Inter-point distance calculations for forest inventory metrics

[GordonWhitmore]

I have recently experimented with the automated delineation of trees from high resolution imagery and also from Lidar.  Since I have also derived the approximate height of each point from the LiDAR derived height layer, I would now like to get the inter-tree distance between each tree and its nearest five neighbors.  My intent is to then use this as a tree specific measure of density/spacing to derive a more accurate estimate of each tree's diameter.  Since there are from 500 - 5000 stems per hectare and I am dealing with approximately 70,000 hectares, the processing may be significant.  While I have some ideas on how this might be accomplished, I don't have a super-computer at my disposal to complete all of the calculations.

If anyone has any ideas on an efficient process for making these calculations, your input would be greatly appreciated.

Gord.

[SepheFox]

HI gordon, I think you could achieve this using the generate near table too​l and specifying closest five. I am curious though, how you will derive the tree diameter from this metric.

[DanPatterson_Retired]

what tools do you have available?  If you have tree point locations, it would be interesting as a test to find out what each space a tree "takes up". 

This could be accomplished by

1. constructing a Voronoi diagram (aka Thiessen Polygon) which would locate each tree within its surrounding "space".  The per tree area and its variation and distribution would be useful metrics to explore.
2. Alternately, one could construct a Delauney triangulation which is the sibling to the Vor Dia.  In this situation, the space located between 3 trees (ergo a triangle) is delineated and the triangle metrics could also be examined.
3. Spatial allocation can also be done in raster world as well and are akin to those outlined in 1 and 2 above

options for 1,2 and 3 can easily be done in Arcmap if you have the appropriate license level and #3 requires the spatial analyst extension.

I have code to accomplish 1 and 2 using vector point data (shapefile inputs and VorDia, ThiPol as outputs) that requires no special license level or extension.  If you are interested I could post/attach it later when I get off this iThingy and onto my real computer.

Just several thoughts to get the thought processes going.

[BruceCarroll1]

Gordon,

I agree that finding the nearest 5 trees would be very compute intensive. Another idea might be to use a "moving window" type of analysis.  You could center the moving window on each tree, and get a count of how many trees are inside that window which would give you a measure of density/spacing. Alternatively you could use one of the neighborhood analysis tools in Spatial Analyst (which is quite efficient) to calculate density/spacing measures across the entire study area. Those density/spacing measures could then be imputed to individual trees.

That begs the question that Sephe asked, how will you use density/spacing to calculate diameters. Crown Area is very strongly correlated with diameter. But density/spacing and crown area may or may not be related due to canopy gaps, especially in younger forests. You may be better served finding a method to calculate crown area from your LiDAR data

[EarlKlug]

Has anyone been able to replicate the pit-free Canopy Height Model (CHM) proposed by Khosravipour et al. (2014).  I know CHM is talked about on another tool though haven't heard of it via ARCGIS.  Thanks...