Benefits and Complications
of Spatial Analysis
Diana Ter-Ghazaryan, PhD, GISP
SSCI 583: Spatial Analysis and Modeling
Today’s Topics
What We Can Do With Spatial Analysis
Key Concerns:
Conceptual Models of Space and Data Models
for Spatial Data
Spatial Autocorrelation and the Spatial
Structure of Data
The Modifiable Areal Unit Problem and
Related Issues
TODAY’S TOPICS | 2
What We Can Do With Spatial Analysis
WHAT WE CAN DO WITH SA | 3
What Is Spatial Analysis?
• “A geographical analysis which seeks
to explain patterns of human
behaviour and their spatial
expression in terms of mathematics
and geometry.” Mayhew, A Dictionary
of Geography.
• The study of “regularities in the
spatial distribution of economic
activity, populations, land use, and
other dimensions of human activity.”
Calhoun, Dictionary of the Social
Sciences.
Poorthuis, A. and Zook, M., 2015. Small stories in
big data: Gaining insights from large spatial point WHAT WE CAN DO WITH SA | 4
pattern datasets. Cityscape, 17(1), pp.151-160.
Multiple Criteria
Decision Analysis
Multiple criteria decision
analysis for answering
spatial questions using
many types of spatial
data
Perez, O.M., Telfer, T.C. and Ross, L.G., 2005.
Geographical information systems‐based
models for offshore floating marine fish cage
aquaculture site selection in Tenerife, Canary
Islands. Aquaculture Research, 36(10),
pp.946-961.
APPLICATIONS 5
Finding Patterns in Spatial Data
Measuring road
density to assess
landscape
fragmentation
Cai, X., Wu, Z. and Cheng, J., 2013.
Using kernel density estimation to
assess the spatial pattern of road
density and its impact on landscape
fragmentation. International Journal
of Geographical Information
Science, 27(2), pp.222-230.
WHAT WE CAN DO WITH SA | 6
Networks and Accessibility
Analyses
Accessibility of health
care, emergency, or other
social services
Huerta Munoz, U. and C. Källestål. 2012.
Geographical accessibility and spatial coverage
modeling of the primary health care network in
the Western Province of Rwanda. International
Journal of Health Geographics, 11:40.
WHAT WE CAN DO WITH SA | 7
Spatial Regression
• Spatial regression models
incorporate the relative
location of dependent
and independent
variables in a regression
model to identify the
different relationships
across space
Figure 1 in: Weir, R. (2019). Using geographically weighted regression to explore neighborhood ‐level
predictors of domestic abuse in the UK. Transactions in GIS 23:1232–1250.
MODELING SPATIAL PROCESSES | 8
Spatial Modeling with Machine Learning
Modeling suitable
habitats for endangered
plant and animal species
Kumar, S. and Stohlgren, T.J., 2009. Maxent
modeling for predicting suitable habitat for
threatened and endangered tree
Canacomyrica monticola in New Caledonia.
Journal of Ecology and the Natural
Environment, 1(4), pp.094-098.
WHAT WE CAN DO WITH SA | 9
Working with Fields
Interpolation of data
points at unmeasured
locations
Pavão, C.G., França, G.S., Marotta, G.S.,
Menezes, P.H.B.J., Neto, G.B.S. and Roig, H.L.,
2012. Spatial interpolation applied a crustal
thickness in Brazil. Journal of Geographic
Information System, 4(2), pp.142-152.
WHAT WE CAN DO WITH SA | 10
Key Concern:
Conceptual Models of Space and Data
Models for Spatial Data
CONCEPTUAL AND DATA MODELS | 11
Object Model
Aspects of the physical
world are
conceptualized as
discrete objects
In the mindset of
Euclidean space, we
conceive of the world
as the empty backdrop
in which objects exist
CONCEPTUAL AND DATA MODELS | 12
Field Model
But some aspects of the
real world do not seem like
individual, discrete objects
We conceive of some
aspects of the world as
continuous features of
varying values
CONCEPTUAL AND DATA MODELS | 13
Difficulties with Objects
Objects can also represent
fields (contour lines)
Objects move and change
Objects don’t have simple
geometries (networks, e.g.)
Objects depend on scale
Objects can be fuzzy or
indeterminate
Objects can have a fractal
dimension (see figure)
CONCEPTUAL AND DATA MODELS | 14
From Conceptual to Spatial Data Models
Vectors and Rasters
*Modified from Exploring GIS: Spatial Data Representation
CONCEPTUAL AND DATA MODELS | 15
Two Data Models: Vector and Raster
The vector data model
stores real world
phenomena as individual
objects in a data layer
The raster data model stores
real world phenomena as
fields in a grid, much like
pixels in an image
CONCEPTUAL AND DATA MODELS | 16
Transformations Between Data Types
*Modified from Exploring GIS: Spatial Data Representation
CONCEPTUAL AND DATA MODELS | 17
Key Concern:
The Spatial Structure of Real-World
Processes versus the Spatial Structure
of Data and Models
SPATIAL STRUCTURES | 18
Tobler’s First Law of Geography
“Everything is related to everything else, but near things
are more related than distant things.”
