3

QGIS 2.14 (Win) comes with SAGA 2.12, which Ordinary Kriging supports 6 variogram model options.

SAGA 2.12

However, if I run SAGA Ordinary Kriging from QGIS Processing Toolbox, there is no drop down menu. Only an entry box in which an equation a + bx is shown.

QGIS Processing SAGA Ordinary Kriging

QGIS 2.14 user guide Kriging explains these options, so I can expect them coming. (future...not now).

Do you know any documentation about how to set variogram models in SAGA on QGIS Processing Toolbox? (Ideally, is there any working example?)

At the moment I am playing with Meuse river dataset, so a reference to Spherical model is desirable. (It is not important for now, though.)

enter image description here

So far I tried to manually put equation in the box (like above), which gave me a Kriging-like map. But I have zero confidence in what I am doing.


I installed SAGA 3.0 on QGIS 2.18 Processing, which was the same.

  • In order to perform a good Kriging Model you will have to make a variogram that matches your dataset (or which comes closest to it). Without actually visualizing your model you will have a hard time picking the right model from a drop-down menu. I therefor prefer using Arcgis or even software like Surfer to first make the variogram and than apply the used setting/parameters to make your Kriging. – Konan Pruiksma Jan 17 '17 at 12:55
  • @KonanPruiksma Ugh, your comment reminded me I have forgotten to mention that the parameters came from {gstat} package in R. Sorry, and thanks pointing out that. I cannot agree with you more about importance of getting right model. My question, however, was not about those parameters, but how I can put them in the User Interface in the Processing Toolbox. Just "a+b*x" does not give me much clue. – Kazuhito Jan 17 '17 at 13:11
  • Aah sorry I did not understand your question. After looking at the Kriging guide from QGIS you provided the parameters you put for the Variogram Model within the SAGA toolbox should not be an equation but a number corresponding to the model you wish to apply. In the guide it says this: 0 — [0] Spherical Model 1 — [1] Exponential Model 2 — [2] Gaussian Model 3 — [3] Linear Regression 4 — [4] Exponential Regression 5 — [5] Power Function Regression – Konan Pruiksma Jan 17 '17 at 14:02
  • @KonanPruiksma Thanks checking out the QGIS 2.14 documentation. If I put such numbers, say [0] in the box, what I get is a grid file with cells of all "nan". I can be totally wrong, but I think I need to somehow provide variogram model and parameters altogether in an equation style. I feel (hopefully) it will be equipped with drop-down menus sometime in the future. – Kazuhito Jan 17 '17 at 14:18
3

Finally, I found this template in SAGA GUI (2.0.1), in the interactive mode of Ordinary Kriging (global).

equation template

So a Spherical model can be like: new semivariogram model

And I got this; result

[Note] If anybody kindly volunteer to test this workflow, this sample point data and parameters are reproduced by R commands:

library(sp)
data(meuse, package= "sp")
coordinates(meuse)= ~x + y
proj4string(meuse) <- CRS("+init=epsg:28992")

library(rgdal)
writeOGR(meuse, ".", "meuse_points", driver = "ESRI Shapefile")

library(gstat)
vgm1 <- variogram(log(zinc)~1, meuse)
fit.variogram(vgm1, vgm("Sph"))       # Nugget= 0.214, Range= 514.4
plot(vgm1)                            # Sill ~ 0.85 by visual
  • I am new to understanding krigging, Can you please tell me how to find this parameters like above 0.214 , is this values changes from set points (region to region) – sudhanshud Sep 29 '17 at 6:19
  • Hi @sudhanshud In R {gstat} package, fit.variogram() function gives us Nugget, and Range for designated variogram model. I needed to plot it to find Sill (Psill). For some reason it returned different values now, so please do not trust the above values (0.214, 514.4).It may be my stupid mistake. – Kazuhito Sep 29 '17 at 9:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.