El Niño phenomenon is a key factor for Latin-American climate prediction. The different activities included in climate prediction can be roughly classified in numerical model simulation and forecast production, and data access/exchange and data analysis. In modern terms, this last topic is referred to as data mining.

Modern climate science deals with different sources of data both from oceanic, surface and upper air observing networks, and also from satellites. Moreover, this data is geographically distributed in various locations and it is stored in different kind of systems and formats. On the other hand, an increasing number of global climate predictions are available from numerical atmospheric and oceanic model integrations (reanalysis projects, ensemble model and multi-model experiments, etc.), which anticipate the ocean and atmospheric description of future climate. These sources of data can jointly help to solve many important problems, such as regional climate change projections, i.e., the effects of climate change on different regions of interest. To this aim, efficient problem-driven statistical analysis tools are required for discovering knowledge, or useful information, within the huge amount of information. Data mining and machine learning techniques have been developed in the last decades to deal with this task, and different alternatives have been studied to make easier the process in a distributed environment such as the GRID.

The main goal of this task is the deployment of data access and mining applications oriented to the Latin-America area and El Niño phenomenon. The applicability of Grid in this topic is clearly manifested in the FP5 funded projects DataGrid and CrossGrid, both including atmospheric research packages for data access, data sharing and data mining.

The first goal of this task is the migrations of pre-existing data access/sharing tools appropriate for environmental data files (climate simulations and observations). Installation and deployment will be performed by the European partners providing the necessary assistance for the customisation. Induction on the use of the applications will be given if necessary, sharing the effort with the partners involved in the training tasks and the local partners.

A catalogue of available simulations and observations data will be defined, focusing on the tropical El Niño area, which is a main interest of the two partners of this project in South America.

A second goal is the migration of data mining clustering algorithms to GRID (both standard methods and neural network Self-Organised Maps SOM). These algorithms will relate both climate simulations and local observations performing what is called a downscaling process. This will allow users to obtain not only raw data, but also spatial climatic patterns of interest associated to their problems and climatic variables of interest. These techniques will be problem-oriented, driven by the users’ input requirements.

To access the data and run the data mining services in a transparent form, a web service will be developed, where users can ask for specific climatic parameters in areas of interest, obtaining an estimation of them using different climate simulations.

CAM

CAM/CCM3. Global climate atmospheric model. This model simulates the evolution of the climate globally at a resolution of 100-300 km, starting with the present sea surface temperature conditions and producing the atmospheric evolution (temperature, humidity, etc.) for a prescribed future period. It is an open source model widely use for climate simulations, including climate change, that will be gridified by SENAMHI. It can be run independently from any opther application or model, but it will also be the first link in a joint work done by all climate partners.

MM5/WRF

Regional atmospheric model. This regional model runs over a reduced area and takes as boundary contions the output of a global circulation model, producing a high-resolution (1-10 km) simulation of the atmospheric evolution. Open source model widely use for weather forecast that will be gridified by UDEC. It can be run independently from any opther application or model, but it will also be the second link in a joint work done by all climate partners.

SOM

(Self-Organizing Maps). Data minig clustering algorithm. The parallel code for this data-mining algorithm was developed in the CROSSGRID proyect using MPI. It finds homogeneous clusters within data and orgnizes them according to their neigborhood relationships. The straightforward application to climate is obtaining a set of weather types from a database of atmospheric states. These weather types can be used to statistically adapt the output of global models producing local forecasts. It will be gridified by UC and besides it can be run independently from any opther application or model, it will also be the third and final link in a joint work done by all climate partners. SegHidro

SegHidro stands for "Segurança Hídrica" (Hydrological Security) and aims at improving the water management of the Brazilian Northeast, a semi-arid region. The project provides a cyber-infrastructure that allows scientists and decision makers to better cope with the variations of water availability of the Brazilian northeastern region, enabling them to better manage the water resources. SegHidro enables, through a grid portal, collaborative work via the coupling of hydrometeorological computer models, providing access to massive grid-based computer resources using a simple infrastructure in both computation and data. From the models available, BRAMS (Brazilian Regional Atmospheric Model), maintained by CPTEC/INPE is of particular importance. It simulates atmospheric circulations on limited geographical areas from large scale down to large eddy simulations (LES) on the planetary boundary layer. What is most important in this collaboration is the fact that the middleware used by OurGrid and EELA are different, so a special emphasis on the interoperability between them will be the main objective.

The EELA proposal focuses on transferring the organisational model of the European project EGEE into Latin-American Countries. Thus the objective of the Climate Prediction package will be the migration and deployment of applications developed in previous EU founded projects (DATAGRID and CROSSGRID), focusing on an important problem with many social and economical implications: El Niño climate regional projection. This problem is a main objective of the two partners of this project in South America.

Thus, climate application in EELA will exploit the grid, as a new framework in which user-friendly data mining climatologically tools can be successfully deployed. Nowadays, there are different databases of extensive multi-model seasonal and climate simulations suitable to perform climate regional projection studies when combined with local information. For instance, ENSEMBLES is a new FP6 project funded by the EU, which will produce an ensemble of climate simulations, stored on different mass storage facilities. Thus, relevant data sets for El Niño areas have to be identified, accessed and processed in a coherent and transparent way. Moreover, databases of local observations (precipitation or temperature in a network of oceanic or surface stations) have to be browsed and accessed in a transparent form. To this aim we shall use results and technology from previous 5th FP EU founded projects, where different applications for user-friendly satellite data access have been developed (DataGrid WP9. Earth Observation Science Applications), proving the efficiency of GRID technologies for sharing and accessing this type of data.

