0

I'm trying to figure out if buffers I have created in ArcMap (measuring percent forest cover in a 1 km distance from a land use map within the buffer) around points at which I will be collecting presence/absence data of mammals exhibit spatial autocorrelation. Basically I want to ensure each of my buffers of forest cover is a distinct unit of measurement even if they are close together in space. Each buffer measures percent forest cover as the explanatory variable for the response variable of mammal presence/absence gathered at the central point of the buffer. I want to make sure these variables of forest cover in each buffer are actually independent variables that can be used as explanatory variables for each individual buffer (and thereby each individual point where I will measure the response variable, mammal presence/absence) without having spatial autocorrelation. I'm a bit new to this so I'm not sure exactly how this should be done with Moran's I. For example, what the input field for this tool in ArcMap should be? Is there a way I could use the values for percent forest cover for this? I determined the amount of total land use in each buffer with the tabulate area tool, and then actually determined percent forest cover in excel, so I don't know if there's a way to merge the land use values from each buffer derived from the tabulate area tool to the corresponding buffer polygon file so that these can be included into the Moran's I tool? I'm assuming the input feature class should be all the buffers of interest merged together?

Example of the land use table after I used tabulate area within the buffer on the land use polygon within it to determine the amount of each land use type within the buffer area--it is in Spanish, but this shows Primary forest, secondary forest, pasture and bare ground

This is an example of one of my 1 km buffers with the central point being where I will collect mammal presence/absence data, the green is primary and secondary forest cover and the yellow is pasture all from the land use polygon that I use in the tabulate area tool to determine the amount of each land cover type in each buffer

  • 1
    'I want to ensure each of my buffers of forest cover is a distinct unit of measurement, even if they are close together in space'. I don't think Moran's I will tell you that. – Martan Mar 26 at 6:37
  • Can you provide a small example of your data, say 5 polygons along with the underlying raster? – Jeffrey Evans Mar 27 at 18:51
0

You are convolving what your experimental unit actually is. For information contained in "buffers" you are going to have within unit and between unit variation. You cannot collapse the within unit variability and quantify autocorrelation nor can you evaluate independence between units without addressing the underlying distribution(s).

For evaluating the independence of your experimental units, I would recommend addressing the distributional equivalency between all pair-wise experimental units using a statistic such as a Kolmogorov-Smirnov test or for nominal data, a Chi square. You can even assign pair-wise distances, or constrain by distance, to the units to explore the distance effect of the equivalency test. Autocorrelation is going to tell you nothing of value here and would not at all be a valid statistic.

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.