engineers, physicists, end
-
users in general and some mathematicians, has become also a
field o
f interest for computational geometers, computer scientists and interdisciplinary
teams both in academic and applied research centers. Several techniques for the
refinement and improvement of meshes in two and three dimensions have been
considered in the l
ast 20 years. In particular, the use of two related mathematical
concepts (the longest
-
edge propagation path of a triangle and its associated terminal
-
edge), have allowed the development of algorithms for dealing with general aspects of
the triangulation p
roblem: triangulation refinement problem, triangulation improvement
problem, and automatic quality triangulation problem~
\
cite{RivaraHitschfeldSimpson}.
These mesh concepts have been later applied for the improvement of obtuse angles
~
\
cite{HitschfeldRivar
a2002a,Hit03a}, as well as for the generation of approximate
quality triangulation~
\
cite{Sim01a}.



The allocation of an appropriate spatial mesh is key to the successful numerical solution
of Partial Differential
Equations ({
\
sc PDEs}).

The spatial discre
tization of the
integration
domain should

be fine enough to represent accurately both the geometry of
domain and all the relevant physical quantities varying in its interior. On the other
hand
,
redundant mesh points in inactive regions of the
domain should

be avoided, since a
a
larger number of mesh points implies longer CPU times and larger memory
requirements to solve the discretized equations. This is particularly important for three
-
dimensional ({
\
sc 3D}) calculations, since typical applications, e.g. f
or computational
fluid dynamics and semiconductor device simulation, run several hours and months on
state
-
of
-
the
-
art supercomputers.


Mixed element meshes formed
by cuboids
, rectangular pyramids, rectangular prisms
and some kind of
tetrahedral

have shown

to be

very successful in the numerical
simulation of 3
-
D
semiconductor devices

using the control volume method.
The control
volume method is a hybrid method of the finite difference method and finite element
method.
It combines the flexibility offered by f
inite difference methods and the
flexibility offered by a more general finite element mesh. The 3
-
D mesh generator
\
Ome~
\
cite{Hitschfeld92}
and its

Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


7

/
43

successors~
\
cite{Hitschfeld2000b,HitschfeldRound96} generate

these mixed element
meshes for control volume m
ethods. The theoretical foundations of these approaches
can be found in~
\
cite{Hitschfeld92,nancyDiss}. These mesh generators can fit complex
domain geometries specified through a boundary representation and
can fulfill

point
density
requirements specified

by the user.
A mesh

can have regions
with a

high point
density and regions with low point density. Between them a graded mesh is built.


Currently, most
limited area
weather forecast
and climate
simulators work with regular
Cartesian meshes because they
use finite difference methods. Each grid element is a
parallelepiped and each region of the domain has the same point density.
These
kinds

of
meshes do not permit to have regions with high and low point density. These
kinds

of
meshes give good
results when

the topography is not too sharp or when the topography
does not need to be exactly modeled.

2.1.3

Data mining


The rapid progress in data acquisition and storage technologies has created a problem of
how to turn measured raw data into useful information. Fo
r example, in diverse Earth
Science domains, the volume of interesting data like time
-
series, images, and
heterogeneous, distributed data sets are already measured in terabytes and will soon
reach petabytes. For example, projections have been made that EOS

(Earth Observing
System) data volumes will reach a terabyte/day by the time all the planned satellites are
flown. This growth by far exceeds human capacities to analyze the databases in order to
find implicit similarities, regularities, rules or clusters
hidden in the data. Therefore,
automated knowledge discovery and efficient similarity searching techniques becomes
more and more important.



Data mining, which is a technology that provides the ability to analyze massive amounts
of information, could be a

solution to this problem. Types of data manipulation include
searches for correlations and patterns in large data sets, data classification, clustering,
change and deviation detection, summarization, and dependency modeling
.
In the
multidisciplinary earth

science context, where intelligent use of data is critical, data
mining can be used to improve our understanding of the environment. The wide variety
of available data in different fields includes: economic statistics, clinical patient records,
government

records, scientific simulation results, remote
-
sensing data, etc. For example,
advanced data mining techniques are very important to merge sensor information from
multiple sensing devices orbiting the earth. Correlations detected in the data might help
pr
edict weather patterns or catastrophic events (such as wildfires or hurricanes).
Intelligent data mining techniques will also be critical in applications such as onboard
science selection, intelligent assistants for scientific discovery, and condition
-
base
d
maintenance of vehicles.

2.2

E
ARTH
S
YSTEM
M
ODELING

AND REGIONALIZATION


It is difficult to unambiguously define climate and the concept itself has evolved over
time. In its Greek origin, the word climate refers to the tilting of the Earth’s rotation
axes. Th
is tilting determines the existence of seasons, i.e., periods of the year with
different characteristics regarding precipitation, temperature, insulation, etc.,
phenomena which in turn result in different status of biotic and abiotic systems. By the
1800,
the concept of climate was defined in an operative manner based on the
Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


8

/
43

classification of different geographical areas in terms of their biotic characteristics.
Köppen’s classification of climate zones, in which terrestrial vegetation was identified
as the
synthesis of the different climatic factors, is still at use. Over the last 10 to 20
years, the evident coupling and interactions among the many factors and subsystems,
e.g., atmosphere, biosphere, cryosphere, hydrosphere,
lithosphere
, has lead, again, to
the
need of synthesis and integration but going far beyond the use of terrestrial vegetation
as synthesis index. The extraordinary complexity of the climate system, understood as
the parts and its interactions is driving science to new paradigms. These cha
nges are not
only driven by the only seek of understanding but more importantly due to the
challenges posed by a rapidly changing planet in which the conditions for survival are at
risk. This enormous endeavor is referred to as Earth System Science. Steffe
n et al (2004)
presented recently a synthesis book of global change research, indicating the main
findings so far:



The Earth is a system that life itself helps to control.



Global change is more than climate change. It is real, it is happening now and
in m
any ways it is accelerating.



The human enterprise drives multiple, inter
-
acting effects that cascade through
the Earth System in complex ways.



The Earth’s dynamics are characterized by critical thresholds and abrupt
changes. Human activities could inadve
rtently trigger changes with catastrophic
consequences for the Earth System.



The Earth is currently operating in a no
-
analogue state.


Furthermore, the unprecedented extent and significance of the role of human kind in
driving global change has suggested

the idea of the new geological era, namely the
Anthropocene. Quantifications of the extent and significance of anthropogenically
driven global change are the following (Steffen et al, 2004):



In the last 150 years humankind has exhausted 40% of the known

oil reserves
that took several hundred million years to generate;



Nearly 50% of the land surface has been trans
-
formed by direct human action,
with significant consequences for biodiversity, nutrient cycling, soil structure,
soil biology, and climate;



M
ore nitrogen is now fixed synthetically for
fertilizers

and through fossil fuel
combustion than is fixed naturally in all terrestrial ecosystems;



More than half of all accessible freshwater is appropriated for human purposes,
and under
-
ground water resour
ces are being depleted rapidly



The concentrations of several climatically important greenhouse gases, in
addition to carbon dioxide (CO
2
) and methane (CH
4
), have substantially
increased in the atmosphere;



Coastal and marine habitats are being dramaticall
y altered; 50% of mangroves
have been removed and wetlands have shrunk by one
-
half;



About 22% of
recognized

marine fisheries are overexploited or already depleted,
and 44% more are at their limit of exploitation;



Extinction rates are increasing sharply i
n marine and terrestrial ecosystems
around the world; the Earth is now in the midst of its first great extinction event
caused by the activities of a single biological species (humankind).


The international scientific community reflects the new challenge
s by new
organizational structures, to some extent reflected in the new programs advocated by
funding agencies all over the world, including Chile. The overarching organizational
Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


9

/
43

structure is the so called Earth System Science Partnership (ESSP), in which
all
international scientific programs, dealing with global change are collaborating, i.e.,
World Climate Research Programme (WCRP), International Geosphere
-
Biosphere
Programme (IGBP), International Human Dimension Programme on Global
Environmental Changes
(IHDP)
, International Programme on Biodiversity Science
(Diversitas). Thus, global change research is a growing body of understanding, both
international and transdisciplinary in essence.


All these changes have as a pragmatic corollary, new needs and deve
lopments regarding
modeling tools. Traditionally, climate models are computational representations of a
myriad of processes mathematically expressed in term of systems of differential
equations, either partial or ordinary or both. The huge complexity of su
ch systems of
equations and their computational implementations, in addition to the growing need
regarding accuracy and precision, have lead, as stated earlier, on the one hand to
increased transdisciplinary and transinstitutional collaboration efforts, an
d on the other
hand to new computational approaches, namely parallel and grid computing. Also, new
approaches based no longer on differential equations but on “Agent
-
based modeling” or
“Cellular Automata” or “Complex Adaptative Systems” have been pointed o
ut and are
being explored (e.g., Finnigan, 2003). All in all, the scientific paradigms, either new or
renewed
old

will
require

of a far more efficient use of computational resources.


In the next sub
-
section we discuss the most used technique so far in ord
er to improve
the accuracy and precision of climate change scenarios.

2.2.1

Downscaling of climate outputs and scenarios


The analysis of the output that will result from the numerous simulations, will

allow
scientists to evaluate model performance, model sensi
tivity and model

response to the
various climate change scenarios on a global basis. However,

most question of climate
change requirer impact assessment that are usually

of interest of a particular country or
region. This means that the horizontal

scales a
t which these information is needed are
much smaller than those at

which the Global models are run, this is of the order of few
100s of km.


