Dear Rok,

the determination of a search radius for point, line or kernel density estimation depends always on your research question and the feature distribution. If your example with caves has an archaeological background the walking distance (maximum per day) would be a possibility.

Another possibility would be (as you suggested yourself) to use spatial statistics. This can be for example Incremental Spatial Autocorrelation or you could also use a factor on the standard deviation on the feature distance.

Best regards,

Felix

Felix, thanks for your reply. What did you mean by calculating standard deviation on the feature distance?

Rok

Hi Rok,

The idea is to calculate the standard distance for your points or the standard deviation of the distance to the nearest point and use for example with a factor of two for your search radius.

If you use Incremental Spatial Autocorrelation just take a distance value where the Z-value peaks within the diagram.

Both methods should provide good estimates for a search radius.

best, Felix

Ok, thanks for your suggestions. I found Incremental Spatial Autocorrelation in ArcGIS tools. I'll try it.

cheers!

Rok

Hello Rok,

Please go through this,

Density Analysis

There are two basic approaches to exploring the density of values in GIS. You map simple calculations of the value of a given attribute by area, or you can create density surfaces for points or line vector data using the Spatial Analyst extension. Note that to use the Spatial Analyst extension, you must first enable it. To do this in ArcMap, click on the Tools menu, go to Extensions, and check mark Spatial Analyst. You can then load the Spatial Analyst toolbar (View – Toolbars) or open Arc Tool box– Spatial Analyst and access the tools from there.

Note: if you need to create a new field for area or length, you can open the attribute table (polygons or lines only), add a new field (type=double),

right-click on the field name, and use the Calculate Geometry function to calculate area or length in the units of your choice. If you want a permanent density field for further analysis, create a new field and use the Field Calculate or to add the results of a calculation dividing your value by area.

Thank your for answer, Suresh.

I tried Incremental spatial autocorrelation, but realised at the very beginning that I have a point data that represents presence (location of caves). The only goal is to draw a density map and I do not have any field data. For inc.spat.autocorrelation I should have, for example, temperature or soil moisture for point data. Then it would be possible to use these method.

There is an option to aggregate data but there is question again how big aggregation area should be...

My research area is 500 km2, and most of the data layers have 12,5 - 30 m resolution...

Thanks for any further thoughts...

Rok

Dear Rok,

Incremental spatial autocorrelation works only if your input data is somehow classified to multiple values. With just one value the method won't work.

If you have just one type of caves go with the std. dev. and the nearest point tool. You will at least get a good estimation for your search distance. There is not one correct value for your search distance. It always depends on your research question.

Wish you the best,

Felix

Yes, I agree. It depends on a research topic. SInce I try to find differences according to bedrock, I'll make many densities of rivers, caves and other phenomena with different radius and then I'll analyse correlation between each one of the density map with bedrock map. Then it might be seen which radius works the best...

Rok