Issue with time variable for Monte Carlo "scenarios"

[danielbwright]

Hey everyone! New poster here.

I am the creator of a software, which, among other things, generates really large numbers of gridded precipitation “storm scenarios” in netcdf format, which can then be used for things like monte carlo simulation to estimate 100-year floods. I have a question I am hoping someone could weigh in on. The question has real-world significance, because I am working with the US Army Corps of Engineers Hydrologic Engineering Center to figure out how to get output from my RainyDay software into their flood simulation software.

Anyway, here’s the issue:
Because the software generates large numbers (think 10,000+) of rainstorm scenarios, each one has a distinct time (e.g. 72 hours of hourly precipitation), as well as a 2D (lat/lon) precipitation field for each time period. Of course it isn’t ideal to create 10,000+ separate netcdf files, one for each scenario. So instead I can create one or several files, which contain multiple storms. Consider this example: Say a file contains N=500 storms, and each storm is comprised of 72 hours of hourly gridded data, but each storm has a different 72 hour period. My software currently writes the a 2-dimensional ‘time’ variable, specifically of size 500x72. This can be done, but as far as I can tell this violates CF standards.

So the question is, can I easily create a file that meets CF standards, but that can contain a large number of gridded “scenarios” each with a different time?

[sethmcg]

This sounds like it's analogous to a series of forecasts at different lead times, which is covered by section 6.2. The main thing is that coordinate variables have to have distinct names, so you can have multiple dimensions that have a meaning of time, as long as they are named differently. So you could call them something like "time" for your 500 scenario times and "event_time" for your 72 hours and that would be CF-compliant.