Using a Bayesian Belief Network for classifying valuation methods

By: David Barton, Anita Bayer, Diana Tuomasjukka, Genevieve Patenaude, James Paterson, Martin Karlsen and Anders L Madsen
November 4, 2016


Bayesian Belief Networks (BBNs):


Oppla is an online platform offering advice on selection of ecosystem service appraisal methods. This note demonstrates the use of a Bayesian belief network (BBN) as an expert system to support method selection.


The BBN is populated with method characteristics collected from ecosystem service assessment and valuation practitioners initially in OpenNESS. This was then extended to encompass also Operas methods for the purposes of Oppla. The templates build on previous methodological expertise which has been shared and tested with case studies during several training occasions organized within OpenNESS (January 2014: MCDA, April 2014: non-monetary methods, October 2014: monetary methods and mapping & modelling tools). The templates also underwent an internal reviewing process. Templates were commented by two reviewers, one having an in-depth knowledge of the specific methodological approach and another one with expertise in other methodological fields, to collect both insider and outsider views to the templates. Templates for monetary methods were reviewed by only one expert of the field. Based on reviewers' comments the templates were checked and revised to increase clarity and coherence.

Table 1: Data matrix for OpenNESS and Opera's ecosystem service method classification

Methods were classified according to requirements and value types they addressed. See method templates below for further details.
Download the Excel sheet with methods characteristics here


The BBN was structured as a classification model where the method selected is a parent node of the classification criteria (Table 1). There are 26 method characteristics.

The network can be used in two ways.

Figure 1: Classification BBN for valuation method selection

Decision Context








OPERA methods

OpenNESS methods


Updating the method characteristics in the model

The correlations of the methods classification model have been quantified using a CSV file corresponding to a Data Matrix similar to the one shown in Figure 1.

It is possible to change the methods characteristics using your own Data Matrix. The Data Matrix has to follow a certain format and should be uploaded as a CSV file.

The buttion below can be used to change the classification represented by the model by uploading a CSV file.

To ease the preparation of the CSV file three templates are provided (example 1, example 2 and example 3 listed below). To use one of the templates, you download the file to local disk, make the required changes and upload the file using the button above. It is important that only the values are changed (there is no test for types, spelling errors, etc). Also, if the file is not loaded correctly, then it may be necessary to use the Excel function Text to Columns under the Data tab.

Download example 1: template CSV file. Download example 2: template CSV file Download example 3 (empty file): template CSV file

Example 1 corresponds to the default version of the model created when the page is load, example 2 is created from example 1 by changing a ones to zeros and vice versa, and in example all values are deleted.

Notice that it is not possible to update the methods or the characteristics of the model. It is only possible to change the values, i.e., how the methods are classified according to requirements.

Here is the XLS file corresponding to the example 1 CSV file Download example 1: template XLS file. It is important to save this file as CSV before uploading to the web site.

The below button can be used to download the classification model to local disk such that it can be open in the HUGIN Graphical User Interface.


Kelemen, E., D. N. Barton, S. Jacobs, B. Martin-Lopez, H. Saarikoski, M. Termasen, G. Bela, L. Braat, R. Demeyer, M. Garcia-Llorente, E. Gomez-Baggethun, J. Hauck, H. Keune, S. Luque, I. Palomo, G. Pataki, P. Tenerilli, and F. Turkelboom. 2015. Preliminary guidelines for integrated assessment and valuation of ecosystem services in specific policy contexts. EU FP7 OpenNESS Project Deliverable 4.3., European Commission FP7.

Kjærulff, U.B and Madsen, A.L. (2013): Bayesian Networks and Influence Diagrams: A Guide to Construction and Analysis. Second Edition. Springer.

Haines-Young, R., Barton, D.N., Smith, R. and A. Madsen (2013): Bayesian Belief Networks, a cross-cutting methodology in OpenNESS: Briefing note version 2.2, 26 September 2013.

Contact information

For further details contact: David Barton @ NINA.