4

I need to convert about 100 different sets of points into percentage volume contours. These are currently in vector format.

What would be the most efficient way to do this? I'd rather do it in ArcGIS but free tools / software is also an option, provided they are reasonably easy to use (I'm not too familiar with coding / prefer user-friendly interfaces).

I'm hoping there's something out there I've missed - so far my only option is to do this in GME after converting the points to raster within ARC. Even with batch conversion it will take me a while to insert each of the files in, one by one (plus I don't fully understand what some of the conversion options actually involve).

As a side problem, I also have in-group overlapping points (points in the same location) which I'd prefer to include in the PVC calculation as they are.

Any help will be much appreciated -

Thanks

EDIT: Do see your point about explaining a little more. Basically I'm looking at naming conventions for different places. For each name, there are a bunch of points which correspond to where people perceive this place to be. Ideally: assuming this follows a normal distribution in space, I'd like to do kernel density estimates of these (gaussian) and then do contour lines which reflected the % distribution (so 95%, 50% and a couple of others maybe). It's exploratory, I can't code, but I'm looking to see if I can do some contours of a bunch of different places and where people think these places are - a sort of vernacular gazetteer. I have at least 100 places though, & each of these should have their own set of contours. Hawth's KDE tool seemed perfect for this as it did both gaussian KDE and PVC's at the same time in batches. Unfortunately it won't work in ArcGIS 10.1 and I can't afford the days it would take me to figure out GME at the moment (would like to though). So I'm a little stuck & need an efficient way to do this.

  • 1
    Could you please explain what a "percentage volume contour" of a set of points is? – whuber Sep 5 '13 at 16:39
  • Yes - basically a contour line (or series of lines) that hold a user-defined percentage of points that belong to a group within that space. – Joules Sep 5 '13 at 18:09
  • Thanks. There has to be more to it than that, because these lines--and even the points they comprise--are not uniquely defined. For instance, you could sweep across the set of points in some random direction and pick up the first 10% of the points, the next 10%, and so on, generating contours that are perfectly parallel lines. Maybe if you explained why you are doing this we could identify additional criteria that would produce solutions that work well for your problem. I am wondering whether your question is similar to stats.stackexchange.com/questions/63447. – whuber Sep 5 '13 at 18:13
  • Yes - I think he's pretty much looking at doing something very similar to me. As I mentioned though, I'm rubbish at coding. Also my end result is exploratory - I'm not looking to predict where further points might be placed but looking to come up with a decent exploratory set of % contours. I've edited the above a little further. – Joules Sep 5 '13 at 22:02
2

I've researched this problem for a while and to my knowledge it cannot be easily done with tools at Your disposal. Although GME is not as difficult as You think. Give it a try.

But for me the perfect option is R. There are many packages that can generate KDE (ks for example) and You can get PVC for any value You want with wild1 package. Then it is a simple question of looping through all the datasets and waiting for the result - this is computationally intensive.

1

If all 100 sets of points have the same attributes/data type structure, you could bring all 100 to ArcMap select all and copy everything in one shot to a new point feature class. Then use spatial analyst tools to create a raster and contours.

  • Yes - they originate from the same file. I split them up when I discovered I couldn't batch process them with the arc kernel tool unless they were. I can't turn them into rasters in arc as there are instances where there's more than one points in the same location and I'd lose over a third of my data. I could try offsetting them. Would I be able to do the PVC calculation on the contours? I don't know if there's any way of doing this. – Joules Sep 5 '13 at 15:53
  • Also I need to keep all associated data for each of the points and the points. Sorry about all the questions - hawth's batch kernel was perfect : ( – Joules Sep 5 '13 at 16:05

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.