POINT PATTERN

ANALYSIS
Lecture 4
 Introduction to Point Pattern Analysis
Simplest Spatial Data
• A set of point locations within a defined region.
• A point pattern is a set of events in a study region.
• Each event is symbolized by a point object.
• Data are the locations of a set of point objects.
Applications
-Hot-spot analysis (crime, disease)
-Vegetation, archaeological studies
 Examples
• Position of trees on a landscape.
• Location of crime incidents
• Single type of crime, interaction between types of crime,
• Banks, Distribution of grazing animals in a field
• Study the evolution of the point-pattern over time
• Interaction between two species
 Introduction to Point Pattern Analysis
Requirements for a set of events to constitute a point pattern
-Pattern should be mapped on a plane (flat).
-Study area determined objectively
-Pattern is a census of the entities of interest
-One-to-one correspondence between objects and events
-Event locations are proper
Why Analyze Pattern Analysis?
 Point Pattern
Analysis
Centrography
A very basic form of point
pattern analysis involves
summary statistics such as
the mean center, standard
distance and standard
deviational ellipse.
 How Point Pattern Analysis is Different?
• From Centrographic Statistics :
• Centrographic Statistics calculates single summary measures,
• PPA analyzes the complete set of points
• From Spatial Autocorrelation :
• with PPA, the points have location only; there is no “magnitude” value.
• With Spatial Autocorrelation points have different magnitudes; there is an
attribute variable.
 Use of Point Pattern Analysis
• These point pattern analysis techniques were popular before
computers were ubiquitous since hand calculations are not too
involved, but these summary statistics are too concise and hide far
more valuable information about the observed pattern.
• More powerful analysis methods can be used to explore point
patterns
 Introduction to Point Pattern Analysis

Point Density Point Separation (distance-based)

First-order effect: Second-order effect:
-Variation of intensity of a process across Interaction between locations based on
space. distance between them.
-Number of events per unit area -Relative location
 Density-based Point Pattern Measures
First-order effect
• Sensitive to the definition of the study area.
 Point Pattern Analysis Methods
These methods can be classified into two groups:
• Density based approach and
• Distance based approach.
 Density based techniques

• Density based techniques characterize the pattern in terms of its

distribution over the study area–a first-order property of the
pattern.
• Density measurements can be broken down into two
categories: global and local.

we’ll make a distinction between the intensity of a spatial process and the
observed density of a pattern under study
Global Density
 Local Density

• A point pattern’s density can be measured at different locations

within the study area.
• Such an approach helps us assess if the density–and, by
extension, the underlying process’ local (modeled) intensity ˆλ is
constant across the study area.
• Several techniques for measuring local density are available, here
we will focus on two such methods: quadrat density and kernel
density.
 Density-based Point Pattern Measures
Quadrat count methods
• Record number of events of a pattern in a set of cells of a fixed size
• Census vs. Random
 Quadrat Density
• The study area can be divided into sub-regions (quadrats).
• Then, the point density is computed for each quadrat by dividing the
number of points in each quadrat by the quadrat’s area. (simple density
formula)
• Quadrats can take on many different shapes such as hexagons and
triangles, here we use square shaped quadrats to demonstrate the
procedure.
• The choice of quadrat numbers and quadrat shape can influence the
measure of local density and must be chosen with care.
 Quadrats
• Quadrat regions do not have to take on a uniform pattern across
the study area.
• Quadrats can be based on a covariate.
For example, if it’s believed that the underlying point pattern process is driven by
elevation, quadrats can be defined by sub-regions such as different ranges of
elevation values. This can result in quadrats having non-uniform shape and area.
 Tessellation
• Converting a continuous
field into discretized areas
is sometimes referred to
as tessellation. The end
product is a tessellated
surface.
• If the local intensity changes
across the tessellated covariate,
then there is evidence of a
dependence between the process
that generated the point pattern
and the covariate.
• In our example, sub-regions 1
through 4 have surface areas of
17.08, 50.45, 26.76, 5.71 map units
respectively. To compute these
regions’ point densities, we simply
divide the number of points by
the respective area values.

3/17.08 = 0.17
 Relationship of Elevation and Density
• We can plot the relationship between point density and elevation regions to help assess any
dependence between the variables.
•

Though there is a steep increase in density at the highest elevation range, this increase is not
monotonic across all ranges of increasing elevation suggesting that density may not be explained by
elevation alone.
 • It’s important to note that how one chooses to tessellate a surface can have an
influence on the resulting density distribution. For example, dividing the elevation
into equal area sub-regions produces the following density values.

