Fitting a Variogram to Data

From Dots to a Model

The experimental variogram is a handful of noisy points, one per lag bin. Kriging and simulation need a smooth, valid model that can be evaluated at any lag, so the final step of variogram analysis is to fit a model: choosing the shape and the nugget, sill, and range that best match the points. This widget puts you in the modeling seat.

Weight the Reliable Lags

Not all experimental points deserve equal trust. The short lags are computed from many sample pairs and are reliable; the long lags rest on few pairs and are noisy. A good fit, and the weighted score in the widget, weights each point by its pair count, so the model is pinned to the trustworthy short-lag behavior and is not dragged around by a single wild far-lag point.

Why a Model, Not the Points

A fitted model is not just smoothing for its own sake. Kriging requires a valid variogram, one that yields a stable, solvable system at every lag, and the raw experimental points do not guarantee that. The model also lets you impose geological judgment, setting the range from known body sizes or the nugget from repeat-measurement scatter, even where the data are thin. Press auto-fit to see the least-squares best, then judge whether it matches what you know of the rock.