In order to address this question, over the last decade several downscaling

technics have
been developed. There are

3 main approaches to produce

regionally relevant scenarios,
i.e., with enough spatial and temporal

resolution to address regional problems (e.g.,
CLARIS WP 3.1, 2003 and

references therein...??? Other references). These are:



1) High resolution and varia
ble resolution Global Circulation Models. This

technic
consists in identifying periods of interest (a time
-
slice) within a GCM

simulation and
rerun this period at a higher resolution. Or use a Global scale

models that has been
adapted for having variable
horizontal resolution, so

that one can choose to, effectively,
zoom over a particular region. This

technic has the advantage of having one model, and
therefore a physically

internally consistent simulation for a particular region. A
disadvantage is

that fo
r the whole domain the same physical parameterization packages
are used,

and these are resolution dependent.



Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


10

/
43

2) Regional climate modeling. In this technic a Regional Climate Model (RCM)


is run
over a certain region of interest and forced at boundaries

and at


initial time with fields
taken from observations (Reanalysis) or Global


model output. The use for RCM for
climate change studies
are approached

by choosing a particular time
-
slice from the
output of a long continuous run by

a AOGCM. Careful

consideration of both the forcing
fields errors and regional model errors has to be given in order to interpret correctly any

results (Seth and Rojas, 2003; Rojas and Seth, 2003).


3) Statistical downscaling. In this method a statistically relevant


rel
ationship is seeked
between a ``predictor'' field (eg, SSTA in central

Pacific, 500mb geopotential height
field) and a variable of interest (eg, percipitation, 2m temparature). The main disavetage
of such a technic for

future climate change studies is the
hypothesis that the relationship


between the predictor and predicant also holds under future climatic

conditions.


The two former techniques are known as dynamical downscaling as dynamical

equations are solved in a coupled mode. Hybrid approaches combin
ing two of

these
techniques (statistical
-
dynamical, high resolution global model driving

regional climate
models, etc.) are also possible.


For ths project we will use a regional climate model to dynamically downscale

AOGCM output for future climate change

scenarios for the region of Chile and

the
adjacent ocean.



In fact, in the third IPCC report the need for a wider use of RCMs to

provide
information for regional climate change scenarios was

expressed. Especially because
results in that report showed

inc
onsistency in the signs for the temperature and
precipitation

changes among models for the region of South America south of 15S

(Giorgi and Fransisco, 2000).

2.3

A
TMOSPHERIC FORECAST
(LGK,

J
ORGE
,

E
NRIQUE
,

J
UAN
,

R
ICARDO
,

M
AISA
,

R
ODRIGO
)


The atmosphere is a rap
idly evolving dynamical system in which physical quantities
(energy, momentum, etc.) are transferred and exchanged. Matter in different states of
aggregation is imbedded in it and constitutes itself the evolving fluid. Transferences and
exchanges of physic
al quantities occur simultaneously at a very broad range of scales
both in time and space. The fluid at discussion, i.e., atmosphere has a mass of about 10
21

g, which consists, in the first 100 km above the ground, of a rather homogenous mixture
with about

78% of molecular nitrogen (N
2
) and 20% of molecular oxygen (O
2
). Despite
its overwhelming predominance in mass and volume these species can be seen as mere
carriers of species, typically found in often almost undetectable amounts (i.e., trace
species), wh
ich in fact define the state of the atmosphere. The movements of N
2
and O
2

results of energy and momentum exchanges at the upper and lower boundaries of the
atmosphere as well as of energy transfer processes in which trace species are involved.
However, lo
oking at N
2

and O
2

as “mere carriers” is not a trivial task. In fact, the
equations which “merely” describe the movement of the atmosphere (Navier
-
Stockes)
remain up to date one of the most challenging areas of research in theoretical and
applied mathemati
cs and physics.


Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


11

/
43

Navier
-
Stokes is highly non
-
linear, it shows chaotic behavior, and it can only be solved
by numerical approximations that are intensive in computer resources. Hence, it is not a
surprise that predicting the evolution of the atmosphere, ev
en if only referring to its
inert and most abundant constituents, is a major challenge for scientists. But the interest
in resolving such a mathematical system of equations is not purely academic. Resolving
Navier
-
Stockes for the atmosphere is in fact what

we know as “weather forecast”, with
all daily experienced implications in everybody’s daily life. Over the last 40 of 50 years,
weather forecast has evolved from a rather playsome (ludic) activity of a few
enthusiastic scientists, to a world
-
wide professi
onal activity in which scientists are
definitely a part of the team in charge, and in which enormous computer resources are
invested. In fact, weather forecast has evolved hand in hand with the evolution of
computer and communication technology, often lead
ing such developments (e.g.,
Bengtson, 1999). Besides a more complete description of the physics and a better
numerical solution of the equations, the use of data assimilation has been a fundamental
factor for the improvement of weather forecast.


Lately,
weather forecast has become even more complex and resource demanding. On
the one hand, as the required accuracy in the operational forecast has increased, the
explicit inclusion of processes involving trace species has become necessary and crucial.
On the
other hand, air quality degradation has become a serious and sensitive problem
in and around urban centers where more than 50% of the world’s population lives,
leading to the establishment of new missions and requirements for weather services. In
fact, the

largest and leading weather centers in the world have already or about to
expand its activities to “chemical weather forecast”. In practice this means that in
addition to Navier
-
Stockes, one must solve equations that describe the spatial and
temporal dist
ribution of trace species. These equations include the effects of transport
and mixing processes, chemical and physical transformations, emissions and deposition
processes. Again, this is taking place in close connection with the development of
computer sc
ience and technology, data assimilation, and of course fundamental science.



Due to the chaotic nature of the atmosphere, the prediction of weather is

inherently un
uncertain science and no deterministic forecast can be

made. This means that any effort
i
n knowing in advance the weather conditions

in the future has do be done a
probabilistic manner. There are several

technics that have been developed and applied
for numerical weather

prediction. The idea is to estimate the uncerenties of the initial
state
by

performing ensemble simulation with slightly different initial conditions and

see how these small difference grow in time. This is a way of estimating the

probability
density function of the atmosphere, that gives the probability

associated that a given

atmospheric state may occur. It is therefore important

to understand that any experiment
intended to predict the response of the

earth
-
system to anthropogenic forcing involves
constructing the present day

PDF by this means, and estimating the future PDF a
s close
as possible to

compare with.


2.4

D
ATA ASSIMILATION
(J
AIME
,

A
XEL
)


Inverse modeling techniques are becoming key tools in global change science,
particularly atmospheric science. This is due to the increasing amounts of data that are
becoming available,

for instance through remote sensing and satellites, and also to the
need of objective assessments of the uncertainties in current atmospheric chemistry
Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


12

/
43

transport models (CTMs), which are very complex systems that combine different
sources and types of inf
ormation, represent numerous processes at various interacting
scales. The increase in observational data is

expected to result in improvements of the
performance of available three
-
dimensional chemistry transport models (3
-
D CTMs), a
better bracketing of t
he strength and location of emissions of relevant compounds,
optimization of the design monitoring networks, and a better quantification of model
errors and sensitivities, etc.. In turn, this will require of an extensive use and a better
comprehension of i
nverse modeling techniques, including use of adjoint and data
assimilation. These endeavors are all functional to the consolidation of Earth System
Modeling tools for a better basis on atmospheric chemistry and climate research, and the
establishment of lo
ng
-
term sustainable policies.


Data assimilation in
climate models in general, and in atmospheric model in particular,
requires knowledge of a wide range of variational and control techniques

(e.g. Bennet,
2003)
. Most of the classical techniques are based
on
: 1)

optimal control approaches with
weigthed least squares error fun
ctionals and adjoint techniques; 2)

estimators based on
Kalman filters or simply best linear unbiased estimators
(BLUE), which use

a priori
statistical information

of the parameters and

measures; 3) proj
ection methods
such as
finite dimensional representation of solutions by means of Green functions or
construction of finite d
imensional basis using backward

solutions. All
these

methods are
related
with each other,
and
are
in fact complem
entary. Some new techniques arising
from actual theoretical studies are
: 1)

estimators which are a priori insensible with
respect to some perturbations as sentinels methods or

exact controllability methods; and
5
)

Carleman estimates (asymptotics) in which

a priori estimates of the error
????
while

locally recovering coefficients in the equations
.



From the point of view of the mathematical models
at play in climate and atmospheric
models, there are the following sort of
equations: a) advection
-
difussion eq
uations
(
continuity

equation
) coupled with a system of ordinary differential equation (chemical
and deposition
precesses); b) kinetic transport equations (aerosols or interacting
particles, radiative transfer equations)
;

and c) specific fluid
-
dynamics equa
tions
model
ed by adapted versions of Navier
-
Stokes equations (Coriolis term for quasi
-
geostrophic models, shallow water equations, free boundary liquid surface equations,
etc.). Even if the standard data assimilation problem is linked to source, boundary
or
initial conditions recovering, it is still not well understood from the practical and
theoretical point
s

of view. The problem of recovering coefficients in the equation is
much more difficult and it is the subject of important theore
tical developements
nowadays. Part of
these issues will be addressed through this project, namely.....

Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


13

/
43

3

S
TATE OF THE ART

AND OPPORTUNITIES IN

C
HILE


In this section we provide a brief review of the state of the art in Chile regarding
computer science and technology
,

forecast
and
climate
modeling
, as well as science
networking and capacity building activities
. Also, we highlight the opportunities
we
foresee

through the realization of this project.

3.1

C
OMPUTER
S
CIENCE AND TECHNOLOG
Y


something....to bolderly go where man/woman nev
er has gonde before...conexión a lo
de Florencio Utreras


At the time being, we have developed the first AOP kernel for Java [Tanter05]. This
kernel is an extension of our reflective platform for Java, Reflex [Tanter03]. Reflex
operates via load
-
time trans
formation of Java bytecode, by using the Javassist library
from Prof. Shigeru Chiba, Tokyo Institute of Technology [Chiba00], also used in the
new JBoss 4 application server. This implementation technique is particularly well
-
suited for distributed systems
. We have already applied Reflex and our kernel approach
in a non
-
distributed setting, by the development of the Sequential Object Monitors
(SOM) library, for concurrent programming [Caromel04].





