Accurate reservoir model requires complete information of subsurface properties, specifically porosity and permeability. Reservoir heterogeneity is the fundamental challenge for geoscientists to predict these properties which may affect the reservoir performance and their well productivity. Porosity is one of the key parameters for accumulation of hydrocarbon but its prediction is difficult due to significant variation over a reservoir volume. A spatial distribution of porosity can be investigated by integrating the 3-D seismic and well log attributes which may help in determining such reservoir variation. In addition, nonlinear multivariable regression techniques such as multivariable transform, Genetic algorithm, and Probabilistic Neural Network analysis have also been implemented to achieve high correlation coefficients between well log properties and seismic data. Results from nonlinear regression have better correlations than linear regression. In this study, a 3-D low frequency model (LFM) is proposed which can be estimated by kriging interpolation of resultant impedance values from well log data. Furthermore, �seismic inversion� is adopted for extracting correlated attributes to merge with the LFM so as to better construct a pseudo log volume. A polynomial neural network (PNN*1) is utilized to convert resultant acoustic impedance values into a distinct reservoir property such as porosity. PNN* is trained, tested and validated by using gammaray and resultant acoustic impedance values as input and effective porosity values as a target. The trained PNN* is then applied over the whole reservoir volume to generate a pseudo log volume. In the proposed low frequency model, an attempt has been made to achieve high correlation between the predicted and measured porosity logs. It will improve the reservoir characterization and lead to better estimation of hydrocarbon reserves. This low frequency model achieves better correlation between the predicted and true porosity log even with a minimum number of measured well logs. © 2016, The Association of Korean Geoscience Societies and Springer-Verlag Berlin Heidelberg.