How can one (statistically) compare the feature values of one set of points against the feature values of a different set of points (within the same spatial extent)?

I am aware of methods of comparing the spatial distributions of spatial point patterns and for testing for spatial randomness, but I would like to find out in which ways I could statistically compare two sets of feature values for two sets of points within the same spatial context.

I have quickly done a rough and ready comparison of my points visually so I'm really interested in computing the significance and/or (dis)similarity between point feature values of different surveys of the same area.

Here's a little more background on my specific problem. I am working on a project to develop a method for monitoring subtidal seagrass. Part of this involves sonar surveys of seagrass beds using an echosounder. Each point pattern corresponds to a different survey of the same seagrass bed. The only difference between surveys is either the tidal state (low tide or high tide) or the speed of the survey vessel. The feature values of each point pattern corresponds to a metric of the seagrass bed e.g. bioheight.

Here's a plot of a survey to illustrate. Here's a plot of a survey to illustrate.

So I want to compare the similarity of feature values for the same metric for different surveys.

I mainly work in R but I am open to solutions using any GIS package.

  • 2
    Are the survey points a different set of locations each time? Or are the locations precisely enough established that they are essentially the same location being measures? Note also this isn't a "point-pattern analysis" but a geostatistical problem.
    – Spacedman
    Nov 10 '17 at 13:03
  • 2
    Like any statistics problem you need to construct a model for the thing you are asking a question about (total sea grass amount, regions of seagrass loss, etc etc) and then how that relates to the measurements you are taking and other covariates (time of year, weather, etc etc).
    – Spacedman
    Nov 10 '17 at 13:05
  • Does each survey capture the same information you are trying to know with varying results based on the variable survey conditions?
    – JMT2080AD
    Nov 10 '17 at 22:50
  • @Spacedman Yes, the points are always in the same locations - the values for each point actually represent an average for a 5m radius around the point.
    – Tom Newton
    Nov 13 '17 at 13:28
  • 2
    Please take heed of @Spacedman advice in that this is not a PPA problem, as he is an expert in the field. You really need to specify a spatial-temporal model. There are numerous routes that you can take here in spatial/geostatistics, temporal analysis or parametric statistics but, the first step is to dig into the literature. As to R, you may want to start with the spacetime and gstat packages. There is also a Taks View on this topic: cran.r-project.org/web/views/SpatioTemporal.html Nov 13 '17 at 17:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.