The public Network evolution is in an appropiate phase to p
ermit the grid computing
in Chile. Fast WAN links with reazonables cost ables to transport bigs ammount of
data between several nodes geographically spread.



There is a common thought about grid computting is the way to face large
computting modelling. Se
veral reasech groups are building clusters.



There are a weakness about the know
-
how in grid computing operations, as much in
instrumentation as in modelling. Opportunitie to develop such competences.



There are several climate phenomena which would requir
e an exact representation of
the topography and/or the cities. In this project we are interested in studying the
modeling of the pollution distribution through the cities under several weather
conditions.




Our hypothesis is that mixed element meshes are
flexible enough to allow the
modeling of the shape of the buildings and houses, and then allow us to solve
the problem in which we are interested. Our mesh generator can generate
meshes with the following characteristics:



Meshes are mostly formed by parall
elepipeds. Then in these regions, the
problem can be modeled as in Cartesian meshes



Meshes combine high point density and low point density regions according to
the application requirements. The there is no need of interpolation of the
boundary conditions
among different nested meshes. The equations must be to
handle different



Elements types with their respective control volumes.



M
eshes use prisms, pyramids and tetrahedral when the domain geometry
requires it and in the transition from high point density
to low point density
Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


14

/
43

regions. As before, the equations must be discretized to handle different
elements types with their respective control volumes.


In this project we take advantage of our research experience in the field of meshing,
which includes the d
evelopment of algorithms and of several prototypes tested in an
academic setting~
\
cite{HitschfeldRivara2002a,Hit03a,Sim01a}, as well as the
development of an object
-
oriented mesh generator for semiconductor device
simulation~
\
cite{Hitschfeld92}.


3.2

F
ORECAST
AND CLIMATE MODELING

3.2.1

Climate modeling


High
-
resolution regional climate models or data are not yet available in Chile.
This
imposes severe limitations to the quality and relevance of
the information available for
undertak
ing

mid
-

and long
-
term planning, an
d for decision making regarding strategic
issues such as availability and use of water, energy and food resources.

This in turn
undermines
the negotiation grounds for Chile in a globalized world.

To illustrate this we
present two examples:


1)
Seosonal for
ecast of El Niño events
.
The Chilean Weather Service has as a mandate
to provide seasonal climatic forecast. This is done so far
largely

based on
diagnostic
s,
analyse
s and projections of the behavior of El Niñ
o Southern Oscillation (ENSO). A
l
arge portion
of the information needed on this matter
,

is
produced by different foreign
meteorological Services and research centers,
is gathered via

the

internet. The model
products on seasonal prediction are analyzed along with the local behavior of
atmospheric varia
bles. Simple statistical relationships between ENSO and precipitation,
and temperature are used for a rather qualitative forecast
s
. Unfortunately, the products
available over the internet are too coarse
in spatial resolution
to provide accurate
forecasts.


2)
Climate

change

scenarios and planning
.
Chile may be seriously affected by climate
change (e.g. IPCC, 2001). It is probable, for example, that arid zones will continue
advancing southward into the Central zone where the population and important
agricu
ltural activities are concentrated (e.g. CONAMA, 1999). Changes in the intensity
and frequency of phenomena such as El Niño and La Niña are also expected along with
increasing severity of droughts. Unfortunately, the scenarios predicted by global models
a
re imprecise and uncertain for Chile given its geography and topography, for example
with respect to changes in cloud cover and their effects. All of this makes it imperative
that Chile generates the capacity to understand and make use of tools such as
mat
hematical models to identify and evaluate different scenarios in time to act
accordingly.


This project will provide tools, infrastructure and new human resources to address,
among others, the problems indicated here. In other words, we see it as an oppor
tunity
to establish regional climate tools and data, and most importantly local expertise in this
area allowing an appropriate basis for science development and decision making in
regarding with strategic issues such as availability and use of water, energ
y and food
resources.

Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


15

/
43

3.2.2

Atmospheric Forecast


A substantial improvement in weather forecast in the last 5 years has been the
incorporation of regional scale models.

A meso
-
scale weather forecast model, namely
the
Penn
-
state Mesoscale Model version 5 (MM5)
w
as implemented in the CWS in
1999 with a spatial resolution of 50 km, which later was improved to 10 km.
This model
is at operational use over Central Chile. Other weather forecast models have been also
implemented but they not in operational use.
These dy
namic atmospheric models have
permitted to improve the forecast in term of the precipitation distribution, intensity and
beginning/ending of the events.
Of course, these efforts have been done in connection
with capacity building activities.
However,
the s
taff with enough expertise to manage
these systems is still insufficient. Furthermore,
the CWS is facing new challenges in
relation with weather forecast in order to satisfy the always increasing needs of the
public and in general the different activities
of the country, m
ainly now in this global
economy
.
As all weather services in the world,
the
weather
forecast
can
no longer be a
purely “meteorological forecast” but a more bro
adly “environmental forecast”.
This
change means a new development in the numeri
cal model
s

currently used
at the CWS.
This also implies
new a more powerful computer capacity, and
enhanced
training

of the
staff and even the incorporation of highly qualified personnel, able not inly to use but to
modify and adapt the tools
.


One of the

main objectives of this project is the implementation of the operational
numerical chemical weather forecast at the
Chilean Weather Service
. This means having
the capability of not only forecast the weather in term
s

of the traditional meteorological
varia
bles as precipitation, air temperature, wind, etc; but also of the chemical elements
present in the air, its concentration and distribution. This new information will be
essential for different activities and for polic
y

decisions. In fact, an additional
im
provement in the performance of the model is expected by introducing the chemical
parameters which until now are parameterized or not included at all.

This will take place
in practice through the adoption of a chemical weather forecast system that combines

continental scale forecast provided by the models at use at CPTEC in Brazil, the MM5
model, already used operationally at the CWS, and the dispersion model POLAIR.

3.2.3

Data assimilation


To the best of our knowledge, the only data assimilation research
applie
d
to
atmospheric and oceanographic
problems
in Chile is that developed at the Center for
Mathematical Modeling (CMM) at the University of Chile.
Over the last two or three
years, several applications have been developed. For example, the problem of
estimat
ing the location and strength of quasi
-
static pollution sources has been assessed
by means of the so
-
called retroplume method using measurements of quasi
-
passive trace
species that are transported over complex terrain as constraints (e.g., Issartel, 2003;

Quiroz et al, 2003). Also, using several inverse techniques, including BLUE, adjoint,
and sentineles, the problem of improving urban emission inventories, in particular
mobile sources, has been addressed

(Cerpa and Osses, 2004; Muñoz et al, 2004)
.

Connect
ed with this applied work there has also been significant theoretical work
developed regarding.....

3.3

S
CIENCE NETWORKING AN
D CAPACITY BUILDING


Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


16

/
43

The proposed project relies heavily on research networks and, and also on the increasing
availability and use of c
omputer grids and communication systems.
Moreover, these
networks and the graduate and post
-
graduate programs will allow the formation of
scientists and professionals.

3.3.1

Technological networks

PRAGMA connection

Physical networks

Clusters at CMM and DMC


In
the Distributed Programming Languages group in the Computer Science Department,

we have been working on different aspects of Programming Languages for Distributed
Systems: Cluster Concurrent Programming, Distributed Garbage Collection, Reflection
and Aspec
t
-
Oriented Programming (AOP).


We are members of a proposal for a Network of Excellence (NoE) named
CoreGrid
, of
the European Community, which includes 46 institutions. In particular, we have an on
-
going collaboration with the OASIS group from the INRIA S
ophia
-
Antipolis France,
for the last 2 years, which has been highly fruitful in terms of development and research.


Eric Tanter is also involved via its co
-
affiliation at the Ecole des Mines de
Nantes/INRIA OBASCO France, in the
Aspect
-
Oriented Software D
evelopment

Network of Excellence (NoE) of the European Community. After his PhD is finished,
collaboration will go on with this institution.


We also have on
-
going collaboration with Tokyo Institute of Technology, Japan, in
particular with Professor Shige
ru Chiba, which will come to the workshop organized in
the DCC in November.


3.3.2

Scientific collaboration


Researchers and institutions involved in this project participate actively in a number of
collaborative research effort. A brief summary of those efforts

is presented hereby.

3.3.2.1

Computer science

PRAGMA connection again??

Other related projects??

3.3.2.2

Climate and atmospheric research

1.

Urban Mobile Emissions in South American Mega cities (
UMESAM
, IAI
SGPII 03SGP211


211, 2004
-
2005).
This is a coordination network.
T
he main
goal of UMESAM is to promote the development of new collaborative research
in the Americas, able to connect local air quality initiatives and global change
issues, involving researchers from Argentina, Brazil, Chile, Colombia, Peru, and
the United
States of America. This research focuses on two work packages: (a) a
methodology to estimate urban emissions from mobile sources, and (b) inverse
modeling techniques both on the regional and local (city by city) scales. This
work is developed within the fr
amework of an international network, financed
via the Interamerican Institute for Global Research (IAI), which is expected to
Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


17

/
43

expand its scope to other pollutants, including greenhouse gases, and the
establishment of coordinated regional networks of observ
ations, which are all
functional to the consolidation of Earth System Modeling tools for a better basis
on atmospheric chemistry and climate research, and the establishment of long
-
term sustainable policies.

2.

Application and development of inverse modeling
tec
h
niques to air quality
modeling and network d
e
sign (
CONICYT
-
INRIA
, 2002
-
2004)
. Inverse
modeling techniques for source estimate have been the central issue of this
collaboration
, which consists mainly of short
-
term exchanges of researchers and
students
.

The motivation emerges from the fact that there are increasing amounts
of data that are becoming available, for instance through remote sensing and
satellites, and also to the need of objective assessments of the uncertainties in
current atmospheric chemi
stry transport models (CTMs), which are very
complex systems that combine different sources and types of information,
represent numerous processes at various interacting scales.

