بررسي كارايي مدل ماشينهاي بردار پشتيبان در پيش بيني غلظت نيترات در آبهاي زير زميني و مقايسه آن با مدل MODFLOW

چكيده انگليسي :

Performance Evaluation of Support Vector Machines in Prediction of Nitrate Concentration in Groundwater and Its Comparsion to MODFLOW Raheleh Arabgol r arabgol@cv iut ac ir Date of Submission Department of Civil Engineering Isfahan University of Technology Isfahan Iran Degree M Sc Language FarsiSupervisor Dr Majid Sartaj msartaj@cc iut ac ir Abstract Nowadays Nitrate contamination of groundwater resources is one of the most important environmental issues Nitrate is both soluble and mobile so it is prone to leaching through soil with infiltrating such as rainfall orirrigation water On the other hand discharge of wastewaters from industrial and agricultural activities has resultedin degradation of water resources Elevated nitrate concentrations in drinking water can cause Methemoglobinemiain infants and stomach cancer in adults As such the US Environmental Protection Agency US EPA hasestablished a maximum contaminant level MCL of mg l NO N Therefore access to reliable water resourceswill be a real challenge for many communities especially in semi arid and arid regions Considering the above assessment and prediction of water quality will be a first step in planning and management of water resources Mathematical modeling and simulation is one of the tools used by researchers for this purpose Groundwaterprovides one third of the world s drinking water Importance of identifying groundwater pollution and increasingdemand for water quality demonstrate the need for creating powerful reliable and predictive models In this field intelligent and data driven models are new methods that are rapidly expanding in various fields of science Thesemodels are able to train and generalized they could be used for estimation prediction and management in variousaspects of water resources Support Vector Machine is a method for classification and regression Even if there arelimited available data SVMs have good performance and a greater ability to generalize compared with othertraditional methods Lack of data for validating the model and using simple assumptions to simulate groundwaterflow will lead to undesirable results in physically based models Therefore this study presents a Support VectorMachine SVM model to predicting concentration of nitrate in groundwater of Arak Plain For this purpose samples from 56 observation wells over a period of one year for 5 consecutive seasons were collected and used Easily measurable parameters such as temperature electrical conductivity groundwater level total dissolvedsolids dissolved oxygen pH and well location were used as input parameters in the SVM based nitrate prediction Different models were trained and tested based on the real available data and different input parameters Then theoptimum model were selected and used to prediction nitrate concentration in Arak Plain Prediction results wereassessed using different efficiency measures including R R8 E and RMSE They are 6 9 6 6 and 6 respectively Prediction results show the ability of SVM to build accurate model with strongpredictive capabilities compared to MODFLOW results R 6 99 Keywords Data Driven Model Support Vector Machines Nitrate Concentration Groundwater Resource

عربگل، راحله

بررسي كارايي مدل ماشينهاي بردار پشتيبان در پيش بيني غلظت نيترات در آبهاي زير زميني و مقايسه آن با مدل MODFLOW

مدلسازي داده محور