OSAPOL uutisia ja tapahtumia

ERAD16, European Conference on Radar in Meteorology and Hydrology, 10-14.10, 2016. "Classification of Non-Meteorological Echoes from Finnish and Turkish Dual-Polarization Weather Radar Data". "Consistency-Driven Optical Flow Technique for Nowcasting and Temporal Interpolation" [PDF]




8th IPWG (International Precipitation Working Group) & 5th IWSSM (International Workshop on Space-based Snowfall Measurement), 3.-7.10.2016, "Dual-polarization C-band Radar Measurements and Characterization of Hail in Finland between 2012 and 2015" [PDF]




T. Mäkinen & L. Holmström (2016). Modeling Probability Density through Ultraspherical Polynomial Transformations. Accepted for publication in Communications in Statistics - Simulation and Computation. DOI: 10.1080/03610918.2016.1186181

We have deviced a method based on a fast L2 measure, for modeling the probability density (PDF) of arbitrarily high-dimensional data sets with any pointwise evaluable function. This allows the construction of statistically optimal and computationally efficient models of data sets that cannot be easily modeled through other methods, and thus also allows fast and efficient statistical analysis of such data. In this project we are applying the new method to radar data analysis and the construction of an automatic classification system.


Figure 1: A synthetic set of data points drawn from a distribution approximating a 5/7 epicycle.
Figure 2: A statistically optimal analytic (polynomial) model of the PDF of the test set, constructed with the new method.