This document outlines Andrew Crooks' research using GIS and agent-based modeling to study various social and environmental problems. It provides examples of applications including modeling border security, disease spread, refugee camps, slum formation, and wildfire evacuations. New sources of crowdsourced data from social media and the internet are discussed as a way to supplement traditional data and obtain near real-time information to inform models. The integration of GIS and agent-based modeling is presented as a way to explore complex systems and human behaviors at fine spatial and temporal scales.

1. Department of Computational Social Science
GIS and Agent-
based Modeling
Andrew Crooks
Center for Social Complexity
George Mason University
acrooks2@gmu.edu, www.gisagents.org, @AndyCrooks

2. Presentation Outline
• GIS and ABM: A Gallery of Applications
– Border Security
– New Sources of Data:
• Fusing New Data into Models:
–Disaster Relief, Diseases, Wildfires
• Summary, Challenges & Opportunities

3. Smuggling Corridors in Arizona
Year 2007 2008 2009 2010
Persons
apprehended 378,323 317,709 241,673 212,202
Pounds
of
narcotics
seized 1,360,200 1,045,621 1,204,702 1,128,960*
Magnitude of the problem *
Extrapolation
of
940,800
pounds
seized
by
August
31

4. Landcover
Terrain relief
Roads and tracts
Roughness
Population density and
background traffic
Points of interest
Night time exposure
Land ownership
Data layerAnalysis type
Viewshed analysis
Routing
Behavioral routing
Border Security: GIS layers

5. Zones
of
control
Types
Capabilities
Alliances
Locations
Border Security: Social Layers - Red Organizations

6. Zones
of
control
Types
Capabilities
Alliances
Zones
of
control
Border Security: Social Layers - Red Organizations

7. Zones
of
control
Types
Capabilities
Alliances
Organizational
specialties
Border Security: Social Layers - Red Organizations

8. Zones
of
control
Types
Capabilities
Alliances
Penetration
modes
Border Security: Social Layers - Red Organizations

9. Zones
of
control
Types
Capabilities
Alliances
Alliances,
Rivalries
and
Enmities
Border Security: Social Layers - Red Organizations

10. Blue Organizations
• Blue force mix and CONOPS:
– Areas of responsibility.
– Patrol mixes (mobile surveillance systems, forward bases, force rotation
and force posture).
– Saturation of video surveillance systems, IFT, UGS and their placements.
A hypothetical blue sensor placement scenario

11. Łatek, M.M., Mussavi Rizi, S.M., Crooks, A.T. and Fraser, M. (2012), 'A Spatial Multiagent Model of Border Security for the
Arizona-Sonora Borderland', The Computational Social Science Society of America Conference, Santa Fe, NM.
Border Security: System Architecture

12. Border security
Link to Movie

13. • Further Work:
• Expand decision making capabilities of the Red
side, including more modes of penetration.
• Expand economics of the borderland, both licit
and illicit components
• Validate the model using 2008—2009 data.
Border Security
Rocky
Mountain
Information
Network
seizure
data
Humane
Borders
migrant
death
compilation
“Breadcrumbs”
approach.
Trash
removal
reports
as
another
layer
under
preparation.
Łatek, Mussavi Rizi, Crooks & Fraser, M. (2012), Social Simulations for Border Security. European Intelligence and Security Informatics
Conference,

14. Sample Runs
Border Security: Sample Runs

15. Fusing New Data into Models

16. Harvesting Crowdsourced
Information
• Web 2.0 and Social Media:
• Volunteered Geographical
Information (VGI) and Ambient
Geographical information (AGI).
• Provides a new lens to study the
urban systems as a living, evolving
social organism:
• Advanced situational awareness.
• Unique opportunities for knowledge
discovery and modeling
Stefanidis, Crooks, & Radzikowski. (2013), Harvesting Ambient Geospatial Information from Social Media Feeds, GeoJournal, 78, (2): 319-338.

