Comparison continues...

Kriging is based on the concept that nearby data is more relevant than data that is further away. A semivariogram cloud was used to confine the data control points. The farthest distance between two points was 941,138m. This distance was used to calculate the appropriate lag size. The known points are weighted more heavily that the estimated points. I tried each of the mathematical models to see which one would produce the lowest RMS. The best RMS was achieved with the pentaspherical model: 5.52. The rest of the process was the same as for the Part 1 map including clipping and creating isolines.

Comparing the Two Maps

The two maps look somewhat different, compare the range of rainfall as shown in the legends. The Inverse Distance Weighted map has a high rainfall of 42.3188. Whereas, the Kriging map has a high of 36.1078. (These values should be rounded to the nearest tenth).

Evaluation Criteria

Honoring the Control Point Data

Both methods should ‘honor the control point data,’ however, because the Kriging method weights the known control points higher than estimated points, Kriging is better at honoring the control point data. Although it is a standard ccriteria, if I were to name this evaluation criteria, I would call it “protection of control point data” or “protecting” rather than “honoring” which to me, erroneously personifies the data.

Correct at Non-control Points

Both methods use cross-validation to increase the accuracy of the estimated data at non-control points, however Kriging uses a more sophisticated method. Kriging is based on the concept that nearby data is more relevant than data that is further away and therefore nearby, known data points are weighted more heavily. A semivariogram cloud was also used to confine the data control points.

Ability to Handle Discontinuities

An example of a discontinuity is an unusual geologic feature compared with the surrounding area. The Kriging method is able to account for those features whereas the Inverse Distance Weighted Method is not. Since the Inverse Distance Weighted Method uses a grid, it is not able to capture those unusual features. The Kriging method does not use a grid.

Execution Time

Simple triangulation is the fastest method for larger data sets. Inverse Distance Weighting and Kriging are then better for smaller data sets, and where a smaller geographic area is shown, and thus a need for a more accurate map.