4. the nature of geographic data
TRANSCRIPT
Geographic Information Systems and ScienceSECOND EDITIONPaul A. Longley, Michael F. Goodchild, David J. Maguire, David W. Rhind© 2005 John Wiley and Sons, Ltd
4. The Nature of Geographic Data
© John Wiley & Sons Ltd
© 2005 John Wiley & Sons, Ltd
Overview: Spatial is SpecialThe 'true' nature of geographic dataThe special tools needed to work with themHow we sample and interpolate (gaps)What is spatial autocorrelation, and how can it be measured?Fractals and geographic representation
© 2005 John Wiley & Sons, Ltd
Why GIS?Small things can be intricateGIS can:
Identify structure at all scalesShow how spatial and temporal context affects what we do
Allows generalization and accommodates errorAccommodates spatial heterogeneity
© 2005 John Wiley & Sons, Ltd
Building RepresentationsTemporal and spatial autocorrelationUnderstanding scale and spatial structure
How to sample How to interpolate between observations
Object dimensionsNatural vs. artificial units
© 2005 John Wiley & Sons, Ltd
Spatial AutocorrelationSpatial autocorrelation is determined both by similarities in position, and by similarities in attributes
Sampling intervalSelf-similarity
© 2005 John Wiley & Sons, Ltd
© 2005 John Wiley & Sons, Ltd
Spatial SamplingSample framesProbability of selectionAll geographic representations are samplesGeographic data are only as good as the sampling scheme used to create them
© 2005 John Wiley & Sons, Ltd
Sample DesignsTypes of samples
Random samplesStratified samplesClustered samples
Weighting of observations
© 2005 John Wiley & Sons, Ltd
© 2005 John Wiley & Sons, Ltd
Spatial InterpolationSpecifying the likely distance decay
linear: wij = -b dij
negative power: wij = dij-b
negative exponential: wij = e-bdij
Isotropic and regular – relevance to all geographic phenomena?
Inductive vs. deductive approaches
© 2005 John Wiley & Sons, Ltd
Spatial Autocorrelation MeasuresSpatial autocorrelation measures:
Moran; nature of observations
Establishing dependence in space: regression analysis
Y = f (X1, X2 , X3 , . . . , XK)Y = f (X1, X2 , X3 , . . . , XK) + εYi = f (Xi1, Xi2 , Xi3 , . . . , XiK) + εi
Yi = b0 + b1 Xi1 + b2 Xi2 + b3 Xi3 + . . . bK XiK + εi
© 2005 John Wiley & Sons, Ltd
© 2005 John Wiley & Sons, Ltd
© 2005 John Wiley & Sons, Ltd
© 2005 John Wiley & Sons, Ltd
© 2005 John Wiley & Sons, Ltd
Functional formThe assumptions of inference
Tobler’s LawMulticollinearity
© 2005 John Wiley & Sons, Ltd
Discontinuous VariationFractal geometry
Self-similarityScale dependent measurementRegression analysis of scale relations
© 2005 John Wiley & Sons, Ltd
ConsolidationInduction and deductionRepresentations build on our understanding of spatial and temporal structuresSpatial is special, and geographic data have a unique nature