GIS and Simulation

1
GIS and Simulation

• Matthew White
• EVSC 468
• March 20

2
Models and Simulation
3
Simulation

• A model is any representation of a process,
  event, system, object, etc.
• Flowcharts
• Systems of equations
• Action figures
• Simulation uses models to predict (or at least
  help to understand) behavior in the real world.

4
Modeling

• Deterministic
• Behavior is predictable, no randomness involved
• Example a system of equations
• Stochastic, or probabilistic
• Variability in inputs
• Probability distributions identified through
  existing data or experience
• BestFit is an add-in for Excel that will match
  (or attempt to) data to various distributions
• Example rainfall

5
Stochastic Simulation

• Monte Carlo
• Repeated random sampling of inputs from a given
  distribution to calculate expected values of some
  output
• Stochastic inputs and deterministic computations
• _at_Risk, another add-in for Excel by Palisade
  Decision Tools
• Example calculating pi
• Discrete Event
• Models process flows that rely on multiple random
  events
• Stochastic inputs and computations
• Arena, by Rockwell Software
• Example queue time at banks

6
Simple Arena Model
7
Distributions

• I promise you that Im normal, and I have a
  distribution to prove it.
• Normal
• Exponential
• Poisson
• Binomial
• Triangular
• Uniform

8
Temporal Variability

• Static
• No change over time
• Example water level in a lake
• Dynamic
• Continual change over time
• Most systems are dynamic
• Example water level in a river

9
Why Simulate?

• Simulation allows the exploration of unknown
  processes
• Helps understand how systems work, events occur,
  etc.
• Increased predictability of events mitigates
  potential damages

10
GIS SimulationModelBuilder
11
Capabilities

• Looping
• Allows multiple time steps and iterations
• Branching
• Controls process flow
• Intermediate outputs
• Important to save results from each iteration to
  see changes over time
• Can be saved in a list

Looping
Branching
12
Adding Randomness

• Use a random number generator (RNG) with any of
  the following tools
• Parameters
• Calculate value
• Feature Classes
• Calculate field
• Create random points
• Raster data
• Create random raster

13
Example Calculate Value

• Tools
• Surface toolset
• Spatial Analyst
• Slope
• General toolset
• Data Management
• Calculate Value
• The output of Calculate Value is added to the Z
  factor in the Slope tool. Within the Slope dialog
  box, select the variable name in the Z factor
  field.

arcgis.rand("Normal 2 2") 2
14
Real World ExampleFire Spotting

• example taken from ArcGIS online help

15
Fire Spotting Model

• Understanding the process (or not)
• Wind speed direction (A)
• Temperature (B)
• Slope and location (C)
• Fire resistance (D)

16
Fire Spotting Model

• Early stages of the simulation model.

17
Fire Spotting Model

• Later stages of the model.

18
Fire Spotting Model

• Starts with a general growth model.
• Decides whether or not knew spots will be
  generated based on a calculation using a random
  number.
• If spotting occurs, generates random raster
  points to catch on fire.

19
Fire Spotting Model

• Also need a branch where fire spotting does not
  occur.
• Branches merge and the new locations are looped
  back to the starting point.

20
Other Examples

• Flood prediction
• Irrigation analysis
• Predictive logistic models based on demographics
• Political support
• Retail tracking
• And more!

21
Final Thoughts
22
Final Thoughts

• Simulation is a supplemental decision tool
• Models are always simpler than the real world
• Assumptions may not be accurate
• Variables can be ignored or misunderstood
• Always conduct sensitivity analysis on results
• Change distributions or variable values
• ModelBuilder can be used for sensitivity and
  error analysis

23
References

• ArcGIS Online Help
• http//webhelp.esri.com/arcgisdesktop/9.2/index.cf
  m?TopicNameAn_overview_of_advance_modeling_throug
  h_simulations
• 2007 Special Topic, Gavin Schmidt
• http//www.vcrlter.virginia.edu/jhp7e/advgis/loca
  l/students/2007/schmidt_simulation_and_gis07.pdf