3.

A Europe
-
South America Network for Climate Change Assessment and Impact
Studi
es (
CLARIS
, 2003
-
2006). 6th Framework program, European Union.

The
CLARIS project aims at strengthening collaborations between research groups
in Europe and South America to develop common research strategies on climate
change and impact issues in the subt
ropical region of South America through a
multi
-
scale integrated approach (continental
-
regional
-
local). First, CLARIS will
favour the transfer of knowledge and expertise on Earth System Models, their
different components and coupling procedures. Moreover,
CLARIS will provide
to European and South American scientists involved in climate modelling in
South America the framework to compare and exchange their methodologies.
This framework will also be completed by an easy
-
access database compiling
the observed
and simulated climate data required for models to be both validated
and properly forced. Second, complementary to that modelling aspect, CLARIS
will facilitate access to large scale climate data sets and climate simulations, and
it is a major goal for CLAR
IS to initiate the setting
-
up of a high
-
quality daily
climate database for temperature and precipitation. The European expertise
acquired through the European Climate Assessment Project will be essential to
meet this objective. The resulting database will
be of great value to validate and
evaluate the model skills in simulating climate trends and extreme event
frequency changes. Finally, at a local scale, CLARIS aims at creating a bridge
between the climate research community and stakeholders in the framewo
rk of
three pilot actions designed to integrate multi
-
disciplinary components and to
demonstrate the potential and feasibility of using climate information in the
decision
-
making process. Three major areas are adressed: agriculture, health and
pollution. T
he CLARIS framework will facilitate the participation of European
researchers to IAI (Inter American Institute) projects and the submission of new
common research proposals. Moreover, its opening towards stakeholders (e.g.
agriculture, reinsurance, hydroel
ectricity), associated to the project through an
expert group, will promote future initiatives on climate impact analysis, thus,
contributing to related sustainable development strategies.

4.

Impacts of natural and anthropogenic aerosols on the stratocumulus
deck off
Chile (
ECOS Syd
, 2004
-
2006)
. This collaboration, betwen researcher at the
University of Chile and researchers from climate laboratories en France, is
intended to better understand the aerosol
-
cloud
-
climate interactions and to
provide tools for ass
essing and predicting these effects by means of global and
Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


18

/
43

regional scale modeling. Also, it aima at strengthening the links with other
research activities, particularly remote sensing of aerosols and clouds and
dynamical modeling of the marine boundary la
yer over the Eastern Pacific.


In addition,
a new initiative has been proposed to the Inter American Institute for Global
Change (IAI) in order to improve air quality and climate relevant emission scenarios for
and to provide air pollution scenarios South
American megacities. This is the natural
continuation for the UMESAM pilot project, and it is to be considered complementary
to this initiative.

3.3.3

Capacity building


As stated repeatedly, this project emphazises capacity building actions.
A significant
porti
on of the resources will be devoted to graduate and post graduate students. We
foresee that the students and professional, particularly at the Chilean Weather Service,
will follow one of the following programs:



Professional M. Sc. program in Atmospheric Sc
ience and Meteorology at the
Department of Geophysics, University of Chile



M. Sc program in Informatics and Computer Science, University Federico Santa
María, Chile



Ph. D program in Computer Science, University of Chile



Ph. D program in Fluid Dynamamics, U
niversity of Chile



Ph. D proram in Oceanography, University of Concepción, Chile

Of course, students from other programs in other institutions are not excluded. Also,
combined programs between Chile and other countries may be of particular interest,
especi
ally if co
-
guidence is provided by a partner institute.


A

part of the
project’s
resources are located in traveling and per
-
diem costs that will
allow research and training stays
of these people
hosted by
our international
counterparts. Also, stays of for
eign researchers and professionals are considered under
that item.
This will permit the realization of hands
-
on training, which is fundamental
when implementing sophisticated climate and atmospheric models, and grid computing.

3.4

C
ASE STUDY
:

A
NTHROPOGENIC IMP
ACTS OF SULFATE AERO
SOLS OFF THE
C
HILEAN
COAST
UNDER PAST
,

PRESENT AND FUTURE S
CENARIOS


The world’s most extended and persistent stratus deck around the world is that located
under the Pacific high off the coast of Chile and Southern Peru. This cloud deck

has a
large impact on the regional and global circulations (e.g., Hartmann et al, 1992). Chile
has a long tradition of exploiting mineral resources, particularly copper. The copper
smelters in Chile stand for about 1 % of the ca. 70 TgS/yr oxidized sulfur

(SO
x
) emitted
worldwide. There is evidence that these emissions are transported off
-
shore in
connection with coastal lows, which are subsynoptic features that often occur in this
area (Gallardo et al, 2002; Hunneus et al, 2004). In addition, in the coast

of Central and
Northern Chile exchanges of biogenic trace species are significant (Scholes et al, 2003),
particularly dimethylsulfide (DMS), which is an important precursor of natural aerosols
and cloud condensation nuclei (CCN). Volcanoes appear to be a
significant source of
sulfate aerosols (Amigo et al, 2004). Finally, Santiago, one of South America’s
megacities is also a significant source of polluted air, containing both organic and
inorganic aerosols that also may affect the properties of clouds.
Desp
ite the relevance of
Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


19

/
43

these processes for the regional and global climate, our quantitative understanding and
our predictive capabilities to assess potential future changes are scarce.


To improve our quantitative understanding of aerosol
-
cloud
-
climate int
eractions off the
Chilean coast, we will implement and run a regional
-
scale climate model, namely the
Rossby Center Atmospheric and Oceanic Model (RCAO, Rummukainen et al, 2003), a
model provided in kind by the Swedish Hydrological and Meteorological Insti
tute
(SMHI) for reserach purposes to the Center for Mathematical Modeling.
RCAO is a
regional climate model that represents the sulfur cycle and its interactions with
stratocumulus clouds (e.g., Ekman and Rodhe, 2004). It will be run as a dynamical
interpo
lation tool, i.e., it will produce regional simulations using as boundary conditions
global scale climate simulations provided in kind by European centers

via the CLARIS
network
. Past (1850
-
1860), present (2000
-
2010) and future (2100
-
2110) scenarios will
b
e simulated.


We choose to focus on sulfur aerosols because of the importance of both anthropogenic
and natural emissions of sulfur in this area of the world, and because at least
anthropogenic sources are relatively well constrained.
In the case of anthro
pogenic
sources, we will rely on available estimates. In the case of volcanic emissions we will
use results from ongoing research performed in Chile and else where (Clavero/Amigo ,
personal communication??).
Here we will focus on biogenic emissions of sulf
ur
aerosols, and the overarching question of the coupling between the ocean and the
atmosphere.


DMS (di
-
methyl
-
sulphide) is an ubiquitous trace gas, derived from the precursor b
-
dimethylsulphoniopropionate (DMSP), which is produced by many marine algae.

The
production of DMS by phytoplankton is the dominat source of atmospheric sulfur in the
Southern Hemisphere (IPCC, 2001) and constitutes an important forcing function of
climate by its role in cloud formation and acidity (e.g.Charlson et al.1987). The
v
entilation of DMS (i.e the exchange of DMS from the ocean to the atmosphere) has
been thoroughly (??) described as a function of wind speed and the DMS concentration
in seawater. However, while wind speed over global ocean is now available thanks to
satel
lite borne scatterometer technology, DMS in seawater is not. Thus the prediction
the DMS in seawater (i.e. realistic inputs data for atmospheric scientists) is a key step
for a good understanding global and regional sulphur cycle, including atmospheric
pro
cesses, and for our developing capabilities to predict climatic changes at relevant
time scales.




Fig.1 Schematical representation of processes and factors that regulate DMS in seawater (modified from
Liss et al., 1993)


Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


20

/
43

Several biologic and chemical p
rocesses and factors (Fig.1) reported to be link with
DMS and precursors in the ocean, shows variability in virtually all spatio
-
temporal
scales. Thus a prediction of DMS in seawater involve the understanding of the
functioning of the marine ecosystem tha
t produces DMS (e.g. for quantify the
production and loss of DMS) and its sensitivity to forcing functions. In order to give
realistic input data for atmospheric scientists to use in theirs climate models,
biogeochemical
-
ecological modeling (e.g. Gabric e
t al., 1996) and correlation of DMS
to suitable variables (e.g. Chlorophyll a, species) can be used, and has been target for
several research programs (e.g. SOLAS).


W
e consider this
case study as
a pilot action
that could help triggering the merging of
several research
initiatives

and groups

in Chile, and elsewhere

into a long
-
term
scientific
program with a much broader scope.

3.5

C
ASE STUDY
:

C
HEMICAL WEATHER FORE
CAST FOR URBAN CENTE
RS IN
C
HILE



In the early 1990’s, only Gaussian models were available

and u
sed

in Chile. There were
two main factors impairing the development, implementation and use of more complex
dispersion models.
The first reason was that expertise in the area was nearly absent in
the country except for the pioneering work developed by Ulri
ksen and collaborators at
the University of Chile (Ulriksen et al., 1992).

The second, reason was the lack of
appropiate data to run and evaluate the models.


Over the last 5 to 10 years substantial changes have occured.

For instance, under the
National C
ommission for the Environment (CONAMA), strong efforts were made

by
the Chilean State

to establish emission inventories, meteorological and air quality
networks. This has happened mainly but not solely in the Santiago Metropolitan area in
connection with a
ttainment and prevention plans. Also, these problems and the
resources invested in them by stake holders, both public and private, have dr
a
wn the
attention and interest from scientists in various disciplines, at various places along the
country
. Some of th
ese initiatives are summarized in
Table X
.

In addition, and
complementary to this, meteorological models have been implemented both in
universities (e.g, University of Chile) and public services (
e.g.,
Chilean Weather
Service), providing suitable inputs fo
r dispersion modeling applications.

