Landscape metrics from rasters

168 views
Skip to first unread message

James Doyle

unread,
May 9, 2012, 2:54:50 PM5/9/12
to Maxent
Hi everyone,

Has anyone tried using landscape metrics as predictor variables in
Maxent? I am trying to use some of the metrics for land cover that
fragstats has but the output from fragstats is not conducive to use in
Maxent. Specifically, I am trying to generate a continuous map of
distance from any given cell to a different classification or distance
to a cell of specific classifcation (e.g. distance to edge; distance
to forest or distance to crops); and edge length/density or shape
indices within a neighborhood (e.g. 10 cell x 10 cell window). I was
wondering if anybody had some suggestions on ways to extract these
types of metrics from my land cover raster within ArcGIS so I have a
continuous map to use as a background layer, rather than doing it
within fragstats and having several disconnected landscapes as the
background (which I think will preclude generating the output map?).

Thanks in advance,

Jimmy

John Clark

unread,
May 10, 2012, 11:41:23 AM5/10/12
to max...@googlegroups.com
Howdy,

Not sure this is what you had in mind, but for my analysis I extracted 'developed' from NLCD to create a mask of developed areas, then ran a distance to produce a continuous surface.  For edge complexity you might do standard deviation within a moving window.  Haven't done this, just an idea, but if you really want to look at edge density you could vectorize your patches, convert the boundaries back to raster, and run a neighborhood analysis (..10x10) to count all edge pixels.

Best Regards,

John Clark



--
You received this message because you are subscribed to the Google Groups "Maxent" group.
To post to this group, send email to max...@googlegroups.com.
To unsubscribe from this group, send email to maxent+un...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/maxent?hl=en.


James Doyle

unread,
May 16, 2012, 8:42:47 AM5/16/12
to Maxent
Hi John,

Thanks for the input. I thought about vectorizing then re-rasterizing,
but it turned out I just needed to read the Fragstats manual more
thoroughly. Fragstats has a moving window analysis that kicks out a
continuous raster for each metric.

Cheers,

Jimmy
Reply all
Reply to author
Forward
0 new messages