[stem-dev] RE: Another fix to air transport

[Stefan Edlund <edlund@xxxxxxxxxxxxxxx>]

Right, there was a couple of reasons why there's an absolute number arriving/leaving a region or pipe transport system in the code now. If we simply use a migration rate and multiply with the base node population, it wouldn't work for edges that goes between two pipe transportation systems like the US system and CA system, since the node population for a pipe transport system are travellers already flying. Also airports are infrastructure that has a certain fixed capacity that doesn't easily change for instance if a population grows.

I make sure I don't remove more people from a region than is available. But I agree we might still need to do something for "disasterous" diseases that kills of most of the population and reduce the flow of people through air transport systems.

Thanks,
/ Stefan

Stefan Edlund
Public Health and Computer Science Research
IBM Almaden Research Center
(408) 927-1766 edlund@xxxxxxxxxxxxxxx


"Lessler, Justin T." ---04/09/2009 10:56:31 AM---Stefan – It all sounds pretty much in line with what I was thinking, except one thing. Namely, I was calculating a relative r

"Lessler, Justin T." <jlessler@xxxxxxxxx>

04/09/2009 10:56 AM

To	James Kaufman <kaufman@xxxxxxxxxxxxxxx>, Stefan Edlund <edlund@xxxxxxxxxxxxxxx>
cc	Daniel Ford <daford@xxxxxxxxxxxxxxx>, "stem-dev@xxxxxxxxxxx" <stem-dev@xxxxxxxxxxx>, Yossi Mesika <MESIKA@xxxxxxxxxx>, "judyvdouglas@xxxxxxxxxxx" <judyvdouglas@xxxxxxxxxxx>
Subject	RE: Another fix to air transport

Stefan – 
It all sounds pretty much in line with what I was thinking, except one thing. Namely, I was calculating a relative rate of travel out of an airport, not absolute numbers (i.e., #leaving/base population size).  The reason is that if we are modeling a disease or other process that substantially changes the number of people in an area, we would want the number travelling from that area to go down proportionally. Otherwise everything sounds right
 
Thanks for seeing this through.
-Justin
 
From: James Kaufman [mailto:kaufman@xxxxxxxxxxxxxxx] 
Sent: Thursday, April 09, 2009 1:38 PM
To: Stefan Edlund
Cc: Daniel Ford; Lessler, Justin T.; stem-dev@xxxxxxxxxxx; Yossi Mesika; judyvdouglas@xxxxxxxxxxx
Subject: Re: Another fix to air transport
 

Thank you Stefan!! 

When we get back let's extend this writeup and add something to our wiki on how air transport is implemented. 
Awesome job!!! 

Best Regards,
Jamie

IBM Almaden Research Center, 650 Harry Rd.
San Jose, CA 95120-6099
email: kaufman@xxxxxxxxxxxxxxx
phone: (408) 927-2477  (tie 457-2477)

Stefan Edlund/Almaden/IBM

04/09/2009 07:01 AM

To	Daniel Ford/Almaden/IBM@IBMUS, James Kaufman/Almaden/IBM@IBMUS, stem-dev@xxxxxxxxxxx, Yossi Mesika/Haifa/IBM@IBMIL, jlessler@xxxxxxxxx
cc
Subject	Another fix to air transport

Hi all, 

I just checked in another fix to air transportation that fixed an issue when running an integrating solution simulation. Please get latest and run update.xml on ...internal.data.geography.infrastructure.transportation. 

Here's how air transportation is implemented. Justin, let me know if this is how you originally envisioned this since you wrote the first implementation. 

Under resource/data/relationship/airtransport2006/ in the internal.data.geography.infrastructure.transportation project you'll find the .properties files that defines the air traffic flow. For instance, in USA_1_USA_2.properties you'll find one row for counties in the US with air transportation infrastructure. For instance, 
this line: 

# US-CA-06081 =  San Mateo County 
63 = US-CA, US-CA-06081, 22242, 22242, 707161 

This means that from San Mateo county there are 22242 arrivals and departures each day (the time period is defined by TIME_RATE_PERIOD at the top of the file). Arrivals and departures are always the same. 707161 is the population in San Mateo. 

