Objective: To explore the various techniques for spatial interpolation
General description: Visiting every location in a study area to measure the height, magnitude, or concentration of a phenomenon is usually difficult or expensive. In stead, one can select strategically dispersed sample input point locations and interpolate the surface to assign an estimated value to all the other locations. Input data points can be either irregularly (somewhat randomly) or regularly spaced points of the measurement. The resulting grid is an attempt to provide a best estimate of what quantity is on the actual surface for each location. The surface interpolators make certain assumptions about how to determine the best estimated values. Based on the phenomena the values represent and on how the sample points are distributed, different interpolators will produce different estimates in relation to the actual ‘true’ values.
A series of different interpolators will be discussed: polynomial trend surface, LOESS, inverse distance weighting, bilinear, cubic splines and kriging. Each technique will be explained within the context of how they are implemented in the R environment. In addition some discussion will be devoted to cross validation. A tutorial on spatial interpolation using the gstat package primarily will follow the lecture/discussion portion of the afternoon session.
Required back-ground knowledge: basic statistic and some knowledge of kriging
Software / R packages required: gstat; akima; fields; raster; tripack; stats; lattice.
- chapter 8 "Interpolation and Geostatistics" in Bivand, R., Pebesma, E., Rubio, V., (2008) Applied Spatial Data Analysis with R. Use R Series, Springer, Heidelberg, pp. 378.
- Hengl, T. (2009) A Practical Guide to Geostatistical Mapping, 2nd Ed. University of Amsterdam, www.lulu.com, 291 p.