Geographic Information Systems 

Intro  Techniques  Spatial  Geostatistics  Geocoding  GIS software

 Geostatistics

Using Geostatistics to predict fields from points. Point pattern analysis. A way of looking at the statistical properties of spatial data. What makes it unique from other kinds of statistics is the use of graph theory and matrix algebra to reduce the number of parameters in the data being analyzed. This is necessary because it is actually the second-order properties of the GIS data that need analyzing.

When we measure any phenomena, our observation methods dictate the accuracy of any subsequent analysis. Whether our study is concerned with the nature of traffic patterns in an urban core, or with the analysis of weather patterns over the Pacific, there will always contain a variable or a degree of precision which escapes our measurement; this is determined directly by the scale and distribution of our data collection, or survey methods. In order to apply statistical relevance to spatial analysis, an 'average' must be determined so that points, or gradients, outside of any immediate measurement may be included as to their predicted behavior. Limitations in statistics and data collection mean that it is impossible to directly measure a continuum without the inferential methods of analysis, of which, several forms of interpolation are used in order to predict the behavior of particles and locations not directly measured.


Hillshade model derived from a Digital Elevation Model (DEM) of the Valestra area in the northern Apennines (Italy)Interpolation is the process by which a surface is created, usually a raster dataset, through the input of data collected at a number of sample points. There are several forms of interpolation, each which treats the data differently, depending on the properties of the dataset. In comparing interpolation methods, the first consideration should be whether or not the source data will change (exact or approximate). Next is whether the method is subjective, a human interpretation, or objective. Then there is the nature of transitions between points: are they abrupt or gradual. Finally, there is whether a method is global (it uses the entire dataset to form the model), or local where an algorithm is repeated for a small section of terrain.

Digital Elevation Models (DEM), Digital terrain models (DTM), triangulated irregular networks (TIN), Edge finding algorithms, Theissen Polygons, Fourier analysis, Weighted moving averages, Inverse Distance Weighted, Moving averages, Kriging, Spline, Trend surface analysis.

Regionalized variable theory

Spatial Autocorrelation Principle: Data collected at any position will have a greater similarity to, or influence on, those locations within its immediate vicinity.