Storing ADCP data as CF 1.6 timeSeriesProfile

This is a summary of many of the important details in how our CF-1.6 data generated by collect.py meets the convention

Global attributes added:

• Conventions: CF-1.6
• standard_name_vocabulary: CF-1.6
• FeatureType:timeSeriesProfile
• geospatial bounds
• Time coverage
• Title
• Project summary
• Keywords
• Keywords_vocabulary:

other components

• Variables with CF-names have a ‘standard_name’ attribute containing the cf name- example: u_1205 gets ‘sea_water_eastward_velocity’
• A ‘z’ coordinate replaces our old ‘depth’ coordinate variable
• The dimensions of the variables need to be (time, z) or (time)
• Both types may be in a timeSeriesProfile file
• To map the ‘height’ variable to the ‘z’, coordinate, variables need to have the attribute:
• coordinates = time height latitude longitude
• If present, each variable’s valid_range attribute must fit the data. If the range is too restrictive the all the data will be set to NaN. The valid_range attribute will be removed, and the minimum and maximum attributes will be renamed to ‘actual_min’ and ‘actual_max’, enabling the same functionality less restrictively.

More details: The EPIC depth coordinate is replaced by z, and a height variable In the CF-1.6 compliant files, there's a "z" coordinate (no z variable) and a height variable dimensioned just by "z". The secret to getting the CF client to map height into 'z' is the attribute 'coordinates = time height lat lon'.

The number of coordinates do not have to match the number of dimensions in HURRIRENE_BB/9141wh-a.nc, u_1205 has dimensions (time, z), and these (and more) attributes:

coordinates: time height latitude longitude
standard_name: eastward_sea_water_velocity

In theory a client that can plot CF will map ‘height’ into Z before plotting. ncBrowse doesn’t know how to use the coordinates attribute so the y axis is 0-n, not the actual depth values.

Single depth variables (like P_1294), are presumed to be at the elevation of the variable with the attribute ‘standard_name’: 'surface_altitude' In our data, sensor_depth has standard_name = "surface_altitude" ;, so CF assigns the elevation in sensor_depth to P_1294. NcBrowse displays the single depth variables correctly.