NSCPHD1050 Uncertainty Propagation in Spatial Environmental Modelling

PLEASE NOTE         

The PhD course database is under construction. If you want to sign up for this course, please click on the link in order to be re-directed. Link: https://phdcourses.ku.dk/nat.aspx

Scope

Input data for spatial environmental models may have been measured in the field or laboratory, derived from remotely sensed imagery or obtained from expert elicitation. Data are also often digitised, interpolated, classified or generalised prior to submission to a model. In all these cases errors are introduced. Although users may be aware that errors propagate through their models, they rarely pay attention to this problem. However, when the accuracy of the data is insufficient for the intended use then this may result in inaccurate model results, wrong conclusions and poor decisions.

The purpose of this course is to familiarise participants with statistical methods to analyse uncertainty propagation in modelling, such that they can apply these methods to their own data and models. The emphasis is on Monte Carlo simulation methods. Attention is also given to geostatistics for quantification of spatial interpolation errors and on the effects of spatial auto- and cross-correlations on the results of an uncertainty propagation analysis. The course also addresses methods to determine the relative contribution of individual sources of uncertainty to the uncertainty of the final result. Quantification of model parameter uncertainty is covered using Bayesian calibration techniques.

The methodology is illustrated with real-world examples on heavy metal pollution of the soil, on the flooding of a river area in the Netherlands after a dike break, and on calculation of soil hydrological parameters using a combination of linear regression and kriging. The uncertainty analysis of the examples is largely carried out by the course participants themselves in computer practicals. For this, use is made of the R language for statistical computing.

After completing this course, participants will have a clear understanding of how uncertainties in spatial information can be represented statistically using probability distributions, how uncertainties propagate through spatial analyses, and how to apply uncertainty propagation techniques in their own work.

The target group for the course are PhD students with an interest in uncertainty assessment of environmental data and models. In addition, researchers and professionals with the same interests may also benefit from following the course.

Background of participants

Participants are expected to have an intermediate understanding of statistics and basic understanding of environmental science and geo-information science. Familiarity with the R‑programming language is preferred but not required.

Study material

Relevant literature, lecture materials and computer practical exercises and answers will all be made available digitally to the participants.

Programme

DAY 1, Tuesday 10 November

9.00      – 9.15       Welcome, introduction

9.15      – 10.45     Lecture 1 (problem definition, what is uncertainty, how can it be represented statistically, identification of uncertainty model)

10.45     – 11.15     Coffee/tea break

11.15     – 12.30     First halve of course participants present their research in max five slides or poster

12.30     – 13.30     Lunch

13.30     – 15.30     Lecture 2 (geostatistics, semivariogram, kriging, spatial stochastic simulation)

15.30     – 16.00     Coffee/tea break

16.00     – 17.30     Computer practical geostatistics (apply theory and methods from the lecture to a real world case on soil pollution in the Geul river valley)

19.00     – 22.00     Joint dinner

DAY 2, Wednesday 11 November

9.00      – 10.00     Lecture 3 (Uncertainty propagation with the Taylor series method)

10.00     – 10.45     Computer practical Taylor series method (apply theory and methods from lecture 3 to case of lead ingestion in Geul river valley)

10.45     – 11.15     Coffee/tea break

11.15     – 12.30     Second halve of course participants present their research in max five slides or poster

12.30     – 13.30     Lunch

13.30     – 14.30     Computer practical Taylor series method continued

14.30     – 15.30     Lecture 4 (Uncertainty propagation with the Monte Carlo method)

15.30     – 16.00     Coffee/tea break

16.00     – 17.30     Computer practical Monte Carlo method (apply theory and methods from lecture 4 to case of soil lead concentration in Geul river valley)

DAY 3, Thursday 12 November

9.00      – 10.45     Assignment on Monte Carlo uncertainty propagation of a flood event following a dike break in the Betuwe area in the Netherlands

10.45     – 11.15     Coffee/tea break

11.15     – 12.30     Assignment continued

12.30     – 13.30     Lunch

13.30     – 15.30     Assignment continued

15.30     – 16.00     Coffee/tea break

16.00     – 16.30     Assignment feedback

16.30     – 17.30     Uncertainty game, with drinks and snacks

DAY 4, Friday 13 November

9.00      – 10.45     Lecture 5 (Bayesian calibration for model parameter uncertainty assessment, including explanation of Markov chain Monte Carlo)

10.45     – 11.15     Coffee/tea break

11.15     – 12.30     Computer practical Bayesian calibration

12.30     – 13.30     Lunch

13.30     – 15.00     Computer practical Bayesian calibration continued

15.00     – 16.00     Remaining topics, Evaluation, Closing

Teachers

Dr. Gerard B.M. Heuvelink holds an MSc in Applied Mathematics of Twente Technical University and a PhD in Environmental Sciences of Utrecht University. He was assistant professor in Geostatistics and Stochastic Simulation with the University of Amsterdam and worked as senior researcher in Geostatistics with Alterra, Wageningen. He is currently employed as associate professor in Geostatistics with the Soil Geography and Landscape group of Wageningen University and senior researcher Pedometrics and Digital Soil Mapping with ISRIC World Soil Information. He is also a visiting professor of the Institute of Geographical Sciences and Natural resources Research, Chinese Academy of Sciences. Dr. Heuvelink has written over 220 scientific publications on geostatistics, spatial uncertainty analysis and pedometrics, about 95 of which appeared in peer-reviewed international journals. He has been involved in many research projects dealing with spatial uncertainty in environmental modelling and spatial analysis and is worldwide recognised as a leading scientist in pedometrics and spatial uncertainty analysis. Dr. Heuvelink is associate editor of Spatial Statistics and the European Journal of Soil Science, and editorial board member of Geoderma, Environmental and Ecological Statistics, International Journal of Applied Earth Observation and Geoinformation and Geographical Analysis. In 2014 he was awarded with the Richard Webster medal of the Pedometrics Commission of the International Union of Soil Science.

Dr. Sytze de Bruin holds an MSc in Soil Science and a PhD in Geo-information Science of Wageningen University. After obtaining his MSc he worked four years in Central America (Costa Rica and Nicaragua) as an applied soil scientist. Since his return to The Netherlands he has worked for Wageningen University where he is currently employed as associate professor in Geographical Information Science within the Laboratory of geo-information science and remote sensing. His research and education focus on using sound methodology for transforming spatial data into useful geo-information. He is particularly interested in uncertainty analysis to assess fitness-for-purpose, data acquisition, including spatial and temporal sampling and sensing, spatio-temporal interpolation and other quantitative methods used in spatial and temporal analysis (e.g. time series analysis). He has been involved in several research projects with applications ranging from land degradation assessment to precision agriculture. Sytze de Bruin has written over 40 papers published in peer-reviewed international journals. He is associate editor of the International Journal of Geographical Information Science and he serves on the editorial board of Spatial Statistics.