Mining spatio-temporal patterns of periodic changes in climate data

The climate changes have attracted always interest because they may have great impact on the life on Earth and living beings. Computational solutions may be useful both for the prediction of the climate changes and for their characterization, perhaps in association with other phenomena. Due to the cyclic and seasonal nature of many climate processes, studying their repeatability may be relevant and, in many cases, determinant. In this paper, we investigate the task of determining changes of the weather conditions, which are periodically repeated over time and space. We introduce the spatio-temporal patterns of periodic changes and propose a computational solution to discover them. These patterns allows us to represent spatial regions with same periodic changes. The method works on a grid-based data representation and relies on a time-windows analysis model to detect periodic changes in the grid cells. Then, the cells with same changes are selected to form a spatial region of interest. The usefulness of the method is demonstrated on a real-world dataset collecting weather conditions.