Geographic Information Systems
Analytical Modelling in GIS

Mark Foley
mark.foley@dit.ie

Analytical Modelling in GIS

Learning Objectives
- Explain what a 'model of spatial process' is
- Identify the most common types of spatial model
- Describe how process models are implemented in GIS
- Give examples of how GIS has been used in the modelling of physical and human processes
- Define diffusion modelling and describe where it can be used
- Describe multi-criteria evaluation and how it is implemented in GIS
- Outline the main problems associated with the use of GIS to build process models
- Describe how public input can be incorporated in GIS analysis

Spatial form and spatial process

Form
- How the world looks
Process
- How the world works
What factors (processes) act to change form?
Examples
- Population change
- Consumer spending patterns
- Soil erosion
- Climate change

Examples of types of process models

Physical process models
Human process models
Decision-making models
Any model simulates real world processes
Aids understanding of real world processes
Cheaper (and safer) to simulate rather than do the real thing
A priori models
- Used where a body of theory is yet to be established
- Investigate whether a phenomenon is actually happening
- Example: Global Warming
A posteriori models
- Designed to explore an established theory

Natural and scale analogue models
- Uses actual events or real-world objects as a basis
- Scale models
  - Maps
  - Images
Conceptual models
- Usually expressed in verbal or graphical form
Mathematical models
- Statistical techniques
- Stochastic or deterministic

Natural analogue model for predicting avalanche hazard

Simplified conceptual model of avalanche prediction

Regression model of slope against avalanche size

Modelling physical and environmental processes

Forecasting
- Trends
- Conditions
- Changes
- Outcomes

Methodology for estimating emissions within a GIS framework

A simplified conceptual forest fire model

Derivation of catchment variables using a DEM

Modelling human processes

How do people move through space?
How do they make decisions such as
- Where to live
- Which shops to buy from
- Where to go on holiday
Spatial interaction models
- Predict flow
Supply, demand and distance
- Distance decay functions

Residuals (percentage difference between actual and predicted sales) for The Specialty Depot’s store network in Toronto, Ontario, Canada

Modelling the decision-making process

Simplified MCE algorithm
- Weighted linear summation technique
Steps
- Define problem
- Selection of criteria
- Standardisation of criterion scores
- Allocation of weights
- Application of algorithm
Problems
- Choice of algorithm
- Specification of weights

Applying a simple linear weighted summation model in raster GIS

Weighting data layers in the house hunting case study

Problems with using GIS to model spatial processes

Quality of source data for model calibration
Availability of real-world data for validation
Implementation within a GIS
Complexity
- Data tends to be large and problematic
- Long-term data may not exist
- May have to wait a long time to prove or disprove the model

User-entered area of perceived ‘high crime’ together with attribute data. (left) Simple area and comment about why the specific areas were chosen; (right) Output showing all user areas averaged, together with ranked comments for one area

Total crime densities for Leeds for all crimes committed in 2002. (a) Blacker areas are higher in crimes. The circular high is genuine and mainly reflects the location of the inner ring road around the city. (b) Areas sprayed as ‘high crime’ areas by users in a pilot running from August to September 2002. Blacker areas are those felt to be higher in crime. (c) The difference between (a) and (b), generated after stretching the highest perceived crime area levels to the highest real crime levels and the lowest perceived crime levels to the lowest crime levels. Red areas will tend to have higher crime levels than expected, blue areas lower. UK Census Wards are shown for reference

Soil erosion models

Source: (b) SciLands GmbH, www.scilands.de

Example of a model builder interface (Idrisi32 Macro Modeler) showing a dynamic urban growth model based on land use and suitability map inputs used to produce a map of urban growth areas