17. A GeoSocial Modeling Approach
GeoSocial data mining:
The combination of geospatial, social
network, and content analysis, to
understand the human landscape.

18. Thematic Spaces: Neighborhood Example
Crooks et al., (2015), Crowdsourcing Urban Form and Function, International Journal of Geographical Information Science, 29(5): 720-741.

19. Traffic Speeds From GPS Taxi Data
Changing traffic situation as
detected by floating car data –
Berlin, Germany (only major
roads shown).
(a) 16 December 2013 – 1 am.
(b) 8 am.
(c) 5:30 pm.
Crooks et al., (2015), Crowdsourcing Urban Form and Function, International Journal of Geographical Information Science, 29(5): 720-741.

20. Opportunities: Supplement Traditional Data
Crooks et al., (2015), Crowdsourcing Urban Form and Function, International Journal of Geographical Information Science, 29(5): 720-741.
Crowdsourced Data

21. Agent-Based Modeling
• How can we use the crowd here?
– New sources of data.
– Near “real time” information.
– New ways to explore how people
perceive & use the space.
– Connections between people &
organizations.
– Insights into human behavior?
– Rob Axtell: “… there is a large
research program to be done over
the next 20 years, or even 100
years, for building good high-
fidelity models of human behavior
and interactions”
Weinberger (2011), 'Web of War: Can Computational Social Science Help to
Prevent or Win Wars?' Nature, 471: 566-568.
Mobile agents
Immobile agents
Artificial World
If <cond> then
<action1> else
<action2>

22. • Instant reports from media and Web 2.0 technology
(e.g. Twitter, Ushahidi etc..)
• Data released over the internet:
Haiti Earthquake 12th January 2010
- Mostly from the “bottom-up” via
crowdsourcing and VGI
- E.g. Google Map Maker, OpenStreetMap
etc...
– Ground damage, tent cities etc...
• Can ABM and GIS be integrated
to assist post-disaster relief
operations rather than just
evacuations?
Crooks & Wise (2013), GIS and Agent-Based models for Humanitarian Assistance, Computers, Environment and Urban Systems, 41: 100-111.
ABM and GIS for Disaster Relief

23. • Roads (green primary, red secondary).
• Refugee camps emerge (blue).
Source: http://vimeo.com/9182869
Haiti Earthquake 12th January 2010

24. Model Inputs: All Geo-referenced

25. The Environment
• 8km by 6km area of Port-au-
Prince
• Cell Resolution 100m2
– Multiple agents per cell,
can move 100m per tick
(~2m/s)
• Tick ~1min
• Agent population derived
from LandScan (~1.3 million)
– 20 agents max per cell
• Vector roads used for
navigation
– Roads are of different
types
• Centers are hypothetical
• Destruction (red: most
damage, grey: no data
Model Inputs: All Geo-referenced

26. The Agents
• Motivated by their energy levels (initially set by destruction)
• Seek to maximize their energy over the course of the
simulation:
• Agents can choose to move toward a food distribution
center (based on their knowledge of available centers)
or to remain at home.
• Prefer a closer center to a farther one.
• If the agents believe that getting the food will cost as
much energy as the food itself can provide, they will not
move.
• Agents expend energy to move.
Haiti Earthquake 12th January 2010

27. Spread of Information
Link to Movie

28. Comparison of Results for Different Aid Centers
Random Good Bad

29. Slums: Global Context
• 1 Billion people living in more than 200,000 slums on
the face of this planet.
Source: Davis (2006), Planet of Slums, http://tinyurl.com/dkjkeg
30 Largest slums in the world

30. Source: maps.google.com/Source: http://mapkibera.org/
VGI for Slum Mapping

31. Ahmedabad, INDIA
• Population: 3.5 million
• 1.5 million people living in slums
• 1668 slums and chawls
• 1.5 million were migrants
• Area: 192 sq km
• Density: 23000 per sq km
Ahmedabad, India

