Geostatistics and GIS
This Phd course is an introduction to the analysis, description and
modelling of geographical referenced data or spatial data in
short. The practical applications of software tools, underlying
theory, and the correct application of these tools is emphasized
through the use of software to analyze and model spatial data. The
software used is Geostatistical Analyst and Spatial Analyst in ArcGIS and packages geoR and gstat for R.
Prerequisites for the course are an introductory statistics course,
an introductory GIS course, and a matrix algebra course.
The course covers nine days, see schedule below.
Location and dates
The course start 12 April 2010 and takes places over four weeks in April+May 2010.
The location is the DJF building in
Aarhus. Willemoesgade 15, 8200 Aarhus N, building 2114, rooms 168 and 166
(see here and here).
The following subjects will be covered during the course in greater or
lesser details as time permits. See below for a detailed course
Overview, Course Topics and Case Study: Overview
of applications and techniques to be covered in the course: univariate
and multivariate statistics, spatial correlation analysis, modelling,
Exploratory Data Analysis: Statistical analysis,
vs. summarization, mapping, representing spatial data, continuous vs.
categorical data, histograms and probability distibutions,
cross-correlation in multivariate data, data transformations
(logarithmic, square root, etc), software use and
Spatial Correlation: Spatial
correlation and associated statistical measures, calculation of
experimental variograms, fitting model autocorrelation functions,
variogram anisotropy and nested structures, indicator variograms,
crosscorrelation (spatial co-variability of multiple variables).
Spatial Estimation (Kriging): Techniques for
spatial estimation, 'best' linear unbiased estimation, the kriging
system of equations, use and misuse of kriging variance, sensitivity
of kriging estimation to variogram structure, kriging strategy,
cross-validation of spatial data, co-kriging.
Stochastic Simulation: Simulation vs. kriging,
Geographical Information Systems and
Geostatistics: The use of GIS and geostatistics via
ArcGIS Geostatistical Analyst.
Model based geostatistics: Maximum likelihood, REML,
the Gaussian assumption (important or not), kriging with external trend,
generalised least square (GLS) estimator.
Probability map of Threshold exceedance.
Most of this course is based on the book by
Webster and Oliver (2007).
Material about exercises in Arc-map and geoR will be handed out.
- Webster, R. and Oliver, M. A. (2007), Geostatistics for
Environmental Scientists (second edition), J. Wiley & Sons, New York.
- Højsgaard, S. and Halekoh, U. (2009-2010). An introduction to R in the biological sciences
- Diggle, P. J. and Ribeiro Jr, P. J. (2007), Model-based Geostatistics, Springer, New York.
Computers and Software
The room we are using contains computers with dual screans.
Arc-map (and Arc-GIS) will be used for the exercises on the first 5 days.
Possibility to install on your own labtop.
The statistical programming environment R with add-on
package geoR will be used in the later part of the course.
Participants will get detailed information on how to
use the software.
The Lecture Plan
In general the individual course day starts at 9:00 in the
morning and ends at 16.00 with a lunch break between 12.00 and 13.00.
- Mon 12 April. Morning 9-12 (room 168): Introduction to the course, presentation of participants and access to computers. Arc-map exercise about spatial
data visualisation. Afternon 13-16 (room 166): Lectures about interpolation methods. Reading: Webster and Oliver (WO) chapter 1, 2 and 3.
- Tue 13 April. Morning 9-12 (room 168): computer exercises in ArcGIS about spatial interpolation methods,
and finishing of the spatial data visual tools. Afternon 13-16 (room 168). Lectures, which introduces spatial covariance, variogram fitting and kriging.
Reading: WO, chapter 4.1-4.6, 4.9, 5.1-5.2 (except 5.2.1), 8.1-8.2
- Mon 19 April. Morning 9-12 (room 168): ArcGIS exercises about variogram fitting and kriging. Afternon 13-16 (room 166) :
Lectures about spatial covariance, variogram fitting and kriging (more in depth). We will start with the kriging exercise handed out on tuesday.
Reading: WO, chapter 4.1-4.7, 4.9, 5.1-5.6 (except 5.2.1), 8.1-8.7
- Tue 20 April. Morning 9-12 (room 168):
Lectures about cokriging. Reading: WO, chapter 10.1-10.4. Afternon (room 168) ArcGIS exercises about cokriging.
- Mon 26 April. Morning 9-12 (room 168): Computer exercises: Introduction to R and to package geoR.
Reading, R-intro (Højsgaard and Halekoh) chapter 1 and 3.
Afternon 13-16 (room 166) :
Lectures about cokriging (continued). Reading, WO chapter 10.4.
- Tue 27 April. Morning 9-12 (room 168):
Lectures about kriging and the normal distribution. Reading: WO, chapter 8.10 and 12. Afternon 13-16 (room 168): start on ArcGIS exercise (to be handed in).
- Mon 3 May. Morning 9-12 (room 168): Computer Exercises using geoR. Afternon 13-16 (room 166) : Regression kriging, REML and kriging with trend. Reading, WO, chapter 9.1-9.3.
- Tue 4 May. All day (room 168): geoR exercise (to be handed in).
- Wed 5 May. Morning (room 162): Course evalueation, continue the hand-in exercise. Afternon: cancelled
The workload amounts to 5 ECTS.
Evaluation of participants
To pass the course the participants need to take part in the lectures
by active discussion. Doing two written exercises to be
evaluated by the teachers.
The teachers on the geostatistics module are:
Ole F. Christensen
Mogens H. Greve
Enrollment can be done by sending an e-mail OleF.Christensen@agrsci.dk containing your name and affilitation.
A list is given below :
Korhan Ozkan (Bio, AU)
Lars Dalby (DMU/Bio, AU)
Roy Mathew Francis (DJF)
Kristian T. Nielsen (Bio, AU)
Kabindra Adhikari (JPM, DJF)
Gerardo Ojeda (IMAR, Universidade Coimbra, Portugal)
Allan Timmermann (DMU/Bio, AU)
Fan Deng (JPM, DJF)
Philipp Trenel (AgroTech)
Anne Mette Kjeldsen (AgroTech)
Mette Vestergaard Odgaard (JPM, DJF)