topic Re: Interpolation with Barrier in ArcGIS GeoStatistical Analyst Questions

Interpolation with Barrier

yanma — Wed, 12 Feb 2014 15:35:32 GMT

[ATTACH=CONFIG]31354[/ATTACH]

So as shown in pic I have a bunch of stations with water sample data and polygons representing the water in harbor.

The objective is to interpolate data over the water body.

To best simulate peninsula's barrier effect, I understand that I am better off with Diffusion Interpolation with Barriers and/or Kernel Interpolation with Barriers in Geostatistical Analyst Tool.

For those who are experienced in this kind of endeavor,
1. Should I aggregate the polygons first to connect the waters?
2. If I set the polygons as barriers, will it be considered islands instead of waters? With this concern, how should I best specify the parameters for the interpolation tool?

These are some really powerful tools in ArcGIS. I would love to engage in related discussions not limited to this forum.

Will check for updates frequently. Any input is appreciated!

my

Re: Interpolation with Barrier

yanma — Wed, 12 Feb 2014 17:39:26 GMT

update:

I tried diffusion interpolation with barriers on the data and it turned out as below:
[ATTACH=CONFIG]31364[/ATTACH]

There is large areas of blank which I do not understand..

Any suggestion?

my

Re: Interpolation with Barrier

SteveLynch — Thu, 13 Feb 2014 14:29:13 GMT

Trying increasing the bandwidth value.
I would rather suggest you first use KernelInterpolationWithBarriers, same concept, larger bandwidth brings in more data, however, it may then over smooth the data.

Trial and error and look at using the CrossValidation tool or it can all be done in the Geostatistical Wizard.

You might want to move this post to http://forums.arcgis.com/forums/100-Geostatistical-Analyst

-Steve

Re: Interpolation with Barrier

yanma — Thu, 13 Feb 2014 16:02:56 GMT

Trying increasing the bandwidth value.
I would rather suggest you first use KernelInterpolationWithBarriers, same concept, larger bandwidth brings in more data, however, it may then over smooth the data.

Hi Steve

thanks for replying.

So the larger the bandwidth, the more calculation will be done right?

Since it does not have a guidance value, how do I decide what order of magnitude to start with as the bandwidth?

my

p.s. to move this thread to other forum, is there a way other than posting it again?

Re: Interpolation with Barrier

SteveLynch — Thu, 13 Feb 2014 16:07:26 GMT

for initial default see Geostatistical Wizard or

[ATTACH=CONFIG]31410[/ATTACH]

Re: Interpolation with Barrier

MarcoBoeringa — Thu, 13 Feb 2014 16:29:10 GMT

p.s. to move this thread to other forum, is there a way other than posting it again?

Yes there is, I have moved it for you based on Steve's suggestion. Forums MVPs have basic moderation rights and can do this.

Re: Interpolation with Barrier

yanma — Thu, 13 Feb 2014 17:49:55 GMT

Yes there is, I have moved it for you based on Steve's suggestion. Forums MVPs have basic moderation rights and can do this.

Thanks!

mymymymymymymy

Re: Interpolation with Barrier

EricKrause — Thu, 13 Feb 2014 19:01:18 GMT

Hi Steve

thanks for replying.

So the larger the bandwidth, the more calculation will be done right?

Since it does not have a guidance value, how do I decide what order of magnitude to start with as the bandwidth?

my

p.s. to move this thread to other forum, is there a way other than posting it again?

The default bandwidth is calculated such that it minimizes the root-mean-square crossvalidation error. However, any location that does not have any points within the bandwidth distance will be calculated as NoData. As you increase the bandwidth, the areas of NoData will get smaller, but the quality of the predictions will decline. Keeping these two issues in mind, you need to decide on a bandwidth that works for your needs.

Re: Interpolation with Barrier

yanma — Thu, 13 Feb 2014 20:16:30 GMT

The default bandwidth is calculated such that it minimizes the root-mean-square crossvalidation error. However, any location that does not have any points within the bandwidth distance will be calculated as NoData. As you increase the bandwidth, the areas of NoData will get smaller, but the quality of the predictions will decline. Keeping these two issues in mind, you need to decide on a bandwidth that works for your needs.

Hi Eric

Thanks for clarifying. That is crystal clear explanation. Very helpful.

On a side note, I am still not very sure about difference between mask and processing extent. When I select extent as display, I get result as if the data points outside the screen are also accounted. In terms of display and calculation respectively, what kind of role does each play?

Thanks
my

Re: Interpolation with Barrier

EricKrause — Thu, 13 Feb 2014 20:21:45 GMT

Please read through the documentation on Mask and Extent. They do a good job of explaining the concepts. If you still have questions after reading the topics, I'm happy to answer them.

Mask:
http://resources.arcgis.com/en/help/main/10.2/index.html#//001w0000001t000000

Extent:
http://resources.arcgis.com/en/help/main/10.2/index.html#//001w00000009000000

Re: Interpolation with Barrier

yanma — Fri, 14 Feb 2014 17:08:20 GMT

Please read through the documentation on Mask and Extent. They do a good job of explaining the concepts. If you still have questions after reading the topics, I'm happy to answer them.

Mask:
http://resources.arcgis.com/en/help/main/10.2/index.html#//001w0000001t000000

Extent:
http://resources.arcgis.com/en/help/main/10.2/index.html#//001w00000009000000

correct me if I'm wrong.

ArcGIS selects those features that appear in the processing extent, process them, and generate a feature that not necessarily fit the extent. Then it clips the result to the shape of the mask.

(I think another way to put my question is that, in the process of input-process-output, if I were to insert Processing Extent and Mask to where they come into play, where should they be respectively? )

Thanks,
-my

Re: Interpolation with Barrier

EricKrause — Fri, 14 Feb 2014 17:41:11 GMT

Extent does not affect which data points are used in the model. Even points outside the Extent will be used in the calculations. Extent refers to the extent of the output raster or geostatistical layer. If you want to control which points are used to build the model, you can do that with Selection.

Mask simply sets some areas of the output to NoData. Specifically, anything outside the Mask and within the Extent will be set to NoData. Note that Mask is only supported for raster output; this environment does not work with geostatistical layer outputs.

Re: Interpolation with Barrier

yanma — Fri, 14 Feb 2014 18:22:43 GMT

Extent does not affect which data points are used in the model. Even points outside the Extent will be used in the calculations. Extent refers to the extent of the output raster or geostatistical layer. If you want to control which points are used to build the model, you can do that with Selection.

Mask simply sets some areas of the output to NoData. Specifically, anything outside the Mask and within the Extent will be set to NoData. Note that Mask is only supported for raster output; this environment does not work with geostatistical layer outputs.

However, in the link you shared, the first line goes:"Tools that honor the Output Extent environment will only process features or rasters that fall within the extent specified in this setting."

A little confusing.

my

Re: Interpolation with Barrier

EricKrause — Fri, 14 Feb 2014 18:28:33 GMT

Sorry for the confusion. That text is referring to the output raster or output features, not the inputs.