[M-GI30] Near Surface Investigation and Modeling for Groundwater Resources Assessment and Conservation

[MGI30-P01] Muliple-point geostastics for Simulation of Lithological Classification

*TING-AN LIN¹、Hwa-Lung Yu¹ (1.National Taiwan University )

キーワード：Multiple-point geostatistics, Lithological Classification, Training Image

In the applications and studies of undergroudwater, it is important to understand the geological lithological composition in the study area. In order to find out the lithological distribution in the study area, many geological spatial statistical methods used to analyze the lithological composition on unknown points. One of the shortcomings in the traditional geostatistical methods like two-point based method(e.g., Kriging) is that they based on variogram, therefore, not able to handle complex and heterogeneous spatial structures. Additionally, their results are too smooth. Multiple-point geostatistics is a general statistical framework to model spatial fields with complex structures. The goal of Multiple-point geostatistics is to overcome the limitations of the variogram. It uses training image(TI) instead of variogram to estimate the conditional probability at interpolation location by the observed data and the already interpolated data. Take advantage of TI helps extracting spatial structure information and precisely describing more complex structures. This study focuses on Choshui river alluvial fan, using multiple-point geostatistics method to do simulation of lithological classification.