5 Replies Latest reply on Feb 6, 2013 2:25 PM by laura_mendenhall

    Estimating Kernels Using Geospatial Modeling Environment

    ljcwalston
      Hi,

      I recently downloaded Geospatial Modeling Environment (GME) to support some analyses of ecological telemetry data. I need to run some KDE analyses. I have upgraded ArcGIS desktop to v10 (SP4), updated R (2.15.0) and installed the data libraries and statconn. I can calculate basic MCPs for my data using the genmcp command, no problem. So I know that at least some of the software is working. However, I try to run the kde for the same data and I get null (empty) output regardless of the input, or other parameters. I've tried the default kde settings, as well as other values, and the result is the same. The output also lacks geographic reference, which was something that was retained for other analyses like genmcp. So it makes me think that the kde command requires something else. Anyone experience this?  Troubleshooting ideas?

      ArcGIS Desktop 10 (SP4)
      GME 0.6.2
      R (2.15.0)
      StatConn

      Thanks for any info.  http://www.spatialecology.com/gme/index.htm
        • Re: Estimating Kernels Using Geospatial Modeling Environment
          vsfoote
          Lee,

          Syntax is pretty straight forward, is the point data projected and did you set an extent?

          [ATTACH=CONFIG]13736[/ATTACH]

          Stuart
          • Re: Estimating Kernels Using Geospatial Modeling Environment
            ljcwalston
            Yes. The same point data (projected UTM coordinate system) that was used successfully to produce MCPs (genmcp) was used in kde. I also tried setting extents (even though that is optional). Null output each time, and the geographic reference was lost.
            • Re: Estimating Kernels Using Geospatial Modeling Environment
              vsfoote
              Lee,

              Had to find an ArcGIS Desktop 10.0 system to work from -- GME 0.6.2 is compiled against Desktop 10.0 and chokes when installed on a 10.1 seat.

              Anyhow, kde worked from the GME GUI command builder interface but has a couple of quirks.  It didn't like pulling point features from an old 9.3 File GeoDatabase, but it consumed Point Shape files without issue. Also, for the output raster, had to be sure a full path to the raster was specified and not try to put the raster back into a File Geodatabase.

              I had to adjust the CELL size and BANDWIDTH (e.g. Radius) to complete processing of a useable raster.  But the resulting surface did show KDE, and the raster was registered into the projected coordinate system.

              For your UTM projected point data, for starters maybe try a cell size of 20 (meters) and a bandwidth of 200 (meters)--other than input, output, cell size and bandwidth the rest were defaults. And the extent becomes that of the input Point Feature class.

              Stuart
              • Re: Estimating Kernels Using Geospatial Modeling Environment
                lstaff
                I guess it all depends on the version of StatConn you have.  According to the GME website if you had the newest version there were issues.  There is now an updated GME and you don't need to have StatConn.  I suggest you uninstall the old version and try the new version.  I was able to make KDEs.
                • Re: Estimating Kernels Using Geospatial Modeling Environment
                  laura_mendenhall
                  Lee,

                  I was getting null values when using unprojected data and/or a forced extent. Once I projected my point shapefile, the KDE ran smoothly but each animal ID has a different extent. I need to force the extents to be the same and when I include that in my for loop GME either runs into an ambiguous error or, if it does run, my output is null. Here is my for loop:

                  for(i in 1:length(ids)){
                  kde(in="animal_points_prj.shp", out=paste("KDE_",ids[i],".img"), bandwidth="SCV", cellsize=30,
                  ext="study_area_prj.shp", where=paste("BirdID='",ids[i],"'"));
                  }

                  Anyone else having trouble forcing the extents across multiple animal IDs? The old Hawth's Tools had a simple button to force extents.

                  Thanks,

                  LM