Abstarct: Grade modelling and reserve estimation of ore deposits is one of the most important parameters for designing and product scheduling in mining operation. Generally reserve estimation methods are divided to geometric methods and distance baxsed methods. geometric method include polygon, triangular and section baxsed, comprise inverse power of distance, geostatistics. In this thesis, reserve of Zarshooran gold mine located in 25km north of takab county, in west azarbayjan province, has been estimated with inverse power of distanc, geostatistics and artificial neural networks methods. At first, statistical Analysis carried out with the use of exploration data and trend analysis was investigated. Then block model of this deposit was created. Next, for grade of each block was estimated with the use of inverse power of distance. In the next step after carring out variography, grade of each block was estimated by kriging. In next step, reserve estimation was done by using artificial neural networks. This method was created from inspiration of human neural system and is capable of learning and can distinguish the relationship between non-linear input and output data. In reserve estimation by using artificial neural networks, after training the network with exploration data, the relationship between input data (coordinate of samples) and output (Au grade) are determined by neural network. Then this network can be used to estimate our grade. In this thesis For estimation of blocks grade, two kind of networks used; BP neural network and RBF network. At the end sequential Gaussian simulation was used to produce 100 realizations for this deposit. The result shows that the reserve of this deposit with 0.4% cut-off grade by inverse power of distance, kriging, artificial neural networks and sequential Gaussian simulation with 95% confidence are 149, 137, 151 and 195 tonns respectively.