Indicator Methods

Indicator methods are a way to model what is happening in a rock formation between the wells where we have data. Instead of trying to guess the exact value of a property like porosity at every point, indicator methods convert the data into a series of yes-or-no questions ("is the porosity above 10 percent here, yes or no?") and then build a probability map for each question. The maps are stitched together into a full reservoir model. The technique handles complicated, lopsided, or two-peaked distributions of rock properties that simpler methods cannot describe well.

Key Takeaways

Indicator methods are nonparametric, which is a fancy way of saying they do not assume the property values follow a tidy bell-curve distribution. Real reservoir properties often do not.
The core trick is the indicator transform: pick a threshold (say, 10 percent porosity) and convert every measured value into a 1 (above the threshold) or a 0 (below). Repeat for several thresholds. The result is a stack of yes-or-no maps that together describe the full distribution.
Each threshold needs its own indicator variogram, the math that describes how spatial similarity changes with distance for that yes-or-no question. Modeling K thresholds means modeling K variograms, which is more work than the single variogram needed for Gaussian methods.
Indicator methods are the right choice when the property has a bimodal or multimodal distribution (vuggy carbonates with both tight matrix and big vugs), when the irreducible water saturation has a sharp threshold, or when facies bodies need to be simulated with discrete categories.
The biggest practical headache is order-relation correction. The independently estimated probabilities at multiple thresholds sometimes do not line up into a valid probability distribution at every grid cell, and the inconsistencies have to be fixed before the model can be used.

Fast Facts

Indicator geostatistics was built in the early 1980s at Stanford University by André Journel, building on the foundational kriging math developed by Georges Matheron at the École des Mines de Paris in the 1960s. The free GSLIB software library released by Deutsch and Journel in 1992 made the technique broadly available to the petroleum industry without requiring proprietary software, and most modern reservoir modeling tools (Petrel, RMS, JewelSuite) include indicator simulation as a standard option.

What Indicator Methods Do, Explained Simply

Imagine you have porosity measurements at three wells. Well A reads 22 percent. Well B reads 4 percent. Well C reads 23 percent. You want to fill in a map of the entire field between those wells, including a fourth location where you might drill next.

The traditional approach (sequential Gaussian simulation, the most common alternative) assumes the porosity in this field follows a bell curve, and it picks a value for the new location somewhere near the average. If the average is 14 percent, the simulation tends to predict porosity around 14 percent at the new location. The problem is that real reservoirs often do not have a bell-curve distribution. In a vuggy carbonate, the rock is either tight matrix at 4 percent porosity or vuggy dolomite at 22 percent. There is very little rock at 14 percent. The Gaussian model averages those two populations into a fictional middle that does not exist.

An indicator method handles this differently. Instead of asking "what is the porosity here," it asks a series of yes-or-no questions: "is the porosity above 5 percent here? Above 10? Above 15? Above 20?" Each question becomes a separate map. At a location like Well B, the answer to all four is "no." At Well A or C, the answer to all four is "yes." Between the wells, the maps capture the probability of each "yes" at each point. Stack the maps together, and you get a full porosity model that respects the actual two-peaked distribution rather than smoothing it into a misleading average.

Where Indicator Methods Earn Their Keep

Vuggy carbonates are the classic case. Reefal limestones in many Middle East fields, the Smackover dolomites of the US Gulf Coast, and the Devonian carbonates of the Western Canadian Sedimentary Basin all show bimodal porosity. Indicator methods preserve the two populations in the simulation. A reservoir engineer running flow models on a vuggy carbonate built by Gaussian simulation often finds the simulator predicts uniform moderate flow. Reality shows early water breakthrough through the high-porosity vuggy network. The flow model built on indicator-simulated porosity matches the observed behavior much better.

The other strength is facies modeling. When a reservoir contains discrete rock types (sand, shale, coal, carbonate), each can be its own indicator class. The simulation places facies bodies according to their indicator variograms, which capture the size and shape of the bodies in each direction. The output is a 3D model of facies that respects the geological architecture rather than smearing it into continuous gradients.

Indicator methods are also called indicator geostatistics, sequential indicator simulation (SIS), or indicator kriging. The categorical version used for facies modeling is sometimes just called facies simulation or indicator-based facies simulation. Related terms include sequential Gaussian simulation (SGS, the most common alternative for continuous properties; assumes a bell-curve distribution and is simpler to implement, but smears bimodal distributions into a fictional average), variogram (the spatial correlation model in geostatistics; indicator methods need a separate variogram at each threshold, which is more modeling work than Gaussian simulation requires), kriging (the linear estimation method underlying both Gaussian and indicator simulation; uses spatial correlation to weight nearby data points when predicting at an unsampled location), facies (a body of rock with a particular combination of lithology, texture, and depositional setting; indicator methods are commonly used to simulate the spatial arrangement of facies in a reservoir model), and multi-point geostatistics (MPS, a newer approach that uses training images to capture complex geological patterns indicator simulation cannot represent, such as sinuous fluvial channels).

Why the Two-Peaked Histogram Tells You to Switch Methods

A reservoir engineer working on a Middle East carbonate field plots the porosity histogram from 280 core plugs. The histogram has two peaks: one centred around 4 percent (tight limestone matrix) and another centred around 22 percent (vuggy dolomite). Almost no plugs fall in the middle.

The default reservoir modelling workflow uses sequential Gaussian simulation. The engineer runs it. The output model has porosity averaging around 13 percent across most of the field, with smooth gradients connecting the wells. The flow simulation built on this model predicts uniform sweep efficiency and water breakthrough at year 12. The actual field saw water breakthrough at year 4 in three of the seven producers, with the others still producing relatively dry oil at year 9.

The engineer rebuilds the model using sequential indicator simulation with a threshold at 10 percent porosity. The vuggy network shows up as discrete connected bodies in the model, with the tight matrix filling the spaces between. The new flow simulation predicts early water breakthrough through the vuggy zones, late water in the matrix-dominated zones. It matches the field's actual production behaviour within a year on every well. The two-peaked histogram was the signal that the wrong tool was being used. Indicator methods saw what the Gaussian model could not.