While the high density in the western part of the study area remains, the density values to the east are
no longer consistent across the other three regions.
 Density-based Point Pattern Measures
Kernel-density estimation
• -Pattern has a density at any location in the study region
• -Good for hot-spot analysis, checking first-order stationary process,
and linking point objects together geographic data.
 Kernel Density Function
• Kernel density estimation (KDE) is a non-parametric way to
estimate the probability density function of a random variable.
• Kernel density estimation is a fundamental data smoothing
problem where inferences about the population are made, based
on a finite data sample.
• The Kernel Density tool calculates the density of features in a
neighborhood around those features. It can be calculated for
both point and line features.
 Kernel Density
• The kernel density approach is an extension of the quadrat method: Like the
quadrat density, the kernel approach computes a localized density for subsets
of the study area, but unlike its quadrat density counterpart, the sub-regions
overlap one another providing a moving sub-region window.
• This moving window is defined by a kernel. The kernel density approach
generates a grid of density values whose cell size is smaller than that of the
kernel window.
• Each cell is assigned the density value computed for the kernel
window centered on that cell.
 Kernel Density
• A kernel not only defines the
shape and size of the window,
but it can also weight the points
following a well defined kernel
function. The simplest function is
a basic kernel where each point
in the kernel window is assigned
equal weight.
Example:
X= 1.5, Y = 8.5
1/9 = 0.11
 Kernel Density
• Some of the most popular kernel
functions assign weights to points
that are inversely proportional to
their distances to the kernel
window center. A few such kernel
functions follow
a gaussian or quartic like
distribution function. These
functions tend to produce a
smoother density map.

Closer points are weighted

more heavily.
 Kernel Density Adjusted for Covariate
• Here, instead of dividing the study region into discrete sub-regions (as was done with
quadrat analysis), we create an intensity function that is dependent on the
underlying covariate. This function, which we’ll denote as ρ, can be estimated in one of
three different ways:
– by ratio,
– re-weight and
– transform methods.
We will not investigate into the differences between these methods, but note that there is
more than one way to estimate ρ in the presence of a covariate.
• In the following example, the elevation raster is used as the covariate in the ρ function
using the ratio method. The right-most plot maps the modeled intensity as a function of
elevation.
 Scatter Plot
• We can compare the modeled
intensity function to the kernel
density function of the
observed point pattern via a
scatter plot.

A red one-to-one diagonal line is added to the plot. While an increase in predicted intensity is
accompanied with increasing observed intensity, the relationship is not linear. This can be
explained by the small area covered by these high elevation locations which result in fewer
observation opportunities and thus higher uncertainty for that corner of the study extent. This
uncertainty is very apparent in the ρ vs. elevation plot where the 95% confidence interval envelope
widens at higher elevation values (indicating the greater uncertainty in our estimated ρ value at
those higher elevation values).
 Distance Based Methods
• Distance based methods for pattern analysis whereby the interest
lies in how the points are distributed relative to one another (a
second-order property of the point pattern) as opposed to how the
points are distributed relative to the study extent.

Three distance based approaches are covered next: The average nearest
neighbor (ANN), the G, F and K function.
 Average Nearest Neighbor

• An average nearest neighbor

(ANN) analysis measures the
average distance from each
point in the study area to its
nearest point. In the following
example, the average nearest
neighbor for all points is 1.52
units.
 Distance-Based point Pattern Measures
Second-order effect
• Nearest-neighbor distance
-The distance from an event to the nearest event in the point pattern.
• Mean nearest-neighbor distance
-Summarizes all the nearest-neighbor distances by a single mean value
-Throws away much of the information about the pattern.
 Distance-Based point Pattern Measures
• G Function
-Simplest
-Examines the cumulative frequency distribution of the nearest-
neighbor distances
-The value of G for any distance tells you what fraction of all the
nearest-neighbor distances in the pattern are less than that distance.
 Distance-Based point Pattern Measures
• F function
-Point locations are selected at random in the study region and
minimum distance from point location to event is determined.
-The F function is the cumulative frequency distribution.
-Advantage over G function: Increased sample size for smoother
curve
 Distance-Based point Pattern Measures
• K function
-Based on all distances between events
-Provides the most information about the pattern
 Distance-Based Point Pattern Measures
• Problem with all distance functions are edge effects
• Solution is to implement a guard zone.
 First and 2nd
Order Effects
• The concept of 1st order
effects and 2nd order
effects is an important one.
It underlies the basic
principles of spatial
analysis.
 References
• https://mgimond.github.io/Spatial/chp11_0.html
• https://slideplayer.com/slide/9107423/
• https://www.slideshare.net/Tolasa_F/spatial-analysis-and-modeling