Finally, but
most importantly, there have been several professional and scientific theses dealing with
dispersion modeling at various universities, expanding the available expertise and
know
-
how in Chile.



In general, a
ll dispersion modeling applications have been u
sed for diagnostic
assessments, and not for operative daily use.
The closest approach to operative
dispersion modeling and forecast, is to the best of our knowledge developed at the
Department of Geophysics, U
niversity of Chile, with focus on air pollution in Santiago,
as part of a collaborative work with research centers in North America and Europe.
However, this application is not a part of the daily management of air quality
undertaken by environmeental auth
orities.

In fact, the only air quality forecast system at
operational use is the one provided by the National Center for the Environment
(CENMA). This prediction system is based on multivariate regression techniques, not
dispersion modeling.


We propose i
n this work to establish an operational chemical weather forecast system
,
Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


21

/
43

within the Chilean State apparatus, at the Chilean Weather Service, linking boundary
meteorological and chemical data provided by the Brazilian Center for Climate and
Weather Forecas
t

(CPTEC
,
www.cptec.inpe.br/meio_ambiente
), and high
-
resolution
(
~2 km) meteorological and chemical weather forecast tools at the CWS.

This
developement and new service is intended to fulfil a gap in
satisfying the requirements
of environmental management, not only in Santiago, but also in other cities suffering
severe air pollution problems (e.g., Temuco). We envisage, in fact, a complementary
use of the new chemical weather forecast tools and the sta
tistical forecast tools at use up
to now. Furtheremore, we foresee that such a developement will provide a substantial
basis for research in atmospheric science, computer science, applied mathematics, etc..

Tabla
x
.

Dispersion modeling activities in Chile
since the mid 1990’s.

Problem addressed

Model Type

Institution and contact person

Dispersión de monóxido
de carbono y material
particulado respirable
de origen primario

Modelo de transporte a
escala urbana para trazas
inertes

National Commision for the En
vironment
(CONAMA)


Flores et al., 2000

http://www.conama.cl

Olivares et al, 2002; Gallardo et al, 2002

(
lgallard@dim.uchile.cl
)

Gidhagen et al, 2002.

lars.gidhagen@smhi.se

Dispersión regional de
azufre oxidado y
arsénico en Chile
Central

Modelo de dispersión
para escala regional para
trazas químicas acoplado
a las salidas de un
modelo de pronóstico
numérico del tiempo
(HIRLA
M
-
MATCH).

Análisis de trayectorias
en la cuenca de
Santiago durante
episodios de
contaminación invernal

Modelo diagnóstico de
campos de viento a partir
de observaciones
superficiales
-

acabello@cenma.cl


CENMA, 20
01a

acabello@cenma.cl

Determinación de áreas
de influencia de fuentes
en la macrozona central
de Chile

Modelo de trayectorias
para campos interpolados
cinemáticamente y
forzados por topografía
compleja.

Dispersi
ón de oxidantes
fotoquímicos en
Santiago

Modelos de dispersión
de mesoescala para
trazas químicas acoplado
a las salidas de un
modelo de pronóstico
numérico del tiempo
(MM5) o a modelo de
diagnóstico.

Rainer Schmidt (
schmitzr@dgf.uchile.cl
)




Héctor Jorquera (
jorquera@puc.cl
)




Sonia Montecinos, Temuco



Melitta Fiebig, La Serena



?? USACH



Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


22

/
43

4

O
BJECTIVES

AND APPROACH

4.1

O
BJECTIVES

4.1.1

General objectives



To transfer and esta
blish front line p
arallel and distributed computing
technology for chemical weather forecast and climate in Chile both for scientific
and operational use



To increase the critical mass of professionals and scientists in the areas of
parallel
computing, clim
ate, and chemical weather forecast


COMPLETAR TODOS!!!

4.1.2

Specific objectives




Design of cluster/communication network for optimal use of available resources
(clusters, networks+ models)??



Parallelization of some un
-
parallelized models (e.g., POLAIR)



Data mi
ning….structure and use large
-
data bases (e.g., satellites) for on
-
line data
assimilation



Implementation of flexible/adaptative grids for more accurate calculation…???



COMPUTINES TODOS!!!




Implementation of flexible/adaptative grids for more accurate calcu
lation…???



Create the know
-
how in parallel computing and grid computting.



Create a testbed to develop in new research in grid computing.




To provide accurate regional emissions and climate change scenarios for Chile,
with emphasis on the evolution of air
quality in large urban centers, and on the
impacts of anthropogrnic and natural emissions on the extensive stratocumulus
deck off the coas
over Central and Northern Chile



To establish operational chemical weather forecast for Chile, particularly for
urban
centers
(e.g., La Serena

Coquimbo, Santiago, Temuco, etc.).



To establish active research networks with leading centers, particularly in South
America
,
functional to Earth System Modeling.



to construct a conceptual model of the key biogeochemical process
es that are
involves in DMS production/consumption, which will emphasize the inclusion
of relevant characteristic of Humboldt current system and its variability (e.g.
coastal upwelling; interannual variability; phytoplankton succesion);



to explore/adequa
te models set it with particular regional characteristic (using
e.g. available bio
-
oceanographic field/satellite data and information available for
these region) in order to study the sensitivity of DMS to biotic and abiotic
factors (e.g. variability of al
gal biomass/production associated with el ENSO);
and



to produce a “best guess” of DMS exchanges in the ocean
-
atmosphere boundary
for the region based in available information and/or modeling.




Construct a model regional climatology for North/Central Chil
e and

Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


23

/
43



adjacent ocean.




Obtain model estimate of future climate for various climate change



scenarios for the same region.




Fulfill the commitment acquired by the state of Chile, upon becoming a



active member of the IPCC. This is, advancing in the science o
f climate change



by providing regional climate change information.



To estimate the distribution of oxidized sulfur, both natural and anthropogenic,
over the stratatocumulus area over the Eastern Pacific



To estimate the potential impact on the properties of

the persistent stratus deck
(direct and indirect effects)



To estimate the relative importance of natural and anthropogenic sources on the
burden and distribution of oxidized sulfur



To evaluate the sensitivity of the estimates with respect to model resolut
ion with
emphasis on relevant weather conditions such as coastal lows and cut
-
off lows.



To improve, in close collaboration with other research groups, the accuracy of
the simulations by means of assimilation of satellite data.



To improve, in collaboration
with concurrent research the parametrizatio
ns of Sc
clouds in current models






COMPLETAR TODOS!!!

4.2

M
ETHODOLOGICAL APPROA
CH AND ORGANIZATION


Our activities will be developed under four technichal and two administrative work
packages

(WPs)
.

Each of these WPs
will be coordinated by a senior researcher who will
also participate in the WP responsible for the whole coordination and realization of the
project. This will facilitate the communication among the groups and disciplines
represented in each WP. Also, this

will allow a more efficient control of the project’s
development.

In addition to the WPs and to the review process defined by FONDEF, we
will establish a panel of external reviewers. This external panel should act as an
advisory group that, beyond technic
al and scientific aspects, provides guidance for a
better use and application of the tools and data to be produced by this project. Thus, we
will invite senior scientists and stake holder to be a part of that panel.

A diagram of the
WPs and their interconn
ections is sketched in Figure
x
. A brief description of the
activities to be developed

by each WP is described in the next subsection.


Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


24

/
43


Figure X.

Organizational structure.

4.2.1

Work packa
ge 1 (WP1): Computing


This work package is pivotal for all activities a
s it provides the theoretical and
technological basis for the implementation of parallel and distributed computing
, which
is

needed for the applications in climate and chemical weather.

There are two work
-
lines within this WP. On the one hand, theoretical
and design questions, and on the
other hand, technical aspects and deployments of software and tools.


a)

Theory and development
.

In this sub WP, research and capacity building
actions will study the following themes:



Design and development of communications

and networks



Parallelization algorithms for complex systems
.



Mixed element meshes over complex topographies and cities
. In order to
generate mixed element meshes for the modeling of the pollution distribution
in cities, we will have to develop new modul
es to:

o

Read the geometry of the cities

o

Read new user requirements and boundary conditions

o

Extend the current mesh generation algorithms to handle geometries
that have not been considered yet

o

Write the mixed element mesh with the control volumes as require
d
by the prototype software simulator.



Data mining for environmental data



Object oriented programing and coupling engines


There exists an inherent relationship between the distributed programming

paradigm (like CORBA, RMI). these technologies are used to

distribute objects

in a Object Oriented (OO) schema. Thus, by using object name servers, IDL and

Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


25

/
43

ORBs can be implemented a object distributed services. These shema normally

are applied in Inteligent Agents, but ligthly explored in parrallel

computing.


Therefore, the project will explore the posibility to apply these

kind of technology in a high
-
level framework to parallelize code

encapsulating the secuential code inside a distributed object, and so,

achieve the code parallelization with a lower eff
ort than rebuid the

secuential code.

On the other hand, these framework can be used to buid new

parallel applications throught the OO paradigm, fact that improve the

mantenience and simplicity of the code.


b)

Thechnical support and deployment
. This group

of researchers and technicians
will provide technical support for the development and implemen
tation of
software (programing), including coupling engines, deployment of networks,
etc..


Results and products expected from this WP are:



Pocos, concretos y cu
mplibles

4.2.2

Work package 2 (WP2): Climate modeling




Downscaling


The methodology for obtaining regional climate climate change

information consists
essentially of two parts: 1) The reconstruction

of the present day climate state over
Chile using high resolut
ion

models and 2) evaluation of the response of the climate
system to the

changing atrnopogenic forcing. Accordingly the steps involved in

applying this methodology include the following:


1) validation of the global model output for present day climate

ag
ainst reanalysis and
available observational datasets. Aspects that will be evaluated are: annual cycle,
leading modes of variability, representation of large scale features that determine the
regional climate, such as the south pacific anticyclone.



