Novel Aspects
- A framework for model benchmarking is outlined
- Defining limits of acceptability for models while allowing for data
uncertainties is emphasised
- If all models are rejected then it should instigate a learning process
that will improve uncerstanding