SPATIAL STRUCTURES | 19
Spatial Autocorrelation
A measure of the relationship between the value of a variable
at a location and the same variable but at another location
separated by some specified distance
Put another way: Spatial autocorrelation is measurement of
Tobler’s First Law. Spatial autocorrelations test and measure
this assumption for different types of spatial objects,
processes, and events
How does an attribute’s value change across space? Are values
near to one another (called neighbors or in the same
neighborhood) similar? More so than we would expect in the
case of complete spatial randomness?
SPATIAL STRUCTURES | 20
Spatial Heterogeneity
And yet, everywhere is not the same;
we see an uneven distribution of
features and variables
Does a thermometer 100km away tell
you anything about the temperature
where you are?
The answer differs: across a vast
desert, probably yes; on the other
side of a mountain, probably no. Vitharana, U.W.A., Mishra, U., Jastrow, J.D.,
E.g. soil factor change across Alaska, Matamala, R. and Fan, Z., 2017. Observational needs
for estimating Alaskan soil carbon stocks under
current and future climate. Journal of Geophysical
creating different soil regimes Research: Biogeosciences, 122(2), pp.415-429.
SPATIAL STRUCTURES | 21
Determining Spatial Relationships:
Distance
Distance / nearness is
the spatial concept
called out explicitly in
Tobler’s first law and is
clearly the most basic
way we might consider
phenomena to be
spatially related
SPATIAL STRUCTURES | 22
Determining Spatial Relationships:
Adjacency / Contiguity
Places that are next to each
other; places that share a border
A binary measurement
Areas within adjacent spatial
entities can also be distant, so
this is a different way to measure
a spatial relationship
Rook’s case: features sharing an
edge are neighbors
Queen’s case: features sharing an
edge OR a node are neighbors
SPATIAL STRUCTURES | 23
Determining Spatial Relationships:
Proximity Triangles; Nearest Neighbor
Other methods of assessing spatial relationships is the consider
the closest neighbor or the closest of a different entity
For example, the proximity polygon method
SPATIAL STRUCTURES | 24
Cataloguing Spatial Relationships
• Measures of spatial autocorrelation and spatial statistics
depend upon the creation of a spatial weights matrix
• Expression for spatial relationships (called the spatial
structure or neighborhood structure of a spatial dataset)
• Represents the spatial relationships to be used in spatial
autocorrelation/dependence measures e.g. who are
neighbors
SPATIAL STRUCTURES
SECTION TITLE| |252
Key Concern:
The Modifiable Areal Unit Problem
(MAUP) and Related Issues
MAUP | 26
Aggregation Is Problematic
How do you group or “bin” your data?
This is true for all sorts of data, including temporal data (this
figure) and spatial data
MAUP | 27
The Modifiable
Areal Unit Problem
When (point) data are
aggregated, how the
data are aggregated
will greatly influence
the result and can
change the outcome.
MAUP | 28
Two Types of Effects
Scale Effect: when point data are aggregated at a certain areal
unit, or when data from one scale are analyzed at another
(size of the areal unit changes)
Zonal Effect: when different boundaries of the aggregated
data change the outcome; the way in which the data are
assembled at a given scale changes
MAUP | 29
Scale effect:
Size changes
b>a, c>b
Zonal effect:
Shapes
d & e have
same size,
different
shapes
shapefile, geodatabase
(file storage) Shape f is a
combination
MAUP | 30
Different
aggregation
schemes yield
different results
(and maps)
Xu, P., Huang, H. and Dong, N., 2018.
The modifiable areal unit problem in
traffic safety: Basic issue, potential
solutions and future research. Journal of
Traffic and Transportation Engineering
(English edition), 5(1), pp.73-82.
MAUP | 31
Ecological Fallacy
A value for an aggregated unit may not apply to all individuals
within the aggregated unit
A concept related to the MAUP
MAUP | 32
Edge Effects
Another concept related to the MAUP
Where we draw the boundary of our study area will impact the
results of our analyses inside that boundary
Occurrences on the other side of the study area boundary can
impact our study, but if we limit our data to the study area
boundary we won’t be aware of them.
But we always have to draw our boundary somewhere … so we
must be thoughtful about how to do so and be aware of the
implications
MAUP | 33
Scale
• Scale of analysis
should match the
scale at which spatial
processes operate –
finer resolution data
and larger scale is not
always better
• Recall 581 project
modeling urban Avelino, A.F.T., Baylis, K. and Honey-Rosés, J., 2016.
Goldilocks and the raster grid: selecting scale when
watershed with evaluating conservation programs. PloS one, 11(12),
different scale data … p.e0167945.
MAUP | 34
