ADVANCED GIS

SV425

Spatial Data
Analysis and
Modelling

Spatial Data Modelling

Takudzwa JERA
Dept of Geomatics Engineering
University of Zimbabwe
 Spatial Data Modelling
Spatial Data Modelling
 The ability to generate models simulating real-world
events is one of most significant feature for GIS.
 Although any map is a spatial model as it provides a
representation of the real world, GIS can be used to
analyse multiple variables and then interpret these
variables to simulate or predict potential events. This
capability allows us to interpolate information for large
areas of the landscape.
 Models enable us to represent, in a simplified form,
some aspect of reality.
 They consist of a set of spatial variables tied together
by an arithmetic operation or set of commands, to
represent past, present and future conditions
 Spatial data models can be classified by Purpose, Logic
or Methodology
 Spatial Data Modelling
Classes of Spatial Data Models - Purpose
Descriptive Model Explanatory Model
 A descriptive model describes the current  Explanatory models attempt to explain or
conditions of a real world environment account for the occurrence of existing
phenomena by identifying the factors
 A simple thematic map showing land use
involved and assessing their relative
can be considered a descriptive model
influence.
because represents an environmental
characteristic that actually occurs on the  Examples of explanatory models include
ground. Other soil erosion models, ozone depletion, and
algae blooms.
 Examples include digital elevation models,
land use coverage, meteorological maps,
vegetation cover maps.
 Spatial Data Modelling
Classes of Spatial Data Models - Purpose
Predictive Model Normative Model
 A predictive (or prescriptive) model  Normative models attempt to affirm how
predicts (by using the factors identified in phenomena (especially network) ought to
an explanatory model) where you might operate in the real world; they recommend
find occurrences of a particular optimal solutions for given situations.
environmental phenomena.
 Examples include food aid distribution,
 For example, a site suitability map, derived traffic volume levels, and route planning
from vector and raster datasets, such as
land use, land cover, roads, rivers etc
might be used to predict the capacity of an
area to locate a dump site.
 Spatial Data Modelling
Classes of Spatial Data Models - Logic
Inductive Model Deductive Model
 An inductive model moves from the  Deductive models derive specific
specific to the general, basing its conclusions by using general premises
conclusions on evidence observed in established in scientific theory or physical
previous studies. laws - where the variables and their
interactions are well understood.
 Typically, inductive models attempt to
identify general conditions or rules when  An archaeological potential model that
important themes and relationships are not uses physical constraints (e.g., slope,
well understood. aspect, proximity to water) to predict the
location of sites is an example of deductive
 An archaeological potential model that
logic.
relies strictly on the distribution of existing
sites is an example of inductive reasoning.
 Spatial Data Modelling
Steps in Spatial Data Modelling
 Developing a model involves working
through a series of conceptual steps, from
identifying the problem through to
implementing the results.
 It is critical to the development of a model
to understand the problem at a conceptual
level prior to developing the model and
therefore following the steps detailed
below prior to beginning work on the
computer is highly recommended.
 Spatial Data Modelling
Step 1: Identify the Problem
 In order to establish a model that addresses
a specific problem, it is necessary to clearly
define the problem and goal(s) of the
model.
 Parameters for the analysis have to be Selection of a suitable site for a
determined by bringing up the following Dump site
questions
- What phenomena are being modelled?
- Why is the model necessary?
- What is the spatial scale and extent of the
model
 Spatial Data Modelling
Step 2: Break the Problem Down
 After you have established the problem and
the goals, breaking the problem down into
its constituent parts helps create more
manageable steps.
 This involves identifying the objectives • Determine the requirements
required to reach your goal, the phenomena • Slope, proximity to water,
involved, and the interactions between proximity to road etc
these phenomena. • Compile the required dataset
• define how their associated
 Establishing the phenomena allows the attribute and spatial
modeller to identify and assemble the relationships contribute to
datasets required for processing. Often, a creating
flowchart can be useful in visualizing and
understanding the spatial and attribute
relationships between the constituents.
 Spatial Data Modelling
Step 3: Develop and Calibrate the Model
 Identifying the tools and mathematical
operations required for analysis is addressed
at this stage. Using the data examined in
Step 2, the modeler builds the model from
these tools and operations.
we may then want to refine the results
 Subsequent repeated running of the model based on other criteria.
allows the modeller to calibrate the model. 100m
 Calibration involves comparing the results of
the model with its input data and
subsequently adjusting the parameters or
mathematical operations to arrive at more
accurate results.
 Using a GIS, the modeller would calibrate
the model by repeatedly running the model
and comparing the results to the data.
 Adjustments to parameters and operations
would be used to refine the model results.
 Spatial Data Modelling
Step 4: Validate the Results
 Validation is an evaluation of the model’s capacity for accurately predicting the real world
phenomena.
 This involves comparing the results to field data and/or running the model using a different
set of data representing conditions that are unlike those used in the calibration phase. If
another set of data are not available, a suitable alternative consists of splitting the one dataset
into two subsets: one for developing and calibrating the model, the other for the validation
process.
Error Detection
 Visual Inspection – one method for identifying errors is simply by looking at the output to see if the results
seem logical and consistent and that the output data reflects the input.
 Documentation – metadata for input datasets can be used to ensure you are using the data appropriately,
based on scale, accuracy, attribute values.
 Validation Rules – topological and attribute domain ranges are two useful validation rules that allow you
to check the data against the design.
 Consistency of Results – collecting data again and repeating the conversion and processing steps can be
used to compare the results of a model.
 Ground Truthing – verifying the results of a model in the field is a dependable method forvalidating data.
 Statistics – correlating the model results with a closely associated variable is a statistical approach
 Spatial Data Modelling
Step 5: Implement the Model Results
 Once the model has been validated, the
modeller can then implement the model
results.