NOTICE: This version of the NSF Unidata web site (archive.unidata.ucar.edu) is no longer being updated.
Current content can be found at unidata.ucar.edu.
To learn about what's going on, see About the Archive Site.
NOTICE: This version of the NSF Unidata web site (archive.unidata.ucar.edu) is no longer being updated.
Current content can be found at unidata.ucar.edu.
To learn about what's going on, see About the Archive Site.
Hi Rich: Rich Signell wrote:
Wouldnt each of these just be a subset of "best", discarding the first n points?TDS Folks, The Forecast Model Run Collection feature of the THREDDS Data Server is very cool, but of course now that we've been using it, we have some ideas about how it could be even cooler and we are wondering what the rest of the community thinks. Currently the FRMC virtual dataset we use all the time is the "best time series" dataset, which yeilds a continuous time series as best it can, starting from the analysis time of each forecast, taking only those values up to the analysis time of the next forecast. Starting from the analysis time (tau=0) is not always "best", however, because some models initialize from coarser grid models or ingest data in a way that makes it desirable to wait until the dynamics have developed a bit. So for people using these models, their "best" time series might start at tau=6, 12 or even 24 hours.
So how about having the FMRC create a variety of different continuous time series, one for each tau in an individual forecast, and call them best_0hour, best_3hour, best_6hour (or tau_0hour, tau_6hour or something similar). In fact, if folks had access to all these different time series, it would be easy for them to compare to observed time series to determine which offset time series really WAS best! Do others agree that these additional continuous time series would be useful? Are there other virtual datasets that would also be useful to have FMRC automatically generate? Are their datasets that FMRC currently generates that are not useful? (For example, we have never used anything other than the best time series dataset, and we can't figure out how the "constant forecast offset" and "constant forecast date" datasets might be useful)
Steve Hankin might shed some light on why those are useful.
thredds archives: