GGR Lecture 6 - Spatial Analysis PDF

GGR Lecture 6 - Spatial Analysis Spatial Analysis - Going beyond visually looking at maps and formalizing the approach to exploring relationships and patterns in geographic data - Powerful, answers questions about how access to transit and income correlate, food deserts and soci...

GGR Lecture 6 - Spatial Analysis Spatial Analysis - Going beyond visually looking at maps and formalizing the approach to exploring relationships and patterns in geographic data - Powerful, answers questions about how access to transit and income correlate, food deserts and socioeconomics correlate, compares different aspects of the environment (eg. the difference in wines in different places), redline areas and environmental hazards - Some spatial analyses are concerned with looking at spatial patterns in particular objects or events (eg. are trees clustered around the city?) - Other methods look at whether like objects or events are more near or far from each other (eg. look where things are in space AND how far or close things are) - Methods can examine a single variable or they may be multivariate This can change throughout, you might start with one variable, but then add in a layer into the same work flow Looking for Patterns in Spatial Data - The human is brain is very good at pattern recognition - analysis helps confirm what we see - Eg. Surface Temperature and Income - we see a pattern that there is higher income in areas in the cooler areas and lower income in areas that are water - this starts to suggest there may be a correlation between surface temperature and income, use spatial analysis to determine whether the correlation actually exists 4 Types of Spatial Analysis - Point Pattern Analysis: Analysis of the spatial arrangement of points with a single theme to answer the key question of how locations of objects/events in space vary in relationship to other objects/events Always just a single theme !!!!! Eg. just shark bites, not whale attacks as well Overall we want to know where things happen, so that we can think about why they happened First step of analysis Types of Point Patterns  - Uniform Pattern: points are evenly scattered throughout space, underling process causes regular dispersion (16 or less) - Approx. Random Pattern: No appreciable forcing process, a mix of clustering and dispersion (17 to 45) - Clustered Pattern: Points are clustered in space, underlying process leads to clustering (46 or more) Eg. ballot boxes in Georgia, are they clustered, more spread out, or random? (clustered around bigger cities with more people) Pattern depends on where we look - eg. Global Patterns: look at the entire study area, Local Patterns: test a specific subset of a study area - global may show a clustered pattern, whereas local may show them more dispersed We can also see how point pattern locations change over time and compare if they are more dispersed or clustered than before - Autocorrelation Analysis: Analysis that considers both location and a single attribute to determine if like values are clustered together or if they are more dispersed - is there a pattern where values that are systematically alike are morecloser? Whether or not a pattern is different from what you would expect if the pattern was random Whether or not a data set is in line with Tobler’s First Law of Geography Tobler’s First Law of Geography: everything is related to everything else, but near things are more related than distant things - Exceptions: places where there are hard boundaries, where there is irregularities in geography Spatial Autocorrelation - Negative Spatial Autocorrelation: Similar values are scattered throughout spaces, things closer together likely to have different values, underlying process causes regular dispersion - Approx. Random Pattern: No appreciable forcing process, mix of clustering and dispersion  - Positive Spatial Autocorrelation: Similar values are clustered in space, things closer together likely yo have same values, underlying process leads to clustering Positive spatial correlation, consistent to Tobler’s first law * we can also use polygons, doesn’t necessarily have to be points for spatial autocorrelation Consistent with tobler’s first law *how clustered are they? ← next question using tools not covered in this course - Proximity Analysis: Analysis that focuses on the spatial relationships between two themes or two types of data Eg. relationship between burglaries and police stations Classic Example: John Snow’s Cholera Map: looked at the location of water pump in a neighbourhood, found that cases clustered around one pump → led to the question of “is this disease due to the water pump?” , turned off water pump and cases went down Accessibility Analysis:  - Euclidean Distance: Straight line distance between two points, the “as the crow flies” distance - Network Distance: Distance you must travel over a transportation network to reach somewhere - what paths are available? - Manhattan Distance: based on a grid, a form of network distance where you can move left and right, up and down, but you can’t move diagonal - Correlation Analysis: Analysis that focuses on describing the spatial relationships between multiple attributes or theme Remember correlation does not equal causation Street Tree Density and Household income - there may be a relationship, correlation analysis tell us if there actually is a relationship - Test looks for clusters of extreme variables in both, it did find that there was clusters in areas with high density of trees and higher income and the opposite Limitation to Correlation: Interoperability - Interoperability