Home | english  | Fakultät | Impressum | Datenschutz | KIT

A data fusion method applied to paleoclimatic reconstruction

A data fusion method applied to paleoclimatic reconstruction



Gustavo Soto Muster


The challenge of a data fusion method lies in how to integrate sources of data, redundant and/or complementary, to obtain estimates more robust and reliable. One expects that estimates can be better than if any of these source were used individually. 
If you are planning to design a data fusion method for real-world  applications, it is important to bear in mind several issues, mainly, related to how to combine different types of information at different levels of accuracy (uncertainty) using multivariate data, including prior information (our beliefs) when estimating the model of a stochastic process.
In this presentation, as a case of study, I will focus specifically on addressing the problem of reducing the uncertainty in the estimation of the historical earth temperature curve using ice cores oxygen isotope measurements by means of an (data fusion) technique known as Control Variate, in a Bayesian framework. This technique reduces the variance of the estimator by using redundant information from other data sources. In this case, using measurements of another ice cores located geographically close each other. It will show some preliminary results of applying the proposal bayesian data fusion technique and prospect of the future work.