In the implementation, for each simulation cycle, we take 22242 people from the San Mateo region and move them up to a Pipe Transporation System node for California. We also take 22242 people from the California Pipe transportation system and move them down to the San Mateo region. People are picked evenly no matter what disease state a person is in, so you're not less likely to travel if you're infected. The population of the California Pipe Transportation System node is the sum of all arrivals from each California county. 

Above the California Pipe Transportation System Node is the US Pipe Transportation System  Node (and above that eventually the World Pipe Transport System Node). This node is defined in the USA_0_USA_1.properties file. The arrivals/departures in this file defines how many people to take from each state pipe transportation system node and move up and down between the US Pipe Transportation Node each cycle. Observe that in cases where we're building a scenario using only US states (no counties), people are moved up and down from the state region nodes instead of the state pipe transportation system nodes. Hence, the graph that we build contains 2 edges from a pipe transportation node, one to a region and another to the lower level pipe transportation system node. It is possible that a model will end up with dangling edges when a lower level air transportation graph is not available. Those edges are removed automatically and I fix the code to avoid reporting unresolved edges for such edges since it will confuse the end-user. 

The population of all Pipe Transporation System nodes are initialized before a simulation begins. 

Disclaimer: Right now the rates in USA_0_USA_1.properties are probably not correct, we need to revisit this. The idea is that this file can be generated from the USA_1_USA_2.properties file calculating the probability that a trip is "inter-state" versus "intra-state". 

Regards, 
/ Stefan 


Stefan Edlund
Public Health and Computer Science Research
IBM Almaden Research Center 
(408) 927-1766   edlund@xxxxxxxxxxxxxxx

----- Forwarded by Stefan Edlund/Almaden/IBM on 04/09/2009 06:10 AM -----

Stefan Edlund/Almaden/IBM

04/03/2009 11:19 AM

To	Daniel Ford/Almaden/IBM@IBMUS
cc	James Kaufman/Almaden/IBM@IBMUS, stem-dev@xxxxxxxxxxx, Yossi Mesika/Haifa/IBM@IBMIL
Subject	Re: Air transport now worksLink

Thanks. 

Another fix in this morning so please get the latest and run the ant script again. I forgot to correct for the rate time period of the air transport pipe. For instance, if a pipe shuffles 1000 people up and down every day, if a simulation cycle is only 6 hours we now correctly shuffle only 250 people per cycle. 

Regards, 
/ Stefan 

Stefan Edlund
Public Health and Computer Science Research
IBM Almaden Research Center 
(408) 927-1766   edlund@xxxxxxxxxxxxxxx

Daniel Ford/Almaden/IBM

04/03/2009 07:59 AM

To	Stefan Edlund/Almaden/IBM@IBMUS
cc	James Kaufman/Almaden/IBM@IBMUS, Yossi Mesika/Haifa/IBM@IBMIL, stem-dev@xxxxxxxxxxx
Subject	Re: Air transport now worksLink

Excellent, I cc'd stem-dev. 

Dan

Stefan Edlund/Almaden/IBM

04/02/2009 08:36 PM

To	Yossi Mesika/Haifa/IBM@IBMIL
cc	James Kaufman/Almaden/IBM@IBMUS, Daniel Ford/Almaden/IBM@IBMUS
Subject	Air transport now works

Hi, 

air transportation should work now if you get the latest and run the update.xml ant script (should only be necessary to do the infrastructure.transportation one). You can test by defining a US scenario (state only or down to counties level 2) and dragging the air transporation graph(s) into the model. If you define a disease where the road and common border infectious factor is 0 you should still see the disease spread from region to region via air travel. 

 Yossi, can you do a new build and put up on the site (if you got time maybe make sure everything works first)? 

Thanks! 
/ Stefan 

Stefan Edlund
Public Health and Computer Science Research
IBM Almaden Research Center 
(408) 927-1766   edlund@xxxxxxxxxxxxxxx