Multi-variate attributes analysis to highlight vsh trend as soft guide into facies modelling

Abstrak

The following paper presents a simple method in generating volume of shale (VSH) from seismic attributes. Multi- variate statistic methods were used to analyze correlation of seismic attributes and well-log properties to define VSH. VSH is one of rock properties that is considered to have direct link with rock facies. Regression technique was utilized after the establishment function of high correlation between seismic attribute and VSH by using multi-variate analysis was achieved. Intermediate process was introduced by looking up the highest correlation of seismic attributes with other well-log properties (e.g porosity, density, or water saturation) in the case where correlation value of multi- variate matrices is low. This intermediate process should be performed where other rock properties have a high correlation with VSH. Once the intermediate correlation process is established, the seismic attribute - VSH correlation can be properly completed.

In this study case, example of 2 layers-horizon so called direct method (SS-1 & SS-2) and 2 layers- horizon of indirect method (SS-3 & SS-4) is discussed. Result of multivariate analysis from mean envelope relative impedance and RMS relative impedance of the SS-1 and SS-2 zone could correlate directly with VSH, respectively. Whereas, the SS-3 and SS-4 zone indicates direct correlation of RMS relative impedance to NPHI and attribute of mean sweetness corelate to PHIE. Fortunately, these well-log properties of NPHI and PHIE have high correlation with VSH.

In order to established mathematical relationship between rock properties and seismic attributes selected from multivariate analysis, we generate cross plot of each layer and define sensible envelop of the data distribution as an approximation of low-base-high case scenario related to polynomial function which generated later. The proposed method offers fast yet quite robust solution to generate VSH trend as soft-guide into 3D facies modeling whenever data is limited.