Dear Martin > Take a global model output of temperature, for example. How do you > describe temperature differences between this model and another one? The > resulting quantity is still an "air_temperature" (OK, actually > "air_temperature_difference") with units of "K". Yet, it would be nice to > know that this field is a result of differencing two models. "Difference" > could be accomodated relatively easily with the standard_name modifier (but > how do you describe what has been differenced?). More complicated operations, > such as normalized mean bias (X-Y/(X+Y)) will at some point be impossible to > maintain through standard_name modifiers, I believe.
Yes, I agree, generalised description like this is a tricky issue. Standard name modifiers really only work for describing things done with a single quantity. Cell methods are more powerful because they describe operations on coordinates too. Although there are cases where it seems like a small step to extend this to operations which combine variables, I think that could get very complicated in the end. In general, CF metadata describes what a quantity *is* and not how it was calculated from other quantities. It would probably help if we focussed on some real use-cases where it is essential to provide *systematic* metadata i.e. which can be processed by programs. It is always possible to provide descriptive metadata, useful to humans, in non-standardised attributes such as long_name and history, and this can explain how the quantity is obtained. Best wishes Jonathan _______________________________________________ CF-metadata mailing list CF-metadata@cgd.ucar.edu http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata