The application for the free-of-charge NERC-funded Bayesian training in January and March 2024 has just closed.
(documents related to the NERC-funded course are at the bottom of this page)
UKCEH Edinburgh, Bush Estate, Penicuik, Midlothian, EH26 0QB, Location map & travel directions
UKCEH Wallingford, Maclean Building, Benson Lane, Crowmarsh Gifford, Wallingford, Oxfordshire, OX10 8BB Location map & travel directions
On-demand online course
We are exploring a number of options to continue this training offer beyond March 2024. This includes a paid-for face-to-face and on-demand course
On-demand course £499 (tbc)
Face-to-face course from £799 (tbc)
summer/autumn 2024 (tbc)
The face-to-face courses run from 1 pm on Monday to 1 pm on Friday.
Short course description:
This interactive 5-day course will be a mixture of presentations and practical exercises. It will give you a solid grounding in Bayesian methods that you can use with any kind of model and data set to compare models, estimate parameters, analyse uncertainties and communicate results.
The course will use examples of models and data from natural environment research. However, you will be able to transfer the methods and techniques to any other model, data set or research question. There will be time during practical exercises where you can opt to use the materials provided by the trainers, or use your own model(s) and data.
By the end of the course, learners will be able to
- Understand scientific papers that use Bayesian methods and terminology
- Know how to use Bayesian software
- Design and apply simple Bayesian statistical models
- Know how to combine disparate data sets in a single analysis
- Compare the relative plausibility of different models
- Comprehensively analyse uncertainties associated with model outputs
- Communicate uncertainties clearly
- PhD students
- Early career researchers
- Academics/ researchers, including those from industry or the charitable sector
Learners need to have some practical knowledge of the programming language R [ https://www.r-project.org/about.html ]. No previous exposure to probability theory or Bayesian methods is necessary, but students may want to read some introductory material beforehand. We recommend the reading listed below:
Hardware and software requirements:
All data analysis and modelling will use the freely available language R. You will need to bring your own laptop with R installed. We can provide a limited number of laptops (if you only have a desktop machine). We can provide guidance on how to install R on your machine. Our Bayesian analyses will mostly use the freely available software JAGS (which can be called from R) and some self-written code. The course leaders will prepare R-scripts beforehand. We will briefly illustrate some alternative software to JAGS, including BUGS and INLA.
The cost of accommodation is not included in the course fee of future courses.
Peter Henrys, Statistician, UKCEH
David Cameron, Ecosystem modeller, UKCEH. David has over 13 years of experience in the application of Bayesian methods in environmental and ecological research. For the last three years he has been one of the lecturers at the spring school of COST Action PROFOUND “Bayesian calibration, forecasting and multi-model predictions of process-based vegetation models” in Grenoble, France (http://cost-profound.eu/site/event/3rd-profound-spring-school-bayesian-calibration-forecasting-and-multi-model-predictions-of-process-based-vegetation-models-in-rencurel-grenoble-france-5-10-of-may/)
Peter Levy, Ecosystem modeller, UKCEH. Peter has worked on plant physiology, carbon sequestration, and the exchanges of greenhouse gases between ecosystems and the atmosphere. In recent years, he has applied Bayesian approaches to the analysis of these data, for example in estimating the uncertainty in cumulative gas emissions.
Lindsay Flynn Banin, Statistical Ecologist, UKCEH
Previous course participants said:
"Well-considered course structure and content – interactive and friendly with a relaxed teaching style that meant we could easily discuss issues along the way.Very good at explaining the Bayesian approach and how it can be applied. Also – fantastic resources and R code provided via GitHub!"
(learner, September 2019)
Recommended reading before the course:
1) Hartig et al. (2012. J. Biogeogr. 39: 2240-2252; https://onlinelibrary.wiley.com/doi/epdf/10.1111/j.1365-2699.2012.02745.x)
2) Van Oijen (2017). Curr. For. Rep. 3: 269-280; https://doi.org/10.1007/s40725-017-0069-9 and http://nora.nerc.ac.uk/id/eprint/518404/
Selected publications by the trainers that make use of Bayesian methods:
Levy, P., Van Oijen, M., Buys, G. & Tomlinson, S. (2018). Estimation of gross land-use change and its uncertainty using a Bayesian data assimilation approach. Biogeosciences 15: 1497-1513. https://doi.org/10.5194/bg-15-1497-2018.
Bagnara, M., Van Oijen, M., Cameron, D., Gianelle, D., Magnani, F. & Sottocornola, M. (2018). Bayesian calibration of a simple forest model with a multiplicative mathematical structure: a case study with a Light Use Efficiency model in an alpine forest. Ecological Modelling 371: 90-100. https://doi.org/10.1016/j.ecolmodel.2018.01.014.
Levy, P., Cowan, N., Van Oijen, M., Famulari, D., Drewer, J. & Skiba, U. (2017). Estimation of cumulative fluxes of nitrous oxide: uncertainty in temporal upscaling and emission factors. Eur. J. Soil Sci. 68: 400-411. https://doi.org/10.1111/ejss.12432 .
Höglind, M., Van Oijen, M., Cameron, D. & Persson, T. (2016). Process-based simulation of growth and overwintering of grassland using the BASGRA model. Ecological Modelling 335: 1-15. https://doi.org/10.1016/j.ecolmodel.2016.04.024 .
Reyer, C., Lasch, P., Flechsig, M. & Van Oijen, M. (2016). Integrating parameter uncertainty of a process-based model in assessments of climate change effects on forest productivity. Climatic Change 137: 395-409. https://doi.org/10.1007/s10584-016-1694-1 .
Minunno, F., Van Oijen, M., Cameron, D. & Pereira, J.S. (2013). Selecting parameters for Bayesian calibration of a process-based model: a methodology based on canonical correlation analysis. J. Uncertainty Quantification 1: 370–385 < http://dx.doi.org/10.1137/120891344>.
Cameron, D.R., Van Oijen, M., Werner, C., Butterbach-Bahl, K., Grote, R., Haas, E., Heuvelink, G., Kiese, R., Kros, J., Kuhnert, M., Leip, A., Reinds, G.J., Reuter, H.I., Schelhaas, M.J., de Vries, W. & Yeluripati, J. (2013). Environmental change impacts on the C- and N-cycle of European forests: a model comparison study. Biogeosciences 10: 1751-1773. http://www.biogeosciences.net/10/1751/2013/bg-10-1751-2013.html
Minunno, F., Van Oijen, M., Cameron, D., Cerasoli, S., Pereira, J.S. & Tomé, M. (2013). Using a Bayesian framework and global sensitivity analysis to identify strengths and weaknesses of two process-based models differing in representation of autotrophic respiration. Environmental Modelling & Software 42: 99-115.
Van Oijen, M., Reyer, C., Bohn, F.J., Cameron, D.R., Deckmyn, G., Flechsig, M., Härkönen, S., Hartig, F., Huth, A., Kiviste, A., Lasch, P., Mäkelä, A., Mette, T., Minunno, F. & Rammer, W. (2013). Bayesian calibration, comparison and averaging of six forest models, using data from Scots pine stands across Europe. Forest Ecology and Management 289: 255-268.
Van Oijen, M., Cameron, D.R., Butterbach-Bahl, K., Farahbakhshazad, N., Jansson, P.-E., Kiese, R., Rahn, K.-H., Werner, C., Yeluripati, J.B. (2011). A Bayesian framework for model calibration, comparison and analysis: application to four models for the biogeochemistry of a Norway spruce forest. Agriculture and Forest Meteorology 151: 1609-1621.
Thorsen, S.M., Roer, A.-G. & Van Oijen, M. (2010). Modelling the dynamics of snow cover, soil frost and surface ice in Norwegian grasslands. Polar Research 29: 110-126.