Directed connectivity

[Miranda Gray]

Hi,
I am interested in exploring this toolkit for fire risk connectivity. I have created a resistance, or cost, or capacity surface for fire spread and would like to integrate into that the influence of wind vectors. So I would like to create a directed edgelist or lgf file but am unsure how to go about this. Essentially, to start simple, I would like the wind to exert an equal but opposite resistance or cost going from (i,j) to (j,i). There is one dominant wind direction from the SW, so always from south to north and from west to east would receive the lesser resistance. Do you have any ideas of how I can go about this with existing toolkit capabilities? Thanks in advance.

Miranda

[carlos]

Miranda - I'll have to ponder this a bit, a very interesting application. There might be some previous work on hydrologic connectivity that's relevant.

[Miranda Gray]

Hi again Carlos,

I was hoping you could go into a little more detail about the way that the network flow methods deal with multiple sources and targets, that are not paired 1-1. Is the output a sum of all flow values from all sources and targets, or a maximum? Also, I was thinking of approaching the fire connectivity with an averaging of network flow maps, where the cost differs based on prevailing wind directions but the capacity would stay the same. Do you have any experience mosaicing connectivity maps, similar to this, and have any thoughts on it? 

Thanks so much,

Miranda

[carlos]

Miranda - I'll contact you directly and see if I can offer any tips.

[carlos]

I'm also posting my reply here for future reference:

What you are trying to do is a novel application (to my knowledge),
and I don't know whether graph-based analysis is a good choice of
method, but it may be. I'll attempt some quick answers.

By the way, graph networks have been used to study fire spread in
buildings, there may be some parallels to your problem.
see:
Fire Safety Journal Volume 9, Issue 3, August 1985, Pages 287-300
Modeling of fire spread through probabilistic networks
Wai-Ching Teresa Ling*, Robert Brady Williamson
College of Engineering, University of California, Berkeley, California
94720 U.S.A.
http://dx.doi.org/10.1016/0379-7112(85)90039-6

1) How do network flow methods deal with multiple sources and targets,
that are not paired 1-1?

The network flow centrality metric in the CAT sums solutions for
individual source-sink pairs. It iteratively solves from source node 1
in group A to all sinks in group B, then from source node 2 to all
sinks, etc. and sums the results.
I found the attached pdf dealing with multiple source-sink flow
problems, but I don't know much about this.

2) Is the output a sum of all flow values from all sources and
targets, or a maximum?

Sum of flows from each source-sink pair, see above.

3) I was thinking of approaching the fire connectivity with an

averaging of network flow maps, where the cost differs based on
prevailing wind directions but the capacity would stay the same. Do
you have any experience mosaicing connectivity maps, similar to this,
and have any thoughts on it?

This sounds intuitively reasonable, and you should have any problem
averaging the network flow output (you'd just join the output files
based on the node id). I think that's what you mean by mosaicing.