A data warehousing and data mining approach for analysis and forecast of cloudburst events using OLAP-based data hypercube

Abstract

The multidimensional data model can be effectively utilised for analysing huge and detailed meteorological datasets forecasted by numerical weather prediction (NWP) model. The model cannot predict any weather event directly. The output products of model are interpreted by man-machine mix to infer the idiosyncratic behaviour of weather events. The mathematical tools for analysis and forecasting are able to provide forecast of weather variables only at grid-points. In this paper, the technology of dimension modelling has been adapted for analysing NWP model output datasets corresponding to sub-grid scale events viz. cloudburst, using OLAP technique. The huge datasets of weather variables available directly and derived indirectly, are mined so as to locate the patterns of cloudburst formation. K-means clustering technique has been used to generate clusters of convergence and divergence, for four real-life cases of cloudburst. It has been observed that clustering technique can help in identification of patterns conducive to formation of cloudburst.