Open ktindiana opened 9 months ago
Is this still relevant? Doesn't sphinx figure out what is possible to compute and what is not by inspecting the forecasts?
If a forecast field is present, SPHINX calculates all metrics for that forecast field. Depending on the model approach, only a subset of those metrics might actually make sense (e.g. only the comparison of the maximum value of the peak flux forecasted to the observed peak flux as opposed to comparing all peak flux forecasts to the observed peak flux).
The idea here is to output only the metrics identified as relevant to the model rather than dumping out everything and expecting users to ignore half of it.
Discussed in meeting: a filter on the reports seems to be sufficient. @lukestegeman could take this one?
Not all metrics apply to all models. With broader participation in the validation challenges and the start of cross-model comparisons, it is becoming worthwhile to make the code calculate ONLY the metrics relevant to a given model to avoid confusion or inappropriate comparisons.
A config file can be created that indicates which metrics should be calculated for each model (specified by short name or string within the short name). The config file must be accessible to the user in the base directory.