Geostatistics of petroleum reserves

This thesis discusses the concepts of geostatistics of petroleum reserves. In order to predict the amount of petroleum in a certain area, one must first record the amount of oil at selected well sites in the area. Then one models the variation in the data using a tool known as the semivariogram. Once a semivariogram model has been chosen, then by means of a method known as kriging, the average amount of oil in the area is predicted. If circumstances exist where kriging is not feasible, then alternative methods of prediction such as random kriging or regularization are used. If the intention is to study the dispersion characteristics of the data, then a method known as conditional simulations is used. Throughout this thesis, some of these methods are slightly modified in order to allow greater freedom in fitting the models to the data. As well, some of these methods are applied to real world examples.