Fourier Transformation and Unsupervised Learning for Extracting Hydrogeological Information from Time Series Data

Cognitive sciences related to hydrogeological studies represent interdisciplinary approaches using computer science, artificial intelligence, and anthropology. Several mathematical methods of machine learning can be fruitfully used in the understanding of groundwater systems and structures such as evolutionary computation, artificial neural networks, global optimization methods. Non-hierarchical cluster analysis as unsupervised learning tool is frequently applied to identification of water-bearing formations, rock typing and in the hydraulic characterization of aquifer systems. Hydrology, as a quantitative earth science, uses them in the processing of data collected in the field, e.g., in inverse modeling, statistical analysis as well as description of long-term dynamic processes. It is presented through an Egyptian example of water level analysis in the article, they are also suitable for describing past processes, climate and anthropogenic effects, and a kind of future prediction of climatic impact. Time cycles of water level change are extracted using the combination of improved Fourier analysis and multivariate cluster analysis. With their help, the accuracy and reliability of the estimation results can be increased, or the extreme noises contaminating the data can be effectively suppressed. The applied and other artificial intelligence methods are of particularly important today when analyzing the effects of extreme weather conditions all over the world.