Data analysis of spatio-temporal sensor data as a contribution to the model analysis for water resources

Abstract

The quality of the information is measured by its accuracy and its relevance over time. Therefore, the process of data analysis of the sensor eco-data is of a great importance to the detection and prediction of the eco-hydrology phenomena. The existing models for data mining do not relate to the continuously changing characteristics of the sensor eco-data. Furthermore, most of the monitoring systems are based on event alert services, which do not answer to the continuous variations of the measured parameters. Our approach embeds the nature of system characteristics into one dynamic model for data mining of continuously changing spatio-temporal characteristics of one eco-hydrology system. The continuously gathered sensor eco-data from the region of Lake Prespa consisted of 320 water samples, among them 224 from the lake gauging stations and 96 from the river gauging stations. Considering the recommendations from the Water Framework Directive (WFD), the sensor eco-data were grouped into three types: physical, chemical and biological, corresponding to their aspect of water quality. All of these types convey the same class definition in the form of value, spatial and temporal information. To define our sensor data mining model we contribute to three segments: outlier analysis, pattern analysis, and prediction analysis. The suggested sensor data analysis model should be of a useful asset in obtaining knowledge for certain aquatic phenomena.