Thunderstorms over South
Carolina seen from NASA’s
ER-2 aircraft at 19,800
metres during the Integrated
Precipitation and Hydrology
Experiment. source
commons.wikimedia.org
by NASA on the commons

Spatial analysis methods

(from 50,000 feet)

Multiple attributes methods

Overlay and friends

Regression and friends

Clustering

Single attribute methods

Characterizing patterns

Interpolation

Spatial relations

image Ian McHarg's Design with Nature 1969
source suzanneodonovan.wordpress.com

Multiattribute methods

[ones we’ve already looked at]

Overlay and regression

Various (spatial) regressions

Clustering

source Essentials of Geographic Information Science by Jonathan Campbell and Michael Shin, online textbook

Single attribute methods

For many people, multiple layers are the heart of GIS

But we can also learn a lot within layers

Specifically

• Characterizing patterns

• Inferring unknowns

• Exploring relationships

Characterizing patterns

In the hope of
getting at processes

Libyan Sahara dune fields, image wikimedia.commons by David Stanley

Spatial autocorrelation

We can measure the extent to which “near things are more related than distant things”

Tools: GeoDa ≫ R (spdep) > ArcGIS

Point pattern analysis

A wide ranging set of approaches to characterizing spatial structure in point patterns

Tools: R (spatstat) ≫ ArcGIS

image source commons.wikimedia.org by Famartin

First order trends

Patterns follow a trend, e.g., in plant ecology

Environmental gradients lead to increasing prevalence

...or in urban settings

Population density drives many things

Incidence of crime

Cases of disease in an epidemic

Presence of cafes, dairies, etc.

Second order interactions

Competition or ‘mutualism’

Competition suppresses close neighbours, leads to even spacing

Mutualism leads to clusters

image source commons.wikimedia.org
Adelie penguin colony, Cape Royds
Ross Island, Antarctica

image source i.warosu.org
Namibian savannah

Competition

Mutualism

image source independent.co.uk
Vinyl Junkies record store
Soho, London

Growth Pattern 19 by Paul Merryman

Also

Network analysis methods

Measurement of segregation

A wide range of surface measures

Fragmentation statistics

...

Inferring patterns

Commonly known as (spatial) interpolation

We have control points, but not at all locations

Interpolation methods fill in the gaps

Output from the SYMAP system
source krygier.owu.edu

Simple methods

image source commons.wikimedia.org

Use the ‘first law’ in combination with weighted means

  • Inverse-distance weighting,
  • natural neighbors,
  • Voronoi polygons

Tools: ArcGIS ≈ R (various libraries)

Geostatistics

aka kriging

Use a mathematical/statistical model of spatial
structure found in the control points to enhance results

 

source Figure 10.16 in O’Sullivan & Unwin

image source
openflights.org

Spatial relations

All of the above is concerned with spatial relations of
  • distance,
  • adjacency,
  • connectivity,
  • interaction,
  • etc.

These per se may be our focus

source: Bunge W. 1962. Theoretical Geography. Gleerup, Lund, Sweden.

mapnificent.net

Tools: recently, Arc may have become quite good at this

image source diagrams.org
© Transport for London

This space is wide open with scope for interesting maps and other visualizations

Tools: check out Scapetoad, plugins in QGIS, GeoDa and some R packages

We’ll look at
aspects of this
under the topic of
network analysis
in a later lecture

Orange cranberry cake by Helen Fletcher on flickr.com

Summary

There’s just as much happening within layers as between them

Any of these approaches might be of interest in the mini-projects (just ask!)

These topics are central to one of the core courses in postgrad (in case you are interested!)