Ripley's k-function analysis

AprilNewlander1

I have some questions regarding my project that I am hoping to get answered to make sure my methods are valid.
Here is a summary of my project and methods:
I have data from LiDAR that I made into a Vegetation Height map as determined by LiDAR. I then made those raster cells that were classified as 1-4 m tall plants into a point shapefile to represent tall plants across the landscape to assess global clustering.
The study area is fairly large. (See attachment) I then did a unweighted analysis seperately for 'above' and 'below' the road, in which the observed values showed clustering across all spatial scales, and the expected line falling above the confidence envelope, which I would have expected to fall within the envelope. Why would this happen?
I was then recommended to run a weighted analysis and then use weighted results combined with the unweighted CI and adjust the unweighted expected line to zero. I used a weight of 1, as weight represents the number of coincident features at each feature location. Since these points essentially represent a 1x1 m area covered by 1 m tall vegetation it is more an index of cover, and does not necessarily represent just 1 plant. Is this an accurate way to use the weighted function? Will using a value of 1 be acceptable in this study? My results are very different than those obtained from the unweighted observed values.(see figures for unweighted results and weighted observed with unweighted CI and exp).
I have also read in ArcGIS help that the largest diffk represents the distance where spatial processes promoting clustering is most pronounced. Is this still the case when I use unweighted CI and expectation and the weighted observed values?

pattern analysis.jpg

below.png

above unweighted.png

above.png

below unweighted.png

Anonymous User

Original User: anewlander

I am very sorry to be the bearer of bad news, but you are violating the assumptions of the Ripley's-K. Lidar derived heights, distinctly, represent an intensity process. Because of this the null hypothesis of homogeneity following a Poisson process does not hold and the statistic is incorrect.

Jeffrey,
I appreciate your response.  Does this hold true even though I first classified the DEM into plant size classes and used those 1-4 m, as determined by Lidar, to create a point shapefile for those pixels.  I did not use actual height.  I have a binomial outcome for each pixel, 1 = >1m; 0 = < 1m.  Since all pixels were sampled, can't I assume the spatial sampling is homogeneous?  what do you mean by an 'intensity process'?
Thanks,
April