I am performing Gaussian geostatistical simulation (conditional). But, the mean and minimum output rasters have some negative values. Some simulated maps have negative values. Does anybody have some idea why this is happening? Could it mean that my ajuste

764
2
01-05-2017 11:29 AM
CarineSilva
New Contributor

I am performing Gaussian geostatistical simulation (conditional). But, the mean and minimum output rasters have some negative values. Some simulated maps have negative values. Does anybody have some idea why this is happening? Could it mean that my ajusted model in the simple kriging is not good? Thanks.  

0 Kudos
2 Replies
DanPatterson_Retired
MVP Esteemed Contributor

Gaussian Geostatistical Simulations—Help | ArcGIS Desktop and

How Gaussian Geostatistical Simulations works—Help | ArcGIS Desktop 

makes no mention of equating negative values as being bad... perhaps you could comment on what you found by examining your results.

0 Kudos
EricKrause
Esri Regular Contributor

It is simulating from a Gaussian distribution, and negative values are valid simulations from this model.  The phenomenon that you are interpolating may not be able to be negative (like rainfall or pollution levels), but the model doesn't know this, and some of the simulations will likely contain negative values just due to chance.

I only know one way to guarantee that all simulated values will be greater than zero, but it requires changing your simple kriging geostatistical layer.  In the Geostatistical Wizard, make sure that you are using simple kriging, and use a Normal Score Transformation.  On the transformation page (with the histogram on the left), choose Lognormal, Gamma, or Log Empirical for the Base distribution (found on the right).  To use these options, all your data values must be greater than zero, but these options will guarantee that all predictions and simulations will also be greater than zero.