Lecture 10 - Spatial Data Analysis

Question 1

What is the objective of spatial analysis?

Answer 1

To transform data into useful info to satisfy requirements/objectives of decision-makers at all levels of detail

Question 2

What does spatial analysis help us to do?

Answer 2

• distinguishes GIS from other info systems
• identify trends in data
• create new relationships from the data
• view complex relationships b/w data sets
• make better decisions

Question 3

Why do analyses?

Answer 3

Analysis functions use spatial and non-spatial attributes in database to answer questions about real world

Question 4

Why do we do spatial analysis of massive data sets?

Answer 4

• search for patterns, anomalies, and trends
• semi-automated b/c of data volumes
• widely used in practice in business (unusual credit card usage), epidemiology (detection of disease outbreak), and climate change (data patterns lead to discovering ozone hole)

Question 5

What are descriptive summaries?

Answer 5

• attempt to summarize useful properties of data sets in one or two statistics
• mean/average widely used b/c measures central tendency

Question 6

How do you find a spatial mean?

Answer 6

Partial (2D) equivalent of the mean would be a centre. Several ways of defining centres:

• centroid
• dispersion

Question 7

What is a centroid?

Answer 7

The 2D equivalent of the mean.

• for a set of points, the centroid is found by taking weighted average of x and y coordinates
• balance point: the point about which the 2D pattern would balance if it were transferred to a weightless, rigid plane and suspended
• ex. population centroids in US

Question 8

What is dispersion?

Answer 8

A measure of spread of points around a centre (useful for determining positional error)

• w/o knowing about how data is dispersed, measure of central tendency may be misleading
• measures of dispersion include: range, average deviation, variance and standard deviation

Question 9

What are spatial patterns? How can we identify spatial patterns in the data?

Answer 9

Can be clustered, dispersed, or random.
Techniques for identifying:
- measures only looking at physical location and geometric patterns of features (ex. unlabelled points)
- measures looking at geographic patterns in attribute data of features (ex. labelled points)

Question 10

What is Tobler’s first law of geography?

Answer 10

Everything is related to everything else, but near things are more related than distant things
- fundamental concept for understanding/analyzing spatial phenomena

Question 11

What is spatial autocorrelation?

Answer 11

Correlation of a variable with itself through space

• measures and analyzes the degree of spatial dependency by comparing values of a sample with values of their neighbours
• if there is any systematic pattern in the spatial distribution of a variable, it is said to be spatially autocorrelated

Question 12

What is positive spatial autocorrelation?

Answer 12

All similar values are located close together

- may violate assumptions about independence of residuals (diff b/w observed and predicted value)

Question 13

What is negative spatial autocorrelation?

Answer 13

Dissimilar values appear in close association (neighbouring areas unalike)

Question 14

What is random patterns in spatial autocorrelation?

Answer 14

Exhibits no spatial autocorrelation

Question 15

Why is spatial autocorrelation important?

Answer 15

• most stats based on assumption that values of observations in each sample independent of one another
• most GIS provide tools to measure the level of spatial autocorrelation

Goals:

• to measure strength of spatial autocorrelation in a map
• to test the assumption of independence or randomness

Question 16

What are the main indices of spatial autocorrelation?

Answer 16

• Moran’s I
• Geary’s C
• Ripley’s K
• Join Count Analysis

Question 17

What do indices of spatial autocorrelation do?

Answer 17

Measures level of interdependence b/w variables and nature and strength of that interdependence
- tests whether the observed value of a variable at one locality is independent of the values of the variable at neighbouring localities

Question 18

How does Moran’s I work?

Answer 18

Based on p-values.
Ex. Moran’s I = 0.66, then p-value is not statistically significant and you cannot reject null hypothesis
Ex. 2 Moran’s I = 0.015, then p-value is statistically significant, so you can reject null hypothesis

Question 19

What are spatial modes?

Answer 19

Attempt to represent variation over the earth’s surface (GIS provides suitable platform)
- manipulate geographic info in multiple stages (representing either a single point in time or predictions about future points in time)

Question 20

What are the benefits of the modelling approach?

Answer 20

• allows experiments to be conducted on simulated systems rather than on the real thing (cheaper and less invasive)
• allows alt scenarios to be evaluated (compare diff policy options and their impacts on the future; planners can experiment “what-if” scenarios)

Question 21

What are the different types of spatial models?

Answer 21

1. static models and indicators
2. individual and aggregate models
3. cellular models

Question 22

What are static models and indicators?

Answer 22

• represents a system at a single point in time
• take multiple GIS inputs and compute useful indices
• ex. groundwater protection model

Question 23

What are individual and aggregate models?

Answer 23

Individual: simulates behaviour of every individual in system (ex. every person in a crowd)
Aggregate: used when too many individual elements to model (ex. impossible to model every molecule of water, grain of sand, etc.)
- ex. simulation of movement of individuals during a parade

Question 24

What are cellular models?

Answer 24

• model a system based on using raster GIS
• each cell in the raster can be in one of a number of states (ex. change through time is represented by change of cell state)
• ex. urban growth simulation where each cell is either undeveloped or developed

Question 25

What are multicriteria methods?

Answer 25

• valuable decision-making tool (most widely used analytical facility in GIS)
• many decisions depend on numerous factors
• these methods attempt to reconcile such differences and to reach consensus on what factors should be combined (stakeholder disagreements)

Question 26

What is decision-making defined as?

Answer 26

The process of choosing between a set of alternatives by analyzing and interpreting geographical info related to alts

Question 27

What are the decision-making process steps?

Answer 27

1. defining decision problem (objective)
2. determining set of evaluation criteria to be used
3. weighting criteria (b/c multiple criteria have varying importance)
4. generating alternative solutions
5. apply decision making
6. recommend the best solution to the problem

Question 28

What is multi-criteria evaluation (MCE)? What are the techniques used?

Answer 28

Involves overlaying thematic layers and finding locations that encompass all desired criteria
Techniques:
- Boolean intersection
- weighted linear combination (WLC)

Question 29

What are Boolean intersection multi-criteria evaluations?

Answer 29

• uses logical operators to find combos of layers

- only locations that are characterized as suitable (value of 1) on all maps will be suitable in the result

Question 30

What are weighted linear combination multi-criteria evaluation?

Answer 30

• assesses the suitability of locations by weighting and combining maps
• multiply thematic layer by factor weight and add across layers
• weights are positive or zero and sun equal to 1
• resulting cell values are in the same range as input maps

Question 31

How do you assign weights for WLC MCE?

Answer 31

• each stakeholder compares each pair of factors (indicating relative importance as a ratio; focus is on inputs rather than alt solutions themselves)
• ratings combined to produce a consensus set of weights along with a measure of the strength of agreement of disagreement