None of the arguments against uncertainty analysis are, in the end, tenable. However, neither is there any readily available guidance about how to do uncertainty analysis in hydrologic and hydraulic modelling. There are, instead, a number of competing methods, within different philosophical frameworks, that might be more or less appropriate in different situations with different model dimensionalities and sources of uncertainty. This makes it difficult for a modeller to choose a method, and for the end-user of modelling results to interpret the resulting uncertainties. We deliberately do not list one key point of uncertainty analysis: cost and resources. Uncertainty estimation is currently considered to be an added component to an analysis, at extra cost. It should, however, be an intrinsic and expected part of any modelling exercise.

Undoubtedly, mature guidance on the use of different methods will develop over time, but it is not too early to suggest that modelling studies might follow a Code of Practice that makes uncertainty analysis an integral part of the modelling process. From the modeller’s point of view, a Code of Practice would (mostly) represent an extension of normal practices for the quality assurance and version tracking of computer codes, with similar documentation requirements. Such a Code of Practice would need to address the following issues:

Taking Account of Model Context

Taking Account of Uncertainty in Model Choice (and implementation details)

Taking Account of Uncertainty in Model Drivers (inputs and boundary conditions)

Taking Account of Prior Uncertainty in Model Factors

Taking account of Dependency in Model Factor

Choice of Uncertainty Estimation methodology

Taking Account of Uncertainty in Observations used in Model Calibration/Conditioning

Taking Account of Uncertainty in Predicting the Future

Communicating Results to Users

If our science is to be meaningful, we should aim to communicate the limitations of the predictions we make in ways that are useful to the wider community. This, in itself, cannot be divorced from the wider socio-political context, as has been demonstrated by the IPCC presentations of uncertainty in future climate predictions, but if the community can start to discuss just what form a Code of Practice should take, these issues should at least begin to be addressed. We suggest that this is now timely. As computer power continues to increase, and remote sensing and the networking of pervasive environmental sensors start to make it easier to obtain conditioning information for new sites, some consensus about good practice in uncertainty estimation and constraint, as routine practice in hydrological science, would undoubtedly be a good thing!

On a practical level, most of this audience will be elected or appointed officials and will have minimal or no training in modeling or risk management. Given the nature of their involvement in the risk management process, prudence requires three assumptions be made:

• Most will want something on the order of “Six Easy Steps to Salvation,” which are easily digestible both by the public and the other decision makers involved.

• Most will not be concerned with disclaimers regarding the fact that uncertainty exists in the world.

• Most will happily point to those responsible for developing the modeling methodology and attempt to deflect blame for a failure to the uncertainty of the model.

Whether these assumptions will, in fact, play a role in the management of an actual event is as much an uncertainty as any; however, the likelihood of any one or more of these assumptions occurring is great enough that they must be noted.

The following documentation should be maintained throughout the research and modeling process:

* Documentation as to the modeling method, the reasons for choosing this modeling method including but not limited to direction provided by the person or organization which requires the preparation of the model

* Documentation as to the choices made in the design of the model and the choices made in the modeling process

* Documentation as to the assumptions used in the design of the model and the modeling process.

* Documentation as to the inclusion or omission of data or data types (most important)

* Documentation as to the method of presentation of the results of the modeling effort.

* Documentation as to the individuals involved in the process.

* Documentation regarding changes dictated by authority or circumstances beyond the scope of the document

Such other documentation as may be deemed necessary to justify the conclusions made in the risk assessment.

This documentation should be presented together with the recommendations as an appendix to the recommendations with appropriate notes on the methods and results of the modeling effort.