Large-scale data processing, made possible by the dissemination of very powerful computers, is a formidable tool for many fields of research. Supercomputers with the capacity to process millions of lines of code per second are being used in the most diverse fields of knowledge to solve problems and make projections.
Brazil’s status among the world ranking of the 500 highest-performing computers is the topic of the report on page 37. Three of the four Brazilian systems on the list are part of a cluster located in a charming building inspired by the hat of Santos Dumont, at the facilities of the National Scientific Computing Laboratory (LNCC), in Petrópolis, Rio de Janeiro State. The fourth is located at the Center for Integrated Manufacturing and Technology of the National Industrial Training Service (SENAI-CIMATEC), in Salvador, Bahia.
The current state of Brazil’s supercomputing is rather gloomy. In 2004, Brazil achieved a 9th place ranking, with nine systems in operation. Today it finds itself in 17th place. Besides the setback in terms of international comparisons, another source of concern identified in the report is the under-utilization of supercomputers that are currently operating in Brazil, due in part to the high costs involved in their operation and maintenance. Essential for advancing science and developing technology, these systems require a well thought-out national policy of acquisition and use.
Supercomputers are used in a broad array of fields such as defense and health as well as in climate analysis – from forecasts of daily rainfall to the possibility of near-future temperature changes on Earth. This month’s cover story discusses the findings of simulations conducted using two climate models processed on high-performance computers.
Initial computer models created to simulate earth’s environment used only data on the physical elements of climate, such as water, air and sunlight. With advances in the models’ data processing and programming capacity, it became possible to increase the degree of detail, estimating for example, how changes in the oceans’ acidity caused by the increase in carbon dioxide in the air affect marine food chains.
One of the difficulties faced by researchers who work with climate modeling is an apparent paradox: since increasingly more variables are considered in the simulations, and each element is subject to variation, this could lead to greater variation in simulation results. A different forecast of temperature variation or rainfall pattern could result from simulations in models that focus on different elements.
The simulations presented in the cover story used two global climate models to basically predict the increase or decrease in temperature and rainfall volume. Along with these two models, a regional-scale model developed by the National Institute for Space Research was then run. The objective went beyond projecting an increase or reduction in rainfall and temperature in Brazil: researchers sought to predict the impact of these changes. More precipitation in forests can have negligible effects, while in metropolitan areas similar amounts of rainfall can be catastrophic. This is why data about economic, social and environmental conditions of all Brazilian municipalities were included. The resulting projections by the two simulations are similar for about 80% of the Brazilian territory, lending weight to the results.
An interesting and useful application of the capacity for processing large quantities of data is described in the report about the genealogy of Brazilian academics conducted from information found in the CNPq’s Lattes Platform. The federal data extracted from 4.5 million curricula were supplemented by information from databases compiled by Capes and the Brazilian Academy of Sciences. Besides being of historical and sociological interest, methodologies based on academic genealogies could be used in the future in the field of evaluation, where they can measure a researcher’s influence on the training of younger generations.
Starting with this issue, the magazine will be a little different: the Data and Projects section will now be made up exclusively of Data. Recently-approved research under the grant mechanisms Thematic Projects and Young Investigators will be highlighted on the Virtual Library page of the FAPESP Documentation and Information Center. The BV, as we call it, is a reference source for research supported by the Foundation: as of 11/5/2016, it contained information about 89,841 research grants and 126,804 scholarships in Brazil and abroad awarded by FAPESP (www.bv.fapesp.br).