Kriging resulting in a raster with multiple NoData cells

[PatrãoPedro]

Hello to all,

I???m trying to generate a kriging from a set of points (+28000) I run the kriging accepting all the default values from the system. The resulting raster comes out with quite allot of NoData cells. The majority of these cells have points inside so I can???t find an explanation to the NoData cells. Does anyone faced a similar problem?
I have exported a smaller portion of the original data and run a second kriging and the NoData cells are still in the same position slightly different shape due to the default cell size changed.
Also, I used the pixel inspector to see the value for those cells which is 255  (I think is the default for NoData).

Last question is it possible to run a kriging using polygons as barriers?

Thanks in advance
Patrao

[SteveLynch]

Patrao

Could you attach the smaller subset of the data and the field that you are interpolating on.

Thanks
Steve Lynch

[PatrãoPedro]

Here it is the Subset.
The field used for the kriging is the H_Ping

Cheers
Patrao

[PatrãoPedro]

Forgot to attach the file.
here it is now

[SteveLynch]

The dataset that you provided has coincident data (151 points). The default model that Spatial Analyst fits cannot make predictions in the areas where you see the Nodata values.

Some solutions are;
- get rid of the coincident points (use Dissolve and then predictions are made everywhere in your area)
- model the semivariogram and better understand your input data. If you have Geostatistical Analyst then this will make it much easier.

Steve

[PatrãoPedro]

Thank you for such a quick reply, I cannot dissolve the points as each point as important information that should be used. In alternative I will write a small code in R to jitter all the coordinates to ensure there are no two equal pairs of coordinates.

Once again thank you for your efforts
Patrao

[SteveLynch]

GA does not fail with this data, it correctly uses coincidint data. Spatial Analyst's default kriging model does not produce output in certain areas. If the kriging assumptions are met then output will be produced everywhere.

The data, in it's current form, is not suitable for kriging.

Please look at the histogram of the input data.

Steve Lynch

[SteveLynch]

Bill

My point was that SA has a problem with coincident points and not GA. We will look into that.

In order to do kringing in SA one needs to do the variography manually, which most people do not do, they accept the defaults.

In GA the variography is interactive via the Wizard. However, we have found that most people accept the default values. We have therefore, in version 10, introduced an "optimize" button which produces good defaults through cross-validation.

Bottom line is, analyse the data before using kriging and do not use kriging because you've heard that it is a good interpolator :-).

Steve