[SOLVED] Variogram Map

Here you can report the problems encountered when using a function relative to the variography:
- experimental variography (vario.calc, vario.grid, vmap.calc, ...)
- model fitting (model.auto)

[SOLVED] Variogram Map

Postby Celeste » Mon Nov 12, 2012 10:51 pm

I have 3D data (X, Y, Z, Au) -> Easting, Northing, Elevation and Gold Grade; and I am trying to build a variogram map. I find that when I lower the number of turning bands and nx, ny then it works but it doesn't look good. I tried using your suggestion with nx=c(200, 200, 200) for the grid, then 1000 turning bands for the simulation and then nx = 30, ny=30 to build the map but it runs forever. Any tips and tricks that may help me?
Celeste
 
Posts: 20
Joined: Sun Oct 14, 2012 8:10 pm

Re: Variogram Map

Postby Didier Renard » Tue Nov 13, 2012 9:53 pm

I have tried to perform the same type of operation as the one you did ... except that I reduced the grid to 100x100x100 (instead of 200x200x200),
because of the limited capacity of my personal desktop (running LINUX).
So I created a 1000000 cell grid (8 times smaller than yours) called 'a', by typing:
a <- db.create(flag.grid=T,nx=c(100,100,100))

and I performed a non-conditional simulation of a structure using turning bands method with 1000 bands:
a <- simtub(,a,model=model,nbtuba=1000)
This step lasts a dozen of minutes (on my old labtop)

Finally I run a 2-D variogram map, benefiting from the grid organization of the data and using the Fast Fourier calculation:
dbmap <- vmap.grid(a,nx=30,ny=30)
This is almost immediate.
But one can argue that we only extracted a 2-D variographic information from a whole 3-D cube of data.

But we can also create the 3-D vmap grid beforehand, say:
dbmap <- db.create(flag.grid=T,nx=c(30,30,30),x0=c(-30,-30,-30))

and calculate the 3-D variogram map by typing:
dbmap <- vmap.grid(a,dbmap)
This hardly takes any longer than for the 2-D case.
Didier Renard
 
Posts: 337
Joined: Thu Sep 20, 2012 4:22 pm

Re: [SOLVED] Variogram Map

Postby Celeste » Wed Nov 21, 2012 9:43 pm

Will it be slower with the more data that is used or is it just the grid that takes so long? I will give it a try. Does it matter how big (the grid) is whether 100, 100 etc.?
Celeste
 
Posts: 20
Joined: Sun Oct 14, 2012 8:10 pm

Re: Variogram Map

Postby Didier Renard » Thu Nov 22, 2012 11:20 pm

The variogram Map calculation needs data (whicch can be defined on a first grid that we will call the Data Grid). The variogram map provides the experimental structural quantity (i.e.; the variogram) in a regular grid (called the Vmap grid).
Of course the calculation time depends on the dimension of these two grids.
The Data Grid is not really a parameter as it corresponds to the data set.
The Vmap grid is chosen by the user. Usually, it does not make sense to calculate the variogram map at too large distances, so the extension of the Vmap grid should be rather limited. In any case, it would not make sense to use a Vmap grid extension larger than half of the Data Grid.
In terms of computation time, the FFT approach saves a lot of time: the time consumption varies proportionally to the N*log(N) where N stands as the number of nodes of the Vmap grid.
Didier Renard
 
Posts: 337
Joined: Thu Sep 20, 2012 4:22 pm

Re: [SOLVED] Variogram Map

Postby Celeste » Mon Nov 26, 2012 7:52 pm

For some reason it still won't run (or is taking too long to run). Let me try again tonight. Maybe I am doing something wrong. I will let you know. Thanks for the help.
Celeste
 
Posts: 20
Joined: Sun Oct 14, 2012 8:10 pm

Re: [SOLVED] Variogram Map

Postby Celeste » Tue Dec 04, 2012 4:02 am

Why did you call everything "a" and not use different names for everything? Does it add everything together?
Celeste
 
Posts: 20
Joined: Sun Oct 14, 2012 8:10 pm

Re: [SOLVED] Variogram Map

Postby Didier Renard » Mon Dec 10, 2012 8:12 am

I don't uinderstand your last post. Could you be more specific on "where do I use 'a' systematically ? please.
Didier Renard
 
Posts: 337
Joined: Thu Sep 20, 2012 4:22 pm

Re: [SOLVED] Variogram Map

Postby Celeste » Mon Dec 10, 2012 9:57 pm

Here's where you used a as an identifier:

a <- db.create(flag.grid=T,nx=c(100,100,100))
a <- simtub(,a,model=model,nbtuba=1000)
dbmap <- vmap.grid(a,nx=30,ny=30)
dbmap <- vmap.grid(a,dbmap)

Does it add everything into "a".
Also can you post how you plotted your dbmap? I may be doing it wrong and I think that's why mine isn't working since everything else runs same as yours, in a dozen of minutes.
Celeste
 
Posts: 20
Joined: Sun Oct 14, 2012 8:10 pm

Re: [SOLVED] Variogram Map

Postby Didier Renard » Thu Dec 13, 2012 11:32 am

Hi Celeste

Now I understand your question correctly.
As a matter of fact, my example was slightly too short as far as the explanations are concerned.
I recall my example first:

a <- db.create(flag.grid=T,nx=c(100,100,100))
a <- simtub(,a,model=model,nbtuba=1000)
dbmap <- vmap.grid(a,nx=30,ny=30)
dbmap <- vmap.grid(a,dbmap)

In the first line, I created a NEW (empty) grid Db (called "a") specifying its geometry (number of nodes, grid mesh, ...)

In the second line, I performed a Simulation using the Turning Bands method of a Random Function obeying to the variogram model (contained in the file "model").
This simulation is non-conditional (non conditionned by any data) as I simply wanted to reproduce the texture of the Random Function. The only parameter
that was specified is the number of bands used by the algorithm (argument "nbtuba"). The result of this simulation is atcked within the Grid Db (called "a").
Note that if you run this procedure several times, the different simulation outcomes will be stacked in the same Grid Db.

In the third line, I wanted to compute the variogram map using the fast FFT algorithm. The VMAP result is a regular grid.
I decided to create the VMAP grid (called "dbmap") specifying its geometry (nx=30 and ny=30). The VMAP grid must be "compatible"
with the grid containing the variable from which we wish to calculate the variogram map. This means that they must share:
- the space dimension
- the mesh of the grid in all directions
Only the count of meshes can be different.

Finally in the fourth line, I computed the VMAP of the variable contained in the Grid Db (called "a" and used as the first argument)
and stored it in the VMAP Db (called "dbmap"). Note that, as in many modules of RGeoS, there is no need to specify the variable (within the Grid Db "a")
on which the calculations should be performed: they are ALWAYS performed on the variable which is asigned the locator "z1".

Hope this will help.
Didier Renard
 
Posts: 337
Joined: Thu Sep 20, 2012 4:22 pm

Re: [SOLVED] Variogram Map

Postby Celeste » Fri Dec 14, 2012 4:43 am

And then what do you use to plot it, so that a screen comes up showing the final result? Is it plot0 or plot?
Celeste
 
Posts: 20
Joined: Sun Oct 14, 2012 8:10 pm

Next

Return to Variography

Who is online

Users browsing this forum: No registered users and 1 guest

cron