2)
Perform a control simulation over a period of 10
-
30 years for present day climate
with a regional climate model, forced by global model. Obtain model statistics of the
present day model climate. ie., the mean state, variabilitity, frequency of extreme

events.


3) The same models will be used to perform high resolution simulations

of at least one
future climate scenario, using boundary conditions from a reliable global projection
which includes greenhouse forcing. The new model statistics (mean state,
variability
and extreme events) will be compared with the RCM simulations for


present climate to
assess differences between the present and


future according to the model simulations.


By applying this methodology it is ensured that the inherent

uncertain
ties associated
with the predictions are both minimized and

understood, so as to construct the most
robust the future climate

state, within the existing computer resource limitations.




i
mplementation and testing of models

o

RCAO

Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


26

/
43

o

PRECIS



Statistical tool for s
easonal forecasting (using MOS Jorge, Juan??)




Air
-
sea exchange...checking the DMS parameterization




Results and products expected from this WP are:



Pocos, concretos y cumplibles


4.2.3

Work package 3 (WP3): Chemical weather

Results and products expected from

this WP are:



Pocos, concretos y cumplibles


4.2.4

Work package 4 (WP4): Coordination and synthesis

Results and products expected from this WP are:



Pocos, concretos y cumplibles


4.2.5

Work package 5 (WP5): Administration

Results and products expected from this WP are
:



Pocos, concretos y cumplibles


4.2.6

Review and advice




4.2.6.1

Air
-
sea exchanges


Refrasealo aquí Rodrigo...

4.2.6.2

Chemical weather forecast


CPTEC already has an operational numerical system for regional weather and air
quality forecast (
www.cptec.inpe.br/meio_ambiente
) which provides carbon monoxide,
particulate material concentration and aerosol optical thickness from biomass burning
and urban emissions in South America. This system is based on the Brazilian
deve
lopments of the Regional Atmospheric Modelling System (B
-
RAMS) coupled to an
on
-
line Eulerian transport model. RAMS is a powerful tool for atmospheric simulation
being equipped with a multiple grid nesting scheme which allows the model equations
to be solv
ed simultaneously on any number of interacting computational meshes of
different spatial resolutions. The chemical mechanism contained in the global CTM
MOZART (Model of Ozone and Related Chemical Tracers) is under implementation.
We propose to run this ai
r quality forecast system on a regional scale covering the South
American continent and deliver initial and boundary conditions of atmospheric
Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


27

/
43

dynamical, thermodynamical and chemistry state for the urban models used in the
participating Institutions. An ex
ample of this already exists in Brazil where CPTEC
models are coupled on line with models at the University of Sao Paulo. ^[from the
proposal to IAI`]


4.2.6.3


Aerosol
-
cloud
-
climate interactions


4.2.7

Timeline, milestones and external review


Work plan


Year 1 : acqui
sition of computing resources

validation of Global model for the south
pacific region


Year 2: regional climate model present day climatology. Present day

model statistics


Year 3: regional climate model future state simulation. New model

statistics compa
rison
with present day statistics.


Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


28

/
43


4.2.8

Participants

4.2.8.1

Chilean counterparts


The Chilean partners belong to two institutions:

1)

University of Chile, Faculty for Physical and Mathematical Science
(Engineering)

2)

Chilean Weather Service


Researchers associated to
Uni
versity of Chile

belong to different departments and
institutes, covering a vast range of expertise in science:

1)

Center for Mathematical Modeling

a)

Laura Gallardo (Project’s director) , PhD in Chemical Meteorology,
Atmospheric Modeling

b)

Axel Osses, PhD in Appl
ied Mathematics, Differential equations and control

c)

Jaime Ortega, PhD in Applied Mathematics, Differential equations and control

d)

Gonzalo Hernández, PhD in Applied Mathematics, Parallel Computing and
Cellular automata

2)

Department for Computer Science

a)

José M.

Piquer, PhD Computer Science, Parallel Computing

b)

Nancy Hitschfeld, PhD Computer Science, Geometric computing

c)

Luis
Mate
u, PhD Computer Science, Distributed computing

d)

Carlos Hurtado, PhD Computer Science, Data Mining

3)

Department for Mechanical Engineering

a)

M
auricio Osses, PhD Mechanical Engineering, Mobile emissions

b)

Roberto Corvalán, Mechanical Engineering, Stationary emissions

4)

Department of Geophysics

a)

José Rutllant, PhD Meteorology, Climatology and stratocumulus


In addition, we will count with the collabora
tion of two younger researchers:




Maisa Rojas, PhD Atmospheric Science, Climatology and dynamical downscaling



Rodrigo Torres, PhD Chemical Oceanography, Ocean atmosphere exchanges and
biogeochemical cycles



Researchers associated to the
Chilean Weather Se
rvice
develop their professional and
research activities within the sections of Operational Forecast and Climatology




Jorge Carrasco, PhD Meteorology, Climatology



Juan Quintana , MSc Atmospheric Science, Climatology



Enrique Garrido, Senior Meteorologist,
Numerical Forecast



Ricardo Alcafuz, Senior Meteorologist, Numerical Forecast


In addition, the Chilean team will consider post
-
graduate students and junior
meteorologists. Also, the project will count with two engineers in charge of computer
resources:



Jua
n Carlos Maureira, Informatics at the Center for Mathematical Modeling

Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


29

/
43



Marcelo Cerda, Informatics at the Chilean Weather Office


4.2.8.2

Foreign counterparts


The foreign counterparts are linked to various research projects and networks. A few
individual are iden
tified as key contact people. Also, the common projects or networks
and the research areas are briefly summarized. The expected contributions and
exchanges are also indicated.


South America



Centro de Previsão de Tempo e Estudos Climáticos (CPTEC), Brazil

Collaboration with CPTEC has been established through two projects:

i)

A Europe
-
South America Network for Climate Change Assessment and
Impact Studies (CLARIS, 2004
-
2007)

ii)

Urban Mobile Emissions in South American Cities (UMESAM, 2004
-
2005)

iii)

South American Emiss
ions, Megacities, and Climate (SAEMC, Proposal
2005
-
2008)


These projects involve atmospheric modeling, assimilation techniques and
down
-
scaling of climate scenarios.


Key
-
contacts

o

Dr. Saulo Freitas

o

Dr. Karla Longo

o

Dr. Carlos Nobre


The expected contribut
ions from CPTEC are:

o

Providing boundary conditions for the implementation of an operational
chemical weather forecast system to be operated at the Chilean Weather
Office.

o

Capacity building opportunities through short (1
-
2 weeks) and mid term
(1
-
3 months) s
tays of Chilean students and researchers at CPTEC and of
CPTEC researchers in Chile.




Universidad Nacional de Córdoba (UCOR), Argentina


Collaboration with UCOR has is expected to occur through:

i)

South American Emissions, Megacities, and Climate (SAEMC, Pro
posal
2005
-
2008)

ii)

Air quality modeling modeling and forecast (CONESUD proposal)


These projects involve atmospheric modeling, with emphasis on inverse and
assimilation techniques for air quality models.


Key
-
contact

o

Dr. Germán Torres


The expected contribut
ions from UCOR:

o

Implmentation of data assimilation (Kalman filtering) for the chemistry,
Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


30

/
43

transport and deposition model to be adopted at the Chilean Weather
Office for operational purposes

o

Capacity building opportunities through short (1
-
2 weeks) and mid t
erm
(1
-
3 months) stays of Chilean students and researchers at UCOR and of
UCOR researchers in Chile.


North America



Pacific Rim Application and Grid Middleware Assembly (PRAGMA), USA

Collaboration with PRAGMA has been started in 2004, and it aims at the
in
corporation of CMM’s computers resources to the PRAGMA grid computing
network.


Key
-
contact

o

Dr.
Peter Arzberger (U San Diego)


The expected contributions from PRAGMA:

o

Provide expertise and advice regarding parallel and distributed
computing

o

Capacity buildi
ng opportunities through short (1
-
2 weeks) stays of
Chilean students and researchers at PRAGMA and of PRAGMA
researchers in Chile.





National Center for Atmospheric Research (NCAR), USA


Collaboration with NCAR has taken place through:

o

Urban Mobile Emissio
ns in South American Cities (UMESAM, 2004
-
2005)

o

South American Emissions, Megacities, and Climate (SAEMC, Proposal
2005
-
2008)


Key
-
contact

o

Gabrielle Pétron


The expected contributions from NCAR:

o

Implement appropiate inverse modeling techniques for atmosphe
ric
models, with emphasis in satellite data and products

o

Capacity building opportunities through short (1
-
2 weeks) stays of
Chilean students and researchers at PRAGMA and of PRAGMA
researchers in Chile.

Europe

France



Ecole Nationale de Ponts et Chaussées (
ENPC) & Institut National de
Recherche en informatique et en authomatique (INRIA), France


Projects in commnon:

o

“Application and development of inverse modeling techqniques to air
quality monitoring network design” (2002
-
2006)

o

Air quality modeling modeling

and forecast (CONESUD proposal)


Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


31

/
43

Key
-
contacts

o

Dr. Bruno Sportisse

o

Dr. Isabelle Herlin

o

Dr. Jean
-
Pierre Issartel

o

Dr. Jean
-
Paul Berroir


The expected contributions from ENPC/INRIA:

o

Transference of chemical weather forecast model “POLAIR“ to the
Chilean count
erparts

o

Capacity building opportunities through short (1
-
2 weeks) stays of
Chilean students and researchers at ENPC/INRIA and of ENPC/INRIA
researchers in Chile.




Laboratoire d’Optique Atmosphérique, LOA, Francia


Projects in commnon:

o

Impacts of natural an
d anthropogenic aerosols on the stratocumulus deck
off Chile (2004
-
2006)


Key
-
contacts

o

Dr. Olivier Boucher


The expected contributions from LOA:

o

Transference and support in the development of appropriate
parameterizations of biogenic sulfur emissions to be

applied in regional
dispersion and climate models

o

Capacity building opportunities through short (1
-
2 weeks) stays of
Chilean students and researchers at LOA and of LOA researchers in
Chile.


