Data transformations - parameter settings

Data transformation consists of several mathematical operations that transform (internally and temporarily) the upscaled property before you use it in the next steps of the Rock Property Modeling workflow. Data transformation is necessary for a number of reasons. Firstly, the simulation algorithm applied in geostatistical simulation methods expects data with a standard normal distribution. Typically, measurements of the subsurface do not follow a standard normal distribution or their distribution of values is skewed and, if not corrected, seriously impairs the modeling result. Secondly, the modeling methods require stationary data (the mean and standard deviation should not change in any spatial direction). Data transformation enables you to temporarily ‘adapt’ the input data so that the modeling methods can be applied in a meaningful way. After the modeling has been run, the data is automatically back transformed in the reverse order of the steps during (forward) transformation.

The objectives of data transformation are:

• To remove spatial trends so that the data is stationary.
• To transform the data into standard normal distribution N(0,1) when modeling with SGS.

Rules when applying data transformation

• Data transformation is strongly recommended when working with upscaled continuous logs which show a clear indication of non-stationarity or non-normally distributed data.
• To apply more than one data transformation, think about the order in which the transformations are carried out. For the transformation, the order of operations is used as specified in the transformation scenario and the reverse order for the back transformation. A different transformation order leads to a different result. Some general guidelines:
  • Truncation (if used) should be the first transformation in the scenario.
  • Normal score transformation (if used) should be the last transformation in the scenario.
  • Transformations to reduce skewness (other than normal score), such as logarithmic transformation for permeability, should be in the transformation scenario before any other spatial 1D trend transformation.