32. Integrating Spatial Environment Layers

33. Exploring the Formation of Slums for
Ahmedabad, India
Patel, Crooks & Koizumi, (2012), Simulating Spatio-Temporal Dynamics of Slum Formation in Ahmedabad, India. 6th Urban Research and Knowledge
Symposium - Rethinking Cities: Framing the Future, Barcelona, Spain.
Link to Movie

34. Dadaab Refugee
Camps Complex
• Located near the Kenya-
Somalia border in the Garissa
District of North Eastern
Province of Kenya.
• Established in 1991 to host
90,000 refugees from Somalia
(UNHCR, 2011).
• Currently, it hosts nearly half a
million refugees including some
10,000 third-generation refugees
(UNHCR, 2011).
Crooks & Hailegiorgis (2014), An Agent-based Modeling Approach Applied to
the Spread of Cholera, Environmental Modelling and Software, 62: 164-177.

35. Dadaab Refugee Camps
Source: Google Earth

36. Representation of Dynamics
of Cholera
• Susceptible –Exposed – Infected – Recovered (SEIR)

37. Input
Parameters

38. • Each time step, agents make
decision about where to go, based
on their needs.
Food Dist. C.
Health Post
Mosque
Market
Water
Visit R
Social
School
Hygiene
Model Process: Goal Selection

39. Scenario 1 – Contamination of Fixed Point
Link to Movie

40. Scenario 2 – Contamination through Runoff
Actual Cholera Outbreaks

41. Colorado Wildfires
• June and July of 2012
• Wildfires in northern and central Colorado prompted
the evacuation of over 30,000 citizens
• Research question:
• Can crowdsourced social multimedia be used to
delineate the extent of the wildfire and fused with an
agent-based model for evacuation?
• Case Study: Waldo Canyon

42. Delineating Events
Link to Movie

43. Note: word size normalized relative to the
occurrence of “fire”
Frequently Adopted Toponym Terms
Delineating Events

44. q
Delineating Events: Flickr Images
Panteras, Wise, Lu, Croitoru, Crooks, & Stefanidis, (2014), Triangulating
Social Multimedia Content for Event Localization using Flickr and Twitter,
Transactions in GIS. DOI: 10.1111/tgis.12122

45. Detection of the Wildfire via Crowdsourced Data
Panteras, Wise, Lu, Croitoru, Crooks, & Stefanidis, (2014), Triangulating Social Multimedia Content for Event Localization using Flickr and Twitter, Transactions in GIS. DOI: 10.1111/tgis.12122

46. Source: Wise 2014
Deriving Mood
Crowdsourced Data for ABM

47. Building Agent Populations
• ~~
Source: Wise 2014
Crowdsourced Data for ABM

48. Agent Decision Making
Source: Wise 2014
Crowdsourced Data for ABM

49. Source: Wise 2014
Crowdsourced Data for ABM
Link to Movie

50. Crowdsourced Data for Validating Agent-based Models
Source: Wise 2014

51. Summary
• Patterns at the macro-level emerge from micro-level
interactions of many diverse individuals:
– E.g. traffic jams, crowds, diseases, urban growth etc.
• The integration of GIS and ABM provides new tools
and a way of thinking to explore urban dynamics at a
fine spatial and temporal scales.
– But research is needed with respect to developing
high-fidelity models of human behavior and
interactions.
–Need to leverage the universe of all data.

52. Opportunities & Challenges
• Crowdsourced Data:
• Provides a new lens for understanding of how people
perceive, use and are affected by space over time.
• Provides links across scales: from micro to macro
phenomena.
• Challenges:
• Collection and storage of data.
• Short time scales vs. long term problems.
• Validation (cross source), participation bias etc…..

53. Opportunities
Crooks et al., (2015), Crowdsourcing Urban Form and Function, International Journal of Geographical Information Science, 29(5): 720-741.