The most important quantity in groundwater resources management is groundwater level at various locations of the aquifer. This quantity has been obtained by means of quantitative groundwater modeling . The main problem to develop such a model is determining exactly its parameters, particularly hydrogeologic coefficients which in most cases inexact ones result in unsatisfactory model results. In this thesis, after a survey on system theory, key definitions of this theory and their examples in literature of groundwater modeling was expressed. Then subsequent to a review on data assimilation methods, various versions of Kalman filter as a sequential data assimilation method was explained. Kalman filters for large-scale systems (like large plain aquifers) are discussed. Ensemble Kalman filters as the most common version of Kalman filter in natural sciences were explained. After that, Kalman filter was applied to the groundwater system. Then Deterministic Ensemble Kalman Filter ( DEnKF ) which is a square root of mean preserving EnKF is explained. Then methods of improving filter performance with focus on localization were discussed and finally a type of localization within the framework of DEnKF was applied. Najafabad aquifer (in Isfahan province) with area of approximately 150 km 2 in timeframe of Mehr 1379 (Oct. 2000) to Shahrivar 1386 (Sept. 2007) with monthly time steps was modeled. Water table level of unit hydrograph of this aquifer within 1379-80 water year has a downtrend and after that changes to uptrend are studied and their stochastic and square root versions are explained. The least amount (1697 m above mean sea level) was in Oct. 2001. By division of the aquifer into 5 zones, the first 54 months’ values of hydrogeologic coefficients for each zone were calibrated with the help of monthly data of 32 observation wells. Then the model was verified over the last 30 months. It showed a good performance in comparison with monthly data of other 17 observation wells of the aquifer. The result of this calibrated model is assumed to be true and at some points after adding a Gaussian noise considered as observation data. DEnKF was combined with 45 observations of true run with inexact hydrogeologic parameters (2, 5, and 10 times of calibrated values of hydraulic conductivity and specific yield). Keywords: Groundwater, Data Assimilation, Localization, Ensemble Kalman Filter
