RE: Swarm and GIS libraries

[Frank Lenk]

I am basically a lurker, having no UNIX or C++/Objective C experience.  
However, I am an experienced economic modeler and forecaster who believes we 
will never be able to predict the non-linear behavior we observe until we 
understand how individuals behave and interact to create the aggregate.

My special field of expertise is forecasting how urban areas grow and develop, 
and my interest is in discovering the least intrusive policies that alter the 
dynamics to produce economies that generate growing equality of opportunity, 
rather than growing inequality, and that produce balanced, sustainable growth 
rather than rapid growth outward coupled with decline and disinvestment inward.

The agent-based part of Swarm gets half the picture, if you think of agents as 
human beings interacting with one another.  The other half is the environment 
in which they live, which constrains many of the possible choices agents might 
otherwise have.  Indeed, the difference in environments contributes greatly to 
urban form -- I live in Kansas City, and the vast quantities of farmland that 
surround the metro seem especially conducive to growing outward rather than 
upward.

In my minds eye, I picture the ideal modeling framework as essentially a 
digital aerial image of the Kansas City area upon which my agents are 
wandering, producing, traveling, buying housing, etc.  As I watch, the aerial 
image changes, with urban development spreading outward, housing being 
demolished inward, traffic jams up where growth exceeds capacity and flows 
unimpeded where the demand anticipated to be served by a new highway never 
shows up.  

In essence, I am watching a raster image being built up over time, except it is 
an intelligent raster.  I can ask it (or actually my agents wandering on top of 
it) why the development proceeds as it does.  Ideally, I would calibrate the 
model by starting with one raster image and altering parameters to best match a 
sequence of later images.

GIS becomes very important in representing the physical environment,  from 
everybody on this mailing list (and
anybody in the field, who is not on it, as well!). So I've put to
together my own list, which will hopefully provoke some discussion on
requirements. Some of the things I have suggested here may appear
naive to seasoned GIS-ers - so please bear with me - I am new to this
world.

I'd also like to see this list get a lot longer to get an idea of the
overall "space" of possibilities - before cutting it down again ;-)

- application/research domains:
  
  * ecological (seems a common one at the moment)
  * urban planning
  * agriculture
  * etc ... ??

- spatial analysis tools

  * Fragstats
  * wavelet analysis tools - ??
  * Fast-Fourier transform tools - ??

- GIS tools

  * GRASS
  * Arc/Info

- GIS formats

  * Arc/Info
  * ...

- scenarios for running a future GIS-enhanced Swarm app.

  Here's some ideas totally off the top of my head - in increasing
  order of unlikelihood and vaporousness :-)

  * totally-decoupled: 
    pouring static data from a established GIS format (eg. Arc/Info ?)
    from a single epoch into our Swarm "agents" and letting the 
    simulation run with whatever rules as specified by the researcher 
    (essentially decoupling the initial data from the rest of the 
    simulation). 
  * loosely-coupled:
    pour the static data into Swarm - and then write out the data
    as modified during the dynamic runs (i.e. at the end of a "run")
    - back into the GIS system for further analysis - giving users
    the full power of the GIS system - but not interactively.
  * medium-coupling:

    ... fill in the blank ? ...

  * highly-coupled:
    pouring the same data from a GIS package - but allow Swarm to 
    constantly communicate with the GIS system, and in effect be
    able to dynamically call some of the more sophisticated "top
    level" GIS functions from within Swarm and use that information
    at run-time from within the model. Also enable data streaming 
    whereby data analysis is decoupled from the Swarm machinery and
    "out-sourced" so-to-speak to the GIS system, but the results of
    the data analysis could then be piped back into Swarm - appearing
    on a BLT widget - say. Of all these options this is extremely
    technically dubious and therefore the most interesting ;-)
    
It seems clear to me that there are many free software packages that
already do many of the kinds of things required for spatial analysis.
We don't want to reinvent the wheel, so it seems that the goal could
be to identify a number of packages that cover the kinds of
functionality that users of GIS and spatial tools need, and start to
write some Swarm interfaces to those. 

To my mind, there's no sense in trying to write everything directly in
Swarm, but having Swarm as a kind of loose nucleus - (the simulation
machinery and interface standards essentially) - for people to attach
their custom analysis libraries to.

My feeling is that the 80-20 rule could come in handy here, if we
could find a few packages that the community already find useful in
the context of their current work that cover 80% of the sorts of
things most folks need, then I think we would certainly be on the way
to a reasonably useful set of libraries.

I am presently investigating the use of Fragstats in exactly this
way.

At the same time, we don't want to make it difficult or discourage
users of less popular packages to interface to Swarm - we want to
encourage them as well. 

Please do tell us what you want! Here's to kick-starting an
interesting discussion!

  --- Alex at the SFI Hive
  
-- 
  Alex Lancaster         |   e-mail: address@hidden
  Swarm Developer        |      web: http://www.santafe.edu/~alex
  Santa Fe Institute     |      tel: +1-(505) 984-8800 (ext 242)
------------------------------------------------------------------

                  ==================================
   Swarm-GIS is for discussion of Geographical Information Systems (GIS)
   and their integration or use with Swarm.  For list administration
   needs (esp. [un]subscribing), please send a message to
   <address@hidden> with "help" in the body of the message.
                  ==================================

<<application/ms-tnef>>