Amazonas – Stochastic Method of Modeling Geological Systems with Arbitrary Distribution of Properties, Including Statistically Unsteady Ones, Based on Non-Parametric Statistics
Event name: Geomodel 2019
Session: Математические методы обработки геологической и геофизической информации
Publication date: 17 September 2019
Info: Extended abstract, PDF ( 700.4Kb )
Price: € 20
The paper considers the method Amazonas developed by the author, which is based on the use of non-parametric (rank) statistics in a sliding window of a given form. The use of nonparametric (rank) statistics, which makes it possible to model the spatial distributions of both discrete and continuous properties with arbitrary distributions of values, including multimodal, reproduce sharp transitions of properties without the need to build a facies model, provides resistance to clustering, incompleteness and the presence of individual error values in the source data. The ability to set an offset allows to reproduce both smooth and sharp transitions between the values, with a varying degree of variability to extrapolate the possible behavior of the system outside the studied area. The applied mathematical formalism makes it possible to obtain a realistic stochastic model of the simulated environment, directly determined by the initial data, taking into account the heterogeneity existing in it, which can be associated with lithology and other factors, without the need for preliminary expert identification of these laws. The method is suitable for modeling both stationary and non-stationary distributed spatial variables, and allows to take into account the instrumental error and the natural variation of the simulated parameters.