View image
The Kriging/Collocation gridding method assigns grid node values by weighting neighboring data points based on a covariance model. There are four covariance models available in WinPICS:
Below is an idealized covariance model comparing the four models. The covariance model describes the shape of the weighting-to-distance curves. The curves are calculated using a correlation length of 200 map units. The yellow dot is the half-correlation length, which is 100 map units.
All of the models begin by giving the greatest weight
to the nearest input data values. The models differ in how they weight
data points at a distance. For example, the steeply dipping Gaussian curve
gives little weight to values found at a distance of 150 map units or
greater. Conversely, the Exponential curve gives less weight to close
values (less than 100 map units) than the Gaussian curve, but gives more
weight to values at greater distances (more that 100 map units). View image
The Second-Order Markov Model with a half-correlation length of 1000 and a minimum variance of measurements of 1 is the default model. The default parameters will produce good results in most cases, but can be changed by clicking <More Options> in the Kernel Gridding window (see Kriging/Collocation Parameters for details) to suit your gridding needs.
Kriging/Collocation tends to connect values into ridges instead of isolated closures like ‘bull’s eyes’. It generally produces good results for sparse or irregularly spaced data sets
What do you want to do?