Germany



Max Planck Institute for Meteorology (MPI
-
H), Germany


C
ollaboration with MPI
-
H has been established through:

i)

A Europe
-
South America Network for Climate Change Assessment and
Impact Studies (CLARIS, 2004
-
2007)


Key
-
contacts

o

Dr. Guy Brasseur

o

Dr. Claire Granier


The expected contributions from MPI
-
H:

o

Transference
, advice and assistance for the implementation of the
regional model REMO

o

Transference of global and continental climate scenarios to be used as
boundary conditions for regional scale climate simulations

o

Advice on appropriate inverse modeling techniques fo
r emission
estimates

o

Capacity building opportunities through short (1
-
2 weeks) stays of
Chilean students and researchers at MPI
-
H and of MPI
-
Hresearchers in
Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


32

/
43

Chile.


Sweden



Swedish Meteorological and Hydrological Institute (SMHI), Rossby Centre,
and Depart
ment of Meteorology, Stockholm University (MISU), Sweden


There is a long tradition of collaboration with SMHI, the Rossby Center and
MISU. Projects in common:

o

Strengthening of the Air Quality Information System (Working area 2):
Application of a regional
-
scale model over the central part of Chile (1998
-
2000)

o

CONICYT
-
FONDECYT 1030809, 2003
-
2006: Stratosphere
-
Troposphere
Exchange processes and their impact on the ozone balance in the subtropics
of the Southern Hemisphere: A multi
-
scale int
e
grated study based

at Cerro
Tololo (30ºS, 70ºW, 2200 m.a.s.l).



Key
-
contacts

o

Dr. Annica Ekman

o

Dr. Joakim Langner

o

Dr. Markku Rummukainen


The expected contributions from SMHI/MISU:

o

Transference, advice and assistance in the implementation of the
parallelized version of the
regional climate model RCAO, including
parameterizations of sulfur aerosols and their interactions with
stratocumulus clouds

o

Capacity building opportunities through short (1
-
2 weeks) stays of
Chilean students and researchers at SMHI/MISU and of SMHI/MISU
r
esearchers in Chile.



4.3

E
XPECTED RESULTS AND
PRODUCTS


Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


33

/
43

5

C
OST
S

AND BENEFITS

A very rough estimate of costs, i.e., the contribution expected from this research grant
(FONDEF), is presented below. In
-
kind contributions from the participant Chilean
institutions

are basically infraestracture, computer power, and personnel. The overall
FONDEF contribution cannot exceed 450.000 M$

(ca. 750 kUS$)
, and the in
-
kind
contribution is expected to be similar or larger tothat amount.

Additional fellowships for
students coul
d be obtained via IAI if we succeed in our application. Also, we assume
that ongoing exchange projects will provide complementary resources for trips and per
-
diem.


Items

FONDEF




U
-
Chile

DMC

Total

Salaries

138030

47850

185880

Sub
-
contracts

0

0

0

Cap
acity building

75600

32400

108000

Hardware

18308

1000

19308

Software

2500

2500

5000

Infraestructure



0

Fungibles



0

Trips and per
-
diem



19620

Seminars, publications



0

Intelectual property



0

General costs associated to the project



21957,52

General costs for the institutions



6756,16

Administrtion costs



29321,7344

TOTAL

234438

83750

395843,4144


Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


34

/
43

R
EFERENCES

Albrecht, B., 1989.
Aerosols, cloud microphysics, and fractional cloudiness. Science, 245, 1227
-
1230.

Andres RJ, Kasgnoc AD (1998)

A time
-
averaged inventory of subaerial volcanic sulfur emissions.
Journal of Geophysical Research 103: 25,251
-
25,261.

Boucher, O. and U. Lohmann, The sulfate
-
CCN
-
cloud albedo effect: A sensitivity study with two general
circulation models, Tellus, 47B, 28
1
-
300, 1995.

Charlson, R.J., Lovelock, J.E., Andreae, M.O. and Warren, S.G., 1987. Oceanic phytoplankton,
atmospheric sulphur, cloud albedo and climate. Nature, 326, 655
-
661.

Charlson, R. J., Langner, J., Rodhe, H., Leovy, C. B. and Warren, S. G., 1991. Pe
rturbation of the
northern hemisphere radiative balance by backscattering from anthropogenic sulfate aerosols.
Tellus 4B3, 152
-
163.

Chuang, C. C., Penner, J. E., Taylor, K. E., Grossman, A. S. and Walton, J. J., 1997. An assessment of the
radiative effects

of anthropogenic sulfate. J. Geophys. Res., 102, 3761
-
3778.

Andres RJ, Kasgnoc AD (1998) A time
-
averaged inventory of subaerial volcanic sulfur emissions.
Journal of Geophysical Research 103: 25,251
-
25,261.

Cosme, E., C. Genthon, P. Martinerie, O. Boucher
, and M. Pham, Sulfur cycle at high
-
southern latitudes
in the LMD
-
ZT General Circulation Model,
Journal of Geophysical Research
, 107(D23), 4690,
doi:10.1029/2002JD002149, 2002.

Gallardo, L., Olivares, G., Langner, J. and Aarhus, B., 2002: Coastal lows and
sulfur air pollution in
Central Chile. Atmos. Env. 36/23 pp. 3829
-
3841

Hartmann, D., Ockert
-
Bell, M., and Michelsen, M., 1992: The effect of cloud type on earth’s energy
balance: global analysis. J. Clim., 5, 1281
-
1304.

Hartmann, D., 1994. Global physical
climatology.
Academic Press, 411 pp..

Haywood, J. and Boucher, O., 2000. Estimates of the direct and indirect radiative forcing due to
tropospheric aerosols: A review. Rev. Geophys., 38, 513
-
543.

Huneeus, N., Gallardo, L., Rutllant, J., and Boucher, O., 20
03. Dispersion of sulfur aerosols off the coast
of Northern Chile: an exploratory study.
Simposio de Cambio Global: Hacia una visión Sistémica,
Punta Arenas.

IPCC, 2001: The scientific basis. Summary for policy makers. IPCC WGI Third assessment report
(Ava
ilable from
www.ipcc.ch
).

Jones, A. and Slingo, A., 1996. Predicting cloud
-
droplet effective radius and indirect sulphate aerosol
forcing using a general circulation model. Q. J. R. Meteorol. Soc., 122, 1573
-
1595.

Kiehl,

J. and Briegleb B, 1993: The relative roles of sulfate aerosols and greenhouse gases in climate
forcing. Science 260, 311
-
314.

Lefohn A.S., Husar J.D., and Husar R.B. (1999). Estimating Historical Anthropogenic Global Sulfur
Emission Patterns for the Peri
od 1850
-
1990. Atmos. Environ., 33,21, 3435
-
3444.

Lohmann, U., 2002. Possible aerosol effects on ice clouds via contact nucleation. J. Atmos. Sci., 59, 647
-
656.

Nemesure, S., Wagener R. and Schwartz, S. E.,1995. Direct shortwave forcing of climate by
anthro
pogenic sulfate aerosol size and composition. J. Geophys. Res., 100, 18739
-
18754.

Ramanathan, V., P.J. Crutzen, J. Lelieveld, A.P. Mitra, D. Althausen, J. Anderson, M.O. Andreae, W.
Cantrell, G.R. Cass, C.E. Chung, A.D. Clarke, J.A. Coakley, W.D. Collins,
W.C. Conant, F.
Dulac, J. Heintzenberg, A.J. Heymsfield, B. Holben, S. Howell, J. Hudson, A. Jayaraman, J.T.
Kiehl, T.N. Krishnamurti, D. Lubin, G. MacFarquhar, T. Novakov, J.A. Ogren, I.A. Podgorny, K.
Prather, K. Priestley, J.M. Prospero, P.K. Quinn, K.
Rajeev, P. Rasch, S. Rupert, R. Sadourny,
S.K. Satheesh, G.E. Shaw, P. Sheridan and F.P.J. Valero. The Indian Ocean Experiment: An
Integrated Analysis of the Climate Forcing and Effects of the Great Indo
-
Asian Haze. Submitted
revisions, J. Geophys. Res.
-
At
mos.,January 2001.

Robertson, L., Langner, J., and Engardt, M. 1999.
An Eulerian limited
-
area atmospheric transport model.
J. Appl.Me
t. 38, 190
-
210.

Rodhe, H., Langner, J., Gallardo, L., and Kjellström, E., 1995: Global scale transport of acidifying
pollut
ants. Water, Air and Soil Pollut., 85:37
-
50.

Rodhe, H. 1999. Clouds and climate. Nature. 401, 223
-
224.

Rodhe, H., Dentener, F. and Schulz, M. The Global Distribution of Acidifying Wet Deposition. Environ.
Sci. Technol, 36 (20), 4382
-
4388.

Rutllant J., Fuen
zalida H., Aceituno P., Montecinos A., Sanchez R., Salinas H., Inzunza J. y Zuleta R.,
2003, Climate dynamics along the arid northern coast of Chile; The 1997
-
1998 DICLIMA
experiment. Submitted to Journal of Geophysical Research

Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


35

/
43

Rutllant J., Fuenzalida H.,

Aceituno P., Montecinos A., Sanchez R., Salinas H., Inzunza J. y Zuleta R.,
2002, Climate dynamics along the arid northern coast of Chile; The 1997
-
1998
DICLIMA

experiment. Submitted to Journal of Geophysical Research.

Scholes et al, 2003.
Biosphere
-
Atmo
sphere Interactions (Ch. 2). In “
The Changing
Atmosphere
:
An Integration and Synthesis of a Decade of
Tropospheric Chemistry Research”.
Brasseur et al (Eds.).
Springer
-
Verlag (
ISBN:
3
-
540
-
43050
-
4).

