lack of output using current option

[Ron]

Hi there,

I am just trying out the toolkit for the first time today, seems like
it has a lot of potential. My trial run with the betweenness
centrality option worked fine after a bit of tweaking. Then I tried
using the same habitat hexmap (scaled so that higher values=better for
wildlife) for the current mapping, and have run into a persistent lack
of output.

I used the edge list-conductance option to make a text file for the
graph. Then on the connectivity tab I used the current flow
betweenness centrality option, with "none selected" for the additional
input files. When I hit run, the program goes for about 20 seconds,
maxing out my CPU at 100%, and then stops. In the output tab it says
it wrote the new output text file, but the file doesn't show up. There
are no error messages displayed.

The betweenness centrality option took about 15 minutes to complete my
trial run, for comparison. The hex files have about 6000 nodes.

My computer: Dell Optiplex with Intel Core 2 Quad, Q9550, 2.83 Ghz,
with 4 GB ram, Windows 7 Ultimate 32 bit.

Any ideas?

Also, one suggestion for the users manual: for each final connectivity
option, could you show a summary table of the needed input files, menu
options to choose, etc? For the betweenness centrality, I noticed that
if I made a graph file including a cost hex file in addition to the
habitat hex file (which the program was happy to let me do), the
connectivity run doesn't work. You cover several of the options in the
tutorials but not all.

Also, you might want to more prominently highlight the fact that the
toolkit uses conductance values for the input habitat file - I made a
nice map of urban betweenness before I figured this out!

Ron Sutherland
Wildlands Network

[carlos]

Ron - Thanks for the suggestions for additions to the manual, we'll
incorporate those in the next version.

I'm not sure about the lack of output from current flow betweenness
centrality. One guess is that you're running out of RAM. A 6000 node
graph could use more than 4GB of RAM, or more likely, more than the
~3.2GB allowed for a 32-bit Windows process. Can you track the amount
of RAM used by the pythonflow.exe process in that Windows Task
Manager? If it goes over 3GB then you likely need to try creating a
somewhat smaller graphfile from a coarser resolution hexmap.
I'd be glad to try to run the file as well. You could email it to me
at klamathconservation at gmail dot com.

Also make sure you've downloaded the most recent version of the CAT
(1.12), which at the same download link for version 1.1 that was sent
out to users a few weeks ago. If you downloaded the CAT after 10/18,
you already have the latest version.

[carlos]

Ron - I emailed you the output file produced by the input you emailed
me. The analysis used ~3GB of RAM. That suggests that you need to
slightly reduce the graph size (increase the hexagon size) to run this
with current flow betweenness on a 32-bit windows OS. ie, there's
theoretically just enough physical RAM on your system (4GB) but 32-bit
windows reserves about 1GB of it. we hope to have an improved
algorithm for current flow analyses in CAT version 1.2 that will
require much less RAM.