(Beginner) Geostastical Analyst - Memory Error

1723
5
06-03-2012 02:29 PM
AlanWilliams1
New Contributor
Hi All,

I am relatively new to geostastical analyst and am an archaeologist (generally with nothing more complicated than a shovel normally).

I am trying to undertake a kriging process of about 4,000 data points across Australia to explore human migration. However in the first instance the explore data>semi variogram simply crashes on me every time I use it, and the use of the geostatistical wizard comes up with an out of memory error message. I have reduced the dataset down to about 500 data points (whcih is way to low for my analysis unfortunately), and end up with the same results.

Any advice on where to go from here would be appreciated. Do I need a bigger computer?

Best wishes

Al
0 Kudos
5 Replies
boonejardot
New Contributor
What are your computer specs?  Did you run the system prerequisite check before installing?  Second thing. Kriging is very complex.  Be sure to talk with someone who has done similiar studies on migration with kriging because there are several different ways to krig the same points and get vastly different results. You might want to check other peer reviewed articles that used kriging effectively.  Just a suggestion. Be sure to experiment with several models, and pick one that matches the data that you have. 

If you computer has enough resources to run the program, and enough hard drive and memory, then it may be a simple fact of data preparation.  Is the data projected?  It needs to be if you are kriging((since kriging relies on the semivariogram and interpolation both which rely on correct distances)).  I cant remember the maximum points kriging allows, but I do know the more points the better so feel good about that.  There is a help button in the kriging tool, that will give you alot of information about kriging.

EDIT:: If it turns out that you have enough resources, you are not using too many points, and your data is projected try creating a new geodatabase to store the output of the kriging tool.  From my own experience, setting up the default database under environments saves alot of headaches in the longrun for sure.  Setting this up is easy, just click add folder connection, click the new database icon in the brosing window, It has a star on a cylinder, and put that database where you are able to store it.  Then go under Geoprocessing-environments ->>>>Select currentworkspace and browse to your new database.  Hope this helps.  Let me know.
0 Kudos
AlanWilliams1
New Contributor
Hi Boone,

Thanks for getting back to me so fast and thoroughly. I have a brand new laptop with 4GB of RAM, 283GB hard drive and a reasonable processor. And certainly arcgis appears to run fine on the computer for a range of bits and pieces (just not krigging!?). Not sure what the minimum requirements are, but my gut feel is I need something bigger.

Thanks for the warning - kriging is certainly beyond my comfort zone and I'll be reading everything I can get my hands on about the technique.

I'll give the geodatabase a try and get back to you, although I think I'll need to fix above first....

Thanks
Al
0 Kudos
boonejardot
New Contributor
If its a brandnew laptop, it should be more than enough to run the Program.  I ran arcmap on a singlecore with 1.5 gigs of ram which slowed it down quite a bit.  At my university we used 2-core processors which ran okay.  A new laptop with a multicore IE 2 , 3 , 4 cores should be more than enough.  If  you try setting the default database and it still doesnt work, I would try and contact some senior members on this forum, or contact techsupport if you have the ability.  Wish you luck.
0 Kudos
AlanWilliams1
New Contributor
Will do

Thanks again

A
0 Kudos
SteveLynch
Esri Regular Contributor
Al

These (explore data>semi variogram) are part of the ESDA tools. The point cloud will consist roughly of 4000*4000/2 points (8 million points). I'll be too dense to get any meaningful info from. You did see the warning message that came up when you tried to process the 4000 points?

I would suggest that you rather take a subset of your data (use the SubsetFeatures tool).

I would suggest no more that 500 points.

The 4000 points will be fine to use when doing kriging though.

Steve
0 Kudos