E3MAIP

Principles

E3MAIP has adopted philosophical principles to assess model quality. These include:

1. The model must include a clear statement of what system is being modelled.

2. The model must have a clear description of the system state and how the state should be interpreted.

3. The model must have a set of clearly stated governing mathematical equations.

4. The model must be clear as to what aspects of the system are being modelled by these equations, and under what circumstances they apply.

5. The equations must have sensible behaviour in simple, known, cases.

6. These equations must be coded up in a transparent, maintainable, manner, both for explaining operations and training new users.

7. The model must be testable against a well defined set of conditions on which data are available.

8. Where there are "free parameters" in the model, it is preferable if these parameters can be developed from data independent of the data used to fit the model.

9. The computer codebase must be structured to allow changes as new information and methodologies become available.

The E3 Model Assessment and Intercomparison Project

examines the epistemological foundations of E3

modelling at 4CMR and in Partner research centres to determine the reliability of these models and develop a path forward to improvement so these models can better inform policy. The goal is to create a community of E3 researchers world-wide that systematically explores the underlying theories, computational structure, empirical foundations and statistical methods of these models. This approach has proven highly successful in improving the reliability of similar models in atmospheric science, climate change and environmental risk assessment.

Our assessment methods include:

Logic tree analysis: to understand the strength of relationships between evidence, modes of reasoning and conclusions.
Uncertainty analysis: to qualitatively and quantitatively characterise the degree of uncertainty in models and their "tolerance" in decisions.
Sensitivity analysis: to qualitatively and quantitatively characterise the contribution of parts of a model to overall uncertainty in predictions, and to guide future research to most effectively reduce uncertainty.
Rationality assessment: to understand how well specific models can meet the needs of decision-makers for locating (as Russell said) "Apt means for the highest ends".

We recognise that different research groups grow up under different theories related to E3 processes, different ways of coding these models, different bodies of data for parameterising the models, etc. Policy makers routinely ask why one group produces results different to another. This has led in many areas of atmospheric science to the formation of Communities of modellers, all working together to compare their models, diagnose the differences, and move towards a consensus model over time. We have adopted the same approach here in regards to E3 models, joining with E3 modelling groups around the world to compare our results, compare the models themselves, diagnose differences, and work together to improve all of our models.

If you work in the area of E3 modelling and would like to join this global E3 Model Assessment and Intercomparison Project, please Download the E3MAIP description or Contact Us...