Fractal Theory and Correlation Clustering Techniques to Improve Climate Change Forecast

• Grant number: 11/06655-3
• Support Opportunities: Scholarships in Brazil - Post-Doctoral
• Start date: September 01, 2011
• End date: May 31, 2013
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Agma Juci Machado Traina
• Grantee: Robson Leonardo Ferreira Cordeiro
• Host Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil

Abstract

Since its establishment, the Intergovernmental Panel on Climate Change has been reporting that, probably due to human intervention, the average temperature on earth has been continually raising at least for the last hundred years. The ten warmest years registered in history are within the last twenty years! 2010 is tied with 2005 as the warmest year on record. Climate changes forecast allows us to understand, to prevent and to mitigate such bad consequences of the human activities. The forecast uses numerical models, known as climate change models, that describe the physical and dynamical processes of the climate system to simulate future climates as response to changes in external forces. Such models are usually evaluated by starting the simulation in a given instant in the past, and using statistical comparison of trend analyses to compare the simulated results to the real recorded data. Today, the analyses of statistical significance indicate that the simulated results closely follow the recorded data, thus giving a strong evidence of the models' correctness. However, at GBdI we have been performing such kind of analysis using Fractal Theory techniques, and we have found that those techniques can clearly differentiate the simulated from the real data. Thus, although the current climate change models are appropriate from the statistical point of view, the Fractal Theory shows that they can be improved. Further studies lead us to believe that the so-called correlation clustering techniques are promising tools to help improving such models. Therefore, the question we intend to answer in this post-doctoral project is: How to use Fractal Theory concepts and correlation clustering techniques to improve the climate change models.