Kriging Versus Simulation
Learning objectives
- Contrast the goals of kriging and simulation
- Explain why the kriged map is too smooth for flow
- Explain what a simulation reproduces
- State when to use each
Two Different Goals
Kriging and simulation start from the same wells and variogram but answer different questions. Kriging asks for the single best estimate and returns a smooth map. Sequential Gaussian simulation asks for a realistic rock and reproduces the full variogram and histogram, adding back the variability kriging averaged away.
Why Smoothness Is Wrong for Flow
The smooth kriged map underestimates the extremes: it has no high-permeability streaks and no tight barriers, only gentle gradients. But flow is controlled by exactly those extremes. Running a simulator on a kriged map gives the wrong dynamics. A simulated realization, rough and full of connected highs and lows, behaves like real rock.
One Map or Many
The price of realism is non-uniqueness: every seed gives a different, equally valid realization, so we run many. Kriging is right when you want one best estimate of a smoothly varying quantity; simulation is right when heterogeneity matters or you need to quantify uncertainty, which is almost always the case for flow modeling.