Twomey, S., 1974. Pollution and the planetary albedo.
Atmo
s. Environ. 8, 1251
-
1256.

Undén et al, 2002.
HIRLAM
-
5 Scientific documentation.
(Available at
http://www.knmi.nl/hirlam/)

Wilson J., Cuvelier C. and Raes F., 2001. A modeling study of global mixed aerosol fields. J. Geophys.
Res., 106, 34081
-
34108.


Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


36

/
43


Nasa
Intelligent Systems Project

(
http://is.arc.nasa.gov/
)


"The Role of Data Mining in Earth Science Data Interoperability",

Rahul Ramachandran,
Helen Conover, Sara J. Graves, Ken Keiser, John Rushing,
ASPRS Annual Conference,
Conference on Remote Sensing Education (CORSE), Education for the Next Millennium,
Portland, Oregon, May 17
-
22, 1999



referencias MAISA



[Caromel04] D. Caromel, L. Mateu, E. Tanter, Sequential Object Monitors, European
Conference

on Object
-
Oriented Programming (ECOOP), 2004.

[Chiba00] S. Chiba, Load
-
time Structural Reflection in Java, European Conference on
Object
-
Oriented Programming (ECOOP), 2000.


[Kiczales97] G. Kiczales et al., Aspect
-
Oriented Programming, European Conference

on Object
-
Oriented Programming (ECOOP), 1997.

[Rodriguez04] L. Rodriguez, E. Tanter, J. Noye, Supporting Dynamic Crosscutting with
Partial Behavioral Reflection: A Case Study, SCCC 2004.

[Tanter04] E. Tanter and J. Noye, Motivation and Requirements for a
Versatile AOP
Kernel, European Workshop on Aspects in Software (EIWAS), 2004.

[Tanter04b] E. Tanter, From Metaobject Protocols to Versatile Kernels for Aspect
-
Oriented Programming, PhD Thesis, University of Nantes, France, and University of
Chile, Chile. 2
004.

[Tanter05] E. Tanter, L. Rodriguez, J. Noye, A Versatile AOP Kernel for Java,
submitted to International Conference on Aspect
-
Oriented Software Development
(AOSD), 2005.



Bibliografia


\
bibitem{babuska86}

Babuska I., Zienkiewicz O.C., Gago J., de~A.~
Oliveira E.R.

Accuracy estimates and adaptive refinements in finite element


computations.

John Wiley
-
Sons, 1986


\
bibitem{MeshReview}

Bern M., Eppstein D.

Mesh Generation and Optimal Triangulation.

Palo Alto Research Center. Xerox, March 1992


\
bibitem{R
ivaraHitschfeldSimpson}

Rivara M.C., Hitschfeld N., Simpson R.B.

Terminal edges Delaunay (Small Angle Based) Algorithm for

the Quality Triangulation Problem.

Computer
-
Aided Design, vol.~33, 263
--
277, 2001


Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


37

/
43

\
bibitem{HitschfeldRivara2002a}

Hitschfeld N., Riv
ara M.C.

Automatic Construction of Non
-
obtuse Boundary and/or


Interface Delaunay Triangulations for Control Volume Methods.

International Journal for Numerical Methods in Engineering,


vol.~55, 803
--
816, 2002


\
bibitem{Hit03a}

Hitschfeld N., Villablanca

L., Krause J., Rivara M.C.

Improving the Quality of meshes for the simulation of

semiconductor devices usin Lepp
-
based algorithms.

International Journal for Numerical Methods in Engineering,


vol.~58, 333
--
347, 2003


\
bibitem{Sim01a}

Simpson R.B., Hitsch
feld N., Rivara M.C.

Approximate Quality Mesh Generation.

Engineering with Computers, vol.~17, 287
--
298, 2001




\
bibitem{Hitschfeld92}

Hitschfeld N., Conti P., Fichtner W.

Mixed Elements {T}rees: {A} {G}eneralization of {M}odified


{O}ctrees for the {G}e
neration of {M}eshes for the {S}imulation of {C}omplex


3
-
{D} {S}emiconductor {D}evices.

IEEE Transactions on CAD/ICAS, vol.~12, 1714
--
1725,


November 1993





Te envio tambien las que solo aparecen el que ya te mande


\
bibitem{Hitschfeld2000b}

R.~Farias

N.~Hitschfeld, G.~Navarro.

Tessellations of cuboids with steiner points.

In Proceedings of the 9th Annual International Meshing

Roundtable, pages 275
--
282. New Orleans, U.S.A., October 2
-
5, 2000.


\
bibitem{HitschfeldRound96}

N.~Hitschfeld and R.~Farias.

1
-
irregular element tessellation in mixed element meshes for the

control volume discretization method.

Proceedings of the 5th International Meshing Roundtable,


pages 195
--
204. Pittsburgh, Pennsylvania, U.S.A., October 10
-
11, 1996.


\
bibitem{nancyDiss}

N.~
Hitschfeld.

Grid Generation for Three
-
dimensional Non
-
Rectangular

Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


38

/
43


Semiconductor Devices.

PhD thesis, ETH Zurich, Series in Microelectronics, Vol. 21, 1993.

Hartung
-
Gorre Verlag, Konstanz, Germany.




Charlson, R.J., J.E Lovelock, Andreae, M.O. Warren, S.
G., 1987, Oceanic
phytoplankton atmospheric sulphur, cloud albedo and climate. Nature, 326, 655
-
661.


Gabric, A.J., G. Ayers, C.N. Murray, J. Parslow, 1996, Use of remote sensing and
mathematical modeling to predict the flux of dimethylsulphide to the atmo
sphere in the
Souther Ocean. Adv.Space Res. Vol 18, No7, 117
-
128.


.IPPC, 2001, The scientific basis, Summary for policy makers. IPCC WGI Thrird
assessment report (available from www.ipcc.ch)


Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


39

/
43

A
NNEX
1:

C
URRICULA


SICTI

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!!!!!!!!!!!!!!!!!!!!!!!!!
Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


40

/
43

A
NNEX
2:

I
NFRASTRUCTURE

Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


41

/
43

A
NNEX
3.

M
ODELS

RCAO model


CPTEC’s

POLAIR


Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


42

/
43

CONTENTS

Summary

________________________________
____________________________

1

1

Backgr
ound and rationale

________________________________
___________

2

2

Science and technology aspects
________________________________
_______

4

2.1

Computer Science and technology

________________________________
_____

4

2.1.1

Parallel and distributed computing, networks and communication (Jo, Gonzalo)

________

4

2.1.2

Mixed element meshes over complex topographies and cities.

______________________

6

2.1.3

Data mining

________________________________
_____________________________

7

2.2

Earth System Modeling and regionalization

_____________________________

7

2.2.1

Downscaling of

climate outputs and scenarios

________________________________
__

9

2.3

Atmospheric forecast (LGK, Jorge, Enrique, Juan, Ricardo, Maisa, Rodrigo)

10

2.4

Data assimilat
ion (Jaime, Axel)

________________________________
_______

11

3

State of the art and opportunities in Chile

_____________________________

13

3.1

Computer Science and technology

________________________________
____

13

3.2

Forecast and climate modeling

________________________________
_______

14

3.2.1

Climate modeling

________________________________
_______________________

14

3.2.2

Atmospheric Forecast

________________________________
____________________

15

3.2.3

Data assimilation

________________________________
________________________

15

3.3

Science networking and capacity building

______________________________

15

3.3.1

Technological networks

________________________________
___________________

16

3.3.2

Scientific collaboration

________________________________
___________________

16

3.3.2.1

Computer science

________________________________
___________________

16

3.3.2.2

Climate and atmospheric research

________________________________
______

16

3.3.3

Capacity building

________________________________
________________________

18

3.4

Case study: Anthropogenic im
pacts of sulfate aerosols off the Chilean coast
under past, present and future scenarios

________________________________
______

18

3.5

Case study: Chemical weather forecast for urban centers in Chile

__________

20

4

Objectives and approach

________________________________
___________

22

4.1

Objectives

________________________________
________________________

22

4.1.1

General objectives

________________________________
_______________________

22

4.1.2

Specific objectives

________________________________
_______________________

22

4.2

Methodological approach and organization

_____________________________

23

4.2.1

Work package

1 (WP1): Computing

________________________________
_________

24

4.2.2

Work package 2 (WP2): Climate modeling

________________________________
____

25

4.2.3

Work package 3 (WP3): Chemical weather
________________________________
____

26

4.2.4

Work package 4 (WP4): Coordination and synthesis

____________________________

26

4.2.5

Work package 5 (WP5): Administration

________________________________
______

26

4.2.6

Review and advice

________________________________
_______________________

26

4.2.6.1

Air
-
sea exchanges

________________________________
__________________

26

4.2.6.2

Chemical weather forecast

________________________________
____________

26

4.2.6.3

Aerosol
-
cloud
-
climate interactions

________________________________
_____

27

4.2.7

Timeline, milestones and external review

________________________________
_____

27

4.2.8

Participants

________________________________
____________________________

28

4.2.8.1

Chilean counterparts

________________________________
________________

28

4.2.8.2

Foreign counterparts

________________________________
________________

29

4.3

Expected results and products

________________________________
________

32

5

Costs and benefits

________________________________
________________

33

Parallel and distributed computing and its application to chemical weather forecast and climate in Chile

FONDEF 2004


43

/
43

References

________________________________
__________________________

34

Annex 1: Curricula

________________________________
___________________

39

SICTI

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Annex 2:
Infrastructure

________________________________
________________________

39

Annex 2: Infrastructure

________________________________
_______________

40

Annex 3. Models

________________________________
_____________________

41

CONTENTS

________________________________
_________________________

42

CONTENTS

________________________________
_________________________

42