Hi Greg, I was able to fix your example file using NcML (attached to this message). However, there were definitely some hacks: 1. The CDM requires that when there are "station" and "time" dimensions, they should appear in the following order when used to dimension variables: "station time". This is not a CF requirement--merely an (undocumented) requirement of our own implementation. I believe we have a JIRA ticket in place to address this in the future. Your original dataset has the order "time station". I was able to fix this in the NcML using 'shape="station time"', however THAT'S NOT GENERALLY POSSIBLE. I was only able to do it in this instance because your dimension lengths were "time=7, station=1", so I could swap the dims without needing to rearrange the data. 2. When using the Orthogonal multidimensional array representation of time series [1], the latitude and longitude variables are expected to be dimensioned by "station" only, not "time station" or even "station time". I noticed that all your latitudes are the same and all your longitudes are the same, so I just dropped the "time" dimension from their shapes in the NcML. FYI, the easiest way to check whether a DSG dataset is CF-compliant is to open it in toolsUI [2]. Go to the FeatureTypes->PointFeature tab. If you can open the dataset there and the rows in the tables look okay, you should be good to go. Cheers, Christian [1] http://cfconventions.org/Data/cf-conventions/cf-conventions-1.6/build/cf-conventions.html#Example%20H2.1.1 [2] ftp://ftp.unidata.ucar.edu/pub/netcdf-java/v4.6/toolsUI-4.6.jar Ticket Details =================== Ticket ID: XIU-305285 Department: Support THREDDS Priority: Normal Status: Closed
Attachment:
spec.55020.2015051806.ncml
Description: Binary data