Taheri,Hamed

Data Mining Approaches on Groundwater Resources Vulnerability Assessment

in Integrated Water Resources Management

For a long time, human activities were not controlled adequately in terms of supply and demand policies; this leading to the situation that sustainability of water resources was harmed. It is thought to be due to the lack of appropriate knowledge for decision makers. On the other hand, huge amounts of data are being generated from different resources which can be a great source are necessitated to become in usable form for managers. Making use of suitable techniques to discover knowledge in form of patterns and rules from the distributed databases similar to water resources databases with noticeable complexity, multiplicity and amount of data are seriously required. Data Mining (DM) can be viewed as a result of the natural evolution of information technology. Data Mining refers to extracting or mining knowledge from large amount of data, searching for hidden pattern and identifying the relationship between the data.

This dissertation presents Data Mining (DM) approaches for groundwater resources vulnerability assessment in terms of knowledge towards integrated water resources management. Two case studies are presented to demonstrate the effectiveness of building DM models in water resources management. First application is developed on estimating relative humidity in a climatological database. The proposed approach uses linear regression data mining technique to estimate the relative humidity by 94% accuracy. The results obtained are very promising and shown the usability of the Data Mining techniques for estimating important climatological parameters.

The second application focuses on predicting and discovering knowledge of groundwater behavior due to different supply and demand polices. Najafabad aquifer’s data in the upstream of Zayandeh Rud river basin are used for this part of research. After preparing a conceptual model and preprocessing data, variety of models based on different scenarios, are designed and implemented. The best developed model predicts the groundwater surface changes approximately by 90% accuracy. Moreover, post processing of the results also leads to valuable knowledge about the groundwater behavior in different conditions. The results show significant contributions of Data Mining techniques to predict and discover hidden patterns between different parts of water resources system.