On the other hand, to automatically extract “patterns” from the climate simulations, it is necessary to implement appropriate data mining techniques. In particular, we consider the implementation of efficient clustering techniques, which provide both visual interpretable classifications of weather types, and downscaled values of climate prediction to local stations. However, “data mining GRIDS” involve a great number of challenges. An efficient deployment of this category of applications for the GRID environment requires extension of the GRID software resources by new components for application-performance monitoring, efficient distributed data access, and specific resource management. This requires the implementation of new middleware components for data mining algorithms, and also the development of user-friendly portals and personalised environments for job submission and data access. Some preliminary work has been carried to deploy data mining applications in the GRID (CrossGrid WP1.4 Data mining in Meteorology). Moreover, the project “Data Mining Grid” of the 6th FP further analyzes the requirements for deploying data mining application in the GRID. Therefore, in this project we will exploit all the available achievements of DataGrid, CrossGrid and results from other ongoing related projects in a way that enables their interoperability.

The main result of this task will be the integration of both climate simulations and observations in a common GRID, granting access to raw and processed data suitable for our dissemination activities, allowing users from different application domains to obtain the specific climatic input variables used by their models in a transparent form.

Therefore, the final expected result of the task is an application to obtain climate predictions for different local variables in the area of El Niño phenomenon.

• Two Information Sheets
• Madrid (Spain), EELA KoM and 1st Workshop, 1-2 February 2006
• A presentation of the EELA Climate Applications
• Mérida (Venezuela), EELA 2nd Workshop, 24-25 April 2006
• A presentation of the status of the EELA Climate Applications
• Santander (Spain), Workshop On Complex Systems: New trends in Technology, 5-9 June 2006
• A presentation of the status of the EELA Biomed and Climate Applications
• Santander (Spain), InterGrid KoM, 19 June 2006
• A presentation of the status of the EELA Climate Applications
• Itacuruçá (Brazil), EELA 3rd Workshop, 24-25 June 2006
• A presentation of the status of the EELA Climate Applications
• Santiago (Chile), 1st EELA Conference, 4-5 September 2006
• A presentation of the status of the EELA Climate Applications
• Geneva (Switzerland), EGEE06 Conference, 25-29 September 2006
• A presentation of the deployment of the EELA Applications in the SEE-GRID Regional Grids Workshop
• Popayán-Cauca (Colombia), 2nd International Seminar on Genomics, Proteomics and Bioinformatics, 25-27 October 2006
• Invited talk "Computación de Alto Rendimiento en GRID. Actividades del Proyecto EELA en Bioinformática y Clima"
• Granada (Spain), Jornadas Técnicas RedIRIS 2006 y XXII Grupos de Trabajo, 13-17 November 2006
• A presentation of the EELA Applications
• Madrid (Spain), CONAMA8 Congress, 27 November-1 December 2006
• A poster with the climate applications
• Paper in Proceedings of CONAMA Conference 8, (CD format)
• Lima (Peru), EELA 4th Workshop, 11-12 January 2007
• A presentation of the status of the EELA Climate Applications
• Abarca, et al. Building a Network in Latin America: e-Infrastructure and Applications. Proceedings of the Spanish Conference on e-Science Grid Computing 1, 83-96 (2007)
• Bogotá (Colombia), EELA 5th Workshop, 5 March 2007
• A presentation of the status of the EELA Applications
• La Plata (Argentina), EELA 6th Workshop, 29-30 March 2007
• A presentation of the status of the EELA Applications
• Vienna (Austria), EGU General Assembly 2007, 15-20 April 2007
• A presentation of the status of the EELA Climate Applications
• Aveiro (Portugal), 4ª Reunión de la Univ. de Aveiro, 26-27 April de 2007
• A demo of the Climate applications
• Manchester (UK), EGEE User Forum, 9-11 May 2007
• A presentation of the EELA Climate CAM+WRF
• A presentation of the EELA applications
• Maputo (Mozambique), IST-Africa, 9-11 May 2007
• A presentation of the EELA applications
• Paper in Proceedings of IST-Africa Conference
• Santiago de Compostela (Spain), IBERGRID Conference, 14-16 May 2007
• A presentation of the EELA applications
• Paper in Proceedings of IBERGRID Conference 1, 29-35 (2007)
• Rio de Janeiro (Brazil), LAGrid Conference, 14-17 May 2007
• A presentation of the EELA applications
• Paper in Proceedings of the LAGrid conference
• Santander (Spain), UC seminar, 16 May 2007
• A presentation of the Climate Applications
• Cofiño, A., San Martín, D. and Gutiérrez, J.M. A Web Portal for Regional Projection of Weather Forecast using Grid Middleware. Lecture Notes in Computer Science, 4489, 82-89 (2007).
• Acapulco (Mexico), American Geophysica Union Joint Assembly, 22-25 May 2007
• A presentation of the status of the EELA Climate Applications
• Varadero (Cuba), EELA Workshop, 29-30 May 2007
• A presentation of the status of the EELA Applications
• Barcelona (Spain), WCRP Workshop on Seasonal Prediction, 4-7 June 2007
• A presentation of the Climate EELA Applications
• Paper in Proceedings of the NETTAB Conference 7, 145-156 (2007)
• Rio de Janeiro (Brazil), XXVII Congresso de la SBC, 30 Jun-06 Jul 2007
• A presentation of the status of the EELA applications
• San Lorenzo de El Escorial (Spain), 7th European Meteorological Society Conference, 1-5 October 2007
• A presentation of the status of the EELA Climate Applications
• Budapest (Hungary), EGEE Conference, 1-5 October 2007
• A presentation of the status of the EELA applications
• La Antigua (Guatemala), 7th EELA Workshop, 17 Oct 2007
• A presentation of the status of the EELA applications
• Mexico City (Mexico), 8th EELA Workshop, 22 Oct 2007
• A presentation of the status of the EELA applications

• UDEC: http://www.udec.cl/concepcion.php